The J2 Universal Tool-Kit Supporting Accident Investigation AIRCRAFT MODELLING AND PERFORMANCE PREDICTION SOFTWARE Key Aspects INTRODUCTION PA-31-325 C/R Navajo Accident Objectives MODEL BUILDING Aircraft Description Geometry Assembling the Model TESTING AND QUALIFYING THE MODEL Sanity Checks Reference Checks PRELIMINARY ACCIDENT MODELLING Updating the Model Initialising the Scenario Preliminary Dynamic Investigation Adding the Detachment Force AAIB INSPECTOR WORKSHOP Additional Information Matching the AAIB Findings CONCLUSIONS Supporting and Supplementing the Evidence Pilot Evaluations.
INTRODUCTION The AAIB asked the question whether the j2 Universal Tool-Kit, a full flight sciences design, modelling and analysis tool, could be used to investigate the characteristics of crash scenarios. When this was confirmed, j2 were requested to investigate the scenario of a particular accident that had occurred. The results were compared to the findings from the report which is available in the public domain. PA-31-325 C/R Navajo Accident The AAIB investigator provided the following case information and a series of questions/objectives. The problem calculate the response of the aircraft due to a section of the outer left wing becoming detached due to tree strikes. The outer 2.4m of wing detached due to impacts with trees, which coincidentally ties in with the outboard end of the left flap i.e. no left aileron remaining. Representation of Accident - Note that the scan shows a PA-31-350 Chieftain with a longer fuselage than the Navajo, but the wing is the same on both aircraft.
Aircraft details Weight at time of accident = 5,800lbs Fuel on board at time of accident = 300ltr CG position = 27.5 aft of datum Landing gear was DOWN Flaps set 10 degrees Engines set to low power, props fully fine Airspeed at point of impact with trees = 115 KIAS Objectives 1. Assuming that the outer 2.3m of the left wing is instantaneously detached, and the right aileron floats at zero deflection (aileron cable circuit now no longer continuous due to damage to left wing), demonstrate the roll due to the imbalance 2. Apply a Point Load at the severed end of the left wing, due to the detachment forces of the severed portion of left wing. Can J2 calculate the aircraft response in pitch, roll and yaw to the left yawing moment? 3. Combine the above two effects to create a simulation of the flight path of the aircraft following detachment of the left wing section?
MODEL BUILDING As the information provided was limited, the first stage of any project is to try to gather as much information as possible. Whilst the investigator would be able to access some data from the manufacturers directly, this was not available to j2 at this point, so all information had to be sourced from data that was available in the public domain. Aircraft Description On 30 September 1964 Piper flew the prototype of a new twin-engine executive aircraft which was then the largest built by the company. Identified at first as the Piper PA-31 Inca, the aircraft had been re-designated as the PA-31 Navajo when deliveries began on 17 April 1967. A six/eight-seat corporate/ commuter transport of cantilever low-wing monoplane configuration with retractable tricycle landing gear, it was powered by two 224kW Avco Lycoming IO-540-K flat-six engines, and was available in optional Standard, Commuter and Executive versions with differing interior layouts. Made available at the same time was the optional PA-31T Turbo Navajo, which differed only by having two 231kW TIO-540-A turbocharged engines, and the range was extended in 1970 by introduction of the PA-PA-31P Pressurized Navajo with a fail-safe fuselage structure in the pressurised section and two 317kW Avco Lycoming TIGO-541-E1A engines. Production of the PA-31 Navajo ended during 1972 and at the same time the company introduced for 1973 the PA-31-350 Navajo Chieftain which, by comparison with its predecessor, had the fuselage lengthened by 0.61m and was powered by two 261kW TIO-540-J2BD turbocharged engines driving counter-rotating propellers. A significant advance in the Navajo family came on 22 October 1973 when Piper flew the first production example of the PA-31T Cheyenne, which combined an airframe generally similar to that of the Pressurized Navajo with two 462kW Pratt & Whitney Aircraft of Canada PT6A-28 turboprop engines. In the following year an additional model of the Turbo Navajo was made available, the PA-31-325 Turbo Navajo C/R, which introduced a 242kW version of the counterrotating engines installed in the Chieftain. Production of the PA-31P pressurized Navajo ended during 1977, at which time a total of 248 had been built, but at the same time the company introduced a new version of the Cheyenne, the PA-31T-1 Cheyenne I, the original Cheyenne then becoming re-designated PA-31T Cheyenne II. Deliveries of the new Cheyenne I, which differed primarily from its predecessor by having 373kW Pratt & Whitney Aircraft of Canada PT6A-11 turboprop engines, began towards the end of April 1978. The Cheyenne range was extended for 1981 by introduction of the PA-31T-Cheyenne IIXL, with the fuselage lengthened by 0.61m and 559kW Pratt & Whitney Aircraft of Canada PT6A-135 engines flat-rated to 462kW. In 1982 production of the PA-31 Navajo terminated after 1,317 had been built. Later production versions of the Navajo family include the PA-31-325 Navajo C/R, PA-31-350 Chieftain and the PA-31T-1 Cheyenne I, PA-31T Cheyenne II and PA-31T-2 Cheyenne IIXL. However, the loss of the Navajo was compensated for in 1983 by introduction of the PA-31P-350 Mojave, which basically combined the airframe of the Cheyenne II with the powerplant of the PA-315-350 Chieftain.
Summary Aircraft Dimensions Wing span 40 ft 8 in Length 34 ft 7 in Height 11ft 3in Cabin Dimensions Height: 4.25 ft Width: 4.15 ft Length: 8.6 ft Volume: 151 ft 3 Door Height: 3.5 ft Door Width: 2.25 ft Baggage External: 63 ft 3 Occupancy Crew: 1 or 2 Passengers: 5-7 Weight and Balance Empty weight 4,250 lb Maximum Takeoff Weight (MTOW) 8,130 lb Max Landing Weight: 6,500 lb Fuel Capacity: 1,122 lb Payload with Full Fuel: 1,284 lb Payload Max: 2,070 lb Propulsion Two Lycoming engines Left TIO-540-J2BD Right LTIO-540-J2BD Horsepower 350 @ 2,575 RPM ea. Propeller Manufacturer Hartzell Hub Model Left HC-E3YR-2A Hub Model Right HC-EY3YR-2AL Number of blades 3 Performance Altitude Limits Maximum Operating Altitude 27,000 ft Service Ceiling 28,300 ft
Speeds Normal Cruise Speed: Max Cruise Speed: Climb Rate All Engines: Climb Rate One Engine Inop: V s Stall Speed V A -Manoeuvring Speed V Mo -Maximum operating Speed V FE -Flaps Extend Speed Mid: Full: V LE -Maximum Landing Gear Operating Speed Range Normal Range: Max Range: Full Range (vfr): Max Range (vfr): Landing Distance Landing Distance: Balanced Field Length: 178 kts 226 kts 1,395 fpm 115 fpm 63.5 kts (flaps down) 162 KIAS 187 KIAS 152 KIAS 130 KIAS 130 KIAS 640 NM 810 NM 757 NM 927 NM 2,902 ft 1,750 ft Geometry The geometry is obtained from various sources (including Jane s AWA) and aircraft 3-Views. Where data is not available, it is estimated from aircraft of a similar class. Wing Span 12.4m Area 21.3m 2 AR 7.37 Chord Root - Jane s, at fuselage intersection 2.610m Inboard Section Projected to 3.135m Outboard Section Projected to 2.465m Tip 0.97m mac 1.826 Airfoil Root Tip NACA 63A415 NACA 63A212
Dihedral 5 Twist Root +1.5 Tip -1 Sweep LE 4.14 25% 0.693 30% 0.0 Wing Geometry for Use in Model Build
Wing Geometry for use in Model Build Flaps and Aileron Geometry for Use in Model Build
Horizontal Tail Span 5.52m Area 5.603m 2 AR 5.438 Chord Root 1.380m Tip 0.650m mac 1.059m Airfoil Root NACA 0011 Tip NACA 0011 Dihedral 0 Twist Root 0 Tip 0 Sweep LE 9.53 25% 5.81
Horizontal Tail Geometry for use in Model Build Vertical Tail Span 2.1m Area 2.898 m 2 AR (for symmetric wing) 3.043 Chord Root 2.070m Tip 0.690m mac 1.495m Airfoil Root NACA 0011 Tip NACA 0011 Dihedral 0 Twist Root 0 Tip 0
Sweep LE 43.46 25% 38.073 Vertical Tail Geometry for use in Model Build Assembling the Model Once the research had been performed and the geometry identified the next stage was to establish the aerodynamic characteristics for the lifting surfaces, fuselage and undercarriage. Lifting Surfaces The lifting surfaces were calculated by putting the airfoil section information into JavaFoil and running analyses over a range of angle of attack values at the cruise Reynolds Number. The analysis for the wing was performed for Root and Tip Airfoil Sections with a range of flap deployments, and a range of aileron deployments. Although the Flaps do not extend to the tips, and the ailerons are not up to the fuselage section, the airfoil is only known at these sections. By putting the values of the root and tips in, and specifying the limits for the flaps and ailerons, the JavaFoil software will interpolate the information only over the region specified. The Ailerons were calculated as a plain flap at 25% chord. For the Flaps, these were identified as single slotted at 20% chord. As no geometry was available for these, a set of coordinates and pivot point were calculated for the root and tip airfoil sections.
Flap Geometry for NACA 63A212 Airfoil Section Flap Geometry for NACA 63A415 Airfoil Section
Detail of Slot Geometry for Flaps Flow Field with Flaps Deployed Notes To partially account for the spanwise lift distribution, the values for the coefficients calculated at the root were calculated using double the aspect ratio of the wing whilst those for the tips were calculated using half the aspect ratio.
From here we were able to establish the clean airfoil section details, and the increments due to the flaps and the ailerons. 3.5 3 2.5 2 1.5 1 0.5 0-20 -15-10 -5 0 5 10 15 20-0.5 Flaps 0 Flaps 5 Flaps 10 Flaps 15 Flaps 25 Flaps 35 Flaps 40-1 -1.5 α [ ] Lift Coefficient for Root Airfoil Section 0.35 0.3 0.25 0.2 0.15 0.1 0.05 Flaps 0 Flaps 5 Flaps 10 Flaps 15 Flaps 25 Flaps 35 Flaps 40 0-20 -15-10 -5 0 5 10 15 20 α [ ] Drag Coefficient for Root Airfoil Section
0-20 -15-10 -5 0 5 10 15 20-0.1-0.2-0.3-0.4 Flaps 0 Flaps 5 Flaps 10 Flaps 15 Flaps 25 Flaps 35 Flaps 40-0.5-0.6 α [ ] Pitching Moment Coefficient for Root Airfoil Section A similar approach was used for the NACA0011 Airfoil with a plan flap for Elevator (37%) and Rudder (40%).
Fuselage To calculate the fuselage characteristics, the fuselage profile along with a simple lifting surface geometry was run through DatCom. The resulting data is then collated into tables and the fuselage contribution extracted. Fuselage and Lifting Surface Geometry used with DatCom Undercarriage The undercarriage data was calculated from methods found in Roskam. Propulsion The propulsion system was broken into 2 sections. Engines The engines were simple powerplants that provided a direct link between the pilot throttle and the output power of the engine. The engines were limited in their maximum power. Propeller The propeller takes the power from the engine and converts it into a thrust using the true airspeed and the propeller efficiency. The efficiency was fixed at 70%. Whilst this is not the most accurate approach, the variation in thrust that will occur over the crash profile will be very minimal and as such an almost constant thrust can be assumed calculated from the thrust requirements to maintain steady flight prior to the crash event. Mass & Inertia Whilst the Mass and CG position were specified in the initial request, this did not include the inertia for the aircraft. The inertias were calculated using Radius of Gyration and the approach outlined in Roskam.
Assembly With all the data and geometry defined, the final stage is to put it all together. The Aircraft model is assembled in j2 software using the structural components and layout of the real aircraft. Each item was also given its reference location relative to their parent. The Centre of Gravity (C of G) CG, Location of the Lift (NP) and in the case of Propulsion Items the Centre of Thrust (C of T) was defined. The Z-Location of the undercarriage NP is variable dependant up the deployment of the gear. The aerodynamics for the fuselage can be added at this stage because they are independent upon anything else. Structural Items on the Aircraft Model X Locations of Structural Items Relative to the Centre of the Nose (-ve Aft)
Z Locations of Structural Items Relative to the Centre of the Nose (-ve Up) With the structure laid out, the next stage is to add the lifting surfaces. This is done through adding in the stripped items. Each stripped item contains the geometry of the lifting surface, including dihedral and twist. With each horizontal item a maximum and minimum Y value is included to account for the start and stop point. In this way we can use the same geometry for all the wing components. The j2 software will only calculate the values for the appropriate spanwise positions. Different Spanwise Max & Min for Each Surface
The final stage is to add the aerodynamics. This is taken from the values calculated earlier. For the wing, the coefficients are functions of the local angle of attack and the spanwise locations. This takes into account the change in airfoil section. With the Ailerons and Flaps, there is an additional parameter in the lookup tables for the relative surface deflection. For the clean wing the values with no surface deflection are added. For the Ailerons and Flaps the values are the increments in the coefficients due to the surface deflection only. Adding in the Aerodynamics When considering the Horizontal and Vertical Tail, the Elevator and Rudder were assumed to be full span. In this case the total values for the clean surface and the contributions due to the surface deflection are included in a single coefficient. One aspect that needs to be considered is the downwash from the wing to the Horizontal Tail. This uses the methods outlined in Roskam, using the geometry and relative locations of the Wing and Horizontal Tail. To account for the fact that there is downwash at zero angle of attack, an alpha equivalent value is used. This is calculated from the lift curve slope of the clean wing, and the total lift for the wing, including flaps. This alpha equivalent value is used in the downwash gradient calculations. The value of the downwash is then applied to the horizontal tail through applying a constant deflection (twist) to the whole tail. Calculating the Downwash and Applying a Twist to the Horizontal Tail With the strip aerodynamics complete, we add in the propulsion characteristics, the undercarriage drag contributions and the mass and inertia. The complete aircraft model can now be qualified.
TESTING AND QUALIFYING THE MODEL With a completed model, the first stage is to run a series of tests. These tests serve two purposes. The first is to ensure that the model has been assembled correctly, and that there are no unusual characteristics present. The second purpose of the tests is to check that the values the model is outputting are in line with expected results. Two types of tests can be run: Sanity Checks This is where the j2 Universal Tool-Kit is used as a virtual wind tunnel. A series of alpha and beta sweeps will be run along with angular velocities to identify the total static characteristics for the aircraft, and the dynamic derivatives. These numbers can be compared to values from an aircraft of a similar type and class to ensure that they are not wildly different. Reference Checks The reference checks are when there are real data for the aircraft under test. This could be wind tunnel data, or flight test data or whatever scenarios/data are available. In these situations, the model can be set up to mimic the same condition and the resultant values calculated and compared. Sanity Checks When running the sanity checks on the piper Navajo, the results for the static/dynamic derivatives and surface contributions were found by running a range of sweeps for alpha and beta, and then for each of the control surfaces. Additional roll, pitch and yaw rate cases were also run over a range of angle of attack values. The derivatives were calculated and results were then compared to those for Airplane B in Roskam s Airplane Design: Part VI. These analyses can be set-up and re-used for different aircraft with the results plotted on a template chart. This provides consistency and speed in the analytical process.
Results to Initial Sanity Checks
From the Sanity checks, the majority of characteristics were within the expected ranges for this class of aircraft. Where there was a discrepancy, in the Roll Damping ( ), where this was slightly higher than expected, this was easily corrected by adding a correction factor to the overall model. Reference Checks Having got the model to a point that it is showing the appropriate characteristics for an aircraft of this class, the next stage was to try some Reference Checks. For these, information was found from the University of Tennessee Space Institute who have their own Piper Navajo that is used for research and experimentation. From this aircraft various tests have been run than in turn have produced some aerodynamic characteristics. By replicating the tests and comparing the results from j2 Universal Tool-Kit to those from the real aircraft, further corrections can be found. In this scenario, we have already built an aircraft model with the mass and cg characteristics relating to the crash to be investigated. As expected, these are not the same conditions as for the Experiment. Therefore, a new Delta Model was created where the mass and cg information matched the UTSI Test. The Aircraft was then initialised to the same flight conditions and a further set of tests performed to enable the data to be compared (classic flight matching). Once the initial tests had been performed and the results compared, it was found that whilst the Lift Curve Slope ( ) was comparable, there was an initial lift offset ( ) error. Small Drag Coefficient errors were also found. Static corrections to the Lift and Drag for the Clean Wing were added to the Clean Wing strips local coefficients and the test cases re-run. The corrections identified were shown to produce acceptable results. Corrections to Mass & CG to Match UTSI Test Point
Comparison of Lift Coefficient (Before and After Correction) to the UTSI Experimental Values Comparison of Drag Coefficient (Before and After Correction) to the UTSI Experimental Values Once the corrections were shown to produce acceptable results, these were added back to the baseline model. At the same time a small correction to the roll damping was also added as a global coefficient on the Total Airframe.
Corrections to,,, Applied to the Baseline Model Correction to Applied to the Complete Airframe From here the sanity checks were re-run to establish the impact of the changes to Wing Lift and Roll Damping to the Overall Dynamic Derivatives. These now fit into the acceptable ranges.
Updated Roll Rate Characteristics
PRELIMINARY ACCIDENT MODELLING Updating the Model When modelling the crash, we did not want to impact directly on the baseline model that we have tested and qualified. The best approach was to create a delta model where the adjustments could be made to account for the tree strike and the resulting loss of the outboard section of the wing. This was done instantly within the j2 software On the delta model, 2 new inputs were added. Wing Lost (m) This enabled us to define how much of the wing, in metres, was severed as the analysis is taking place. Detachment Force (N) This enabled us to add a force, in Newtons, to the model to represent the impulse when hitting the tree. New Inputs on Crash Delta Model These were then connected into the model. The value from the Wing Lost (m) input was added to the left outboard value (Min.y) of the clean wing geometry thus reducing the span of the left wing. As this is driven from an input, the value of the Wing Lost (m) can be modified at any time during the analyses
Changing the span of the Left Wing When considering adding a force to account for the tree strike, a Propulsion Item was used as this could have a force added directly. The engine was given a tilt value of 180 - so that it was always horizontal to the ground regardless as to orientation of the aircraft. The force to be applied would be at the new wing tip position, and would be equal to the Detachment Force (N) value that can be entered as an input. Initialising the Scenario Adding the Tree Strike Force to the Model The next stage was to initialise the aircraft. As there was no information given regarding the flight path angle, the aircraft was initialised in straight and level flight over a range of approach angles. Using the trim rules built into the j2 Universal Tool-Kit, it was very easy to set the aircraft up in any configuration. The aircraft was initialised with: Flaps at 10 Landing Gear Down Elevator Trimming out Any Pitch Acceleration Wings Level Constant Airspeed of 115 KCAS Flight Path Angle from -7 to 0 Zero Forward Acceleration Altitude aligned to flight path angle so after 5s the altitude is 60ft
Trim Model Defining the Start Points for the Analyses From here the initialisation was performed using the delta model.
Completed Analyses Initialising Model Preliminary Dynamic Investigation The first objective was to identify the rolling characteristics of the aircraft with the wing section removed. To do this a response model was set up that allowed the aircraft to travel unaffected for 5s and then at 5s a step input of +2.4 was added to the Wing Lost (m) input. As shown before, this will result in the Left Wing having a semi-span of 3.8m. The aerodynamics within j2 will now ignore any contributions outboard of 3.8m on the left wing, but will include all the contributions from the right wing. Defining the Instantaneous Loss of 2.4m of the Left Wing Starting from the Level Flight Initial Condition (gamma=0 ), the response of the aircraft to this event was analysed. The analysis automatically terminates when the aircraft hits the ground.
Lateral Response to Instantaneous Loss of 2.4m Left Wing As we can see from the above chart, when the left wing section is lost, there is an imbalance in the lift with increased lift on the right wing and reduced lift on the left wing. This results in a negative rolling moment (to the left). In this scenario, the aircraft reaches a maximum roll rate of 24 /s, to the left, and a bank angle of almost 90, left wing down, before impact with the ground. What can also be seen here is the time from tree-strike to impact is approximately 4.5s. Whilst charts are very useful at showing us the behaviour of the aircraft, it is often still difficult to fully interpret the complete behaviour. It is especially difficult in a situation such as this where the aircraft is moving in all 3 axes. In this situation, engineers will typically resort to using small models to follow the attitude values of the aircraft to try to fully understand what is happening. This is not necessary with j2, as it is possible to instantly take the time history generated and display it in a 3-D Virtual Environment Playback. Within this playback, it is possible to move around the aircraft, viewing it from different angles, zoom in/out, and speed up/slow down time to get a more detailed and highly visual understanding of what actually happens.
3-D Playback of Preliminary Study Adding the Detachment Force Having satisfied Objective 1, the next stage was to see what would happen should a detachment force be applied. A method for applying a force to the model has already been created, so the next stage was to add in the detachment force through the response model. It was estimated that the time taken to sever the wing at the flight speed would have been approximately 0.03 seconds. So to account for this, an initial force was applied that then dissipated over 0.03s and then ceased. This force was combined with the wing loss to build the scenario. Combining the Detachment Force and the Severed Wing in a Response Model No value for the detachment force was provided, so the mechanism used allowed for different forces to be applied to help to identify what was probable. An initial estimate for the force was calculated by j2.
The Detachment Force Applied to the Wing at the Severed Location We can now re-run the response analysis from the same start point as before and chart the new values, without writing any code. Lateral Response to Instantaneous Loss of 2.4m Left Wing and Detachment Force We can see from the new chart that the reaction is much more violent. There is an instantaneous yawing moment due to the impact with the tree that causes the aircraft to yaw up to -50 /s. This Yaw rate increases the lift over the right wing which gives an even greater rolling moment such that the aircraft now reaches a maximum roll rate of 140 /s, to the left, and a maximum bank angle of 130, left wing down, before impact with the ground.
3-D Playback of Wing Loss plus Detachment Force At this point the primary 3 Objectives had been satisfied, and the next stage was to discuss our findings with the AAIB inspector.
AAIB INSPECTOR WORKSHOP Once the primary objectives had been satisfied, a short workshop was arranged, where we could present our results and discuss the findings with the AAIB Inspector. The Initial presentation of method and results was understood, and various key points of the modelling and scenario we agreed upon. However, there was viewed to be some discrepancy between the evidence found by the AAIB inspector and the orientation of the aircraft, from the modelling, as it hit the ground. The evidence trail from the crash indicated that the aircraft hit the ground in the following sequence, Right Wing Tip, Right Engine, Nose. Whereas the modelling indicated, that the left wing would strike the ground first. Additional Information When discussing the discrepancies, further information was provided that was not presented in the original data supplied to j2. In the first instance, the tree strike took place on a hill, and the impact with the ground was 120ft below that location. Secondly, the aircraft was on a gradual descent, approx. 3, as shown from the radar track. Below is an extract of the AAIB Report, available in the Public Domain. The aircraft s left wing had struck two 80 ft pine trees approximately 20 ft below treetop height, causing the outer 2.2 m of the left wing, outboard of the left flap, to fragment and detach. Witness marks on the severed tree trunks indicated that the roll attitude of the aircraft at impact with the trees was wings level. The wreckage trail, from the point of the initial tree strikes to the aircraft s final resting position, was 230 m in length and was orientated on a heading of 298 M. Parts of all major sections of the structure and flying controls were identified at the accident site. The first ground impact scar was 185 m beyond the initial tree strike and had been made by the right wingtip; it was 20 cm wide and the narrowness of this mark indicated that the roll angle at ground impact was approximately 90 right wing low. The ground impact marks and distribution of the wreckage indicated that following the right wingtip strike, the right engine hit the ground and detached, shortly after which the aircraft impacted heavily on its nose. The absence of any significant ground scars between the nose impact crater and the main wreckage indicated that the aircraft then bounced a distance of 34 m, before finally coming to rest facing uphill. G-BWHF Radar Track (Map terrain elevations are shown in metres)
Adding an Offset (Disturbance) to the Initial Altitude Whilst this information was different to the original information presented, it was easy to accommodate these changes instantly. To account for the change in height, the response was simply offset by 120ft and as we had already initialised the aircraft over several flight path angles the analysis was run from the 3 descent initial conditions. Response Results to Updated Altitude and Approach What can be seen when we change the initial altitude and flight path angle is that there is very little difference between the Initial Conditions that matched the AAIB Report and the baseline case. As expected, the approach angle has very little effect on the maximum Roll Rate (P) and despite the flight lasting a little longer due to the increased altitude, the aircraft does not achieve any larger a bank angle as the roll rate dropped considerable by the time the aircraft is approaching the ground such that there is very little increase in bank angle.
Comparison of Bank Angle for the Original Case and Using the Initial Conditions from the AAIB Report Matching the AAIB Findings Part of the Workshop with the AAIB Inspector was to look at evaluating different scenarios in real time based on additional information as it is presented. Due to the flexibility and power of the j2 Universal Tool-Kit we were able to study numerous and varied what-if scenarios to identify what conditions may have occurred to present an impact similar to that found through the debris trail. We had already established that the physics model and aerospace knowledge dictates that the aircraft will have a negative (left wing down) roll rate for an incident of this type. Thus we knew that the only way for the right wing to hit the ground first was for the aircraft to rotate beyond the inverted position (180 ) before impact. Therefore the first objective was to identify what areas of the analysis may need adjusting to achieve this scenario within reasonable limits. Three areas were investigated: Aircraft Inertia The inertia estimated for the model may be slightly high. In addition, losing part of the wing would reduce the roll inertia further. A trade study could be performed to identify the effect of reducing the inertia. Roll Damping The roll damping had already been found to be slightly high for the class of aircraft, and a correction had been applied. A study of the effect of a further reduction in the roll damping coefficient could be performed. Detachment Force This had been estimated from simple calculations. A trade study could be performed to assess the bank angle when changing the values of the impact force.
Control Surface Deflection The original briefing explained that the aileron cables were severed but the aileron remained at 0 deflection. However, previous Accident Studies performed by j2 have found that there is a major impact when a control surface is able to float during a dynamic manoeuvre. If the cables had been severed, then it was highly likely that the right aileron would move freely as a result of the aircraft dynamics and inertia. All the studies were quickly and easily performed and the resulting flights charted and visualised during the 2 hour workshop with the AAIB Inspector watching and commenting. This meant we were able to perform live analysis with a qualified inspector adding knowledge and experience into the process concurrently. Roll Characteristics The model was first updated by reducing its roll inertia (I XX ) and the scenario ran again. It was quickly found that to get the aircraft to roll beyond 180 the roll inertia would need to be reduced to a level that it was deemed to be unrealistic. The next stage was to look at the roll damping. As with the roll inertia, the level of change required to achieve the desired result was viewed to be too severe to be realistic. A final study was run to see if a realistic reduction in roll inertia and roll damping would cause the desired effect. This was shown not to be the case. Detachment Force When looking at the detachment force, it was increased by 10% to see what sort of impact this may have on the overall bank angle of the aircraft. It was found that this increase would be sufficient to roll the aircraft over to the right wing. Lateral Response to 10% Increase in Detachment Force
Right Wing Hits the Ground first with 10% increase in Detachment Force Control Surfaces From previous accident studies performed by j2, it has been identified that when left to float (i.e. Control Cables Snapped) control surface are unlikely to remain at zero deflection. This is especially true when there are highly dynamic manoeuvres. As such, it was decided that the right aileron may move following the crash. The motion of the aileron was dependent upon the roll acceleration and damped out by the roll rate. Driving the Ailerons using the Roll Dynamics
Lateral Response to Floating Aileron Left Aileron Deflection when Floating It can be seen that the aircraft rolls beyond the inverted position and the aileron deflects up to 8. However, when looking at the visualisation we can see that the rotation is insufficient for the right wing to hit first.
Right Engine First with Aileron Floating Combinations With a small increase in the Detachment Force or a small deflection of floating aileron both increase the bank angle beyond the inverted. Whilst the detachment force increase can achieve the desired situation alone, a 10% increase was viewed to be too much of a correction. Similarly, any further aileron deflection was assumed to be too large. As such a combination situation was put together. The Detachment Force was increased by 5% and the impact of the roll acceleration (P ) on the aileron was reduced slightly. The aileron characteristics were adjusted and re-flown several times and each time the flight was visualised instantly to see what the aileron deflection was, and how the aircraft impacted with the ground. After only a few attempts, a response was found that appeared to fit the evidence very closely. The Detachment Force was within limits, and the resultant Floating Right Aileron Deflection was within limits. The aircraft orientation upon termination of the flight was 95 Right Wing Down. Thus the right wing tip hits the ground first, followed by right engine and finally nose.
Lateral Response to Combination Scenario Aileron Deflection during Combination Scenario
Ground Impact with Combination Scenario One final sanity check was to look at the predicted distance the aircraft travelled from impact with the tree to the first impact with the ground. Flight Path Profile
What was found from the analysis was that the analyses terminated when the aircraft was 191m from the tree. If we look at the first impact with the ground, this is calculated at 184.5m from the impact with the tree. This compares very well to the 185m presented as evidence from the crash site.
CONCLUSIONS Supporting and Supplementing the Evidence The model build, analysis performed was able to satisfy all the objectives presented by the AAIB inspector and provide dynamic video footage of the scenario. The subsequent workshop with the AAIB inspector was able to fill in the additional information provided and use this to identify how the aircraft was able to rotate almost 270 from impact with the trees. These results were consistent with the AAIB findings including the location of the first impact with the ground. AAIB Photo of Crash Site and Corresponding j2 Flight Path
From being presented with the project to the final workshop and identifying a scenario to match the AAIB findings was 10man days. This shows that the j2 Universal Tool-Kit is not only a very powerful and capable tool for investigating air accidents, but the results can be obtained in a fraction of the time of more traditional approaches. Automatic Generated Playback of Flight from j2 Virtual
Pilot Evaluations One further capability available when working with the j2 Universal Tool-Kit is the ability to fly the aircraft in a manned simulator without the need to write and develop any code. This means that as the aircraft model is being developed, it can be tested and evaluated by a pilot concurrently. Finally, when the delta model was constructed to add the tree strike force and loss of wing, it is possible to have these automatically added into the Instructor/Operator Station and can be injected at any time. This means the controllability of the aircraft after the tree strike can be fully evaluated. Configurable Simulator provided by j2 Pilot Providing a Fully Integrated System with Engineering Workstation Directly Integrated with Desktop or High Fidelity Simulator For further information please go to or contact info@j2aircraft.com