A SPS Comparison Graphs

Transcription

1 A SPS Comparison Graphs This section of the specification document provides either an example of the default graph for each case or instructions on how to generate such a graph external to the program for the SPS Comparison graph sets. These graphs are done using the agency classification scheme. There are no yearly or multi-year graphs in this graph group. Figure A.1-1 SPS graph availability (1.7) The first section of this document contains the tables that identify the various purge codes and their definitions for these graphs. If a code applies to a specific vehicle class, that is included in the description. Some codes may have more than one reason to be used for a graph type or may apply to multiple types of graphs with different reasons. In checking unusual patterns, pay attention to the dates relative to federal and local holidays, severe weight conditions and work stoppages. Days/weeks with marginal conditions for acceptance should be checked for volume, distribution and loading where applicable before finalizing a purge recommendation. Throughout this document the terminology 4-card and 7-card appears. 4-card is used for classification data, hourly records of counts by vehicle class. 7-card is used for weight data,

2 individual vehicle records with axle weights, axle spacings and gross vehicle weight. The data may have been submitted in either U.S. customary or S.I. units. The database and all graphs are in U.S. customary units. A.1 Graph intervals Graph intervals define the amount of data aggregated for a single graph page. For SPS comparison graphs the intervals are weekly and monthly. No other periods are provided because the data for SPS-1, -2, -5 and -6 graphs is supposed to be reviewed at least every other week with a comparison data set as the basis for determining the reasonableness of the data. The various graph types, other than 4-card vs. 7-card have different minimum data requirements to exist. For vehicle distribution graphs the data set used to create an individual line or bar must have 50 vehicles. For GVW and axle distribution graphs the data set must have 100 vehicles or axles as applicable for an individual line to appear. The minimum data requirement may cause days or weeks to vanish from a graph series even when vehicles have been observed. A.1.1 Weekly Weekly graphs are used in review for SPS-1, -2, -5 & -6 sites that meet the following criteria: Comparison data based on a Sheet 16 that includes data quality values At least 125 vehicles per day in a class to be reviewed At least 125 vehicles per day in a group of vehicles to be reviewed Continuous data is expected for the data type They are used for diagnostic purposes without having to meet the above criteria. There are no default weekly graphs. A.1.2 Monthly These graphs are done in agency classification. There are four default graphs defined, one for each graph type. The defaults provide basic review capability for SPS WIM data. Monthly graphs are used to review sites that have the following characteristics: More than 125 vehicles in a class/group per month More than 125 vehicles in a class/group per week for Month but Week alternatives At least one type of continuous data Monthly graphs may also be used for diagnostics when pattern problems or expected value issues from Weekly graphs need to be investigated. A.2 Purge Code Tables The tables in this section provide a list of codes and some basic decision criteria by graph type. Purge decisions are not absolute but must consider the type of comparison data being used. If the comparison set if from a Sheet 16 validation then the reasonableness of that data and the amount of time elapsed must be considered. If the review data is from a surrogate data set, somewhat greater variation from expectations is expected. Additionally when there are several year between the review data and the comparison data, particularly when the review data is for a

3 period prior to a comparison the potential for site changes must be considered. These include equipment relocation, replacement, modification, validation or calibration. Not all of those activities may have been reported to LTPP for inclusion in SITE_EQUIPMENT_INFO. Codes were added in version to provide better descriptive information in the TRAFFIC_PURGES table. They simplify entering comments in that table. In the tables that follow some purge identification criteria have associated limiting values. This means that unless the data element being checked meets that condition, that particular criteria should not be used for purge identification. This eliminates purge conditions that don t make sense due to variability in small sample sizes. The limiting value relates to a site characteristic such as an average or minimum value derived from a valid data set. The data is valid because of either consensus review or a validation or calibration. The limiting value criterion is the same whether the element being reviewed is a date (day), day of week, single class or group of classes. When applying the purge selection criterion it is not necessary to compute the difference. The differences can be assessed by inspection. Generally, determining what a percentage means in terms of increments on the vertical (y-) axis is sufficient. A valid monthly graph with the limiting values marked on it for comparison is another way to apply the criteria. In this case, attention must be paid to the scale of the vertical axis. New comparison graphs for volume in particular will be needed as volumes increase over time. The values generally needed for 4-card vs. 7-card review ( Table A.2-1) are: Average weekly volume, average daily volume and minimum daily volume. The values may be computed or obtained by inspection of a valid monthly graph. They are different for each class or group to be evaluated. They are defined as follows: Average weekly volume Sum of the averages for each day of the week Daily average volume For sites with more than 1000 per week averages are computed on a by day of week basis and rounded to the nearest 50 For sites with 250 to 1000 per week averages are computed for weekdays as a group (the average of the weekday averages) and weekend days as a group (the average of the weekend averages) and rounded to the nearest 10 For sites with less than 250 per week the average is the average of the day of the week averages and rounded to the nearest 5 Minimum volume percent of the average volume for sites with 100 or more per week 60 percent of average value for sites with less than 100 per week Values under 100 rounded down to the nearest 5 Values from 100 to 1000 rounded down to the nearest 10 Values over 1000 rounded down to the nearest 50 The process is discussed in detail in SPS-1, -2, -5 and -6 Traffic Data Processing. Included in the discussion is how to pick the groups for graphic evaluation. Table A.2-1 Purge Codes for 4-card vs. 7-card Review

4 Code Description Limiting Value Selection Criteria 64 Zero data Minimum volume at least 10 Class or group with zero volume (or effectively zero) given the expected minimum. 65 Zero daily volume For reviews including all classes only, total volume for day is zero and it is not attributed to construction or weather 68 Daily 7-card volume significantly greater than 4-card volume 69 Daily 4-card volume significantly greater than 7-card volume 70 Lower volumes than expected, possible sensor problems Average value at least 50 Average value at least 50 Minimum value at least Atypical pattern Average volume at least 250 per week The 7-card volume is more than ten percent larger than the corresponding 7- card value And 1) Review is for a single vehicle class, OR 2) Review is for a group where the same classification algorithm is used for both data types. The 4-card volume is more than ten percent larger than the corresponding 7- card value And 1) Review is for a single vehicle class, OR 2) Review is for a group where the same classification algorithm is used for both data types. Volume below the minimum. See 136, 137, 138, 139 1) Shift in location of a graph line minimum or maximum where the choice of minimum or maximum is obvious (ten percent or greater below or above the other candidate days.) (Should use 101 instead.) 2) Pattern that has demonstrated a relatively constant volume retains its shape (low to high to low) but has a distinct increase or decrease in volume over time. (More than 5 percent a year requires a documented explanation.) 3) The shape changes or a distinct shape appears where the two lines have previously coincided. The disappearance of a shape (not just a

5 Code Description Limiting Value Selection Criteria little bit of differentiation between the lines) should be investigated. 100 Higher volume than expected 101 Change in day of week pattern 136 Shift in location of maximum volume/ percentage 136 Shift in location of minimum volume/ percentage 137 Disappearance of volume difference 138 Larger than expected volume difference Average volume at least 50 Average volume at least 250 per week Average volume at least 250 per week Average volume at least 250 per week Average value for smaller volume at least 50 Average value for smaller volume at least 50 4) For a new site without validation data a pattern that does not conform to typical expectations i.e. weekdays having larger volumes than weekends, classification volumes being greater than or equal to weight volumes. Volume above the maximum expected for any characteristic 1) The day of week the minimum falls on changes day of week. The difference between the minimum and the next larger volume should be at least 5 percent. (Use 136 for by date graphs) 2) The day of week the maximum falls on changes day of week. The difference between the maximum and the next smaller volume should be at least 5 percent. (Use 135 for by date graphs) 3) A one peak weekly volume cycle becomes a two peak cycle or vice versa. Location of maximum in a recurring pattern shifts by more than one day in the week for 3 or more consecutive weeks. Location of minimum in a recurring pattern shifts by more than one day in the week for 3 or more consecutive weeks. A recurring gap pattern changes to lines on top of each other An expected volume difference exists of any magnitude and the observed difference is at least 110 percent of the expected difference. Both data types must use a consistent by not necessarily the same definition for a vehicle class or group of vehicles.

6 The values needed for GVW review (Table A.2-2) are: tare weight, legal maximum weight, average weekly volume, the volume represented by the curve and the volume represented by the graph. The average weekly volume is same as that for the 4-card vs. 7-card review. To determine the average number in a curve use the following Weekly, by Week; Monthly, by Month use volume in header Weekly, by Day of Week Divide the volume in the header by the number of days present plus 1. (The plus 1 is to offset the potential for missing days at low volume sites and the tendency to have sharp differences between weekday and weekend volumes particularly for Class 5 and Class 9 vehicles.) Monthly, by Week divide by number of weeks shown plus 1 not to exceed 5 (or the number of weeks if shown by week) While expected values for 5-axle tractor semi-trailers are incorporated in the software, expected values for other classes that are needed to evaluate site loading are not. For all other classes unloaded is within 15 percent of the typical tare weight of the class. The tare weight is the unloaded weight of the truck as defined by a manufacturer of that type. The typical tare weight should be for the predominant vehicle in the class if multiple configurations are assigned the same classification. An example is buses where both two axle and three axle versions exist. Loaded is between 80 percent and 100 percent of the legal maximum for the class. Where multiple vehicle types are included in a class, multiple peak values may need to be established. Table A.2-2 Purge Codes for GVW Review Code Description Limiting Value Selection Criteria 72 Atypical distribution (See 149, 150, 151) 1) A pattern variance from the expected at the site not described by one or more other codes 73 Over-calibrated 1000 or more per week - 1 bin 500 or fewer per week 2 bins 73 Over-calibrated 1000 or more per week 10 percent of legal maximum to nearest bin 500 or fewer per week 20 percent of legal maximum 2) A pattern for a new site without validation information that does not match the expected typical pattern for the class Class 9s Unloaded and loaded peaks shifted equally to the right AND outside of their expected bounds. Other classes Curve has its reference shape and is shifted to the right from the expected location.

7 Code Description Limiting Value Selection Criteria 74 Under calibrated 1000 or more per week - 1 bin 500 or fewer per week 2 bins 75 Large percentage of tractor trailers over 80 kips 76 Large percentage of tractor trailers under 12 kips 104 GVW distribution inconsistent with history 1000 or more per week 10 percent of legal maximum to nearest bin 500 or fewer per week 20 percent of legal maximum Class 9s only vehicles or more in curve being assessed Class 9s only vehicles or more in curve being assessed Class 9s Unloaded and loaded peaks shifted equally to the left AND outside of their expected bounds. Other classes Curve has its reference shape and is shifted to left from the expected location. 1) The proportion of the curve to the right of the 80 kip bin mark is greater than five percent. 2) The value for the distribution is non-zero beyond 100 kips. The proportion of the curve to the left of the 12 kip mark is greater than two percent. 1) Class 9s only - Either of the solid bars representing the current data is shifted one bin from the center of the two thin bars. 2) Other classes A dominant peak is shifted more than one bin from the expected curve in either direction. 3) The shape of a curve is different from the expected curve where Different is o A uni-modal (single peak) distribution becoming bi-modal (two-peaks), (Use 149) o A distinctly loaded or unloaded site shifting to a partially loaded site (and remaining uni-modal) (Use 150), o Doubling of the size of either the tail below the lower bound of the unloaded peak or above the upper bound of the loaded peak. (Use 152 or 151 respectively)

8 Code Description Limiting Value Selection Criteria This code is preferred to 72 (atypical distribution) because it is a more explicit description of the irregularity. 105 Large shift in loaded peak 106 Large shift in unloaded peak 107 GVW peaks outside expected limits 108 Large percentage of underweight trucks 250 vehicles or more in curve being assessed 250 vehicles or more in curve being assessed Class 9s- 250 vehicles or more in curve being assessed 500 vehicles or more in curve being assessed Class 9s - The solid bar representing the loaded peak has shifted two or more bins (is on or outside the pair of thin lines for the loaded peak.) Other classes - The loaded peak has shifted two or more bins or by 15 percent of the legal maximum (rounded to the nearest bin) from its expected position. Class 9s - The solid bar representing the unloaded peak has shifted two or more bins (is on or outside the pair of thin lines for the unloaded peak.) Other classes - The unloaded peak has shifted two or more bins or by 15 percent of the legal maximum (rounded to the nearest bin) from its expected position. Both solid bars are outside the thin lines that represent the two standard deviation limits for the weight of the peaks. 1) More than five percent of the trucks are less than ninety percent of the minimum tare weight for the population, OR, 109 Large percentage of overweight trucks 500 vehicles or more in curve being assessed 2) There is a non-zero value of the distribution for bins that contain weights less than ninety percent of the expected tare weight for the class. 1) More than five percent of the trucks exceed ten percent of the legal maximum for the class, OR, 2) There is a non-zero value for the frequency distribution in bins that contain weights that are ten percent or more than the legal maximum. 149 Change in number of 1000 vehicles or 1) One peak changing to two peaks

9 Code Description Limiting Value Selection Criteria modes for GWV more on graph distribution 2) Two peaks changing to one peak 3) Two peaks going to three peaks where the third peak is at least a third of the next higher peak 150 Change in dominant loading type 151 Large change in loaded tail of GVW distribution 152 Large change in unloaded tail of GVW distribution 500 vehicles or more on graph 250 vehicles or more in curve being assessed 250 vehicles or more in curve being assessed 4) Three peaks going to two peaks 1) Loaded site becomes unloaded 2) Unloaded site becomes loaded 3) Site not clearly loaded or unloaded (Less than five percent difference between peaks) becomes one or the other 4) Partially loaded site (single peak between 40 and 60 kips for Class 9s) becomes loaded or unloaded 1) Loaded tail beyond 110 percent of legal maximum goes from zero to non zero 2) Loaded tail more than two bins from loaded peak doubles in magnitude 3) Loaded tail beyond 110 percent of legal maximum goes to zero 1) Unloaded tail left of 90 percent of tare weight goes becomes nonzero 2) Unloaded tail more than two bins from unloaded peak doubles in magnitude The values needed for vehicle distribution review (Table A.2-3) are: average volume (group), average volume (class), minimum volume and maximum volume. There are several possible average volumes depending on the graph being used. Reference graphs that do not incorporate comparison values can be prepared with the critical values marked. If the values are computed the following guidelines should be used. The graphs are weekly and monthly.

10 Weekly - Average volume is same as that for the 4-card vs. 7-card review Monthly - Average volume is 4.3 times the average weekly volume Rounding for averages is as follows: For sites with more than 1000 round to the nearest 50 For sites with 250 to 1000 round to the nearest 10 For sites with less than 250 round to the nearest 5 Minimum volume is percent of the average volume. Values under 100 rounded down to the nearest 5 Values from 100 to 1000 rounded down to the nearest 10 Values over 1000 rounded down to the nearest 50 Maximum volume is 125 percent of the average volume. Values under 100 rounded up to the nearest 5 Values from 100 to 1000 rounded up to the nearest 10 Values over 1000 rounded up to the nearest 50 Table A.2-3 Purge Codes for Vehicle Distribution Review Code Description Limiting value Selection Criteria 72 Atypical pattern 1) Change in predominant truck types either through additions or removal 102 Too many unclassifieds Average weekly volume > 100 2) Large change in volume either increase or decrease 1) The percentage of unclassified vehicles increase by more than 5 percent. 2) Existence of unclassified vehicles for several weeks where none have been seen before. For classification data, a change of input record format to include unknowns and or unclassifieds is not grounds for making this a purge list entry in the absence of other data irrationality. 3) More than ten percent unclassifieds in total. This must be considered in context with the population being evaluated since the same value is used for unclassified in all groups.

11 Code Description Limiting value Selection Criteria 103 Outside +/-10 percent range Average volume > 250 1) Any class that is twenty percent or more of the heavy truck population and has the top of its bar either above the top of the comparison block or below the bottom of the comparison block. 2) Any class that is about ten percent of the heavy truck population and has an expected weekly volume of more than 150 that effectively disappears from the distribution for several weeks. Note that such an occurrence is entirely possible for low volume sites. 135 Shift in location of maximum volume/ percentage 136 Shift in location of minimum volume/ percentage 137 Disappearance of volume difference 138 Larger than expected volume difference 139 Missing expected vehicle classes 140 Volume distribution inconsistent with history 141 Percentage distribution inconsistent with history Average volume > 250 Average volume > 250 Average volume > 250 Average volume > 250 Average volume for missing class > 50 3) A class other than Class 9 that was not expected to be at least ten percent, becomes fifteen percent or greater (including unclassifieds). The dominant class changes or the percentage in a dominant class is two-thirds or less of its typical value A class (not the unclassifieds) with less than five percent of the population exceeds ten percent of the population A pattern of 4-card volumes being greater or less than 7-card volumes becomes same volume for both 1) Volume difference exceeds ten percent when there is no expected difference set for the site. 2) Volume difference changes by more than twenty five percent A class disappears from the distribution for three or more consecutive weeks Relative volumes do not match the expected pattern Relative percentages do not match the expected pattern

12 Code Description Limiting value Selection Criteria card/7-card difference too large Average weekly volume > 100 Difference in volumes, excluding the unclassifieds/unknowns (Class 15 or 20) is card distribution inconsistent with 7- card card distribution inconsistent with 4- card 145 Volume larger than the expected maximum 146 Volume less than the expected minimum Maximum volume > 300 Minimum volume < 175 greater than ten percent When the 7-card distribution has been defined as valid for the site and the 4-card distribution using the same classification algorithm does not match it. 4-card data is data to be purged When the 4-card distribution has been defined as valid for the site and the 7-card distribution using the same classification algorithm does not match it. 7-card data is data to be purged Volume exceeds expected maximum by more than ten percent of the expected maximum Volume is below the expected minimum by more than ten percent of the minimum value The values needed for axle review (Table A.2-4) are: legal maximum weight by axle group, typical class minimum axle weight, average weekly volume, the volume represented by the curve and the volume on the graph. The average weekly volume is same as that for the 4-card vs. 7- card review. To determine the average number in a curve use the following Weekly, by Week; Monthly, by Month use volume in header Weekly, by Day of Week Divide the volume in the header by the number of days present plus 1. (The plus 1 is to offset the potential for missing days at low volume sites and the tendency to have sharp differences between weekday and weekend volumes particularly for Class 5 and Class 9 vehicles.) Monthly, by Week divide by 5 (or the number of weeks if shown by week) Table A.2-4 Purge Codes for Axle Distribution Review Code Description Limiting Values Selection Criteria 72 Atypical distribution See 153, 154, 155, 156 1) A pattern variance from the expected at the site not described by one or more other codes 2) A pattern for a new site without validation information that does not match the expected typical pattern for the class

13 Code Description Limiting Values Selection Criteria 73 Over-calibrated 1000 or more per week - 1 bin 500 or fewer per week 2 bins 1000 or more per week 10 percent of legal maximum to nearest bin 500 or fewer per week 20 percent of legal maximum 74 Under calibrated 1000 or more per week - 1 bin 500 or fewer per week 2 bins 110 Axle distribution inconsistent with history 111 Large percentage of heavy axles 1000 or more per week 10 percent of legal maximum to nearest bin 500 or fewer per week 20 percent of legal maximum 500 or more axles in curve being Class 9s - Both peaks shifted equally to the right AND outside of their expected bounds. Other classes - Dominant peaks shifted equally to the right from their expected peaks. Class 9s - Both peaks shifted equally to the left AND outside of their expected bounds. Other classes - Dominant peaks shifted equally to the left from their expected peaks. 1) Dominant peaks shifted more than one bin from the expected curve in either direction. 2) The shape of a curve is different from the expected curve. Different is o A uni-modal (single peak) distribution becoming bi-modal (two-peaks), (Use 153) o A distinctly loaded or unloaded site shifting to a partially loaded site (and remaining uni-modal) (Use 154), o Doubling of the size of either the tail below the lower bound of the unloaded peak or above the upper bound of the loaded peak. (Use 156 or 155 respectively) This code is preferred to 72 (atypical distribution) because it is a more explicit description of an irregularity. 1) More than five percent of single axles are in bins beyond 20,000

14 Code Description Limiting Values Selection Criteria assessed pounds. 2) The value of the single axle distribution is non-zero past 22,000 pounds. 3) More than five percent of tandem axles are in bins beyond 40,000 pounds. 112 Large percentage of light axles 113 Shifted heavy axle peak 114 Shifted light axle peak 153 Change in number of modes for axle distribution 154 Change in dominant loading type 500 or more axles in curve being assessed 250 axles or more in curve being assessed 250 axles or more in curve being assessed At least 1000 axles on graph At least 500 axles on graph 4) The value for the tandem axle distribution is non-zero past 44,000 pounds. 1) More than five percent of the single axle distribution is less than 6,000 pounds except for Class 5 where the values are 15 percent and 3,000 pounds. 2) More than five percent of the tandem axle distribution is less than 8,000 pounds. The heavy (loaded) peak is more than two bins or 15 percent of the legal maximum (rounded to the nearest bin) from the expected location. The light (unloaded) peak is more than two bins or by 15 percent of the legal maximum (rounded to the nearest bin) from the expected location. 1) One peak changing to two peaks 2) Two peaks changing to one peak 3) Two peaks going to three peaks where the third peak is at least a third of the next higher peak 4) Three peaks going to two peaks 1) Loaded site becomes unloaded 2) Unloaded site becomes loaded 3) Site not clearly loaded or unloaded (Less than five percent difference between peaks) becomes one or the other

15 Code Description Limiting Values Selection Criteria 155 Large change in loaded tail of axle distribution 156 Large change in unloaded tail of axle distribution 250 or more axles in curve being assessed 250 or more axles in curve being assessed 4) Partially loaded site (single peak between 40 and 60 kips for Class 9s) becomes loaded or unloaded 1) Loaded tail beyond 110 percent of legal maximum goes from zero to non zero. 2) Loaded tail more than two bins from loaded peak doubles in magnitude 3) Loaded tail beyond 110 percent of legal maximum goes to zero 1) Unloaded tail left of 90 percent of tare weight goes becomes nonzero 2) Unloaded tail more than two bins from unloaded peak doubles in magnitude

16 A.3 SPS 4-card vs. 7-card Weekly, Week by Date Tables used: DD_CL_CT, DD_WT_CT Type abbreviation: SPS47WD Default: Class 9s There is no comparison data associated with a 4-card vs. 7-card graph. To identify deviations from typical conditions the user will need expected volumes. Day of week patterns will not be obvious due to the date-based nature of the graphs. Comparison weekly graphs will not be particularly helpful since the graph will not always start on the same day of the week. Graphs of the comparison period will however provide an indication of the expected range from high to low and the expected size of the gap between classification and weight data. The expected output of a 4-card versus 7-card comparison varies with the population being compared and the classification algorithm being used. When classification and weight data is created using different schemes (i.e. the weight data is TMG 13-bin and the classification data is length based class (i.e. 6-bins)), 13-bin graphs should be used for 4-card\7-card volume comparisons even at SPS locations. When vehicle classification is done without a weight trigger, comparison evaluations should focus on classes where classification differences due to similar axle spacings and distinctly different weight characteristics are unlikely to occur. When the population is a single class with continuous data from the same or different equipment at the same location, the values should match on a day-to-day basis when both pieces of equipment are working. When the population is a single class with sampled classification data from the same or different equipment at the same location, monthly graphs should be used. When the population is multiple classes and does not include passenger vehicles, the same expectations exist that apply for single class populations. Weekly 4-card vs. 7-card graphs should not be used for vehicles (1-20) unless cars and trucks are weighed OR a comparison period graph is used to evaluate data reasonability. Weekly 4-card vs. 7-card graphs should not be used to evaluate data from the 29 th of the month or later. The purge codes for a 4-card vs. 7-card review are in Table A.2-1, page 3. The graphs for this section are for the default, Class 9, unless otherwise indicated.

17 MN, 0500, West, Lane 1-11/01-07/ Class 9 4-card vs. 7-card Figure A.3-1 SPS 4- vs. 7- card - Weekly, Week by Date Example 1 Error! Reference source not found. is a typical graph for the Class 9 default. Note that the two graph lines are virtually the same. A cyclic pattern appears to exist but the determination of which day of the week belongs with which volume will require a calendar. Because the 1 st day of the graph week will fall on different days of the week for each graph, the range of values, not the shape of the curve is the basis for acceptance/rejection. MN, 0500, West, Lane 1-12/29-31/ Class 9 4-card vs. 7-card Figure A.3-2 SPS 4- vs. 7- card - Weekly, Week by Date Example 2 Figure A.3-2 illustrates why the 5 th week of the month should not be evaluated by this graph. Note the distortion of the horizontal axis. volume range. Purge candidate Do NOT make a purge decision based on this graph for these days (the fifth week) without knowing expected

18 Figure A.3-3 SPS 4- vs. 7-card - Weekly, Week by Date Example 3 (CASE NEEDED) Data in Figure A.3-3 shows a difference at a site with low volumes of Class 9 trucks. The patterns of the 4-card (class) and 7-card (weight) data diverge. Purge candidate - Yes What weight data (lower line), 18 th to 21 st Reason Lower volumes than expected (70) Expected action: Purge low volume weight data days WA, 0200, South, Lane 1-08/15-21/ Class 9 4-card vs. 7-card Figure A.3-4 SPS 4- vs. 7- card - Weekly, Week by Date Example 4 Data in Figure A.3-4 shows a difference at a site with low volumes of Class 9s. The patterns of the 4-card (class) and 7-card (weight) data are very different. Most importantly, the 7- card volumes are zero.. Purge candidate - Yes What Weight data (lower line), entire week 15 th -21 st Reasons Zero data (64) Expected action: Purge zero volume days that are not the result of weather/holidays (Easter) /strikes.

19 Figure A.3-5 SPS 4- vs. 7-card - Weekly, Week by Date - Example 5 (CASE NEEDED) Figure A.3-5 shows a case where the weight data has a larger number of Class 9s than the count data. This is an unexpected condition. Purge candidate - Yes What Weight data (upper line), entire week Reason Daily 7-card volume significantly greater than 4-card volume (68) [More descriptive than higher volume than expected (100) when volume is higher and 4-card data is valid.] Expected action: Before applying purges identify expected values for 4-card and 7-card information. 4-card data may be wrong instead (If so purge candidate is Class data, entire week, with Lower volumes than expected (70)); Check all Vehicles (1-20) 4-card vs. 7-card graph to see if Class 9s are all trucks or possibly include cars (error through version 1.5 of the software.). If class 9s might include cars, reload the data and review again. MI, 0100, South, Lane 1-12/22-28/ Class 9 4-card vs. 7-card Figure A.3-6 SPS 4- vs. 7- card - Weekly, Week by Date Example 6 (One with both 4 and 7 card would be better) Data in Figure A.3-6 shows a site with a known pattern difference based on a comparison period graph. Note that the dates of the drop are Christmas Eve and Christmas. No action required. Retain data. Purge candidate - No

20 WA, 0200, North, Lane 1-04/15-21/ Trucks (4-20) 4-card vs. 7-card Figure A.3-7 SPS 4- vs. 7- card - Weekly, Week by Date Trucks - Example 1 (better one where follows same pattern, not different pattern or leave in but check rasoning, i.e. 18 th ) The group Trucks (4-20) is used in Figure A.3-7 as an example of the 4-card vs. 7- card comparison as it applies to a population. The relative size of the gaps between the data lines should be compared to the sizes in a similar graph for the comparison period. The (in)ability to distinguish between class 3s and Class 5s could contribute to the gap between the lines. Errors in vehicle records leading to omission from the 7-card file would be another reason. From the comparison data set, determine the typical range of errors in collecting 7-card information. This will be a good source for determining the expected size of the gap. Use the vehicle group option whenever shifts in vehicle distribution are observed to see if vehicles are missing from the population as a whole. Purge candidate No

21 WA, 0200, North, Lane 1-04/08-14/ Vehicles (1-20) 4-card vs. 7-card some sites also have weight data for cars. Figure A.3-8 SPS 4- vs. 7- card - Weekly, Week by Date Vehicles - Example 1 Vehicles (Classes 1-20) are used in creating Figure A.3-8. It shows a typical 4- card 7-card graph on a weekly basis where only trucks (classes 4-20) are weighed (lower line). This demonstrates the limitations of the use of vehicles as an evaluation graph. As shown in Figure A.3-9 Purge candidate Unknown Figure A.3-9 SPS 4- vs. 7-card - Weekly, Week by Date Vehicles - Example 2 (CASE NEEDED) The vehicles group (Classes 1-20) is also used in Figure A.3-9. It illustrates a site where the 7- card weight data includes cars. In this case the data for the 5 th would not be immediately rejected. It is possible that the data loaded for that day is missing non-truck vehicles. A purge determination for this case would need to be made on the basis of discrepancies in counts for individual truck classes. Purge candidate Maybe What the 5 th Reason - Lower volumes than expected (70) Expected action: Review graph for trucks or heavy trucks.

22 A.4 SPS 4-card vs. 7-card Monthly, Month by Date (DEFAULT) Tables used: DD_CL_CT, DD_WT_CT Type abbreviation: SPS47MD Default: Class 9s The Month by Date graph is intended to check that volumes are in line with expectations. It also provides information as to whether continuous or sampled data is present. The graph may be useful even when either weight or class data is not present. The expected output of a 4-card versus 7-card comparison varies with the population being compared and the classification algorithm being used. When classification and weight data is created using different schemes (i.e. the weight data is TMG 13-bin and the classification data is length based class (i.e. 6-bins)), 13-bin graphs should be used for 4-card\7-card volume comparisons, even at SPS locations. When vehicle classification is done without a weight trigger, comparison evaluations should focus on classes where classification differences due to similar axle spacings and distinctly different weight characteristics are unlikely to occur. The following are the expectations for acceptable data: When the population is a single class with continuous data from the same or different equipment at the same location, the values should be similar on a day-to-day basis when both pieces of equipment are working. Generally, a single class population with more than 30 vehicles a day should match within 10 percent at the same location. This is subject to the underlying error in collecting data for 7-card records. When the population is a single class with sampled classification data from the same or different equipment at the same location, the values are not expected to match since the same 24-hour period is not used for aggregation for classification and weight data. Sampled classification data has less than 7 days in a month. Data must be compared against expected values. When the population is multiple classes and does not include passenger vehicles, the same expectations exist that apply for single class populations. When the population includes passenger vehicles the volumes will not match unless passenger vehicles are weighed and included in the weight records. The purge codes for a 4-card vs. 7-card review are in Table A.2-1, page 3. The graphs in this section are for Class 9 vehicles unless otherwise stated.

23 MN, 0500, West, Lane 1 - November Class 9 4-card vs. 7-card Figure A.4-1 SPS 4- vs. 7-card - Monthly, Month by Date Example 1 Figure A.4-1 is a 4- card/7-card comparison showing closely matching class 9 volumes from both inputs. Purge candidate No AZ, 0200, East, Lane 1 - August Class 9 4-card vs. 7-card Figure A.4-2 SPS 4- vs. 7- card - Monthly, Month by Date - Example 2 This 4-card/7-card comparison shows a very close match in class 9 volumes under about 1500 per day. When the volumes exceed that level, gaps begin to appear. Do conditions exist that might affect the equipment s ability to capture a complete data set when congestion is present? Are all Class 9s, even those with invalid axle spacing and loading information included in 4-card records but omitted from the weight data? An in depth review of files output from the equipment might be needed or a field check may be necessary. Purge candidate No This site has a consistent repeating gap pattern. The difference between the two volumes is less than 10 percent (by inspection not computation).

24 Figure A.4-3 SPS 4- vs. 7-card - Monthly, Month by Date - Example 3 (case needed) In Figure A.4-3 it is apparent that either the equipment is working intermittently or samples are being taken due to an error in the downloading that is producing partial days of data. Since enough data can be seen to discern a pattern, elimination of some of the weight data for low volumes might be considered. If no pattern is apparent, expected volumes on a day of week or weekday/weekend basis could be used for purge candidate identification. Purge candidate - Yes What Weight data 7 th, 8 th, 20 th Reason Lower volume than expected (70) Expected action: Check weight data values against allowable minimums before applying purges. AZ, 0100, North, Lane 1 - March Trucks (4-20) 4-card vs. 7-card volume significantly greater than 7-card volume (69) Figure A.4-4 SPS 4- vs. 7- card - Monthly, Month by Date Trucks - Example 1 A trucks (Classes 4-20) graph showing similar patterns but different volumes of trucks for the same day is contained in Figure A.4-4. The differential gap between the graph lines is not constant. Purge candidate - Maybe What Class data 17 th, 18 th, 21st-24 th, 28 th -31 st Reason Daily 4-card Expected Actions: Prior to applying the purge check the following to verify bad data: Not a holiday (spring break/easter here), look at a vehicles (1-20) graph to see if a Class 3/Class 5 differentiation problem is affecting the data or review Class 3 and Class 5 volumes using 4-card vs. 7-card graphs to see if Class 5 volumes are rising and Class 3 volumes are falling for the period of interest.

25 cars/pickups or vice versa. AZ, 0100, North, Lane 1 - March Hvy Trks (6-20) 4-card vs. 7-card Figure A.4-5 SPS 4- vs. 7- card - Monthly, Month by Date Heavy Trucks - Example 1 Figure A.4-5 is a heavy truck (6-20) 4-card vs. 7- card comparison for the same site-month as the previous figure. Note both the changing differences here and the increasing volume over time. This graph shows that the relative difference is not a function of class 5 s (2 axle single unit) vehicles that can also be classified as This becomes a purge candidate subject to review of the individual truck classes that constitute a majority of the heavy truck distribution. Purge Candidate - TBD Expected action: A class-by-class evaluation of data for classes that are 10 percent or more of the truck population and a vehicle distribution review for a purge determination. AZ, 0100, North, Lane 1 - May Vehicles (1-20) 4-card vs. 7-card Figure A.4-6 SPS 4- vs. 7- card - Monthly, Month by Date Vehicles - Example 1 Figure A.4-6 is an all vehicles (Classes 1-20) 4 card vs. 7-card comparison where the element interest is the 7-card (truck data) line on the bottom that shows a recurring pattern. The sloppier pattern in the upper curve for classification data would only warrant investigation if an AADT estimate were expected for this site. Given the end of May date, passenger counts might be influenced by Memorial Day weekend. Purge candidate - No

26 AZ, 0100, North, Lane 1 - January Vehicles (1-20) 4-card vs. 7-card Figure A.4-7 SPS 4- vs. 7- card - Monthly, Month by Date - Vehicles - Example 2 This is an all vehicles comparison (Classes 1-20) with continuous data (more than 7 days) and no issues. The 4-card data shows higher volumes because is includes classes 1-3.This would have the same evaluation as Figure A.4-6. Purge candidate No Figure A.4-8 SPS 4- vs. 7-card - Monthly, Month by Date - Vehicles - Example 3 This is an all vehicles 4-card vs.7- card graph with a precipitous drop in volume during the February Check for holidays or weather issues. This month should have no need for further investigation unless it is the sole basis of an AADT estimate. Purge candidate - No

27 AZ, 0100, North, Lane 1 - March Vehicles (1-20) 4-card vs. 7-card Figure A.4-9 SPS 4- vs. 7- card - Monthly, Month by Date - Vehicles - Example 4 Figure A.4-9 shows rapid increases in vehicle population and an essentially constant truck population. The lower curve, the 7-card data has reasonably consistent cycles to the data. Check this classification data to see if the increase is across all classes or is concentrated among unknowns and unclassified vehicles. Also check if it is due to changes in record format due to handling of 4-card data beyond column 51. Purge candidate - Yes What Class data (upper line) from 13 th or 15th to end of month 28 th -31 st unless it can be shown to be Easter week or a spring break pattern) Reason Higher volume than expected (100) Expected action: Verify that the data is not diverging based on non-truck classes before applying purge. If 4-card format problem affected by entries beyond column 51, reload data and repeat review. AZ, 0600, East, Lane 1 - April Vehicles (1-20) 4-card vs. 7-card Figure A.4-10 SPS 4- vs. 7-card - Monthly, Month by Date Example 4a - Vehicles This is a sampled (less than 7- day) data set. The lower volume for the fourth day of weight data in the lower curve is most likely the result of a partial day. That cannot be verified without looking at the original data. Purge candidate - No

28 AZ, 0600, East, Lane 1 - April Hvy Trks (6-20) 4-card vs. 7-card Figure A.4-11 SPS 4- vs. 7-card - Monthly, Month by Date Example 4b Heavy Trucks The pattern of Figure A.4-10 is repeated here when only heavy truck volumes are compared. Purge candidate - No AZ, 0600, East, Lane 1 - April Class 9 4-card vs. 7-card Figure A.4-12 SPS 4- vs. 7-card - Monthly, Month by Date Example 4c Class 9 Figure A.4-12 is the last in a series that illustrate decision making on purges due to low volumes. Note that this graph uses a single vehicle class for which an allowable minimum value for partial days should be known. Purge candidate - No

29 A.5 GVW Monthly by Month Tables used: DD_GVW, COMP_GVW, STAT_QC_GVW_9_DD Type abbreviation: SPSGVWMMon Default: Class 9s This graph is used to determine whether the distribution of gross vehicle weights (GVW) is rational for a given class. A monthly aggregation is a quick way to check the curve but it may hide weeks of poor data. Monthly graphs are best used for sites that have stable patterns or small volumes of the classes of interest (more than 125 per month but less than 125 per week). A Class 9 GVW graph has two pairs of vertical reference lines, the shape of the expected distribution for the site and two columns. The vertical reference lines are the ± 2 standard deviation limits on the expected average GVW for loaded and unloaded vehicles. The expected weights are based on the average vehicle weight for the mode plus the bin above and below the mode. Unloaded vehicles are defined as those weighing less than 40,000 pounds. Loaded vehicles weigh 60,000 pounds or more. The two columns are the average weights for the vehicles being reviewed. Their values are computed using the same criteria as the vertical reference lines. They are located in the bin containing the mean unloaded and loaded weights. There is no reference information beyond an expected distribution for all other vehicle classes. A month s data should be accepted without further investigation if: For Class 9 with the correct comparison period for the month The unloaded mean falls between the left pair of vertical reference lines The loaded mean falls between the right pair of vertical reference lines At least one of the means is centered between the vertical reference lines All points from 16 kips to the left are 2 percent or less All points from 84 kips to the right are 2 percent or less The x-axis goes from 0 to 100 or less For all other vehicle classes (subject to vehicle type and local regulations) Values below lower limit are 10 percent or less (The lower limit will generally be 4. The lower limit is set 2 bins below expected minimum weight. If lower limit is zero the expected value for that bin plus 5 percent is the allowable maximum.) All points above the upper limit are 2 percent or less (The upper limit is set 2 bins above expected or legal maximum weight, which ever is larger based on validation for the site.) The shape is consistent with the comparison pattern. Consistency is a gap of one y-axis increment or less with the same loaded or unloaded tendency. GVW distribution review purge codes are in Table A.2-2, page 6. Unless otherwise noted the graphs in this section are for Class 9s.

30 MN, 0500, West, Lane 1 - December 2006 vs. 12/14/ Class 9 GVW Distribution - No. weighed: 2830 exact match between the two distribution curves. Figure A.5-1 SPS GVW - Monthly, by Month Example 1 Figure A.5-1 is the comparison of monthly data against the comparison data set that was created from data within the same month. It shows an essentially Purge candidate - No MN, 0500, West, Lane 1 - August 2007 vs. 12/14/ Class 9 GVW Distribution - No. weighed: 3005 Figure A.5-2 SPS GVW - Monthly, by Month - Example 2 Figure A.5-2 shows a comparison data set and a GVW distribution from a different month. The GVW curves show the loaded and unloaded peaks in the locations that have been determined valid for this site. As long as the means are within the limits, this data is acceptable. Purge candidate - No

31 Figure A.5-3 SPS GVW - Monthly, by Month Example 3 (no comp data) In Figure A.5-3 the Class 9 distribution shown by the solid line does not overlap or have the same peaking characteristics for the unloaded portion of the curve. Because the loaded portion of the curves match more closely, the data is considered acceptable. The columns for the mean unloaded and loaded weights are centered between the vertical reference lines and the tails fall within the allowable limits. Purge candidate Maybe Expected action: Revert to Month by Week graph to see if cause of shift (loss of unloaded vehicles) can be determined. The effect may be seasonal or holiday related. Figure A.5-4 SPS GVW - Monthly, by Month Class 5 Figure A.5-4 is an example of a comparison graph for other than Class 9 vehicles. It has an expected curve, the solid line, and the monthly distribution, the dashes. This data should be accepted since the peaking characteristics for an unloaded and a loaded peak are consistent. Additionally the maximum gap between any two points on the line is less than one y-axis increment. Purge candidate - No MN, 0500, West, Lane 1 - January 2007 vs. 12/14/ Class 5 GVW Distribution - No. weighed: 1871

32 A.6 GVW Monthly, Month by Week (DEFAULT) Tables used: DD_GVW, COMP_GVW, STAT_QC_GVW_9_DD Type abbreviation: SPSGVWMWk Default: Class 9s This graph is used to determine whether the distribution of gross vehicle weights (GVW) is rational for a given class. To use this graph there should be at least 125 vehicles in the class expected for the week. A class 9 GVW graph has two pairs of vertical reference lines, the shape of the expected distribution for the site and two columns. The vertical reference lines are the ± 2 standard deviation limits on the expected average GVW for loaded and unloaded vehicles. The expected weights are based on the average vehicle weight for the mode plus the bin above and below the mode. Unloaded vehicles are defined as those weighing less than 40,000 pounds. Loaded vehicles weigh 60,000 pounds or more. The two columns are the average weights for the vehicles being reviewed. Their values are computed using the same criteria as the vertical reference lines. They are located in the bin containing the mean unloaded and loaded weights. There is no reference information beyond an expected distribution for all other vehicle classes. A week s data should be accepted without further investigation if: For Class 9 with the correct comparison period for the month The unloaded mean falls between the left pair of vertical reference lines The loaded mean falls between the right pair of vertical reference lines At least one of the means is centered between the vertical reference lines All points from 16 kips to the left are 2 percent or less All points from 84 kips to the right are 2 percent or less The x-axis goes from 0 to 100 or less For all other vehicle classes (subject to vehicle type and local regulations) Values below lower limit are 10 percent or less (The lower limit will generally be 4. The lower limit is set 2 bins below expected minimum weight. If lower limit is zero the expected value for that bin plus 5 percent is the allowable maximum.) All points above the upper limit are 2 percent or less (The upper limit is set 2 bins above expected or legal maximum weight, which ever is larger based on validation for the site.) The shape is consistent with the comparison pattern. Consistency is a gap of one y-axis increment or less with the same loaded or unloaded tendency. GVW distribution review purge codes are in Table A.2-2, page 6. Unless otherwise noted the graphs in this section are for Class 9 vehicles.

33 MN, 0500, West, Lane 1 - December 2006 vs. 12/14/ Class 9 GVW Distribution - No. weighed: 2830 Figure A.6-1 SPS GVW - Monthly, Month by Week Example 1 Figure A.6-1 is the comparison of the monthly data against the comparison data set that was created from data within the same month. It shows a close match between the distribution curves. The means are centered within the 2 standard deviation limits. Purge candidate No

34 Figure A.6-2 SPS GVW - Monthly, Month by Week Example 2a (case needed with better comparison data) Figure A.6-2 shows the loaded weight has increased enough to shift bins as compared to the comparison data set. Purge candidate No Note that this is a graph of the month containing the comparison data set with that comparison. If this comparison is going to be retained, the tendency for an offcenter loaded mean should be a characteristic for accepted data. Figure A.6-3 SPS GVW - Monthly, Month by Week Example 2b (case needed with better comparison data) Figure A.6-3 shows the same site as Figure A.6-2 but for this month it is the average unloaded weight that has increased enough to shift one bin to the left. This was offcenter for the initial month so the shift is not inappropriate and suggests a mean very close to a bin boundary. Purge candidate No Shifts in unloaded peaks are not considered to be as critical as those for loaded peaks. The curves are consistent. Figure A.6-4 SPS GVW - Monthly, Month by Week Example 2c (case needed with better comparison data)

35 Figure A.6-4 is also for the same site as Figure A.6-2. It shows three changes from the previous month found in Figure A.6-3. The first is the shift back to the left of the unloaded mean. The second is the shift to the left of the loaded mean. (It is now two bins from the month with the comparison data.) The third is the lengthening of the x-axis from 128 to 160. This implies that the heaviest vehicle observed weighs around 156,000 lbs. Purge candidate Surrogate comparison set Probably no Sheet 16 validation comparison data - Maybe What Week with heavy vehicle OR entire month Reason - Large percentage of overweight trucks (108) OR GVW distribution inconsistent with history (104) Expected actions: Weekly by Week or Day of Week graph since sufficient number of trucks exists to see if the pattern is shaped by one or more weeks.

36 Figure A.6-5 SPS GVW - Monthly, Month by Week - Example 3 (case needed with better comparison data) Figure A.6-5 shows a month in which all the curves but one have shifted right. The patterns for all are similar to the comparison shape. history (104) Purge candidate Yes What - All weeks or just week of 2/22 Reason Over-calibrated (73) (marginally both means have shifted right.) OR GVW distribution inconsistent with Expected action: This implies calibration drifts. Inquiries need to be made about validation activities or remote adjustment of auto-calibration factors if used. The non-matching weeks should be investigated for cause. The next month s data should be checked carefully. are dominating the unloaded shape. Figure A.6-6 SPS GVW - Monthly, Month by Week Example 4a(case needed with better comparison-more recent) Figure A.6-6 reflects the different properties of the loaded and unloaded peaks. The loaded peak is heavily concentrated at a single weight resulting in very narrow ±2 SD boundaries. The unloaded vehicles however are much more variable in their weights with very little definite peaking. The flatter curve resulted in wider ±2 SD limits. This month they are very light to the point that empty trucks Purge Candidate Yes 1 st through 21 st Reason: Large shift in unloaded peak (106)

37 Expected Action: Due to number of trucks weighed, a weekly, Week by day of week graph for the 1 st 3 weeks of the month to verify that all inconsistent days are purged.

38 Figure A.6-7 SPS GVW - Monthly, Month by Week Example 4b (case needed with better comparison-more recent) Figure A.6-7 is for the same site as Figure A.6-6 with the data being reviewed tracking the validated pattern. Note that for this case the very narrow band for the loaded peak will make some of the standard decision criteria less meaningful with respect to shifts in the location of the loaded peak. Purge candidate No Figure A.6-8 SPS GVW - Monthly, Month by Week Example 5 (case needed with better comparison-more recent) In Figure A.6-8 both peaks have shifted all the way to the left edge of the ±2 SD bounds. Note also the week with the half empty / half full peak. Purge candidate Maybe What th Reason GVW distribution inconsistent with history (104) What 1 st -28 th Reason Large shift in unloaded peak (106) Expected action: Investigate site for seasonality. The shift in the loaded peak should be investigated by Weekly, week by day of week graphs to find unusual days. Note also the length of time between the comparison and the review data.

39 Figure A.6-9 SPS GVW - Monthly, Month by Week Example 6 (case needed with better comparison-more recent) Figure A.6-9 shows a month with centered means but a variety of curve shapes. Purge candidate - Yes What 22 nd 28 th Reason Large shift in unloaded peak (106) Expected Action: The variability of the patterns should be watched with the week of 22 nd checked more thoroughly due to its distinct bimodal pattern that is at variance with the typical curve. A Weekly week by day of week may not be possible due to small sample size. Query database to determine how many vehicles/days are included in curve for 7/22-28 or use 4-card vs. 7-card graph. Figure A.6-10 SPS GVW - Monthly, Month by Week Example 7 (case needed with better comparison-more recent) Figure A.6-10 shows a month with centered means and very similar curves. The cause for concern in this graph is the thickening tail to the right of the loaded peak. Increasing numbers of truly over weight vehicles need to be verified as either a seasonal affect or a precursor of sensor failure Purge candidate Yes What - 5/1-7 Reason Large percentage of tractor-trailers over 80 kips (75) Expected action: Check for seasonality or special event. Use a Weekly, Week by Day of Week graph to look at individual days.

40 Figure A.6-11 SPS GVW - Monthly, Month by Week Example 8 (case needed with better comparison-more recent) Figure A.6-11 shows a series of curves where the overweight percentage is significantly smaller than that of the comparison data set. Purge candidate No Expected action Determine if a different comparison set is needed or if validation activities have occurred. Figure A.6-12 SPS GVW - Monthly, Month by Week Example 9a (case needed with better comparison-more recent) Figure A.6-12 has insufficient data to make a purge determination. Purge candidate TBD evaluation options. Expected action: Reevaluate period over which comparison data set was created. The period may need to be longer (and created external to the software.) Graph is too irregular to be useful. Converting to a monthly graph with a single week of data may not improve

41 Figure A.6-13 SPS GVW - Monthly, Month by Week Example 9b (case needed with better comparison-more recent) Figure A.6-13 reflects the need to do a review with a monthly rather than by week graph. Purge candidate TBD Expected action Verify the amount of data in the comparison data set. Re-evaluate data using the GVW Monthly by Month graph. Figure A.6-14 SPS GVW - Monthly, Month by Week Example 10a (case needed with better comparison-more recent) Figure A.6-14 is a SPS GVW graph without comparison data. It shows a site with varying weight distributions. However, the bars for the unloaded and loaded means are not inconsistent with expected values for this vehicle class. Purge candidate No

42 WA, 0200, South, Lane 1 - April 1998 vs. <NO COMP DATA for 01/01/1990> - Class 9 GVW Distribution - No. weighed: 2751 Reason Large percentage of tractor trailers under 12 kips (76) Figure A.6-15 SPS GVW - Monthly, Month by Week Example 10b Figure A.6-15 shows another SPS GVW graph without comparison data. Lack of a comparison data set does not eliminate the usefulness of these graphs. Note the position of the unloaded mean is at the critical point for an under weights check 12 kips. Purge candidate Yes What All data in the month Expected action Determine last site calibration date to create a comparison set. Verify that lighter trucks have not been misidentified as class 9s. Reload data if the classification scheme uses the 6-digit TWS format. WA, 0200, South, Lane 1 - December 1998 vs. <NO COMP DATA for 01/01/1990> - Class 9 GVW Distribution - No. weighed: 9506 Figure A.6-16 SPS GVW - Monthly, Month by Week Example 10c Figure A.6-16 is for a site without a comparison data set. This data shows a preponderance of unloaded trucks and a possibility of under-calibration. Purge candidate Yes What 12/22-28 Reason Large percentage of tractor-trailers under 12 kips (76) OR Under calibrated (74) OR Large percentage of underweight trucks (108) (for multiple classes or when Class 9 is not a 5-axle tractor semi-trailer vehicle). Expected action: The unloaded mean is at the limiting value for that peak. If this data was to be retained as valid after investigation for the remaining weeks, a comparison data set should be established to document the reasonableness of this pattern.

43 Figure A.6-17 SPS GVW - Monthly, Month by Week Example 10d (case needed) Figure A.6-17 also has no comparison data set. The issue is the determination of a typical curve between the three basic shapes. None of them are distinctly bimodal. Two have a relatively high percentage of vehicles over 80 kips. Purge candidate Yes What 3/8-21 Reason Large percentage of tractor trailers over 80 kips (75) Expected actions Find out whether a calibration or sensor replacement has occurred. Define a suitable comparison data set. Figure A.6-18 SPS GVW - Monthly, Month by Week Example 11 (NEED CASE) Figure A.6-18 shows a site where a problem is obvious even without a comparison data set. There is a large percentage of Class 9s under 12 kips, peaking at the expected weight of a car. This is indicative of a loading problem for data using the 6-digit truck weight scheme loaded with version 1.5 and earlier. Purge candidate TBD Expected action: Reload data and re-evaluate

44 Figure A.6-19 SPS GVW - Monthly, Month by Week Example 12a (CASE NEEDED) Figure A.6-19 is effectively unimodal. The unloaded average weight has shifted right but the loaded weight is unchanged. The shift implies that given that the period of comparison data and the data being evaluated are for the same period, the value computed is close enough to a bin boundary that rounding influenced the results. Purge candidate No Expected actions Make a two bin shift for unloaded weights a minimum for identifying weeks for as purge candidates at this site. TN, 0600, West, Lane 1 - October 2002 vs. 06/14/ Class 9 GVW Distribution - No. weighed: Figure A.6-20 SPS GVW - Monthly, Month by Week Example 12b Figure A.6-20 peaking has shifted from mostly loaded to mostly unloaded. The comparison distribution is the dotted line. In this case the value of the loaded peak has shifted to the lower limit of the GVW bins used to define loaded vehicles for the comparison data. Purge candidate Yes What All data Reason Atypical pattern (72) OR Large shift in loaded peak (105) OR GVW distribution inconsistent with history (104) Expected actions: Verify that comparison data set is valid. Inquire about validation or change of auto-calibration factors.

45 Figure A.6-21 SPS GVW - Monthly, Month by Week - Example 13 (DIFFERENT REGION-NC) Figure A.6-21 is an example of either a severe calibration problem or more likely a sensor failure. Purge candidate Yes What Month Reason Large percentage of overweight trucks (109) Expected action: Determine how much other data is affected. WA, 0200, North, Lane 1 - February 2005 vs. 01/01/ Class 5 GVW Distribution - No. weighed: 4115 Figure A.6-22 SPS GVW - Monthly, Month by Week - Class 5 Example 1 Figure A.6-22 shows a typical 2-axle single unit truck distribution. It also illustrates that for classes other than 9s there is only a comparison distribution plotted as shown on this figure. Purge candidate No

46 MN, 0500, West, Lane 1 - January 2007 vs. 12/14/ Class 5 GVW Distribution - No. weighed: 1871 Figure A.6-23 SPS GVW - Monthly, Month by Week - Class 5 - Example 2 Figure A.6-23 shows multiple weeks of data versus a comparison curve for a vehicle that is not a Class 9. The variation in the heavier vehicles is not important given the fixed position for 65 percent of the population. Purge candidate No Figure A.6-24 SPS GVW - Monthly, Month by Week - Class 5 - Example 4 Figure A.6-24 is similar to Error! Reference source not found.. In this case the difference comes at the light end of the curve. differentiation problem Yes Reason Atypical pattern (72) Purge candidate Maybe If site has difficulty distinguishing between Class 3s and Class 5s No if there is no 3/5 Expected Actions - If the algorithm has been changed to better differentiate between Class 3s and Class 5s, create a new comparison data set if this is a major vehicle class in the truck population. Otherwise, ignore this class until a new comparison data set is created based on a field validation.

47 A.7 Vehicle Distribution Monthly by Month Tables used: DD_CL_CT, DD_WT_CT, COMP_CL_CT, COMP_WT_CT Type abbreviation: SPSDistMMon Default: Heavy Trucks (6-20) This graph is used to determine whether the mix of heavy trucks in the vehicle population is stable. The distributions for classification data and weight data are graphed on separate figures. To use this graph there should be at least 125 vehicles expected in the month for the group being evaluated. This graph has two types, the Heavy Truck graph and all other groups. The Heavy Truck graph includes vehicles in classes 6 to 20. The graph includes a set of bars for every truck class that is more than ten percent of that population. The interior dark grey portion of the bar is the bounds (±5%) in which the percentage of that class is expected to fluctuate routinely. The light grey outer portions of the bar are the maximum limits (±10%) of allowable fluctuation without further investigation. Both classification and weight data appear on the same graph with the classification on the left and the weight on the right (red and blue respectively). There will also be two comparison bars if there are two types of data present. The classification comparison is the left hand set and the weight comparison the right hand set. The interior portion of the bar is a slightly darker gray for the weight data. A month s data for a data type should be accepted without further investigation if: A heavy truck graph All classes with comparison bars fall within the bars The percentage of unclassifieds is less than the expected value for the site or less than five percent if there are no unclassifieds expected No class without a comparison (gray) bar is more than 15 percent of the week s population For all other distributions The percentage of unclassifieds is less than the expected value for the site or less than two percent if there are no unclassifieds expected. The Class 9 percentage is greater than the Class 8 percentage The purge codes for vehicle distribution review are found in section Table A.2-3 on page 10.

48 MN, 0500, West, Lane 1 - March 2007 vs. 11/25/ Hvy Trks (6-20) Vehicle Distribution (4-card) - No. counted 3, Figure A.7-1 SPS Vehicle Dist. - Monthly, by Month - Heavy Trucks (6-15) Example 1a Figure A.7-1 is an example of a site where the main truck type is the Class 9. It is the only one with 10 percent or more of the heavy truck population. This is the 4-card graph of the set. Purge candidate No MN, 0500, West, Lane 1 - February 2007 vs. 11/13/ Hvy Trks (6-20) Vehicle Distribution (7-card) - No. weighed 3, Figure A.7-2 SPS Vehicle Dist. - Monthly, by Month - Heavy Trucks (6-15) Example 1b Figure A.7-2 is an example of a site where the principal truck type is the Class 9 and it is the only one with 10 percent ofr more of the heavy truck population. This is the 7-card graph of the set. Purge candidate - No

49 Figure A.7-3 SPS Vehicle Dist. - Monthly, by Month Trucks (4-15) - Example 1a Figure A.7-3 is an example of a truck population rather than a heavy truck population graph. There is no comparison data here. Purge candidate Maybe What Entire month Reason Too many unclassifieds (102) Expected action Ask about source of error. Compare percent unclassifieds to its expected value and variability.

50 Figure A.7-4 SPS Vehicle Dist. - Monthly, by Month - Trucks (4-15) Example 1b Figure A.7-4 is an example of a truck population rather than a heavy truck population graph. There is no comparison data here. Purge candidate No

51 A.8 Vehicle Distribution Monthly, Month by Week (DEFAULT) Tables used: DD_CL_CT, DD_WT_CT, COMP_CL_CT, COMP_WT_CT Type abbreviation: SPSDistMMon Default: Heavy Trucks (6-20) This graph is used to determine whether the mix of trucks in the vehicle population is stable. The distributions for classification data and weight data are graphed on separate pages. To use this graph there should be at least 125 vehicles expected in a week for the group being evaluated. This graph has two types, the Heavy Truck graph and all other groups. The Heavy Truck graph includes vehicles in classes 6 to 20. The graph includes a set of bars for every truck class that is more than ten percent of that population. The interior dark grey portion of the bar is the bounds (±5%) in which the percentage of that class is expected to fluctuate routinely. The light grey outer portions of the bar are the maximum limits (±10%) of allowable fluctuation without further investigation. Both classification and weight data appear on the same graph with the classification on the left and the weight on the right (red and blue respectively). There will also be two comparison bars if there are two types of data present. The classification comparison is the left hand set and the weight comparison the right hand set. The interior portion of the bar is a slightly darker gray for the weight data. A week s data for a data type should be accepted without further investigation if: A heavy truck graph All classes with comparison bars fall within the bars The percentage of unclassifieds is less than the expected value for the site or less than five percent if there are no unclassifieds expected No class without a comparison (gray) bar is more than 15 percent of the week s population For all other distributions The percentage of unclassifieds is less than the expected value for the site or less than two percent if there are no unclassifieds expected. The Class 9 percentage is greater than the Class 8 percentage The purge codes for vehicle distribution review are found in Table A.2-3 on page 10. The graphs for 4-card and 7-card data are on separate pages. The data type is contained in the graph header. For the purposes of illustration only one pair of classification and weight data graphs is provided. The review principles are the same independent of the data type so most illustrations will use classification data. The most suitable truck group selection is Heavy Trucks (6-20). This population avoids the complications of site review when there is difficulty differentiating between Class 3 and Class 5 vehicles. Unless otherwise noted, the graphs in this section are for heavy trucks. If there are too few heavy trucks to generate the graphs the next most suitable is the Trucks (4-20) option.

52 MN, 0500, West, Lane 1 - April 2007 vs. 09/18/ Hvy Trks (6-20) Vehicle Distribution (4-card) - No. counted 3,254 Figure A.8-1 SPS Vehicle Dist. - Monthly, Month by Week Example 1a by the color when displayed in color. Figure A.8-1 shows a typical vehicle distribution graph at a site with comparison data. Only Class 9s are a major group in the heavy truck population with more than ten percent of group. This is a 4-card graph. It can be differentiated from a 7-card graph by the heading and Purge candidate No (Unclassifieds are in expected bounds) MN, 0500, West, Lane 1 - February 2007 vs. 09/09/ Hvy Trks (6-20) Vehicle Distribution (7-card) - No. weighed 3,943 Figure A.8-2 SPS Vehicle Dist. - Monthly, Month by Week Example 1b Figure A.8-2 shows a typical vehicle distribution graph at a site with comparison data. Only Class 9s are a major group in the heavy truck population with more than ten percent of group. This is a 7-card graph. It can be differentiated from a 4-card graph by the heading and the color. Note that Class 6 vehicles will be subject to review as a class with almost ten percent of the population. Purge candidate - No

53 NV, 0100, East, Lane 1 - April 1999 vs. 04/01/ Hvy Trks (6-20) Vehicle Distribution (4-card) - No. counted 30,038 Figure A.8-3 SPS Vehicle Dist. - Monthly, Month by Week Example 2a Figure A.8-3 is an example for a site where Class 9s and Class 13s each comprise ten percent or more of the heavy truck population Purge candidate - No NV, 0100, East, Lane 1 - December 2000 vs. 06/01/ Hvy Trks (6-20) Vehicle Distribution (4-card) - No. counted 32, Reason Outside +/- 10 percent limits (103) Figure A.8-4 SPS Vehicle Dist. - Monthly, Month by Week Example 2b Figure A.8-4 is for the same location as Figure A.8-3. As can be seen from the figure the Class 9 vehicles are below the expected proportion of the population and the Class 13s are above their expected proportion. Purge candidate Yes What December 2000 Expected actions: Inquire if the population shifts due to weather or seasonal activity. (It did not however have this element of change in December comparisons for earlier years.)

54 OH, 0100, South, Lane 1 - August 2006 vs. 08/13/ Hvy Trks (6-20) Vehicle Distribution (4-card) - No. counted 34, Figure A.8-5 SPS Vehicle Dist. - Monthly, Month by Week Example 3a Figure A.8-5 is the illustration of a comparison data set against the month for which it was generated. The Class 6 population was not large enough in the comparison data set to generate a comparison block Both 8s and 9s are present in large enough proportions to be watched. Purge candidate - No Expected actions Make a note that Class 6s are at about 10 percent of the population and should be flagged as an error in future months reviews if they go over 15 percent. Figure A.8-6 SPS Vehicle Dist. - Monthly, Month by Week Example 3b (NO COMPARISON DATA) Figure A.8-6 is for the same site as Figure A.8-5 using the same comparison data set. This is an approximation since the comparison set is a surrogate. The comparison data set is also for data some years in the future potentially after one or more equipment replacements or adjustments. Purge candidate Yes What Month of January Reason Outside +/- ten percent limits (103) Expected actions: Determine if a better surrogate data set can be created. Investigate vehicle distribution pattern over time using Agency graphs with by year capability.

55 Figure A.8-7 SPS Vehicle Dist. - Monthly, Month by Week Example 3c (NO COMPARISON VALUES) In Figure A.8-7 both values are outside of the comparison blocks and there is a high percentage of unclassified vehicles. The misclassification of Class 9s as Class 6s with or without 14s is representative of time out failures on loops. Purge candidate Yes What Month of January Reason Too many unclassifieds (102) OR Outside +/- 10 percent range (103) Expected actions: Verify that classification scheme has not changed. Figure A.8-8 SPS Vehicle Dist. - Monthly, Month by Week Example 4a (MI-NO SPS SITES NOW) Figure A.8-8 is an example of a site where there are several classes that are more than 10 percent of the heavy truck population. This site needs to be watched closely as all of the classes are outside the reasonable variability expected (dark portion of comparison block.) Purge candidate No

56 Figure A.8-9 SPS Vehicle Dist. - Monthly, Month by Week Example 4b (MI-NO SPS SITES NOW) Figure A.8-9 is another example of more than one class being over ten percent of the heavy truck population. The distributions are at the lower limits of the expected normal range and there is a substantial percentage of Class 20s. For data collected in the U.S. Class 20 is an unknown/unclassified vehicle when the 6-digit Truck Weight Study (TWS) classification is converted to TMG 13-bin classes Purge candidate Yes. What Month of August Reason Too many unclassifieds (102) Expected action: This is a state (MI) in which multi-unit vehicles with large numbers of axles are common in some areas. Determine by inspection if the software is correctly set up for vehicles with three trailers or if there are routinely vehicles with more than 13 axles; neither are allowed in the software. This involves review of weight records. Use a spreadsheet to get a frequency distribution of the classification codes. Compare them to the interpretation in LTPP Traffic Data Files. If a specific, verifiable configuration is being defined as a 20, submit an SPR for a software change. The fact that the Class 6 and 8 proportions are at the lower end of the expected limits is also of concern.

57 Figure A.8-10 SPS Vehicle Dist. - Monthly, Month by Week Trucks (4-15) Example 1 (Doesn t populate any graph when trucks 4-15 is selected) Figure A.8-10 is a typical vehicle distribution graph where the entire truck population is being evaluated. This graph is most commonly used where heavy trucks (classes 6 and higher) are not a substantial proportion of the truck population. Purge candidate - No Figure A.8-11 and Figure A.8-12 are for the same site, lane and month. They are included to emphasize that vehicle distribution graphs for different data types should not be done completely independently. Figure A.8-11 SPS Vehicle Dist. - Monthly, Month by Week Trucks (4-15) Example 2a(Doesn t populate any graph when trucks 4-15 is selected) Figure A.8-11 has vehicle classes beyond 13. These are unknown or unclassified and have an unduly high percentage. Purge candidate Yes What - Month of January Reason Too many unclassifieds (102) Expected action: Determine whether the data was correctly loaded into the database. Classification data loaded prior to version 1.5 could misinterpret data beyond column 51 for 4- cards. It also could loose vehicles translating between C-card and 4-card formats. If data beyond column 51 was not correctly interpreted, reload the data and redo the review.

58 Figure A.8-12 SPS Vehicle Dist. - Monthly, Month by Week Trucks (4-15) - Example 2b (Doesn t populate any graph when trucks 4-15 is selected) Figure A.8-12 shows the same data. There is a possibility that all unknown or unclassified vehicles were removed from the weight data set before submission. To make a determination this site should be graphed with the Trucks (4-20) group. Purge candidate TBD Expected action: Create new graph with Trucks (4-20)

59 B Agency Graphs This set of graphs is done using the DD* tables for all intervals to preserve the ability to review the data in the scheme in which the data was provided by the agency and initially loaded. As a result there is the potential to view as many as 20 individual vehicle classes. The software does however assume that the agency defines Class 9 as a 5-axle tractor-trailer (3S2 or 2S3) and trucks are in classes 4 and higher. It is not possible to redefine another class number to be 5-axle tractor-trailers for those graphs assuming Class 9 characteristics on the figure. In cases where there is a difference and either Class 9 or truck populations are important characteristics of the review, the user should use the 13-bin series of graphs where data has been converted to the TMG classification scheme. Figure A.8- Agency Graph Availability (2.0) The first section of this document contains the tables that identify purge codes and their definitions for these graphs. Wherever a code is specific to a vehicle class, that fact is noted as part of its description. Note that some codes may have more than one reason for being selected for a graph type or may apply to multiple types of graphs with different reasons.

60 B.1 Graph intervals Graph intervals define the amount of data aggregated for a single page graph. For Agency graphs the intervals are weekly, monthly, annual and multi-year. The various graph types, other than 4-card vs. 7-card have different minimum data requirements to exist. For vehicle distribution graphs the data set used to create an individual line or bar must have 50 vehicles. For GVW and axle distribution graphs the data set must have 100 vehicles or axles as applicable for an individual line to appear. The minimum data requirement may cause days, weeks or months to vanish from a graph series even when vehicles have been observed. B.1.1 Weekly Weekly graphs are used in review for sites that meet the following criteria: At least 125 vehicles per day in a class to be reviewed At least 125 vehicles per day in a group of vehicles to be reviewed Continuous data is expected for the data type OR GPS data collected by RSC Any graph that does not yet exist in software may be generated in Excel using the spreadsheets developed for graph testing. There are no default weekly graphs. B.1.2 Monthly There are four default graphs defined, one for each graph type. The defaults provide basic review capability for Agency data. They are the 4-card vs. 7-card month by date, the monthly GVW, monthly vehicle distribution, and monthly axle distribution. All but the vehicle distribution graph are done for Class 9 vehicles by default. Monthly graphs are used to review sites that have the following characteristics: More than 125 vehicles in a class/group per month by Week alternatives At least one month of data Monthly graphs may also be used for diagnostics when pattern problems or expected value issues from Weekly graphs need to be investigated. B2 Purge Codes The tables in this section provide a list of codes and some basic decision criteria by graph type. Purge criteria are not absolute. Multiple elements must be considered including the amount of monitored data before making a final purge recommendation. In the tables that follow some purge identification criteria have associated limiting values. This means that unless the data element being checked meets that condition, that particular criteria should not be used for purge identification. This eliminates purge conditions that do not make sense due to variability in small sample sizes. The limiting value relates to a site characteristic such as an average or minimum value derived from a valid data set. The data is valid because of either consensus review or a validation or calibration. The limiting value criterion is the same whether the element being reviewed is a date (day), day of week, single class or group of classes.

61 When applying the purge selection criterion it is not necessary to compute the difference. The differences can be assessed by inspection. Generally, determining what a percentage means in terms of increments on the vertical (y-) axis is sufficient. A valid monthly graph with the limiting values marked on it for comparison is another way to apply the criteria. In this case, attention must be paid to the scale of the vertical axis. New comparison graphs for volume in particular will be needed as volumes increase over time. The values generally needed for 4-card vs. 7-card review ( Table A.2-1) are: Average weekly volume, average daily volume and minimum daily volume. The values may be computed or obtained by inspection of a valid monthly graph. They are different for each class or group to be evaluated. They are defined as follows: Average weekly volume Sum of the averages for each day of the week Daily average volume o For sites with more than 1000 per week averages are computed on a by day of week basis and rounded to the nearest 50 o For sites with 250 to 1000 per week averages are computed for weekdays as a group (the average of the weekday averages) and weekend days as a group (the average of the weekend averages) and rounded to the nearest 10 o For sites with less than 250 per week the average is the average of the day of the week averages and rounded to the nearest 5 Minimum volume 25 percent of the average volume o Values under 100 rounded down to the nearest 5 o Values from 100 to 1000 rounded down to the nearest 10 o Values over 1000 rounded down to the nearest 50 The process is discussed in detail in LTPP Monitored Traffic Data Processing. Included in the discussion is how to pick the groups for graphic evaluation. Table B.1-1 Purge Codes for 4-card vs. 7-card Review Code Description Limiting Value Purge Identification Criteria 64 Zero data Minimum volume at least 10 Class or group with zero volume (or effectively zero) given the expected minimum. 65 Zero daily volume For reviews including all classes only, total volume for day is zero and it is not 68 Daily 7-card volume significantly greater than 4-card volume Average value at least 50 attributed to construction or weather The 7-card volume is more than ten percent larger than the corresponding 7-card value And 1) Review is for a single vehicle class, OR 2) Review is for a group where the same classification algorithm is used for

62 Code Description Limiting Value Purge Identification Criteria both data types. 69 Daily 4-card volume significantly greater than 7-card volume Average value at least 50 The 4-card volume is more than ten percent larger than the corresponding 7-card value And 1) Review is for a single vehicle class, OR 2) Review is for a group where the same classification algorithm is used for 70 Lower volumes than expected, possible sensor problems Minimum value at least Atypical pattern Average volume at least 250 per week both data types. Volume below the minimum. See 136, 137, 138, ) Shift in location of minimum or maximum where the choice of minimum or maximum is obvious (ten percent or greater below or above the other candidate days.) (Should use 101 instead.) 2) Pattern that has demonstrated a relatively constant volume retains its shape (low to high to low) but has a distinct increase or decrease in volume over time. (More than 5 percent a year requires a documented explanation.) 3) The shape changes or a distinct shape appears where the two lines have previously coincided. The disappearance of a shape (not just a little bit of differentiation between the lines) should be investigated. 100 Higher volume than expected Average volume at least 50 4) For a new site without validation data a pattern that does not conform to typical expectations i.e. weekdays having larger volumes than weekends, classification volumes being greater than or equal to weight volumes. Volume above the maximum expected.

63 Code Description Limiting Value Purge Identification Criteria 101 Change in day of week pattern Average volume at least 250 per week 1) Minimum changes day of week and the minimum was no more than 95 percent of the next higher value. (Use 136 for by date graphs) 2) Maximum changes day of week and the maximum was at least 105 percent of the next lower value. (Use 135 for by date graphs) 3) Minimum (maximum) changes between weekday day and weekend day and the difference between weekdays as a group and weekend days as a group is greater than 10 percent of the larger average. 136 Shift in location of maximum volume/ percentage 136 Shift in location of minimum volume/ percentage 137 Disappearance of volume difference 138 Larger than expected volume difference Average volume at least 250 per week Average volume at least 250 per week Average value for smaller volume at least 50 Average value for smaller volume at least 50 4) A one peak weekly volume cycle becomes a two peak cycle or vice versa Location of maximum in a recurring pattern shifts by more than one day in the week for 3 or more consecutive weeks. Location of minimum in a recurring pattern shifts by more than one day in the week for 3 or more consecutive weeks. A recurring gap pattern changes to lines on top of each other An expected volume difference exists of any magnitude and the observed difference is at least 110 percent of the expected difference. Both data types must use a consistent by not necessarily the same definition for a vehicle class or group of vehicles.

64 The values needed for GVW review ( The values needed for GVW review (Table A.2-2) are: tare weight, legal maximum weight, average weekly volume, the volume represented by the curve and the volume represented by the graph. The average weekly volume is same as that for the 4-card vs. 7-card review. To determine the average number in a curve use the following Weekly, by Week; Monthly, by Month use volume in header Weekly, by Day of Week Divide the volume in the header by the number of days present plus 1. (The plus 1 is to offset the potential for missing days at low volume sites and the tendency to have sharp differences between weekday and weekend volumes particularly for Class 5 and Class 9 vehicles.) Monthly, by Week divide by number of weeks shown plus 1 not to exceed 5 (or the number of weeks if shown by week) While expected values for 5-axle tractor semi-trailers are incorporated in the software, expected values for other classes that are needed to evaluate site loading are not. For all other classes unloaded is within 15 percent of the typical tare weight of the class. The tare weight is the unloaded weight of the truck as defined by a manufacturer of that type. The typical tare weight should be for the predominant vehicle in the class if multiple configurations are assigned the same classification. An example is buses where both two axle and three axle versions exist. Loaded is between 80 percent and 100 percent of the legal maximum for the class. Where multiple vehicle types are included in a class, multiple peak values may need to be established. Table A.2-2) are: tare weight, legal maximum weight, average weekly volume, the volume represented by the curve and the volume represented by the graph. The average weekly volume is same as that for the 4-card vs. 7-card review. To determine the average number in a curve use the following Weekly, by Week; Monthly, by Month; Yearly, by Year use volume in header Weekly, by Day of Week Divide the volume in the header by the number of days present plus 1. (The plus 1 is to offset the potential for missing days at low volume sites and the tendency to have sharp differences between weekday and weekend volumes particularly for Class 5 and Class 9 vehicles.) Monthly, by Week divide by 5 (or the number of weeks if shown by week) Monthly, by Day of Week; Yearly by Day of Week - Use volumes in 4-card vs. 7-card Month (Year) by Day of Week graph Yearly, Year by Month Use volume from Vehicle Distribution, Trucks, Yearly, Year by Month. While expected values for 5-axle tractor semi-trailers are incorporated in the software, expected values for other classes that are needed to evaluate site loading are not. For all other classes unloaded is within 15 percent of the typical tare weight of the class. The tare weight is the unloaded weight of the truck as defined by a manufacturer of that type. The typical tare weight should be for the predominant vehicle in the class if multiple configurations are assigned the same classification. An example is buses where both two axle and three axle versions exist.

65 Loaded is between 80 percent and 100 percent of the legal maximum for the class. Where multiple vehicle types are included in a class, multiple peak values may need to be established.

66 Table B.1-2 Purge Codes for GVW Review Code Description Limiting value Purge Identification Criteria 72 Atypical distribution (See 149, 150, 151.) 1) A pattern variance from the expected at the site not described by one or more other codes 2) A pattern for a new site without validation information that does not match the expected typical pattern for 73 Over calibrated 1000 or more per week - 1 bin 500 or fewer per week 2 bins the class Class 9s Unloaded and loaded peaks shifted equally to the right AND outside of their expected bounds or more per week 10 percent of legal maximum to nearest bin 500 or fewer per week 20 percent of legal maximum 74 Under calibrated 1000 or more per week - 1 bin 500 or fewer per week 2 bins 1000 or more per week 10 percent of legal maximum to nearest bin 500 or fewer per week 20 percent of legal maximum 75 Large percentage of tractor trailers over 80 kips 76 Large percentage of tractor trailers under 12 kips Class 9s only vehicles or more in curve being assessed Class 9s only vehicles or more in curve being assessed Other classes Curve has its reference shape and is shifted to the right from the expected location. Class 9s Unloaded and loaded peaks shifted equally to the left AND outside of their expected bounds. Other classes Curve has its reference shape and is shifted to left from the expected location. 1) The proportion of the curve to the right of the 80 kip bin mark is greater than five percent. 2) The value for the distribution is non-zero beyond 100 kips. The proportion of the curve to the left of the 12 kip mark is greater than two percent.

67 Code Description Limiting value Purge Identification Criteria 104 GVW distribution inconsistent with history 1) Class 9s - Either of the peaks is shifted more than one bin from the expected curve in either direction. 2) Other classes A dominant peak is shifted more than one bin from the expected curve in either direction. 3) The shape of a curve is different from the expected curve. Different is o A uni-modal (single peak) distribution becoming bi-modal (two-peaks), (Use 149) o A distinctly loaded or unloaded site shifting to a partially loaded site (and remaining uni-modal) (Use 105 Large shift in loaded peak 106 Large shift in unloaded peak 107 GVW peaks outside expected limits (Class 9s) 108 Large percentage of underweight trucks 250 vehicles or more in curve being assessed 250 vehicles or more in curve being assessed 250 vehicles or more in curve being assessed 500 vehicles or more in curve being assessed o 150), Doubling of the size of either the tail below the lower bound of the unloaded peak or above the upper bound of the loaded peak. (Use 152 or 151 respectively) This code is preferred to 72 (atypical distribution) because it is a more explicit description of an irregularity. The loaded peak has shifted two or more bins or by 15 percent of the legal maximum (rounded to the nearest bin) from its expected position. The unloaded peak has shifted two or more bins or by 15 percent of the legal maximum to the nearest bin from its expected position. Both peaks are outside the vertical lines marking the expected location of peaks for 5-axle tractor-trailer combinations. 1) More than five percent of the trucks are less than ninety percent of the minimum tare weight for the population, 2) There is a non-zero value of the distribution for bins that contain weights less than ninety percent of the expected tare weight for the class.

68 Code Description Limiting value Purge Identification Criteria 109 Large percentage of overweight trucks 500 vehicles or more in curve being assessed 1) More than five percent of the trucks exceed ten percent of the legal maximum for the class OR 2) There is a non-zero value for the frequency distribution in bins that contain weights that are ten percent or 149 Change in number of modes for GWV distribution 150 Change in dominant loading type 151 Large change in loaded tail of GVW distribution 152 Large change in unloaded tail of GVW distribution 1000 vehicles or more on graph 500 vehicles or more on graph 250 vehicles or more in curve being assessed 250 vehicles or more in curve being assessed more than the legal maximum. 1) One peak changing to two peaks 2) Two peaks changing to one peak 3) Two peaks going to three peaks where the third peak is at least a third of the next higher peak 4) Three peaks going to two peaks 1) Loaded site becomes unloaded 2) Unloaded site becomes loaded 3) Site not clearly loaded or unloaded (Less than five percent difference between peaks) becomes one or the other 4) Partially loaded site (single peak between 40 and 60 kips for Class 9s) becomes loaded or unloaded 1) Loaded tail beyond 110 percent of legal maximum goes from zero to non zero 2) Loaded tail more than two bins from loaded peak doubles in magnitude 3) Loaded tail beyond 110 percent of legal maximum goes to zero 1) Unloaded tail left of 90 percent of tare weight goes becomes non-zero 2) Unloaded tail more than two bins from unloaded peak doubles in magnitude

69 The values needed for vehicle distribution review ( Table A.2-3) are: average volume (group), average volume (class), minimum volume and maximum volume. There are several possible average volumes depending on the graph being used. Reference graphs can be prepared with the critical values marked. If the values are computed the following guidelines should be used. The graphs are weekly, monthly, yearly and multi-year. The multi-year graphs use yearly averages. Weekly - Average volume is same as that for the 4-card vs. 7-card review Monthly - Average volume is 4.3 times the average weekly volume Yearly Average is 52 times the average weekly volume Rounding for averages is as follows: o For sites with more than 1000 round to the nearest 50 o For sites with 250 to 1000 round to the nearest 10 o For sites with less than 250 round to the nearest 5 Minimum volume is 25 percent of the average volume. o Values under 100 rounded down to the nearest 5 o Values from 100 to 1000 rounded down to the nearest 10 o Values over 1000 rounded down to the nearest 50 Maximum volume is 125 percent of the average volume. o Values under 100 rounded up to the nearest 5 o Values from 100 to 1000 rounded up to the nearest 10 o Values over 1000 rounded up to the nearest 50 Table B.1-3 Purge Codes for Vehicle Distribution Review Code Description Limiting value Purge Identification Criteria 72 Atypical pattern 1) Change in predominant truck types either through additions or removal 2) Large change in volume either increase or decrease

70 Code Description Limiting value Purge Identification Criteria 102 Too many unclassifieds Average weekly volume > 100 1) The percentage of unclassified vehicles increase by more than 5 percent. 2) Any class that is about ten percent of the heavy truck population and has an expected weekly volume of more than 150 that effectively disappears from the distribution for several weeks. Note that such an occurrence is entirely possible for low volume sites. 103 Outside +/-10 percent range Average volume > 250 3) More than ten percent unclassifieds in total. This must be considered in context with the population being evaluated since the same value is used for unclassified in all groups. 1) Any class that is twenty percent or more of the heavy truck population and whose observed percentage in the population is outside of the range of +/- 10% of the expected percentage. 2) A class that is about ten percent of the heavy truck population and has an expected weekly volume of more than 150 that effectively disappears from the distribution for several weeks. It is entirely possible for low volume sites that there will be an occasional missing week. 135 Shift in location of maximum volume/ percentage 136 Shift in location of minimum volume/ percentage 137 Disappearance of volume difference Average volume > 250 Average volume > 250 Average volume > 250 3) A class other than Class 9 that was not expected to be at least ten percent, becomes fifteen percent or greater (including unclassifieds). The dominant class changes or the percentage in a dominant class is twothirds or less of its typical value A class (not the unclassifieds) with less than five percent of the population exceeds ten percent of the population A pattern of 4-card volumes being greater or less than 7-card volumes becomes same volume for both

71 Code Description Limiting value Purge Identification Criteria 138 Larger than expected volume difference Average volume > 250 1) Volume difference exceeds ten percent when there is no expected difference set for the site. 139 Missing expected vehicle classes 140 Volume distribution inconsistent with history 141 Percentage distribution inconsistent with history card/7-card difference too large card distribution inconsistent with 7- card card distribution inconsistent with 4- card 145 Volume larger than the expected maximum 146 Volume less than the expected minimum Average volume for missing class > 50 Average weekly volume > 100 Maximum volume > 300 Minimum volume < 175 2) Volume difference changes by more than twenty five percent A class disappears from the distribution for three or more consecutive weeks Relative volumes do not match the expected pattern Relative percentages do not match the expected pattern Difference in volumes, excluding the unclassifieds/unknowns (Class 15 or 20) is greater than ten percent When the 7-card distribution has been defined as valid for the site and the 4-card distribution using the same classification algorithm does not match it. 4-card data is data to be purged When the 4-card distribution has been defined as valid for the site and the 7-card distribution using the same classification algorithm does not match it. 7-card data is data to be purged Volume exceeds expected maximum by more than ten percent of the expected maximum Volume is below the expected minimum by more than ten percent of the minimum value

72 The values needed for axle review (Table B.1-4) are: legal maximum weight by axle group, typical class minimum axle weight, average weekly volume, the volume represented by the curve and the volume on the graph. The average weekly volume is same as that for the 4-card vs. 7- card review. To determine the average number in a curve use the following Monthly, by Week divide by 5 (or the number of weeks if shown by week) Table B.1-4 Purge Codes for Axle Distribution Review Code Description Limiting value Purge Identification Criteria 72 Atypical distribution See 153, 154, 155, ) A pattern variance from the expected at the site not described by one or more other codes 73 Over calibrated 1000 or more per week - 1 bin 500 or fewer per week 2 bins 1000 or more per week 10 percent of legal maximum to nearest bin 500 or fewer per week 20 percent of legal maximum 74 Under calibrated 1000 or more per week - 1 bin 500 or fewer per week 2 bins 110 Axle distribution inconsistent with history 1000 or more per week 10 percent of legal maximum to nearest bin 500 or fewer per week 20 percent of legal maximum 2) A pattern for a new site without validation information that does not match the expected typical pattern for the class Class 9s - Both peaks shifted equally to the right AND outside of their expected bounds. Other classes Dominant peaks shifted equally to the right from their expected peaks. Class 9s - Both peaks shifted equally to the left AND outside of their expected bounds. Other classes - Dominant peaks shifted equally to the left from their expected peaks. 1) Dominant peaks is shifted more than one bin from the expected curve in either direction.

73 Code Description Limiting value Purge Identification Criteria 111 Large percentage of heavy axles 112 Large percentage of light axles 500 or more axles in curve being assessed 500 or more axles in curve being assessed 2) The shape of a curve is different from the expected curve. Different is o A uni-modal (single peak) distribution becoming bi-modal (two-peaks), (Use 153) o A distinctly loaded or unloaded site shifting to a partially loaded site (and remaining uni-modal) (Use 154), o Doubling of the size of either the tail below the lower bound of the unloaded peak or above the upper bound of the loaded peak. (Use 156 or 155 respectively) This code is preferred to 72 (atypical distribution) because it is a more explicit description of an irregularity. 1) More than five percent of single axles are in bins beyond 20,000 pounds. 2) The value of the single axle distribution is non-zero past 22,000 pounds. 3) More than five percent of tandem axles are in bins beyond 40,000 pounds. 4) The value for the tandem axle distribution is non-zero past 44,000 pounds. 1) More than five percent of the single axle distribution is less than 6,000 pounds except for Class 5 where the values are 15 percent and 3,000 pounds. 113 Shifted heavy axle peak 250 axles or more in curve being 2) More than five percent of the tandem axle distribution is less than 8,000 pounds. The heavy (loaded) peak is more than two bins or by 15 percent of the legal

74 Code Description Limiting value Purge Identification Criteria assessed maximum (rounded to the nearest bin) from the expected location. 114 Shifted light axle peak 250 axles or more in curve being assessed The light (unloaded) peak is more than two bins or by 15 percent of the legal maximum (rounded to the nearest bin) 153 Change in number of modes for axle distribution At least 1000 axles on graph from the expected location. 1) One peak changing to two peaks 2) Two peaks changing to one peak 3) Two peaks going to three peaks where the third peak is at least a third of the next higher peak 154 Change in dominant loading type 155 Large change in loaded tail of axle distribution 156 Large change in unloaded tail of axle distribution At least 500 axles on graph 250 or more axles in curve being assessed 250 or more axles in curve being assessed 4) Three peaks going to two peaks 1) Loaded site becomes unloaded 2) Unloaded site becomes loaded 3) Site not clearly loaded or unloaded (Less than five percent difference between peaks) becomes one or the other 4) Partially loaded site (single peak between 40 and 60 kips for Class 9s) becomes loaded or unloaded 1) Loaded tail beyond 110 percent of legal maximum goes from zero to non zero 2) Loaded tail more than two bins from loaded peak doubles in magnitude 3) Loaded tail beyond 110 percent of legal maximum goes to zero 1) Unloaded tail left of 90 percent of tare weight goes becomes non-zero 2) Unloaded tail more than two bins from unloaded peak doubles in magnitude

75 B.2 4-card vs. 7-card Week by date Tables used: DD_CL_CT, DD_WT_CT Type abbreviation: Ag47WD Default: Class 9 This graph is used to check volume ranges in frequently reviewed sites for classification and weight data. Both data types do not need to be present for the graph to be useful. This graph can be used for individual classes or for groups of vehicles. To identify deviations from typical conditions the user will need expected volumes or a weekly graph from a reference period for the class (group) of interest. Day of week patterns will not be obvious due to the datebased nature of the graphs. A reference graph can be used to store expected volumes for a sevenday period. The expected output of a 4-card\7-card comparison varies with the population being compared and the classification algorithm being used. When classification and weight data are created using different schemes (i.e. the weight data is TMG 13-bin and the classification data is length based class (i.e. 6-bins)) then 13-bin graphs should be used for 4-card\7-card volume comparisons. Only day of week comparisons can be made with 13-bin 4-card vs. 7-card graphs. When the population is a single class with continuous data from the same or different equipment at the same location, the values should match on a day-to-day basis when both pieces of equipment are working. The match is corresponding highs and lows when groups of vehicles are being considered. When the population is a single class with sampled classification data from the same or different equipment at the same location, monthly graphs should be used. When the population is multiple classes and does not include passenger vehicles, the same expectations exist that apply for single class populations. Weekly 4-card vs. 7-card graphs should not be used for vehicles (1-20) unless cars and trucks are weighed OR a reference graph for the expected volumes is used to evaluate data reasonability. Weekly 4-card vs. 7-card graphs should not be used to evaluate data from the 29 th of the month or later. The purge codes for a 4-card vs. 7-card review are in Table A.2-1, page 3. The graphs for this section are for the default, Class 9, unless otherwise indicated.

76 IA, 0100, South, Lane 1-03/15-21/ Class 9 4-card vs. 7-card Figure B vs. 7- card Weekly, Week by Date Example 1 Error! Reference source not found.error! Reference source not found. is a typical graph for the Class 9 default. Note that the values are virtually the same. A probable cyclic pattern exists but the determination of which day of the week belongs with which volume will require a calendar to determine. Because the 1 st day of the graph week will fall on a different day of the week for each month and vary by month over years, the range of values, not the shape of the curve is the basis for acceptance/rejection. Purge candidate Not if typical range for site Figure B vs. 7-card Weekly, Week by Date Example 2 Figure B.2-2 illustrates why the 5 th week of the month should not be evaluated solely by this graph. Note the distortion of the horizontal axis. volume range. Purge candidate Do NOT make a purge decision based on this graph for these days (the fifth week) without knowing expected

77 NE, 1030, West, Lane 1-12/22-28/ Class 9 4-card vs. 7-card Figure B vs. 7-card Weekly, Week by Date Example 3 Figure B.2-3 is an example where knowing the expected volumes AND looking at the dates is critical. The potential issue here for Class 9 vehicles is that the dot in the middle is Christmas Day. sensor problems (70) Purge candidate Maybe What 12/27-28 Reason - Lower volumes than expected, possible Figure B vs. 7-card Weekly, Week by Date Trucks 4-20 Figure B.2-4 shows the truck population. A gap of varying sizes would be expected at a site where there is a 3/5 classification problem. The match between the two types should not be considered exact due to the scale (100s.) Purge candidate No

78 Figure B vs. 7-card Weekly, Week by Date Heavy Trucks (6-20) between and Class 3 and Class 5 vehicles. Figure B.2-5 shows the heavy truck population only. It is useful at sites with problems distinguishing Purge candidate No Figure B vs. 7-card Weekly, Week by Date Vehicles (1-20) Figure B.2-6 illustrates why this graph type should not be uses without knowledge of expected volumes for the population as a whole. This is a somewhat unusual site due

79 to its high percentage of trucks. Purge candidate - No

80 B.3 4-card vs. 7-card Month by date (DEFAULT) Tables used: DD_CL_CT, DD_WT_CT Type abbreviation: Ag47MD Default: Class 9 The Month by Date graph is intended to see that volumes are in line with expectations and to determine whether continuous or sampled data is available from the data provided. The graph shows if there is a recurring pattern in vehicle volumes and whether or not the classification and weight equipment are seeing the same number of vehicles. Both data types do not need to be present for the graph to be useful. This graph can be used for individual classes or for groups of vehicles. The expected output of a 4-card\7-card comparison varies with the population being compared and the classification algorithm being used. When classification and weight data is created using different schemes (i.e. the weight data is TMG 13-bin and the classification data is length based (i.e. 6-bins)), 13-bin graphs should be used for 4-card\7-card volume comparisons. Only day of week comparisons can be made with 13-bin 4-card vs. 7-card graphs. To determine whether the difference in vehicles between pieces of equipment is the result of seeing different numbers or different classification schemes the vehicle distribution graphs must be used. When the population is a single class with continuous data from the same or different equipment at the same location, the values should match on a day-to-day basis when both pieces of equipment are working. When the 4-card and 7-card data comes from different pieces of equipment the patterns should match. In this instance vehicle classes that require weight trigger to move accurately classify vehicles should also be checked. These include buses, 2-axles single unit trucks, and 3- or 4- axle tractor-trailer combinations. For sites where recurring congestion exists multiple trailer vehicle classes weight characteristics should be reviewed. When the population is a single class with sampled classification data from the same or different equipment at the same location, monthly graphs should be used. When the population is multiple classes and does not include passenger vehicles, the same expectations exist that apply for single class populations. A 4-card record may be absent and a 7-card present if less than twenty-four hours of data collected. Retention of the 7-card data will be based on how closely the volume matches previous patterns. The purge codes for a 4-card vs. 7-card review are in Table A.2-1, page 3. The graphs for this section are for the default, Class 9, unless otherwise indicated.

81 IA, 0600, South, Lane 1 - July Class 9 4-card vs. 7-card Figure B vs. 7-card Monthly, Month by Date Example 1 Figure B.3-1 is an example of an ideal comparison graph. However, due to the scale, it should not be assumed that the volumes observed are exactly the same. Purge candidate - No VT, 1683, South, Lane 1 - March Class 9 4-card vs. 7-card Figure B vs. 7-card Monthly, Month by Date Example 2 Figure B.3-2 has gaps between the observed volumes but they are less than 10 percent by inspection. Purge candidate - No

82 11/29/2011 Page 1 of 12 IA, 0100, South, Lane 1 - January Hvy Trks (6-20) 4-card vs. 7-card Figure B vs. 7-card Monthly, Month by Date Heavy Trucks Figure B.3-3 shows the expected match for a heavy truck pattern. This is ideal and a slight gap may occur depending on site conditions. Purge candidate No 11/29/2011 Page 1 of 12 IA, 0100, South, Lane 1 - January Trucks (4-20) 4-card vs. 7-card Figure B vs. 7-card Monthly, Month by Date Trucks Figure B.3-5 shows an ideal match for a truck pattern. Gaps can occur based on site conditions including the ability of the equipment to correctly identify Class 3s and Class 5s. Purge candidate No

83 11/29/2011 Page 1 of 1 IA, 0100, South, Lane 1 - March Vehicles (1-20) 4-card vs. 7-card Figure B vs. 7-card Monthly, Month by Date Vehicles Figure B.3-5 shows how much less variable the truck patterns are than the vehicle pattern as a whole. Unless a site needs an AADT estimate, the vehicle variability is not of particular interest. Figure B vs. 7-card Monthly, Month by Date Vehicles Figure B.9-6 shows a large difference (more than 10 percent) between 4-card and 7-card volumes. Reason: Failure of the equipment OR the data are coming from two different equipments for that site. Purge candidate Yes

84 B.4 GVW Month (DEFAULT) Table used: DD_GVW Type abbreviation: AgGVWMMon Default: Class 9 This graph is used to determine whether the distribution of gross vehicle weights (GVW) is rational for a given class. A monthly aggregation is a quick way to check the curve but it may hide weeks of poor data. This graph is best used for sites that have stable patterns or small volumes of the classes of interest. There are two varieties of this graph type. The Class 9 graph has two pairs of vertical reference lines. The first set at 28 and 36 kips and the second set at 72 and 80 kips. For all the other vehicles types no reference lines exist. A month s data should be accepted without further investigation if: For class 9 o An unloaded peak, if it exists, falls on or between the left pair of vertical reference lines o A loaded peak, if it exists, falls on or between the right pair of vertical lines o All points from 16 kips to the left are 2 percent or less o All points from 84 kips to the right are 2 percent or less o The x-axis goes from 0 to 100 or less. For all other vehicle classes (subject to vehicle type and local regulations) o Values for 0 or 2 are 10 or less o All points for 72 and larger are 2 percent or less o The pattern is consistent with the reference pattern for the class. GVW review purge codes are in The values needed for GVW review (Table A.2-2) are: tare weight, legal maximum weight, average weekly volume, the volume represented by the curve and the volume represented by the graph. The average weekly volume is same as that for the 4-card vs. 7-card review. To determine the average number in a curve use the following Weekly, by Week; Monthly, by Month use volume in header Weekly, by Day of Week Divide the volume in the header by the number of days present plus 1. (The plus 1 is to offset the potential for missing days at low volume sites and the tendency to have sharp differences between weekday and weekend volumes particularly for Class 5 and Class 9 vehicles.) Monthly, by Week divide by number of weeks shown plus 1 not to exceed 5 (or the number of weeks if shown by week) While expected values for 5-axle tractor semi-trailers are incorporated in the software, expected values for other classes that are needed to evaluate site loading are not. For all other classes unloaded is within 15 percent of the typical tare weight of the class. The tare weight is the

85 unloaded weight of the truck as defined by a manufacturer of that type. The typical tare weight should be for the predominant vehicle in the class if multiple configurations are assigned the same classification. An example is buses where both two axle and three axle versions exist. Loaded is between 80 percent and 100 percent of the legal maximum for the class. Where multiple vehicle types are included in a class, multiple peak values may need to be established. Table A.2-2, page 6. Unless otherwise indicated, the graphs in this section are for Class 9.

86 Figure 11-1 Figure 11-2 values. Figure B.4-1 GVW, Monthly, by Month Example 1 This figure shows a distribution with peaks in the middle of the expected Purge candidate No Figure B.4-2 GVW, Monthly, by Month Example 1 (NEED TO FIND A CASE LIKE THIS) This figure shows a distribution with peaks at the upper range of the expected values. Due to the relatively high percentages of vehicles 84 kips and higher and the range of the x-axis this site

87 should be checked for the most recent data of calibrating and yearly trends and the pattern on the same month for previous years. Purge candidate Maybe What TBD Reason Large percentage of overweight trucks (109)

88 Figure B.4-3 GVW, Monthly, by Month Class 5 (NC-different region-needed) The graphs for other than class 9 vehicles do not show expected ranges for unloaded and loaded peaks, the typical weights for other classes or vehicles of interest for other classes or vehicles of interest should be known by the review. Figure B.4-3, the example to the left, shows a small percentage of vehicles under 4,000 pounds indicating that if this represents 2-axles single unit trucks, it is not picking up unusual numbers of pickups or other 2-axle passenger vehicles. Purge candidate - No Figure B.4-4 GVW, Monthly, by Month Example 3 Figure B.11-4 shows that the loaded peak is outside of the vertical limit lines. Purge candidate Maybe Reason: loaded peak has shifted two or more bins form the legal maximum expected position (105). Figure B.4-5 GVW, Monthly, by Month Example 4 (needed) Figure B.11-5 shows a graph with only one peak. Purge candidate Yes

89 Figure B.4-6 GVW, Monthly, by Month Example 5 (NEEDED) Figure B.11-6 shows a large percentage of tractor trailers over 80 kips. Purge candidate Yes Reason: the portion of the curve to the right of the 80 kips bin mark is greater than five percent (75). Figure B.4-7 GVW, Monthly, by Month Example 6 Figure B.11-7 shows a large percentage of tractor trailers under 12 kips. Purge candidate Yes Reason: the portion of the curve to the left of the 12 kip mark is greater than two percent (76).

90 GVW Month by Week Table used: DD_GVW Type abbreviation: AgGVWMWk Default: Class 9 This graph is used to determine whether the distribution of gross vehicle weights (GVW) is rational for a given class. The GVW month by week allows review of consistency of GVW distributions. It is particularly useful if the monthly graph looks unreasonable. The by week graph may show an unusual week, or a drift during the month. To use this graph there should be at least 125 vehicles expected in the class for the week. There are two varieties of this graph type. The Class 9 graph has two pairs of vertical reference lines. The first set at 28 and 36 kips and the second set at 72 and 80 kips. For all the other vehicles types no reference lines exist. A week s data should be accepted without further investigation if: For class 9 o The unloaded peak falls between the left pair of vertical reference lines o The loaded peak falls between the right pair of vertical reference lines o At least one of the peaks is centered between the vertical reference lines o All points from 16 kips to the left are 2 percent or less o All points from 84 kips to the right are 2 percent or less o The x-axis goes from 0 to 100 or less. For all other vehicle classes (subject to vehicle type and local regulations) o Values for 0 or 2 are 10 or less o All points 72 and larger are 2 percent or less o The pattern is consisted with the reference pattern for the class where consistent is within 1 increment on the vertical axis at all points GVW review purge codes are in The values needed for GVW review (Table A.2-2) are: tare weight, legal maximum weight, average weekly volume, the volume represented by the curve and the volume represented by the graph. The average weekly volume is same as that for the 4-card vs. 7-card review. To determine the average number in a curve use the following Weekly, by Week; Monthly, by Month use volume in header Weekly, by Day of Week Divide the volume in the header by the number of days present plus 1. (The plus 1 is to offset the potential for missing days at low volume sites and the tendency to have sharp differences between weekday and weekend volumes particularly for Class 5 and Class 9 vehicles.) Monthly, by Week divide by number of weeks shown plus 1 not to exceed 5 (or the number of weeks if shown by week) While expected values for 5-axle tractor semi-trailers are incorporated in the software, expected values for other classes that are needed to evaluate site loading are not. For all other classes

91 unloaded is within 15 percent of the typical tare weight of the class. The tare weight is the unloaded weight of the truck as defined by a manufacturer of that type. The typical tare weight should be for the predominant vehicle in the class if multiple configurations are assigned the same classification. An example is buses where both two axle and three axle versions exist. Loaded is between 80 percent and 100 percent of the legal maximum for the class. Where multiple vehicle types are included in a class, multiple peak values may need to be established. Table A.2-2, page 6. Unless otherwise indicated, the graphs in this section are for Class 9. NC, 0200, South, Lane 1 - March Class 9 GVW Distribution - No. weighed: Figure B.4-8 GVW, Monthly, by Week Example 1 In Error! Reference source not found. a review of the weekly patterns shows consistent unloaded peaks except for the third week (in green). For this week both the unloaded and loaded peaks have shifted right. The data for this week should be further scrutinized using a day of week graph and inquiring about special circumstances that may have affected loading. Note that questions already exist about the reasonableness of the value of the loaded peak. Purge candidate Yes What All weeks Reason Large percentage of overweight trucks (109)

92 NC, 0200, South, Lane 1 - April Class 5 GVW Distribution - No. weighed: 4962 Figure B.4-9 GVW, Monthly, by Week Class 5 Figure B.4-9 is an example of monthly, weekly graphs for other that class 9 vehicles. It shows somewhat different patterns by week. However, the relative locations of the two peaks and the percentage in the right hand tail are essentially invariant. This group would be accepted Purge candidate - No 11/30/2011 Page 1 of 1 PA, 1597, East, Lane 1 - October Class 9 GVW Distribution - No. weighed: 945 Figure B.4-10 Monthly, by Week Example 3 Figure B.12-3 shows all the loaded peaks are to the left of the maximum expected range. Purge candidate - Yes Reason: loaded peaks have shifted two or more bins to the left. (105) Figure B.4-11 Monthly, by Week Example 4 (CASE NEEDED) Figure B.12-4 shows two peaks changing to one.(149)

93 Purge candidate Yes Figure B.4-12 Monthly, by Week Example 5 (case needed) Figure B.12-5 shows only unloaded peak exists for all the weeks in Oct Purge candidate - Yes Figure B.4-13 Monthly, by Week Example 6 (case needed) Figure B.12-6 shows a large percentage of tractor trailers under 12 kips. Purge candidate - Yes Reason: the proportion of the curve to the left of the 12 kip mark is greater than two percent.

94 B.5 Vehicle Distribution Month by Week (1.6) Tables used: DD_CL_CT, DD_WT_CT Type abbreviation: AgDistMWk Default: Trucks (4-20) This graph is used to determine whether the mix of trucks in the vehicle population is stable. The distributions for classification data and weight data are graphed on the same figure. Classification data is in the left hand column and weight data in the right. A week s data for a data group should be accepted without further investigation if: Unclassified vehicles make up less than five percent of the total. A heavy truck graph (Classes 6-20) o All classes greater than 10 percent of the population are within +/- 10 percent of their o expected value. The percentage of unclassifieds is less than the expected value for the site or less than two percent for the group if there are no unclassifieds expected. For all other distributions o The percentage of unclassifieds is less than the expected value for the site or less than 5 percent for the site if there are no unclassifieds expected. o The Class 9 percentage is greater than the Class 8 percentage (5 times is typical) The purge codes for vehicle distribution review are found in section Table A.2-3 on page 10. Unless otherwise indicated, the graphs in this section are for Trucks (4-20).

95 Figure B.5-1 Vehicle Dist Monthly, Month by Week Example 1a (NEEDED) Purge candidate Yes What Month Reason Too many unclassifieds (100) 11/30/2011 Page 2 of 2 IN, 3003, North, Lane 1 - January Trucks (4-20) Vehicle Distribution (7-card) - No. weighed 21,401 Figure B.5-2 Vehicle Dist Monthly, Month by Week - Example 1b Purge candidate No

96 Figure B.5-3 Vehicle Dist Monthly, Month by Week - Example 2 (NEEDED) Purge candidate Yes What Class data Reason Too many unclassifieds (100) 11/30/2011 Page 2 of 2 IN, 5518, North, Lane 1 - January Hvy Trks (6-20) Vehicle Distribution (7-card) - No. weighed 73,393 Figure B.5-4 Vehicle Dist Monthly, Month by Week Heavy Trucks Purge candidate - No

97 11/30/2011 Page 1 of 2 IN, 5022, North, Lane 1 - September Vehicles (1-20) Vehicle Distribution (4-card) - No. counted 308,870 Figure B.5-5 Vehicle Dist Monthly, Month by Week Heavy Trucks Purge candidate No Figure B.5-6 Vehicle Dist Monthly, Month by Week (needed) Figure B.18-6 shows that the percentage of class 8s is almost equal to class 9s. Purge candidate Yes Reason: percentage of class 8s should be less than class 9s, at least

98 B.6 Axle Distribution Month (DEFAULT) Table used: DD_AX Type abbreviation: AgAxMMon. Default: Class 9 - Singles and Tandems The axle distributions are available for use in when GVW distributions are unavailable. The single axle graphs have an expected maximum line at 20,000 pounds. The tandem axle graphs have a similar line at 36,000 pounds. The use of these limiting values is still under development in the review process. Generally a class 9 single axle graph will be effectively unimodal with a peak around the steering axle weight and a possible second definitive peak to the right of that. The heavier peak will exist when there are a significant percentage of either 2S3 s or class 9 s with split tandems. In the former case the weight is from the single drive axle on the tractor. In the latter, each axle in the split tandem is recorded as a single because the group functions as two singles. This graph is used to determine whether the distribution of loading using axles as a surrogate for GVW is rational for a given class. This graph should be limited to loading investigations when there are too few vehicles to generate a monthly GVW graph. To use this graph there should be at least 60 vehicles in the class expected in a month. There is a single vertical line where the expected legal limit is for the axle group (singles and tandems). A month s data should be accepted without further investigation if: o The curve has the same shape as the reference graph (no more than a y-axis increment s difference at any point) o The peak(s) are within 2 bins of their expected locations o Less than 5 percent of the single axles are in bins beyond 20,000 pounds or the percentage in any bin past 22,000 pounds is effectively 0 o Less than 5 percent of the tandem axles are in bins beyond 40,000 pounds or the percentage in any bin past 44,000 pounds is effectively 0 A determination on data reasonableness can be made even when graphs for all of the selected axle groups are not available. The purge codes for axle distribution review are in Table B.2-4, page B-18.

99 11/30/2011 Page 1 of 1 VT, 1683, South, Lane 1 - March Class 9 Single Axle Distribution - No. axles 4,801 Figure B.6-1 Axle Distribution - Month Vehicle Class 9 - Axle Group 1 - Singles This is a typical single axle graph at a site where few if any class 9 s have split tandems. Purge candidate No 11/30/2011 Page 1 of 1 VT, 1683, South, Lane 1 - June Class 9 Tandem Axle Distribution - No. axles 11,727 Figure B.6-2 Axle Distribution Month Vehicle Class 9 Axle Group 2 Tandems This is a typical tandem axle graph showing a loaded peak at the expected location and few groups beyond 40,000 pounds. Purge candidate - No

100 Figure B.6-3 Axle Distribution - Month Vehicle Class 9 - Axle Group 1 Singles (CASE NEEDED) Figure B.18-3 shows more than 5 percent of the single axles are in bins beyond 22,000 pounds. Purge candidate - Yes Figure B.6-4 Axle Distribution Month Vehicle Class 9 Axle Group 2 Tandems (Case needed) Figure B.18-3 shows more than 5 percent of the tandem axles are in bins beyond 44,000 pounds. Purge candidate - Yes

101 IN, 5022, North, Lane 1 - December Class 9 Tandem Axle Distribution - No. axles 126,906 Figure B.6-5 Axle Distribution Month Vehicle Class 9 Axle Group 2 Tandems In figure B.18-5 the location of maximum peaks shifted to the right. Purge candidate - Yes Figure B.6-6 Axle Distribution - Month Vehicle Class 9 - Axle Group 1 Singles (CASE NEEDED) The location of the maximum peak shifted to the right. Purge candidate - Yes

102 B.7 Axle Distribution Month by Week Table used: DD_AX Type abbreviation: AgAxMWk. Default: Class 9 - Singles and Tandems This graph is used to determine whether the distribution of loading using axles as a surrogate for GVW is rational for a given class. This graph should be limited to loading investigations when there are insufficient number of vehicles to generate a by week GVW graph for a truck in Classes 6 and higher and where a problem is noted within the GVW Monthly By Month graph. To use this graph there should be at least 60 vehicles in the class expected in a week. There is a single vertical line where the expected legal limit is for the axle group (singles and tandems). A month s data should be accepted without further investigation if: o The curves within each graph have the same shape. o The peak(s) are within 2 bins of their expected locations. o Less than 5 percent of the single axles are in bins beyond 20,000 pounds or the percentage in any bin past 22,000 pounds is effectively 0. o Less than 5 percent of the tandem axles are in bins beyond 40,000 pounds or the percentage in any bin past 44,000 pounds is effectively 0 A determination on data reasonableness can be made even when graphs for all of the selected axle groups are not available. The purge codes for axle distribution review are in Table B.2-4, page B-18. VT, 1683, South, Lane 1 - April Class 9 Single Axle Distribution - No. axles 5,103 Figure B.7-1 Axle Distribution Month by Week Vehicle Class 9 - Axle Group 1 Singles This is a typical single axle graph at a site where few if any class 9 s have split tandems. Purge candidate No

103 VT, 1683, South, Lane 1 - February Class 9 Tandem Axle Distribution - No. axles 8,677 Figure B.7-2 Axle Distribution Month by Week Vehicle Class 9 - Axle Group 2 - Tandems This is a tandem axle graph showing a loaded peak beyond the expected location and more than 5% of axle groups beyond 40,000 pounds, particularly during the third week in February. of heavy axles (111) Purge candidate Yes Reason Large percentage Figure B.7-3 Axle Distribution Month by Week Vehicle Class 9 - Axle Group 1 Singles (CASE NEEDED) Maximum peaks shifted to the right and some of the weeks have more 5% beyond 20,000 pounds. Purge candidate Yes

104 Figure B.7-4 Axle Distribution Month by Week Vehicle Class 9 - Axle Group 2 Tandems (CASE NEEDED) Figure B.19-4 shows more than 5 percent of the tandem axles are in bins beyond 44,000 pounds. Purge candidate - Yes Figure B.7-5 Axle Distribution Month by Week Vehicle Class 9 - Axle Group 1 Singles (CASE NEEDED) Figure B.19-5 shows more than 5 percent of the single axles are in bins beyond 22,000 pounds. Purge candidate - Yes

105 C 13-bin Graphs (TMG classification) These graphs are created using the MM_* or YY_* tables as applicable. The data in these graphs have been transformed into the 13-bin classification scheme described in FHWA s Traffic Monitoring Guide. This makes comparisons between sites possible since the classes that are used to define truck groups are the same across all locations. The transformation into this scheme also limits the classes for which loading graphs are available to classes 4-13 because of the way the GVW and axle table contents have been defined and computed. If a user is interested in the loading patterns of passenger vehicles, agency specific classes, unknowns or unclassifieds they will need to use the agency graphs. Every 13-bin graph is replicated in the agency graph options. This document provides an example of the default graph for each case. Additional illustrations are provided where graphs vary by vehicle population or class. For most graphs included in the software, more examples are provided for cases requiring further investigation. Figure C-1 13-bin Availability (1.7) These 13-bin graphs use a different definition of trucks, heavy trucks and vehicles from the SPS, Agency and STAT_QC graphs. Trucks are Classes 4 to13. Heavy trucks are classes 6 to13. Vehicles are either Classes 1 to 13 or all records, Classes 1 to 13 and 15. In the figure captions Vehicles consists only of Classes 1 to 13 unless labeled otherwise.

106 C.1 Graph intervals Graph intervals define the amount of data aggregated to be represented by a single page. For 13- bin graphs the intervals are monthly, yearly and multi-year. C.1.1 Monthly There are no monthly default graphs. Monthly graphs are used to review sites that have the following characteristics: More than 125 vehicles in a class/group per month More than 125 vehicles in a class/group per week for Month by Day of Week alternatives Monthly graphs may also be used for diagnostics when pattern problems or expected value issues from Agency Weekly graphs need to be investigated. They are 4-cards vs. 7-cards. C.1.2 Yearly There are four default graphs defined for this interval. They are the figures expected to be included in an annual review packet. Yearly graphs are used to review sites that have the following characteristics: More than 125 vehicles in a class/group per month More than 125 vehicles in a class/group per day of week for Month by Day of Week alternatives Yearly graphs may also be used for diagnostics when pattern problems or expected value issues from Monthly graphs need to be investigated. C.1.3 Multi-Year No default graphs are defined for this interval. Yearly graphs are used to review sites that have the following characteristics More than 125 vehicles in a class/group per month More than 125 vehicles in a class/group per day of week for Month by Day of Week alternatives At least two years of data Multi-year graphs are used to see when shifts in patterns that are consistent within a year are inconsistent with history. Multi-year graphs can be considered for addition to an annual QC packet when a year is being questioned, particularly a past year, or for information.

107 C.2 Purge Codes The tables in this section provide a list of codes and some basic decision criteria by graph type. Purge criteria are not absolute. Multiple elements must be considered including the amount of monitored data before making a final purge recommendation. These codes are unlikely to be used as the majority of the reviews inudve agency graphs. Duies are applied to daily tables. They will be identifier using 13-bin graphs when the incoming data cannot be easily reviewed using the actual agency classes. In the tables that follow some purge identification criteria have associated limiting values. This means that unless the data element being checked meets that condition, that particular criteria should not be used for purge identification. This eliminates purge conditions that don t make sense due to variability in small sample sizes. The limiting value relates to a site characteristic such as an average or minimum value derived from a valid data set. The data is valid because of either consensus review or a validation or calibration. The limiting value criterion is the same whether the element being reviewed is day of week, single class or group of classes. When applying the purge selection criterion it is not necessary to compute the difference. The differences can be assessed by inspection. Generally, determining what a percentage means in terms of increments on the vertical (y-) axis is sufficient. A valid monthly graph with the limiting values marked on it for comparison is another way to apply the criteria. In this case, attention must be paid to the scale of the vertical axis. New comparison graphs for volume in particular will be needed as volumes increase over time. The values generally needed for 4-card vs. 7-card review (Table C.2-1) are: Average monthly volume, average day of week volume and minimum day of week volume. The values may be computed or obtained by inspection of a valid monthly graph. They are different for each class or group to be evaluated. They are defined as follows: Average monthly volume 4 times the sum of the average day of week volumes for the validation month or first month with continuous data. Daily average volume o For sites with more than 4000 per month averages are computed on a by day of week basis and rounded to the nearest 50 o For sites with 1000 to 4000 per month averages are computed for weekdays as a group (the average of the weekday averages) and weekend days as a group (the average of the weekend averages) and rounded to the nearest 10 o For sites with less than 1000 per month the average is the average of the day of the week averages and rounded to the nearest 5 Minimum volume 25 percent of the average volume o Values under 100 rounded down to the nearest 5 o Values from 100 to 1000 rounded down to the nearest 10 o Values over 1000 rounded down to the nearest 50 The process is discussed in detail in LTPP Monitored Traffic Data Processing. Included in the discussion is how to pick the groups for graphic evaluation.

108 Table C.2-1 Purge Codes for 4-card vs. 7-card Review Code Description Limiting Value Purge Identification Criteria 64 Zero data Minimum volume at least 10 Class or group with zero volume (or effectively zero) given the expected minimum. 65 Zero daily volume For reviews including all classes only, total volume for day is zero and it is not 70 Lower volumes than expected, possible sensor problems Minimum value at least Atypical pattern Average volume at least 1000 per month 100 Higher volume than expected Average volume at least 50 attributed to construction or weather Volume below the minimum. See 136, 137, 138, ) Shift in location of minimum or maximum where the choice of minimum or maximum is obvious (ten percent or greater below or above the other candidate days.) (Should use 101 instead.) 6) Pattern that has demonstrated a relatively constant volume retains its shape (low to high to low) but has a distinct increase or decrease in volume over time. (More than 5 percent a year requires a documented explanation.) 7) The shape changes or a distinct shape appears where the two lines have previously coincided. The disappearance of a shape (not just a little bit of differentiation between the lines) should be investigated. 8) For a new site without validation data a pattern that does not conform to typical expectations i.e. weekdays having larger volumes than weekends, classification volumes being greater than or equal to weight volumes. Volume above the maximum expected.

109 Code Description Limiting Value Purge Identification Criteria 101 Change in day of week pattern Average volume at least 1000 per month 3) Minimum changes day of week and the minimum was no more than 95 percent of the next higher value. 4) Maximum changes day of week and the maximum was at least 105 percent of the next lower value. 5) Minimum (maximum) changes between weekday day and weekend day and the difference between weekdays as a group and weekend days as a group is greater than 10 percent of the larger average. 136 Shift in location of maximum volume/ percentage 136 Shift in location of minimum volume/ percentage 137 Disappearance of volume difference 138 Larger than expected volume difference Average volume at least 1000 per month Average volume at least 1000 per month Average value for smaller volume at least 50 Average value for smaller volume at least 50 6) A one peak weekly volume cycle becomes a two peak cycle or vice versa Location of maximum in a recurring pattern shifts by more than one day in the week for 3 or more consecutive weeks. Location of minimum in a recurring pattern shifts by more than one day in the week for 3 or more consecutive weeks. A recurring gap pattern changes to lines on top of each other An expected volume difference exists of any magnitude and the observed difference is at least 110 percent of the expected difference. Both data types must use a consistent by not necessarily the same definition for a vehicle class or group of vehicles. The values needed for GVW review (Table C.2-2) are: tare weight, legal maximum weight, average monthly volume, the volume represented by the curve and the volume represented by the graph. The average monthly volume is same as that for the 4-card vs. 7-card review. To determine the average number in a curve use the following Monthly, by Month; Yearly, by Year use volume in header Monthly, by Day of Week; Yearly by Day of Week - Use volumes in 4-card vs. 7-card Month (Year) by Day of Week graph Yearly, Year by Month Use volume from Vehicle Distribution, Trucks, Yearly, Year by Month.

110 While expected values for 5-axle tractor semi-trailers are incorporated in the software, expected values for other classes that are needed to evaluate site loading are not. For all other classes unloaded is within 15 percent of the typical tare weight of the class. The tare weight is the unloaded weight of the truck as defined by a manufacturer of that type. The typical tare weight should be for the predominant vehicle in the class if multiple configurations are assigned the same classification. An example is buses where both two axle and three axle versions exist. Loaded is between 80 percent and 100 percent of the legal maximum for the class. Where multiple vehicle types are included in a class, multiple peak values may need to be established. Table C.2-2 Purge Codes for GVW Review Code Description Limiting value Purge Identification Criteria 72 Atypical distribution 73 Over calibrated 4000 or more per month - 1 bin 2000 or fewer per month 2 bins (See 149, 150, 151.) 3) A pattern variance from the expected at the site not described by one or more other codes 4) A pattern for a new site without validation information that does not match the expected typical pattern for the class Class 9s Unloaded and loaded peaks shifted equally to the right AND outside of their expected bounds or more per month 10 percent of legal maximum to nearest bin 2000 or fewer per month 20 percent of legal maximum 74 Under calibrated 4000 or more per month - 1 bin 2000 or fewer per month 2 bins 4000 or more per month 10 percent of legal maximum to nearest bin 2000 or fewer per month 20 percent of legal maximum 75 Large percentage of tractor trailers over 80 kips Class 9s only vehicles or more in curve being Other classes Curve has its reference shape and is shifted to the right from the expected location. Class 9s Unloaded and loaded peaks shifted equally to the left AND outside of their expected bounds. Other classes Curve has its reference shape and is shifted to left from the expected location. 1) The proportion of the curve to the right of the 80 kip bin mark is greater than five percent.

111 Code Description Limiting value Purge Identification Criteria assessed 2) The value for the distribution is non-zero beyond 100 kips. 76 Large percentage of tractor trailers under 12 kips The proportion of the curve to the left of the 12 kip mark is greater than two percent. 104 GVW distribution inconsistent with history 105 Large shift in loaded peak 106 Large shift in unloaded peak 107 GVW peaks outside expected limits (Class 9s) Class 9s only vehicles or more in curve being assessed 250 vehicles or more in curve being assessed 250 vehicles or more in curve being assessed 250 vehicles or more in curve being assessed 4) Class 9s - Either of the peaks is shifted more than one bin from the expected curve in either direction. 5) Other classes A dominant peak is shifted more than one bin from the expected curve in either direction. 6) The shape of a curve is different from the expected curve. Different is o A uni-modal (single peak) distribution becoming bi-modal (two-peaks), (Use 149) o A distinctly loaded or unloaded site shifting to a partially loaded site (and remaining uni-modal) (Use 150), o Doubling of the size of either the tail below the lower bound of the unloaded peak or above the upper bound of the loaded peak. (Use 152 or 151 respectively) This code is preferred to 72 (atypical distribution) because it is a more explicit description of an irregularity. The loaded peak has shifted two or more bins or by 15 percent of the legal maximum (rounded to the nearest bin) from its expected position. The unloaded peak has shifted two or more bins or by 15 percent of the legal maximum to the nearest bin from its expected position. Both peaks are outside the vertical lines marking the expected location of peaks for 5-axle tractor-trailer combinations.

112 Code Description Limiting value Purge Identification Criteria 108 Large percentage of underweight trucks 109 Large percentage of overweight trucks 149 Change in number of modes for GWV distribution 150 Change in dominant loading type 151 Large change in loaded tail of GVW distribution 152 Large change in unloaded tail of GVW distribution 500 vehicles or more in curve being assessed 500 vehicles or more in curve being assessed 1000 vehicles or more on graph 500 vehicles or more on graph 250 vehicles or more in curve being assessed 250 vehicles or more in curve being assessed 3) More than five percent of the trucks are less than ninety percent of the minimum tare weight for the population, 4) There is a non-zero value of the distribution for bins that contain weights less than ninety percent of the expected tare weight for the class. 3) More than five percent of the trucks exceed ten percent of the legal maximum for the class OR 4) There is a non-zero value for the frequency distribution in bins that contain weights that are ten percent or more than the legal maximum. 5) One peak changing to two peaks 6) Two peaks changing to one peak 7) Two peaks going to three peaks where the third peak is at least a third of the next higher peak 8) Three peaks going to two peaks 5) Loaded site becomes unloaded 6) Unloaded site becomes loaded 7) Site not clearly loaded or unloaded (Less than five percent difference between peaks) becomes one or the other 8) Partially loaded site (single peak between 40 and 60 kips for Class 9s) becomes loaded or unloaded 4) Loaded tail beyond 110 percent of legal maximum goes from zero to non zero 5) Loaded tail more than two bins from loaded peak doubles in magnitude 6) Loaded tail beyond 110 percent of legal maximum goes to zero 3) Unloaded tail left of 90 percent of tare weight goes becomes non-zero 4) Unloaded tail more than two bins from unloaded peak doubles in magnitude The values needed for vehicle distribution review (Table C.2-3) are: average volume (group), average volume (class), minimum volume and maximum volume. There are several possible

113 average volumes depending on the graph being used. Reference graphs can be prepared with the critical values marked. If the values are computed the following guidelines should be used. The graphs are monthly, yearly and multi-year. The multi-year graphs use yearly averages. Monthly - Average volume is 4 times the sum of the average day of week volumes for the validation month or first month with continuous data. Yearly Average is 12 times the average monthly volume Rounding for averages is as follows: o For sites with more than 1000 round to the nearest 50 o For sites with 250 to 1000 round to the nearest 10 o For sites with less than 250 round to the nearest 5 Minimum volume is 25 percent of the average volume. o Values under 100 rounded down to the nearest 5 o Values from 100 to 1000 rounded down to the nearest 10 o Values over 1000 rounded down to the nearest 50 Maximum volume is 125 percent of the average volume. o Values under 100 rounded up to the nearest 5 o Values from 100 to 1000 rounded up to the nearest 10 o Values over 1000 rounded up to the nearest 50 Table C.2-3 Purge Codes for Vehicle Distribution Review Code Description Limiting value Purge Identification Criteria 72 Atypical pattern 4) Change in predominant truck types either through additions or removal 5) Large change in volume either increase or decrease 102 Too many unclassified Average monthly volume > 400 4) The percentage of unclassified vehicles increase by more than 5 percent. 5) Any class that is about ten percent of the heavy truck population and has an expected weekly volume of more than 150 that effectively disappears from the distribution for several weeks. Note that such an occurrence is entirely possible for low volume sites. 6) More than ten percent unclassified in total. This must be considered in context with the population being evaluated since the same value is used for unclassified in all groups.

114 Code Description Limiting value Purge Identification Criteria 103 Outside +/-10 Average volume > 3) Any class that is twenty percent or more of the heavy truck population and whose observed percentage in the percent range 250 population is outside of the range of +/- 10% of the expected percentage. 4) A class that is about ten percent of the heavy truck population and has an expected weekly volume of more than 150 that effectively disappears from the distribution for several weeks. It is entirely possible for low volume sites that there will be an occasional missing week. 6) A class other than Class 9 that was not expected to be at least ten percent, becomes fifteen percent or greater 135 Shift in location of maximum volume/ percentage 136 Shift in location of minimum volume/ percentage 137 Disappearance of volume difference 138 Larger than expected volume difference 139 Missing expected vehicle classes 140 Volume distribution inconsistent with history 141 Percentage distribution inconsistent with history card/7-card difference too large card distribution inconsistent with 7- card Average volume > 250 Average volume > 250 Average volume > 250 Average volume > 250 Average volume for missing class > 50 Average monthly volume > 400 (including unclassified). The dominant class changes or the percentage in a dominant class is twothirds or less of its typical value A class (not the unclassified) with less than five percent of the population exceeds ten percent of the population A pattern of 4-card volumes being greater or less than 7-card volumes becomes same volume for both 3) Volume difference exceeds ten percent when there is no expected difference set for the site. 4) Volume difference changes by more than twenty five percent A class disappears from the distribution for three or more consecutive weeks Relative volumes do not match the expected pattern Relative percentages do not match the expected pattern Difference in volumes, excluding the unclassifieds/unknowns (Class 15 or 20) is greater than ten percent When the 7-card distribution has been defined as valid for the site and the 4-card distribution using the same classification algorithm does not match it. 4-card data is data to be purged

115 Code Description Limiting value Purge Identification Criteria card distribution inconsistent with 4- card When the 4-card distribution has been defined as valid for the site and the 7-card distribution using the same classification algorithm does not match it. 7-card data is 145 Volume larger than the expected maximum 146 Volume less than the expected minimum Maximum volume > 300 Minimum volume < 175 data to be purged Volume exceeds expected maximum by more than ten percent of the expected maximum Volume is below the expected minimum by more than ten percent of the minimum value The values needed for axle review ( Table C.2-4) are: legal maximum weight by axle group, typical class minimum axle weight, average monthly volume, the volume represented by the curve and the volume on the graph. The average monthly volume is same as that for the 4-card vs. 7-card review. To determine the average number in a curve use the following Monthly, by Month; Yearly, by Year use volume in header Monthly, by Day of Week; Yearly by Day of Week - Use volumes in 4-card vs. 7-card Month (Year) by Day of Week graph Yearly, Year by Month Use volume from Vehicle Distribution, Trucks, Yearly, Year by Month. Table C.2-4 Purge Codes for Axle Distribution Review Code Description Limiting value Purge Identification Criteria 72 Atypical distribution See 153, 154, 155, ) A pattern variance from the expected at the site not described by one or more other codes 73 Over calibrated 4000 or more per month - 1 bin 2000 or fewer per month 2 bins 4) A pattern for a new site without validation information that does not match the expected typical pattern for the class Class 9s - Both peaks shifted equally to the right AND outside of their expected bounds.

116 Code Description Limiting value Purge Identification Criteria 4000 or more per month 10 percent of legal maximum to nearest bin 2000 or fewer per month 20 percent of legal maximum Other classes Dominant peaks shifted equally to the right from their expected peaks. 74 Under calibrated 4000 or more per month - 1 bin 2000 or fewer per month 2 bins 110 Axle distribution inconsistent with history 4000 or more per month 10 percent of legal maximum to nearest bin 2000 or fewer per month 20 percent of legal maximum Class 9s - Both peaks shifted equally to the left AND outside of their expected bounds. Other classes - Dominant peaks shifted equally to the left from their expected peaks. 3) Dominant peaks are shifted more than one bin from the expected curve in either direction. 111 Large percentage of heavy axles 500 or more axles in curve being assessed 4) The shape of a curve is different from the expected curve. Different is o A uni-modal (single peak) distribution becoming bi-modal (two-peaks), (Use 153) o A distinctly loaded or unloaded site shifting to a partially loaded site (and remaining uni-modal) (Use 154), o Doubling of the size of either the tail below the lower bound of the unloaded peak or above the upper bound of the loaded peak. (Use 156 or 155 respectively) This code is preferred to 72 (atypical distribution) because it is a more explicit description of an irregularity. 5) More than five percent of single axles are in bins beyond 20,000 pounds. 6) The value of the single axle

117 Code Description Limiting value Purge Identification Criteria distribution is non-zero past 22,000 pounds. 7) More than five percent of tandem axles are in bins beyond 40,000 pounds. 112 Large percentage of light axles 500 or more axles in curve being assessed 8) The value for the tandem axle distribution is non-zero past 44,000 pounds. 3) More than five percent of the single axle distribution is less than 6,000 pounds except for Class 5 where the values are 15 percent and 3,000 pounds. 113 Shifted heavy axle peak 250 axles or more in curve being assessed 114 Shifted light axle peak 250 axles or more in curve being assessed 153 Change in number of modes for axle distribution At least 1000 axles on graph 4) More than five percent of the tandem axle distribution is less than 8,000 pounds. The heavy (loaded) peak is more than two bins or by 15 percent of the legal maximum (rounded to the nearest bin) from the expected location. The light (unloaded) peak is more than two bins or by 15 percent of the legal maximum (rounded to the nearest bin) from the expected location. 5) One peak changing to two peaks 6) Two peaks changing to one peak 154 Change in dominant loading type At least 500 axles on graph 7) Two peaks going to three peaks where the third peak is at least a third of the next higher peak 8) Three peaks going to two peaks 5) Loaded site becomes unloaded 6) Unloaded site becomes loaded 7) Site not clearly loaded or unloaded (Less than five percent difference between peaks) becomes one or the other

118 Code Description Limiting value Purge Identification Criteria 8) Partially loaded site (single peak between 40 and 60 kips for Class 9s) becomes loaded or unloaded 155 Large change in loaded tail of axle distribution 250 or more axles in curve being assessed 4) Loaded tail beyond 110 percent of legal maximum goes from zero to non zero 5) Loaded tail more than two bins from loaded peak doubles in magnitude 156 Large change in unloaded tail of axle distribution 250 or more axles in curve being assessed 6) Loaded tail beyond 110 percent of legal maximum goes to zero 3) Unloaded tail left of 90 percent of tare weight goes becomes non-zero 4) Unloaded tail more than two bins from unloaded peak doubles in magnitude

119 New category, need to add: 4-card vs 7-card Month by date New category, need to add: Yearly, year by month by day of week C.3 Month Table used: MM_GVW Type abbreviation: 13GVWMMon Default: Class 9s This graph is used to determine whether the distribution of gross vehicle weights (GVW) is rational for a given class. Since data has been aggregated to the monthly level, irregular patterns at the weekly level cannot be investigated except by using agency graphs. This graph is best used for sites that have stable patterns or small volumes of the classes of interest. There are two varieties of this graph type. The Class 9 graph has two pairs of vertical reference lines. The first set at 28 and 36 kips and the second set at 72 and 80 kips. A month s data should be accepted without further investigation if: For class 9 o An unloaded peak, if it exists, falls on or between the left pair of vertical reference lines o A loaded peak, if it exists, falls on or between the right pair of vertical lines o All points from 16 kips to the left are 2 percent or less o All points from 84 kips to the right are 2 percent or less o The x-axis goes from 0 to 100 or less For all other vehicle classes (subject to vehicle type and local regulations) o Values for 0 or 2 are 10 or less (lower limit is set 2 bins below the expected minimum) o The sum of all values above the upper limit is 2 percent or less (upper limit is set 2 bins above expected or legal maximum, whichever is larger based on validation for the site) GVW review purge codes are in Table C.2-2, page 110. Unless otherwise indicated the graphs in this section are for Class 9.

120 Figure C-2 GVW - Monthly, by Month - Example 1 (CASE NEEDED. NO DATA) This is a typical bi-modal graph for Class 9 with unloaded and loaded peaks within the expected bounds. Sites without this pattern or dominant loaded or unloaded peaks will require further investigation. Purge candidate - No Figure C-3 GVW - Monthly, by Month Example 2 CASE NEEDED. NO DATA) Figure C-3 is an atypical pattern for Class 9 weights. Does it accurately reflect the distribution of class 9 trucks at this site? Check calibration information and review a month by day of week graph and trend graphs. Calibration information is located in the LTPP IMS. that is unlikely. Use a month by day of week graph to see if the pattern is driven by a specific day. In view of the large number of vehicles in this sample Check 13-bin year by Month or 13-bin Years by Year GVW graphs. The former would show whether the changed during the year. The latter would indicate if this were the typical site pattern.

121 Expected action Identify the last calibration and define representative distribution based on following 2 weeks of data. PA, 1597, East, Lane 1 - July 2000 GVW Distribution - No. weighed: 139 Figure C-4 GVW - Monthly, by Month - Example 3a Figure C-4 has a very erratic pattern illustrating a potentially failing site. However, the small number of vehicles in the graph does not support that conclusion. Determine if this is a full month s data or a sample. Use a 4-card/7- card month by DOW graph and additional months or the full year graph for GVW. A sampled data set may or may not be retained. If this pattern is repeated for the entire year, potential sensor failure should be flagged. If it is not, this data may be a candidate for exclusion. Purge candidate Maybe Figure C-5 GVW - Monthly, by Month - Example 3b (CASE NEEDED) The shift and spread of the tail indicates that sensor failure is probable. Purge candidate - Yes This Class 9 graph in Figure C-13 is shifted left with a long tail on the right. This data is from the same site as shown in Figure C-4 but collected 13 months later. Since the amount of data is approximately the same, it might be representative of the entire month rather than merely a sample. Expected Action Check prior graphs to determine when sensor failure probably occurred and what data should remain in the database.

122 needs to be maintained by the analysis for reference purposes. Purge candidate No Figure C-6 GVW - Monthly, by Month Class 5 (Look into PA for this case and odd cases, seems to vary) This is a typical distribution for a Class 5 vehicle population in Figure C-14. Review of vehicles other than Class 9s should be undertaken when they are present in significant numbers, contribute significantly to the site loading or there are insufficient Class 9s to make a determination on the reasonableness of the weight data. For all other vehicle classes there are no expected value markers for loaded, unloaded or maximum allowable values. This information Figure C-7 GVW - Monthly, by Month Class 8 - Example 1a (CASE NEEDED, THIS ONE HAS BEEN CHANGED) Figure C-7 is one example of a Class 8 GVW graph. It needs to be considered in context with other monthly graphs to determine if it is representative or not. In comparison to Figure C-8 this graph is distinctively different in terms of peaking and range of weights. Questions to be asked include whether a change of classification algorithm was implemented, a change in site location or equipment problems. Data in Figure C-8 was collected only four months later than Figure C-7. Purge candidate Maybe Expected action Determine the definition of a Class 8 in the algorithm at the site

123 Figure C-8 GVW - Monthly, by Month Class 8 - Example 1b (CASE NEEDED, NO LONGER EXISTS) Purge candidate TBD Expected action Determine the definition of a Class 8 in the algorithm at the site Figure C-9 GVW - Monthly, by Month Class 8 - Example 1c (CASE NEEDED, THIS ONE HAS BEEN CHANGED) Purge candidate TBD Expected action Determine the definition of a Class 8 in the algorithm at the site

124 C.4 Vehicle Distribution - Month (avg.) Table used: MM_ CT Type abbreviation: 13DistMMon Default: Trucks (4-13) This graph is used to determine whether the mix of vehicles in the population is stable over time. It combines 4-card and 7-card data, with the classification data (4-card) in the left hand column and the weight data (7-card) on the right. A month s data for a data group should be accepted without further investigation if: The percentage of unclassifieds is less than five percent. A heavy truck graph (Classes 6-13) o All classes that are expected to be more than ten percent of the population are within a range that is that value plus or minus 10 percent. o The percentage of unclassifieds is less than the expected value for the site or less than two percent for the group if there are no unclassifieds expected. For all other groups o o The difference between classification and weight data is less than five percent for vehicles present in both data types. The exception is Class 5s where a Class3/5 discrimination pattern is expected to exist. The Class 9 percentage is greater than the Class 8 percentage (5 times is typical) The purge codes for vehicle distribution review are found in Table C.2-3 on page 113. Unless otherwise indicated, the graphs in this section are for Trucks (4-13).

125 use weight triggers. NC, 0200, South, Lane 1 - October Trucks (4-13) Vehicle Distribution - No. counted 21,781 / No. weighed 18,246 Figure C-10 13DistMMon Example 1 In looking at a site with proportionally different volumes several questions arise. One is whether the 4 and 7 card data comes off the same or different equipment. The second is whether the same algorithm and methodology is used in different pieces of equipment. Classification algorithms that do not include weight may produce very different results from algorithms that The example in Figure C-11 is potentially such a case. This is triggered by the larger numbers of Class 4s, Class 5s and Class 8s in the 4-card data than in the 7-card results. Class 4 and 5 vehicles are 2-axle vehicles that may include passenger vehicles. Class 8 vehicles may also be passenger vehicles pulling long trailers and thus on a weight basis be excluded from 7-card information. Another query is whether the amount of congestion during data collection might have affected the recognition of single unit as individual vehicles rather than multiple unit vehicles. As in all cases the size of the sample affects the ultimate decision on the data rationality. Figure C-11 13DistMMon Example 2 Purge Candidate Maybe Expected Action Check Site_Equipment_Info to see what equipment was used for collection of class and weight data. Investigate whether algorithms for 4 and 7-cards differ.

126 NC, 0200, South, Lane 1 - July Trucks (4-13) Vehicle Distribution - No. counted 29,349 / No. weighed 0 Figure C-12 DistMMon - Example 3 Distribution graphs will be produced with one or both data types present. When a single data type is present only general assessments of the relative volumes are possible. Expected Action compare to periods when 7-card data is present NC, 0200, South, Lane 1 - March Hvy Trks (6-13) Vehicle Distribution - No. counted 20,239 / No. weighed 18,194 Figure C-13 DistMMon (6-13) The distribution graph using only heavy trucks, classes 6-13, removes the distraction of inconsistent class 4 and Class 5 volumes.

127 Figure C-14 13DistMMon (6-13) Example 1 This figure shows no 4-card records of heavy trucks. Check the graph for all vehicles to determine whether 4-card records do not include any heavy trucks or whether the 4-card records exist at all. Figure C-15 13DistMMon (6-13) Example 2 Here slightly more trucks were counted than weighed. The relative difference is similar for each class of vehicles. This is typical for many sites.

128 Figure C-16 13DistMMon (1-13, 15) For locations with relatively large truck volumes it may be possible to use this graph to evaluate comparative data collection. In the case of Figure C-16 about one-sixth of the population is trucks. Note that the classification data contains unknowns while the weight data does not. The analyst should know if this is a characteristic of the agencies submissions. A sheet 7 is necessary. NC, 0200, South, Lane 1 - April Vehicles (1-13,15) Vehicle Distribution - No. counted 100,596 / No. weighed 19,131 Figure C-17 13DistMMon (1-13, 15) The all vehicles graph will generally result in a virtual loss of any ability to discriminate between truck volumes as shown in Figure C-17 for sites with relatively low truck volumes in the population.

129 Figure C-18 13DistMMon (1-13, 15) - Example 1a, b When this graph is used with only one data type, the results on a month-to-month basis can be tricky. In this figure, the scale varies between 4 and 7-card data. Additionally, as the population size changes, the relative proportion of significant vehicle classes can change as well. Although they look different, these two graphs show similar relative proportions of volumes by vehicle class for trucks (4-13). Figure C-19 13DistMMon (1-13, 15) - Example 2 Figure C-19 illustrates the benefit of using this particular vehicle population group option. Here the graph illustrates the very large number of unclassified vehicles in the population. This data is highly suspected unless there a class conversion sheet (sheet 7) can be obtained that shows the vehicles consolidated into Class 15 are in actual fact other types. Given that this is classification data, there may be an issue with data recorded beyond column 51.

130 C.5 Vehicle Distribution - Month by Day of Week (avg.) Table used: MM_ CT Type abbreviation: 13DistMDoWa Default: Trucks (4-13) This graph is useful for determining how vehicle distributions vary by both day of week and vehicle type. The graph has classification data in the left hand column for each day of the week and weight data in the right hand column. These graphs allow visual identification of class 3/5 and 3/8 or 6/8 algorithm differences. The following generally indicate an acceptable data set: Average volumes for the group within 10 percent of expected value for any day of the week The same vehicles in both data types. If they are different, the potential for different class schemes must be investigated. Where Class 3s are included. The number of 3s and 5s combined should be approximately the same for the two data types. Class 15s should be less than five percent. The purge codes for vehicle distribution review are found in Table C.2-3 on page 113. Unless otherwise indicated, the graphs in this section are for Trucks (4-13).

131 Figure C-20 13DistMDoWa - Example 1 4-card and 7-card data show similar average volumes except on Thursdays and Fridays. Check 4- vs. 7-card - Monthly, by Day of Week graphs for insight into why average 7-card volumes are lower on Thursdays and Fridays. Purge Candidate maybe Reason - Larger than expected volume difference (138) OH, 0100, South, Lane 1 - February Class Trucks (4-13) Vehicle Distribution - No. counted 50,089 / No. weighed 47,925 Figure C-21 13DistMDoWa - Example 1 Figure C.8-2 shows a typical distribution where both the classification and WIM equipment are seeing the same truck population. Purge candidate - No

132 Figure C-22 13DistMDoWa (4-13) - Example 2 Sunday 4-card records have significantly more class 5 vehicles than any other day. This is possibly a misclassification of class 3 trucks. Purge Candidate Yes Reason Atypical pattern (72) OR Outside +/-10 percent range (103) Figure C-23 13DistMDoWa (4-13) - Example 3 No 7-cards. Within the 4-card records, relative volumes of truck classes remain mostly constant although absolute volumes change with day-of-week. Purge candidate - No

133 Figure C-24 13DistMDoWa (4-13) - Example 4 Similar to Figure C.8-4 except that the proportion of class 5 trucks increases significantly on Mondays. Purge Candidate Yes Reason - Outside +/-10 percent range (103) Figure C-25 13DistMDoWa (4-13) - Example 5 This is a typical distribution where the classification files show similar relative volumes of truck classes from day to day although absolute volumes decreases during weekend days.. Purge candidate - No

134 Figure C-26 13DistMDoWa (4-13) Example 6 This graph has average daily volumes that are based on a single day of data for each of three days-of-the-week. Volumes are significantly different for the three days and only Thursday shows similarity between 4 and 7-card data. Purge Candidate - Yes Figure C-27 13DistMDoWa - Example 7 A situation where more trucks are weighed than counted is unusual. Purge Candidate Yes Reason - 4-card/7-card difference too large (142)

135 Figure C-28 13DistMDoWa (4-13) - Example 8 4 and 7-card files show similar volumes except on Mondays, Thursdays and Fridays Purge Candidate Yes Reason card/7-card difference too large (142) OR Larger than expected volume difference (138) OH, 0200, North, Lane 1 - July Class Trucks (4-13) Vehicle Distribution - No. counted 20,314 / No. weighed 18,512 Figure C-29 13DistMDoWa (4-13) - Example 9-better case? Same thing? Relative volumes recorded within 4 and 7-card files change over the course of the week. Purge Candidate - Yes

136 Figure C-30 13DistMDoWa (6-13) A chart for heavy trucks can be used when high volumes of 4s and 5s otherwise make visual identification of other truck volumes difficult. Here, the 4 and 7-cards show similar volumes and proportions of heavy trucks. Figure C-31 13DistMDoWa (6-13) - Example 1 Here the 4-card record has many more heavy trucks than 7-card files, but only on Sunday. The files match very well on other days. Are congested conditions on Sundays causing misclassifications of smaller vehicles? Do weight triggers prevent this within the 7-card files? Expected Action Investigate the algorithms used classify vehicles for 4-card and 7-card files.

137 Figure C-32 13DistMDoWa (6-13) - Example 2 More trucks are weighed than counted. Check the graphs of all trucks and all vehicles to see if the pattern holds. Expected Action Investigate whether the discrepancy is due to misclassification. Figure C-33 13DistMDoWa (6-13) Example 3 Volumes are much higher on Tuesdays. Expected Action Check the graphs of all trucks and all vehicles to see if the pattern holds. Check agency 4- vs. 7-card Weekly, Week by Date graphs to see if this is due to a single day or if all Tuesdays have such high volumes.

138 OH, 0800, South, Lane 1 - November Class Hvy Trks (6-13) Vehicle Distribution - No. counted 362 / No. weighed 0 Figure C-34 13DistMDoWa (6-13) - Example 4 Very few vehicles were counted (note that this is an SPS-8 site). Check the graphs for all trucks and all vehicles. Check agency 4- vs. 7-card Weekly, Week by Date graphs to see how long a time period is represented by this data. Was data collected during a Monday or Tuesday? Figure C-35 13DistMDoWa (6-13) - Examples 5a, b The above two charts show a disappearance of class 6 and 8 vehicles at a site in Ohio between 1994 and Check the duration of this phenomenon by looking at all charts for 1994 and Purge Candidate Yes Reason - Missing expected vehicle classes (139)

139 Figure C-36 13DistMDoWa (6-13) - Examples 5c, d Here additional charts from the same site confirm that classes 6 and 8 are expected at the site and that example 5b is anomalous. NC, 0200, South, Lane 1 - October Class Vehicles (1-13,15) Vehicle Distribution - No. counted 110,676 / No. weighed 18,262 Figure C-37 13DistMDoWa (1-13, 15) A vehicle distribution chart for all vehicles will typically wash out results from 7-card files since they do not contain records for classes 2 and 3. The graphs can be used to check for 3/5 misclassification within 4-cards.

140 Figure C-38 13DistMDoWa (1-13, 15) - Example 1a Here the numbers of class 2 and 3 vehicles dwarf the number of trucks. Figure C-39 13DistMDoWa (1-13, 15) - Example 1b At the same site in the opposite lane, it can be seen that the 7-card file does not include classes 2 and 3 so the graph for all vehicles will look identical to the trucks graph.

141 C.6 4-card vs. 7-card Year by Month by Day of Week (DEFAULT) Table used: MM_CT Type abbreviation: 1347YMDoW Default: Class 9s This graph is used to provide a consolidated volume graph to look at three elements: The change in volumes over the course of the year The stability of the day of week pattern The consistency between the classification and the weight data reporting. It has classification data reported on the left hand column and weight data in the right hand column by month. The column is divided by days of week. A copy of this graph should be a part of the annual QC packet. This graph is not the sole factor in making a purge decision. It is a screening element. This is particularly important in the case of sampled data where not all days are present in all months. In reviewing the data the following constitutes what is acceptable: When all days are present, there should be the same difference between 4-card and 7-card columns across all months. Same difference means the same relationship (higher/equal /lower) and relative size of difference When only some days are present, the individual day volumes should bear the same relationship between data types. When only one data type is present, volumes should vary less than ten percent within the year. The exception is the vehicles graph where twenty percent is the allowable maximum. The purge codes for a 4- vs. 7-card review are in Table C.2-1, page 108. Unless otherwise indicated, graph examples are for Class 9.

142 Figure C vs. 7-card Yearly, Month by Day of Week Example 1 Average recorded volumes are much lower in December. Relative volumes for each day of the week are similar to other months. Purge candidate TBD What December Reason Lower volume than expected (70) Figure C vs. 7-card Yearly, Month by Day of Week Example 2 Here February shows high proportions of class 9 vehicles on Saturday. Purge candidate TBD What Feb Reason Change in day of week pattern (101) (Large Saturday volumes)

143 KS, 1005, East, Lane Class 9 4-card vs. 7-card Figure C vs. 7-card Yearly, Month by Day of Week Example 3 Monday volumes are near zero in June while July has no volumes for Tuesday thru Thursday. Purge candidate TBD What June, July Reason Change in day of week pattern (101) Expected Action - Check to see if June and July data are present for all days of the week. KS, 1010, West, Lane Class 9 4-card vs. 7-card Figure C vs. 7-card Yearly, Month by Day of Week Example 4 Purge candidate TBD Expected action: Find another year s data for comparison purposes. There is too little data for site in this year alone

144 Figure C vs. 7-card Yearly, Month by Day of Week Example 5 Purge candidate No (Low volume site; possibly does not have an entire week of data for each month) MI, 7072, North, Lane Trucks (4-13) 4-card vs. 7-card Figure C vs. 7-card Yearly, Month by Day of Week Trucks, Example 1 The distribution of day-ofweek volumes are consistent between 4 and 7- cards and do not change throughout the year. Purge candidate No

145 Figure C vs. 7-card Yearly, Month by Day of Week Trucks, Example 2 Purge candidate Yes What November Reason Higher volume than expected (100) Expected actions: See if there was a change in the definition for classes 3 and 5. Verify that the data is in fact for this site. While duplicates are a possibility in the weight file, the software prevents loading duplicate counts for classification. Figure C vs. 7-card Yearly, Month by Day of Week Heavy Trucks 7-card files show fewer heavy trucks than 4-cards. This is not unusual although the inverse would be. Purge candidate - No

146 MI, 7072, North, Lane Vehicles (1-13) 4-card vs. 7-card Figure C vs. 7-card Yearly, Month by Day of Week Vehicles 4-card files include classes 2 and 3 while 7-card files do not. This accounts for the difference between the volumes in the two file formats. Purge candidate No MI, 3068, West, Lane Vehicles (1-13,15) 4-card vs. 7-card Figure C vs. 7-card Yearly, Month by Day of Week Vehicles (All) Here 4-card volumes increase during the summer months but 7-card volumes do not. The increase is due to automobiles, not trucks. Purge candidate - No

147 C.7 Vehicle Distribution - Year by Month (avg.) (DEFAULT) Tables used are: MM_CT Type abbreviation: 13DistYMon Default: Trucks (4-13) This graph is used to determine whether the mix of vehicles in the population is stable. The distributions for classification data and weight data are graphed on separate figures. To use this graph there should be at least 125 vehicles expected in a month for the group being evaluated. A month s data for a data type should be accepted without further investigation if: A heavy truck graph (6-13) o All classes greater than 10 percent of the population are within +/- 10 percent of their expected value. o The percentage of unclassifieds is less than the expected value for the site or less than two percent for the group if there are no unclassifieds expected. o No class expected to be less than 10 percent of the population is more than 15 percent of the month s population. For all other distributions o The percentage of unclassifieds is less than the expected value for the site or less than 5 percent for the site if there are no unclassifieds expected. o The class 9 percentage is greater than the Class 8 percentage The graphs for 4-card and 7-card data are found on different pages. For the purposes of illustration only one pair of classification and weight data graphs are provided for each type. The review principles are the same independent of the data type so most illustrations will be with classification data. The most suitable truck group selection is Heavy Trucks (6-13). If there are too few heavy trucks to generate the graphs the next most suitable is the Trucks (4-13) option because of its ability to identify unclassifieds in agency submissions using 6-digit Truck Weight Study (TWS) classifications. When used it is not necessary to know what the card format or classification scheme is to have information on unclassifieds when making purge decisions. This will not be relevant to most data after 1996 or agencies using W-card format for loading data submission. The purge codes for vehicle distribution review are found in section Table C.2-3 on page 113. Unless otherwise noted the graphs in this section are for Trucks (4-13).

148 Figure C-50 13DistYMona - Example 1 (doesn t exist) Figure C-50 shows a site where the distribution by vehicle class is consistent throughout the year for both class and weight data Purge candidate - No Figure C-51 13DistYMona (4-13) (case needed) Figure C-51 illustrates a case where Class 8 vehicles exceed Class 9s in January and February. Class 4s are present during those months, then all but disappear. Purge Candidate: Yes What: January & February (4-card) Reason: Percentage distribution inconsistent with history (141)

149 MS, 1001, North, Lane Class Hvy Trks (6-13) Vehicle Distribution - No. counted 58,962 / No. weighed 52,341 Figure C-52 13DistYMona (6-13) Class 13 is presented in Jan through Apr but it disappears after that. Purge Candidate: Yes Reason: missing expected vehicle class (139). Figure C-53 13DistYMona (6-13) (case needed) Class 6s volume increased by more than 40% on May. Except class 6 all other classes are very consistent through out the year. Purge Candidate: Maybe Reason: Percentage distribution inconsistent with history (141).

150 CO, 1047, West, Lane Class Hvy Trks (6-13) Vehicle Distribution - No. counted 19,819 / No. weighed 16,046 Figure C-54 13DistYMona (6-13) From 4-card to 7-card class 8s volume decreased by almost 80 percent. But all other classes are consistent. Purge Candidate: Yes Reason: data for 4-card and 7-card is coming from different pieces of equipment. CO, 1053, North, Lane Class Hvy Trks (6-13) Vehicle Distribution - No. counted 0 / No. weighed 57,471 Figure C-55 13DistYMona (6-13) Volume of all classes increased during summer time. Therefore this region is very seasonal. Purge Candidate: No

151 Figure C-56 13DistYMona (6-13) (case needed) Purge Candidate: TBD

152 D Classification Error The Classification graphs show the data that has been rejected for failing one of four pattern checks for 4-card (or C-card) data: Less than 24 hours for a day (Missing Hours), Clock time verification (1 a.m. volume > 1 p.m. volume), Counter not on (8+ Consecutive Zeros), and Counter error (4+ Consecutive Non-Zeroes). The classification graphs use the ERR_CL table. Data is loaded in this table via the Daily Summaries process if a record is flagged with one of these errors. Data is loaded in this table via Load Files if it fails the missing hours check or one of the other checks that are to be used according to the options selected in SHRP_INFO. There are no purge codes applied by the user as a result of these graphs. It may be necessary to change a code if the incorrect one was assigned prior to using the Daily Summaries process. If an error checking option was incorrectly selected in SHRP_INFO when using Load Files, it may be necessary to reload the data. Figure D-1 shows the classification error graph option screen with all four graphs checked. The software default is to graph all check results. Figure C.7 Classification Error Graph Options Screen

153 D.1 Missing Hours Table used is: ERR_CL Type abbreviation: ERR_CL. Default is N/A This is a fatal error in data. The only correction may be to load files as a group rather than individually with Load Files if the agency submits daily data files that do not run midnight to midnight. Missing Hourly Volume (CA, 2040) 07/24/1999, South, Lane 1 Figure D.1-1 Missing hourly volumes - Example 1 Figure D.1-1 is a typical missing hours graph. A missing hours error may apply to the first or last day of data collection of a sampled site where a 24-hour day could not be created. Figure D.1-2 Missing Hourly Volumes - Example 2 (case needed) Data in Figure D.1-2 shows a site where alternate hours are being collected or possibly only reported due to an aggregation interval of two hours. Expected action: Determine if this is a recurring pattern for the site. If it is, contact the agency and ask for hourly data. There is no way to reload the data in LTAS and have it load into DD_CL_CT without changing the input data file.

154 D.2 1 a.m > 1 p.m. Table used is: ERR_CL Type abbreviation: ERR_CL Default is N/A This is not necessarily a fatal error for data. The determination of whether or not the data should be retained is site dependent. 1 AM Volume > 1 PM Volume (CT, 1803) 01/28/2004, North, Lane 1 Figure D a.m. > 1 p.m. - Example 1a Figure D.2-1 is typical of a day that should be omitted from the DD_CL_CT table. Expected action: If this is one of only a few days of data available for the site-year in the LTPP lane, it may be worth inquiring of the agency about an accident of lane closure. It is however and unlikely case given the low volumes in late afternoon as well. 1 AM Volume > 1 PM Volume (AZ, 7079) 02/07/2004, North, Lane 1 days, asking about construction is prudent. Figure D a.m. > 1 p.m. - Example 1b (different region-west) In Figure D.2-2 unlike Figure D.2-1 an accident or lane closure is a distinct possibility. Expected action: None unless the data is very sparse for the site, no additional action should be taken. If the same pattern occurs over consecutive

155 1 AM Volume > 1 PM Volume (NV, 1020) 01/09/2003, South, Lane 1 Figure D a.m. > 1 p.m. - Example 2 Figure D.2-3 is problematic for retention. The low volumes make it difficult to figure out what the day s pattern actually is. It is possible given the site and date that the count was affected by snow. Expected action: None unless very little data exists for the site year. 1 AM Volume > 1 PM Volume (NJ, 0500) 10/24/2004, West, Lane 2 Figure D a.m. > 1 p.m. - Example 3 data is not stored on an hourly basis. Figure D.2-4 shows what the outcome might be if a major event let out around midnight. Considering the difference between the hour beginning at 11 p.m. (hour 23) and the one beginning at midnight (hour 0), realize that they are separated by 24 hours. There is no way to know what the previous day s hour beginning at 11 p.m. looked like since the

156 Figure D a.m. > 1 p.m. - Example 4 The graph in Figure D.2-5 illustrates the case where it might be advisable to load data without the 1 a.m. > 1 p.m. check. All of the hourly volumes are less than ten. Additionally, from the SHRP_ID (0800) it can be seen that this is a location where very little traffic is expected.

157 D.3 8+ Consecutive Zeroes Table used is: ERR_CL. Type abbreviation is: ERR_CL Default is N/A. This is generally considered a fatal error, particularly when it occurs in the middle of the day. There is no firm LTPP policy on when such data should be accepted if it is determined that a lane closure makes zeros valid for the volume at the site. To exclude such cases over estimates the annual loadings. Sporadic lane closures are probably not sufficient to change the loading checks. For an extended lane closure, it may be preferred to change the loading checks and make a note in SHRP_INFO as to what period the conditions were changed for. This is necessary because it will affect the 4-card vs. 7-card review when extended periods are considered. Weeks, months or years after the fact the reason for zero volumes may not be remembered or known to the reviewer. If the file came in without a record for the hour, LTQC and LTAS do not create one with a zero volume. Those hours are treated as missing data.

158 8/2/2011 Page 1 of 1 8+ Consecutive Zero Volumes (WA, 0200) 09/20/1999, North, Lane 1 is somewhat unusual but not impossible. Figure D Consecutive zeroes - Example 1 Figure D.3-1 is a typical case where equipment stopped recording and the file was filled with zero volumes to have a record for every hour. It is also possible the lane was closed but the timing of that closure 8+ Consecutive Zero Volumes (WA, 0200) 10/06/1998, South, Lane 1 Figure D Consecutive zeroes - Example 2 Figure D.3-2 is a case where the equipment most likely stopped recording. The hours are unlikely to be for a lane closure but the possibility does exist.

159 D.4 4+ Consecutive Non-zeroes Table used is: ERR_CL. Type abbreviation is: ERR_CL Default is N/A. The implication for this data type is that the four (or more) hourly volumes are exactly the same. The software does not put a range on similarity but looks for an exact match. As a rule of thumb consecutive non-zero volumes under 20 are probably a low volume site and should be accepted. Consecutive non-zero volumes over 100 are highly unlikely and the error designation should be retained. Days where the consecutive volumes are between 20 and 100 should be evaluated in the context of time of day and the normal site volumes.

160 4+ Consecutive Static Volumes (WA, 0200) 07/22/1998, North, Lane 1 worth retaining this data at a site with little information. Figure D Consecutive nonzero volumes - Example 1 Figure D.4-1 is an example where judgment should be applied on whether the rejection is valid. Given that the volume is at night and the vehicles per hour persists for about 8 hours, it might be 4+ Consecutive Static Volumes (WA, 0200) 04/13/1998, South, Lane 1 Figure D Consecutive non-zero volumes - Example 2 Figure D.4-2 shows a case where the non-zero volume is clearly less than 20. In this case it is 4. However, the volume shown should be considered in relation to the typical volume at the site. In this case, the norm is several thousand vehicles per day.

161 Figure D Consecutive non-zero volumes - Example 3(case needed) Figure D.4-3 is an extreme example of the 4 consecutive non-zero volumes error. Figure D Consecutive non-zero volumes - Example 4(case needed) Not all cases that show up with this error code have been correctly evaluated as shown in Figure D.4-4. In this case the correct code would have been 8 consecutive zero volumes. Whether this is the result of a software error since fixed or an incorrect purge code identification is not known or relevant. Expected action: 1) Query the ERR_CL table to find out what the record shows (select * from err_cl where state_code = SC and shrp_id = SHRP and year = yyyy and month = mm and day = dd;) 2) Review the data to see if the volumes are very small with respect to the scale or in fact zero. 3) If the values are very small, no further action is required. 4) It the values are 8 or more consecutive zeroes update the value of TRF_ORA_ERR (update err_cl set trf_ora_err = 60 where state_code = SC and shrp_id = SHRP and lane_trf = lane and dir_trf = dir year = yyyy and month = mm and day = dd;)

162 Figure D Consecutive non-zero volumes - Example 5(case needed) Figure D.4-5 is a case where there is nothing wrong with the data when judged by the criteria for classification errors. In this case it would appear the wrong purge reason was applied to the data. Expected action: If this is one of several records reload the data with Load Files and apply the correct purge according to the purge list. If the original purge list cannot be found, the data must be reviewed and the decision made again on keeping the data. 4+ Consecutive Static Volumes (OR, 5005) 10/10/2002, South, Lane 1 Figure D Consecutive non-zero volumes - Example 6 Figure D.4-6 is another case where the code and the classification error do not match. Expected action: Same as for Figure D.4-4 except that the code should be 63 rather than 60 in the update statement.

163 Figure D Consecutive non-zero volumes - Example 7(case needed) Figure D.4-7 is another case where the code and the classification error do not match. This should have been 1 a.m. > 1 p.m. Expected action: Same as for Figure D.4-4 except that the code should be 62 rather than 60 in the update statement. Figure D Consecutive non-zero volumes - Example 8(case needed ) Figure D.4-8 is an illustration of the changes in interpretation of this check. While it was originally 4+ consecutive static volumes including zero, the terminology has and incorporation of the check in LTAS has also produced cases

164 E ESALs per Vehicle ESAL (Equivalent Single Axle Load) is a computed measure that converts the loading of the individual axles in a truck population to a common measure, the 18-kip (18,000 pound) single axle. ESALs can be computed for an individual truck, a truck class or for the entire population given the type and number of axles by weight and given the pavement structure. In LTAS, ESALs are computed as monthly averages for a vehicle class for a fixed pavement type. The ESAL values generated in LTAS are not to be used in pavement design or evaluation. Average ESAL per vehicle is useful for assessing the quality of the data. As the examples contained in this section indicate, patterns may exist, but no site examined to date has been found to have a constant ESAL per vehicle value. ESALs computed in LTAS are a diagnostic tool to determine if investigation of GVW or axle distributions is appropriate. There is a single purge code associated with ESAL reviews 134 Inappropriate ESAL estimate. It should not be the principal reason for a purge unless the value is less than 0.1 or greater than 4.0 for Class 9 vehicles or there are too few vehicles or axle to produce a graph for review. If the GVW or axle distribution graphs do exist, those should form the basis for the purge decision. Figure D.4 ESALs per Vehicle Graph Option Screen

165 E.1 Yearly Table: ESAL_PER_VEH Type abbreviation: ESALYr Default: Class 9 The individual year graph is most useful when no other data exists or when the values or pattern are known and merely being checked. For this graph type, the multi-year option is a more valuable tool. CA, 0500, East, Lane Class 9 Average ESALs per Vehicle Figure E.1-1 ESALs Yearly Figure E.1-1 shows a graph of a typical ESAL per vehicle value for a 5- axle tractor-trailer vehicle. The value is not constant.

166 Figure E.1-2 ESALS - Yearly - Example 1 Figure E.1-2 shows a different pattern from Figure E.1-1 in that it peaks in winter rather than summer. Note that the range of values here is 1.5 to 2.5 rather than 1 to 1.2 as illustrated previously.

167 CA, 2040, South, Lane Class 9 Average ESALs per Vehicle CA, 2040, South, Lane Class 9 Average ESALs per Vehicle CA, 2040, South, Lane Class 9 Average ESALs per Vehicle Figure E.1-3 ESALs - Yearly - Example 2 The group of graphs in Figure E.1-3 illustrates the difficulty of relying on a single year s ESAL information to make a determination of data quality. The upper left figure shows the information from 1992 for this site. Due to the sampled nature of the information, it is not obvious whether the scatter is a trend or bad data. The upper right figure has 1993 data for the same site. Both the top graphs have the same relative shape. However the minimum and maximum values have increased from 1992 to There is insufficient information in this graph to determine why. Finally there is the lower, center graph with ESAL information for 1996 at the same site. Here the location of the peak value has shifted from August to May. Additionally, the minimums on this graph (1.2) are equal to the maximums from 1993 graph. The cause of the shift, more heavy trucks or heavier weights overall would have to be determined by means of other graphs and inquiries of the agency.

168 E.2 Multi-year Table: ESAL_PER_VEH Type abbreviation: ESALYrs Default: Class 9 The disadvantage of the multi-year graph is that it may be several years after data is accepted before it can be questioned in relation to other information. This graph should be generated as part of any annual QC packet for reference. 8/2/2011 Page 1 of 1 MN, 1029, North, Lane Class 9 Average ESALs per Vehicle Figure E.2-1 ESALs Multi-year Figure E.2-1 is from a site with a prominent unloaded peak in the GVW graph. The site is in a state with variable load limits due to freezethaw conditions. This graph has a cyclic pattern but not constant values for the same months over time. Figure E.2-2 ESALs - Multi-year - Example 1 Error! Reference source not found. contains perhaps the closest to a constant value graph identified in the CA data. Unfortunately, the average value per vehicle (0.1) is so low for a Class 9 vehicle as to be suspect.

169 CA, 3042, South, Lane Class 9 Figure E.2-3 ESALs - Multi-year - Example 2 Average ESALs per Vehicle Figure E.2-3 has another example of a constant value ESAL graph. However, given that the early years had higher values and something of a pattern, a reason for the change should be determined if possible. However, the length of time over which the change occurred may prevent that Another alternative would be to look for calibration information (Sheet 16) or a change in the equipment installed at the site (SITE_EQUIPMENT_INFO) or Traffic Sheets 14 and 15. It is also possible that auto-calibration was turned on or the site re-calibrated remotely to hit a target ESAL value. Figure E.2-4 ESALs - Multi-year - Example 3

170 AR, 2042, West, Lane Class 9 Figure E.2-5 ESALs - Multi-year - Example 4 Average ESALs per Vehicle Figure E.2-5 is similar to Figure E.2-4 with an apparently constant value in the early years and a very scattered later year pattern. What is deceiving about this figure is the y-axis. The pattern that actually looks constant over 1995 and 1996 is an increase in values from about 1 to nearly 1.5. The 1999 values range between 2 and 4.5. While 2 is not outside the limits of reasonableness for a Class 9 vehicle, values above 3.5 are clearly suspect since they imply a fleet where on average all trucks are over 80,000 pounds, the legal limit on the interstate. investigation. AR, 3059, West, Lane Class 9 Average ESALs per Vehicle Figure E.2-6 ESALs - Multi-year - Example 5 Figure E.2-6 is similar to Figure E.2-5 in that the early years are in a reasonable range if decreasing. The last years are subject for concern. The initial values are at the upper end of a range that would be associated with a site that has predominantly loaded trucks near the legal limit. However the sudden and drastic drop to an exceptionally low value for a class 9 vehicle requires

171 CA, 3019, North, Lane Class 9 Average ESALs per Vehicle Figure E.2-7 ESALs - Multi-year - Example 6 Figure E.2-7 shows a slowly diminishing range between minimum and maximum values in the cyclic pattern over time. The reasons for the trend and the cyclic nature should be determined CA, 8149, East, Lane Class 9 Average ESALs per Vehicle determined from this graph. Figure E.2-8 ESALs - Multi-year - Example 7 Figure E.2-8 shows the same downward trend as Figure E.2-7 but not the same level of variability in the range between the minimum and maximum within a year. In this case it would appear to be about 0.4. The cause of the trend however, should be explained as it is either lighter trucks overall or an increasing percentage of unloaded to loaded trucks, neither of which can be

172 CA, 2040, South, Lane Class 9 Average ESALs per Vehicle Figure E.2-9 ESALs - Multi-year - Example 8 This last figure shows a simple scatter of points and perhaps an upward trend to the data but nothing that would trigger a particular area of further

173 F STAT_QC Trend Statistics Trend statistics include steering axle weights, B-C axle weights and B-C drive tandem spacing, average Class 9 steering axle weights. These graphs are done in agency classification. The data are accumulated as a part of the loading process. Review of vehicle classes or graphs other than the defaults should be limited to screening, troubleshooting, low volume routes, and agency classes that are critical but not the default vehicle. There are no weekly graphs for this group. Since version 1.6 the data for these graphs includes RECORD_STATUS, the graphs may be done for only RECORD_STATUS = E information or for all data without purges applied. Figure E.2-1 STAT_QC graph availability (1.7)

174 F.1 Purge Code Tables With version 1.5 of the software the data supporting these graphs was changed from monthly to daily values (except the daily steering axle graph.) The daily distributions match the by month distributions then in the database. Data loaded after April 2006 has actual by day values. Purges may or may not show in modified graph shapes based on when the data was loaded. Purges applied to data loaded prior to April 2006 will not change the shape of monthly figures, because all days have the same distribution of values. Purges applied to data loaded after April 2006 may result in changes to graph shapes. While purge codes exist for various types STAT_QC irregularities, it is preferred that the codes applied to the data be associated with a GVW or axle distribution problem for loading in the case of weight based evaluations. Checking the GVW or axle distributions is part of the investigation process. Table F.1-1 Purge Codes for A-axle Graph Review Code Description Selection Criteria 116 Shift in A-axle pattern All curves by month should have the same shape and peak location within 1 bin. If any month can be clearly distinguished from the others by either criterion it should be flagged. 117 A-axle distribution light The peak(s) for the A-axle is(are) two or more bins to the left of its expected value(s). 118 Large percentage of light A-axles More than five percent of the distribution is less than 6,000 pounds for single axles. More than five percent of the distribution is less than 8,000 pounds for all other axle groups. 119 A-axle distribution heavy The peak(s) for the A-axle is (are) two or more bins 120 Large percentage of heavy A-axles 121 Excessive number of other than steering single axles to the right of its expected value(s). More than five percent of single axles are in bins beyond 20,000 pounds or the value for the distribution is non-zero past 22,000 pounds. More than five percent of any axle group are in bins beyond 40,000 pounds or the value for the distribution is non-zero past 44,000 pounds. A graph appears for other than single axles AND the percentage of axles exceeds two percent of all axles identified as steering axles for the class and interval. 122 No single A-axles No single axle A-axle graph appears even when a Yearly, Year graph is generated and other axle groups are graphed as A-axles. The data table should be queried to verify that no single axles exist.

175 Table F.1-2 Purge Codes for B-C Axle Weight Distribution Review Code Description Selection Criteria 123 Shift in B-C axle pattern All curves by month should have the same shape and peak location within 1 bin. If any month can be clearly distinguished from the others by either criterion it should be flagged. 124 B-C axle distribution light The peak(s) is two or more bins to the left of its expected value(s). 125 Large percentage of light B-C axles More than five percent of the distribution is less than 8,000 pounds. 126 B-C axle distribution The peak(s) is two or more bins to the right of its heavy 127 Large percentage of heavy B-C axles expected value(s). More than five percent are in bins beyond 40,000 pounds or the value for the distribution is non-zero past 44,000 pounds. Table F.1-3 Purge Code for B-C (Drive Tandem) Space Review Code Description Selection Criteria 115 Unusual B-C space Since B-C axles are by definition tandems for this graph and drive tandems in particular they should have a peak between 4 and 5 feet. Any peak outside those limits should be purged. The standard deviation should be less than 0.5. Table F.1-4 Purge Codes for Class 9 Steering Axle Review Code Description Selection Criteria 128 Shift in steering axle pattern One of three possibilities, a straight line becomes scattered, a scattered pattern becomes a straight line or a distinct shift from one average to another occurs (step function). 129 Mean steering axle weight light Mean steering axle is more than 1,000 pounds lighter than expected OR is less than 7,500 pounds. 130 Mean steering axle weight heavy Mean steering axle is more than 1,000 pounds heavier than expected OR is more than 12,500 pounds. 131 Change in steering axle A continuously increasing or decreasing trend to the weight. mean 132 Increasing variation in steering axle limits The two standard deviation limits on the mean increase in size or go from solid lines to scatter. 133 Steering axle weight zero The mean steering axle weight plots out as zero.

176 F.2 Graph Intervals There are three possible intervals for STAT_QC graphs: Monthly, Yearly and Multi-year. There are no default monthly graphs. There are four default yearly graphs as shown in Figure E.2-1. There are no default multi-year graphs. If any of these graphs are needed they must be done by spreadsheet. Contact the TSSC for the applicable templates and SQLs for the input data. The yearly and multi-year graphs are virtually the same. The yearly graphs scale appropriately for each individual year. The multi-year graphs are scaled based on the maximum value over all the selected years to permit viewing similarly scaled graphs for making comparisons between years

177 F.3 B-C Axle Spacing distribution Yearly, Year by Months (DEFAULT) Table used: STAT_QC_BC_SPACE Type abbreviation: BCSpcYMon Default: Class 9 This graph is used to illustrate the distribution of B-C Axle Spacings by month, by vehicle type. It will include every month for which at least 100 axles exist. The mean and standard deviation are computed using mid-bin spacings. A site with acceptable data for a Class 9 has the following characteristics: Peaks at 4 and or 4.5 feet No spaces at 12 feet The software accumulates only tandem axles in populating the table STAT_QC_BC_SPACE used for this graph. Tandem axles are by definition axle groups with and overall spacing of 3.3 to 8.0 feet. The purge codes for a B-C axle spacing distribution review are found in Table F.1-2 on page 175. IA, 0100, South, Lane Class 9 B-C Axle Spacing - No. counted 90,716 - Mean 4.0 / Std Dev 0.2 Figure F.3-1 B-C Axle Spacing - Year by Months - Example 1 Figure F.3-1 is a typical Year by Months graph of B-C axle spacing. Purge candidate - No Data Months: JAN FEB MAR APR MAY JUN JUL AUG SEP OCT NOV DEC

178 Figure F.3-2 B-C Axle Spacing Year by Months - Example 2 Figure F.3-2 is another typical accepted B-C Axle Spacing, Year by Month graph. This site however, needs watching because the standard deviation, 0.6 feet. A calibrated site should have 95% of the values within 0.5 feet. Purge candidate - No IA, 0100, South, Lane Class 9 B-C Axle Spacing - No. counted 53,531 - Mean 3.7 / Std Dev 0.3 Figure F.3-3 B-C Axle Spacing Year by Months - Example 3 Figure F.3-3 shows a shifting peak. Purge candidate Maybe Reason Unusual B-C Axle Spacing (115) Expected actions: The cause of the shift should be investigated. Data Months: JAN FEB JUN JUL AUG SEP OCT NOV DEC

179 Figure F.3-4 B-C Axle Spacing Year by Months - Example 4 (case needed) Figure F.3-4 shows a series of shifting peaks. Purge candidate Yes What - September Reason Unusual B-C space (115) Expected action: The peaks at 3.5 and 4 while not unreasonable are qualified by the percentages at 12 feet which should not exist. This site must be reloaded as the data loading error that existed prior to version 1.4 eliminated the peak at 12 feet. After reloading verify the peak in September. IL, 0600, North, Lane Class 9 B-C Axle Spacing - No. counted 315,345 - Mean 4.2 / Std Dev 0.4 Figure F.3-5 B-C Axle Spacing - Year by Months - Example 5 This site needs additional investigation since the axle spacing distribution gets flatter in the winter months. Data Months: JAN MAR APR MAY JUN JUL AUG SEP OCT NOV DEC Purge candidate- Yes What January, March, April, May, November Reason Unusual B-C space (115) Expected action: Double check the questionable months with an A-axle Monthly by Month graph. Review vehicle distribution both Year by Months and Month by Week to see if the problem affects classification. If it does not, a purge may not be appropriate.

180 F.4 Class 9 steering axles Monthly, Month by Date (this needs correcting, not included in previous volume!!!!!) Table used: STAT_QC_A_AX_9_DD Type abbreviation: Aax9MD Default: Class 9 scatter. Purge Candidate - no WA, 1002, East, Lane 1 - December All Class 9s Daily Avg. Steering Axle Weights Figure F.4-1 Class 9 steering axles Monthly, Month by Date example 1 These month-by-date graphs are typically used to determine precise dates when steering axle weights become questionable. This figure shows an example of a functional WIM. Average steering axle weights remain approximately constant and the +/-2SD ranges do not Figure F.4-2 Class 9 steering axles Monthly, Month by Date example 2 (case needed) Questionable steering axle weights are typically found from year-by-date graphs. These Month-by-date graphs can be used to identify the precise dates to be purged.