Coriolis Density Error Compensating for Ambient Temperature Effects

Transcription

1 Coriolis Density Error Compensating for Ambient Temperature Effects Presented by Gordon Lindsay Oil & Gas Focus Group December 2018

2 Contents Project aims and objectives Experiment Setup Phase 1 Exploratory Trials Phase 2 Manufacturer Partnership Solution Highlights Phase 3 Further Solution Validation Conclusions

3 Project Aims and Objectives Problem Incorrect data output from Coriolis technology due to ambient temperature effects

4 Project Aims and Objectives Focus on ambient temperature fluctuations that are representative of conditions found in heavy industry Reasons - Research has shown that this is an area of interest for industry end users and meter manufacturers Personal experience in this area Definite gaps in knowledge Current experimental data available largely a by-product of larger body of work and not the focus of targeted experiments

5 Project Aims and Objectives Aims and Objectives Gain a comprehensive understanding through experimentation of how potential ambient air temperature fluctuations can distort data output from Coriolis Flow Meters Identify weaknesses in existing temperature correction methods for Coriolis Technology Develop solution(s) to problem and test of solution through experimentation and validation

6 Experimentation Phase 1 Experiment Design

7 Experimentation Phase 1 Experiment Design Making use of Very Low Flow Facility at NEL Small rig allows for fine control over all variables, which would be difficult on the bigger facilities at NEL Hardest to compensate for temperature effects on small meters so tackling a complex problem 2 identical flow meters sourced Meters installed in series within the test section of the facility Meters designated as a Reference and a Test meter Note Phase 1 tests conducted with NEL acting as potential end user i.e. No access to meter correction algorithms

8 Experimentation Phase 1

9 Experimentation Phase 1 Experiment Design Test Meter Installed within temperature enclosure. Rig piping modified to enter and exit chamber via service ports Enclosure temperature modified in a controlled manner to simulate real world ambient temperature swings All meter outputs logged over time (Modbus, pulses, 4-20mA) Thermocouples attached to meter body in strategic locations to give outer casing temperature indication

10 Experimentation Phase 1 Experiment Design Reference Meter Installed downstream of the test meter. All meter outputs logged over time (Modbus, pulses, 4-20mA) Open to room atmosphere (controlled to 20 ) Thermocouples attached to meter body (same strategic locations as test meter)

11 Experimentation Phase 1 Experiment Design Temperature of fluid maintained by facility temperature bath Fluid temperature across test line monitored by PRTs in direct contact with fluid. Continuous flow rate during experimentation to minimise fluid residency time within oven Data produced from reference and test meter compared and analysed Test repeated for multiple meter designs Tests repeated over 3 fluids Water Kerosene Gas Oil NEL are UKAS certified for fluid property analysis

12 Meter Calculated Density (kg/m^3) Temperature ( ) Summary of Phase 1 Results Test Meter Calculated Fluid Density drift with respect to ambient air temperature increase Primary Axis (Density) Reference Meter Calculated Density Test Meter Calculated Density Time (minutes) Secondary Axis (Temperature) Test Meter Ambient Air Temperature (Enclosure Temperature) Actual Fluid Temperature Actual fluid temperature through both Ref and Test meters is stable Ref Meter Calculated Fluid Density also stable However as air temperature in oven steps up, test meter Calculated Fluid Density drifts and follows trend Test meter density ultimately drifts by 15.3kg/m 3

13 Selection of Phase 1 Results Journal Paper

14 Phase 2 Industry Partnership Phase 1 results presented to director and senior design team of a Coriolis meter manufacturer at their headquarters Partnership agreed for the remainder of the doctorate Meters specified and manufactured for project Full access to meter internal algorithms and associated raw measurements Access to mechanical internals of meters Test matrix agreed upon in collaboration with manufacturer Allows for further exploration of density error demonstrated in Phase 1 Access to raw measurements and algorithms allows for intelligent temperature compensation techniques to be developed through experimentation and validation at NEL

15 Phase 2 Manufacturer Partnership Key objective of Phase 2 was to develop an intelligent density correction techniques that takes account of the meter response at extreme ambient temperature Improvements to Density calculation in small bore meters a top priority for manufacturer 2 Hz difference in the resonant frequency between air & water If crack it for the smallest meter can upscale solution (with appropriate validation)

16 Experimentation Phase 2

17 Phase 2 Manufacturer Partnership

18 Selection of Phase 2 Results Raw uncorrected signals Density response to ambient temperature stepped increase Fluid - Kerosene Actual fluid temperature confirmed to be stable by facility reference PRTs Reference meter (outside of oven is relatively stable considering uncorrected values) Test meter shows considerable drift (as to be expected) of 80kg/m^3.

19 Selection of Phase 2 Results Corrected signals using manufacturer s existing temperature compensation technique Fluid - Kerosene Still a clear error in Density value with respect to ambient temperature. Drifts by total of 10kg/m^3 from reference density over 45 C air temperature increase (1.3% error)

20 Where does the problem lie? Existing harmonic period temperature correction equation assumes:- Uniform temperature profile across meter body (internal sensor positioning) Operating conditions will be similar to initial factory setup conditions Operating conditions with fluid of similar properties to factory calibration Additionally, variation in other process parameters due to ambient temperature fluctuation is not accounted for NEL Slide 20

21 Selection of Phase 2 Results Highlights One of the temperature sensors tracks fluid temperature very well and is an effective indication of fluid temperature Another temperature sensor however is severely affected by ambient temperature changes. This is currently a disadvantage based on current algorithm structure Can be an advantage however if sensor is repurposed i.e. can be used as an indicator as to the difference between fluid temperature and air/casing temperature Weakness in multiple temperature compensation equations that at present feed into each other i.e. Solution will not be as simple as just deriving a single new density calculation Access to all modbus registers and (including raw data has proved very effective in understanding meter operation in these conditions)

22 Principle of Solution Resonant Frequency Data has shown that fluid properties significantly effect efficiency of correction algorithms Solution therefore needs to adapt to work with live data and not rely on factory coefficients Dynamic X term method - Set of coefficients are continually live generated to suit conditions based on following key variables Fluid Properties Fluid Temperature Meter Body Temperature Air Temperature (approximation) Simplified Example X, Y, Z live adjusted and derived from parameters Raw Frequency = Hz Drive Gain = x% Amplifier Efficiency = x% Temperature (T1) = 20 C Temperature (T2) = 27 C Air temperature calc = 45 C... Fluid = Kerosene X = 5.45 Y= 0.67 Z= Feed into new temperature compensation equations

23 Selection of Phase 2 Results - Kerosene Live correcting with new method Density now live corrects and can keep the density within ~2kg/m^3 regardless of ambient conditions Remaining spikes are due to transient effects between tube stiffness and temperature sensor response time Error is now constrained to a temporary 0.2% error while wait thermal equilibrium

24 Selection of Phase 2 Results - Water Live correcting with new method With a different raw period range detected (water now present in meter), algorithm alters X coefficient in harmonic period temperature compensation equation and once again manages to live correct. Error now constrained to 0.1%

25 Selection of Phase 2 Results - Water Manufacturer version - total error -0.45% and stays there New Correction Technique gives temporary -0.1% per 10 C increment before recovering to expected value

26 Phase 3 Device Modification/Further Trials Further tests for academic interest considered Effect of modifying meter design to include additional temperature sensors Solution performance in less controlled controlled conditions i.e. no perfect 10 C air temperature steps Fluid temperature variation trials in combination with air temperature variation

27 Phase 3 Device Modification/Further Trials Phase 3 experiments conducted with prototype design supplied by manufacturer This allowed future proofing trials of the solution to be conducted Phase 3 experiments have been completed Data analysed at this stage looks to support the solution demonstrated in Phase 2 Data analysis will continue for remained of 2018

28 Selection of Phase 3 Results - Water Starting off the new correction method has already kicked in (ambient air 40 C) Sudden drop in air temperature causes temporary loss of balance between sensors Temperature correction levels out at ~996 expected value

29 Selection of Phase 3 Results - Water Manufacturer total error +0.8% and stays there New Correction Technique +0.5% temporary before recovering (higher error due to more extreme temperature differential)

30 Conclusions/Outcomes/Benefits Unique data sets with respect to ambient temperature effects on Coriolis flow meters have been created Covering 4 major manufacturers Over a range of fluid properties Journal Paper published with a second on the way A solution to Meter calculated fluid density error due to ambient temperature fluctuation has been developed Able to self determine fluid properties and approximate fluid/air temperature differential Based on above, appropriate correction measures selected and applied in internal processor to correct Density value output by meter Solution is earmarked for inclusion in manufacturer s future products Solution is versatile enough to be retrospectively added to existing products as an upgrade A productive collaborative partnership has been established with the manufacturer Access to privileged information Expressed a desire to continue partnership beyond EngD V. Low Flow Rig capabilities have been upgraded

31 Measured Fluid Temperature ( ) Meter Ambient Air Temperature ( ) Selection of Phase 1 Results Test Meter Indicated Fluid Temperature drift with respect to ambient air temperature increase Time (minutes) Actual fluid temperature stable ~20 (measured by calibrated reference PRTs) Gradual heating due to pumps Ref Meter Indicated Fluid Temperature also stable indicating ~20 & following same trend as actual fluid temperature However Test Meter Indicated Fluid Temperature consistently drifts with oven air temperature increases, finishing up at 25.5 With ambient air surrounding test meter finishing at Primary Axis (Meter Air Temperature) Test Meter Ambient Air temperature (Enclosure Temperature) Reference Meter Ambient Air temperature (Room Temperature) Secondary Axis (Fluid Temperature) Actual Fluid Temperature Enclosure Temperature (Test Meter ambient air) Reference Meter Indicated Fluid Temperature Linear (Reference Meter Indicated Fluid Temperature) Linear (Actual Fluid Temperature (mid point between test and ref meter)) Test Meter Indicated Fluid Temperature Linear (Room Temperature (Reference Meter Ambient Air))