转子动力学

Optical Measurement of High-Rate Dynamic VehicleRoof Deformation during Rollover

Jack Lockerby, Jason Kerrigan, Jeremy Seppi and Jeff Crandall

UVA Center for Applied Biomechanics

2013-01-0470

Published04/08/2013

Copyright © 2013 SAE International

doi:10.4271/2013-01-0470ABSTRACTThe goals of this study were to examine the dynamic force-deformation and kinematic response of a late model vansubjected to an inverted drop test and to evaluate the accuracyof three-dimensional multi-point roof deformationmeasurements made by an optical system mounted inside thevehicle. The inverted drop test was performed using adynamic rollover test system (Kerrigan et al., 2011 SAE)with an initial vehicle pitch of −5 degrees, a roll of +155degrees and a vertical velocity of 2.7 m/s at initial contact.Measurements from the optical system, which was composedof two high speed imagers and a commercial opticalprocessing software were compared to deformationmeasurements made by two sets of three stringpotentiometers. The optical and potentiometer measurementsreported similar deformations: peak resultant deformationsvaried by 0.7 mm and 3 ms at the top of the A-pillar, and 1.7mm and 2 ms at the top of the B-pillar. The top of the vehicleB-pillar sustained peak resultant deformation of 146.2 mm116 ms after contact, and unloaded to 77.1 mm (47% of peak)at 291 ms. Peak reaction forces at contact were approximately100 kN, and the force-deflection response between the droptest and the IIHS roof crush test on the same make and modelvehicle showed comparable dynamic and quasi-staticstiffness. The results presented in this study showed that theoptical system can be used to measure dynamic roofdeformations, in three dimensions, at high rates, across alarge area of the vehicle structure, from inside a vehiclesubjected to rollover crash test.

location, and magnitude of roof deformations and occupantinjury metrics.

Previous research has presented roof deformationmeasurement techniques using string potentiometers andvideogrammetry during multiple rollover events [2].However, the methods are not described in detail, the resultsare only presented for single points, and the two techniquesare not compared to each other or any other independentdeformation measurement technique.

The goal of this study was to evaluate the accuracy ofmeasurements made with a three-dimensional multi-pointoptical deformation measurement system in a high-ratedeformation experiment. The optical approach relies on theprinciples of stereo videogrammetry to accurately producethree-dimensional displacement time histories for a highresolution multi-point surface. Combined with vehiclekinematics and ground reaction force time histories, theoptical output could provide a powerful tool for validatingvehicle finite element roof structural models as well asoccupant injury metrics.

To achieve these goals, an inverted vehicle drop test wasperformed on a late model passenger van using the Universityof Virginia Center for Applied Biomechanics' DynamicRollover Test System (DRoTS) [5]. For this test, the vehicleroll drives and sled propulsion were disabled. The testparameters were chosen to mimic the loading that occurs inthe IIHS roof strength evaluation (similar to the FMVSS 216test), to allow for a force-deflection comparison betweendynamic and quasi-static roof loading conditions on the samemake and model vehicle.

Deformation measurement was focused on the passenger sideroof rail between the A-pillar and B-pillar, as this was thepoint of initial roof to ground contact and anticipated location

INTRODUCTIONMulti-point deformation information can be an enabler invalidating vehicle finite element model predictions,examining the mechanism of rollover-crash induced injuries,and studying the type of relationship between the timing,

Downloaded from SAE International by Guangxi University of Tech, Friday, December 06, 2013 09:53:55 PM

of maximum deformation. Two methods were used tomeasure the three-dimensional deformation time history ofthe vehicle roof structure. The optical system utilized two on-board high speed imagers and optical processing software.The second method, based on string potentiometertrilateration, was implemented at two discrete points on theA-pillar and B-pillar to validate the optical system's results.The resultant and component displacements for the A-pillarand B-pillar points were compared against the opticalsystem's output for the same locations.

The optical system incorporated ARAMIS v6 2.0 (GOMOptical Measuring Techniques), a commercially availableoptical measurement software package. ARAMIS utilizeddigital images to analyze and report three-dimensionalmaterial deformations along with the surface coordinates ofpoints, displacements and velocities, and strain values andrates.

METHODSTest Conditions

The vehicle's as tested mass was 1911.9 kg. The initialconditions for the vehicle were a pitch angle of −5 degrees(front end down), a roll angle of +155 degrees (passengerside impact), and a drop height of 0.398 m (total verticalmotion of CG prior to impact), (Figure 1). This produced a2.706 m/s vertical velocity at contact. Unpublished finiteelement analysis results suggested that peak reaction forcefrom this drop height would be approximately equal to thepeak force achieved in the IIHS roof strength evaluation withthe same vehicle.

Figure 1. Initial vehicle orientation for the drop testThe IIHS roof strength evaluation, the quasi-static test on thevehicle used for comparison, consisted of a roof crush test toone side of the vehicle. The test system utilized an uprightassembly with attached loading platen fixed at 25 degrees roll

and 5 degrees of pitch [4]. The vehicle roof was crushed to aminimum of 127 mm of platen displacement at a nominal rateof 5 mm/s. Force and displacement data were reported at 100Hz for the test.

Vehicle Preparation

The vehicle's overall mass and wheel distributions weremeasured upon delivery at UVA as the test goal weight, andmonitored throughout the preparation and installationprocess. The vehicle fluids were drained and the seats, floorliners, roof liners, and curtain airbags were removed to allowfor instrumentation installation. Steel plates were bolted tothe vehicle floor to act as mounting locations forinstrumentation and data acquisition equipment (Figure 2). Acontact strip sensor was taped on the roof in the location ofanticipated initial roof-to-ground contact as a trigger. Thepassenger side interior structural surfaces were painted with astochastic pattern of black dots, 6 to 12 mm diameter withapproximate 40-50% coverage of the patterned area, on awhite background to accommodate the optical system.

Figure 2. Instrumentation mounting plates and sensorcube installed location on the floor pan of the vehicle

Balancing ProcedureThe DRoTS cradle [5], which consisted of two verticallyoriented steel tubing towers at the front and rear, connectedby a set of telescoping steel tubes 5″ in diameter, wasinstalled on the vehicle to fix the location and orientation ofthe vehicle's roll axis relative to the vehicle (Figure 3). It isabout this roll axis that the DRoTS fixture rotates the vehiclein a typical rollover crash test (disabled for this test). Onceinstalled, the telescoping tube ran under the vehicle bodyfrom the front to the rear and the towers were orientedperpendicular to the ground. The towers of the cradle werefixed to the vehicle frame rails in the front and rear of thevehicle by removing the fascia and bumper beams and usingcustom hardware to rigidly connect the towers to thelocations where the bumper beams interfaced with the vehicleframe rails. The cradle assembly was centered (left-right)between the bumper beam/frame rail connections at the front

Downloaded from SAE International by Guangxi University of Tech, Friday, December 06, 2013 09:53:55 PM

and rear of the vehicle. For simplicity, the cradle wasinstalled such that the tube was oriented parallel with theground when the vehicle was resting on its own suspension/tires, at the curb weight condition. The DRoTS fixture wasdesigned to hold the vehicle cradle in such a way that thevehicle roll axis was parallel to the cradle. Once the vehiclewas loaded in the test fixture, the location of the roll axisrelative to the vehicle and cradle was adjusted vertically untilthe roll axis passed through the center of gravity (CG) of thevehicle. If the vehicle CG was not aligned in the center of theframe rails (left-right) then ballast weight was added to oneside to ensure alignment. At the completion of the balancingprocedure, the vehicle would not continue to rotate in eitherdirection after being manually rotated and stopped at 45degree intervals from 0 to 360 degrees. After this adjustmentwas completed the exact location of the roll axis could beidentified from a single marker point at the very front andrear of the cradle.

Figure 3. DRoTS cradle installation on front of vehicle

Force and Kinematics Measurement SensorsThe vehicle was outfitted with an inertial measurement cubecontaining three accelerometers (Endevco 7290E-30, MeggittSensing Systems, San Juan Capistrano, CA) measuringaccelerations about the sensor cube's local x, y, and zcomponent directions. Additionally three angular rate sensorswere included measuring angular velocities about the samelocal sensor cube axes (DTS ARS-1500, DiversifiedTechnical Systems, Seal Beach, CA) for the x direction, andtwo (DTS ARS-300, Diversified Technical Systems, SealBeach, CA) for the y and z directions. The sensor cube wasmounted on the floor of the vehicle, on the lateral center line(Figure 2). The DRoTS roadbed was instrumented withtwenty-four uniaxial load cells, spaced evenly across thesurface, and oriented to measure in the vertical direction(global Z′). Two accelerometers measuring in the vertical Z′direction were mounted to the underside of the wood surface

at the left-front corner and right-rear corner of the centersection of the roadbed.

Deformation Measurement SensorsThe passenger side A-pillar and B-pillar were bothinstrumented with a set of three string potentiometers of 2159mm maximum travel (Firstmark Model 62, FirstmarkControls, Creedmoor, NC). Two mounting plates (25 × 50mm) were constructed out of steel stock. To each plate, asingle screw and a triax accelerometer were attached (EntranTriax EGAXT3, PandAuto Technology Co., Ltd). One of thetwo plates was mounted on the passenger-side A-pillar, at alocation approximately 150 mm from the intersection of theA-pillar, roof rail, and windshield header. The second platewas mounted on the roof rail, approximately 100 mm forwardof the B-Pillar roof rail connection (Figure 4).Figure 4. B-pillar string potentiometer attachment plate

The plates were fixed to the vehicle via a single screwinserted into existing threaded holes used for mounting therollover curtain airbags. Rotation of each plate was preventedwith a sharpened screw, inserted in the plate that wastightened against the vehicle interior at the mountinglocation. The strings from each of three string potentiometerswere attached to the screw on each of the two plates (n= 6string potentiometers). The six string potentiometers wereaffixed to the floor of the vehicle via rigid steel platesattached to the driver, right front passenger and second rowseat mounting locations (Figure 5).Downloaded from SAE International by Guangxi University of Tech, Friday, December 06, 2013 09:53:55 PM

Figure 5. String potentiometer mounting and orientation

Optical SystemThe optical system was comprised of two high-speed imagers(NAC GX-1, NAC Image Technology, Simi Valley, CA)with 16 mm focal length ruggedized lenses (SchneiderOptics, Hauppauge, NY). The imagers were fixed at a 21degree horizontal angle with respect to each other, andmounted on a rigid aluminum camera bar along with six highintensity LED, and two laser sights tracing the center pixel ofeach camera for visual alignment. The camera bar was boltedonto a steel framing structure that was welded into the vehiclefloor at the front row seat mounting beam (Figure 6). In themounted orientation, the cameras shared a focal intersectionpoint 915 mm from the lenses, which was positioned 25 mmin front of the intersection of the vehicle passenger side roofrail with the windshield header (Figure 7). Sections of theroof, roof rail, A-/B-pillars, and windshield header above thepassenger seat were included in the shared field of view.

Figure 6.

Figure 6 (cont). Optical system components andmounting structure on the floor pan of the vehicle

Figure 7. Images from the left and right cameras at t=0

ms, structure undeformed

Coordinate MeasurementsA coordinate measurement machine (CMM) (Titanium Arm,FARO Technologies, Lake Mary, FL) was used to determinethe locations of the front and rear roll axis marker points, sidedoor latch plates, wheel centers, the inertial measurementcube origin and orientation markers, the locations of thestring potentiometer attachment screw locations on the A-/B-pillars, the points where the strings come out of thepotentiometer bases, the accelerometers mounted on thestring attachment plates, and four points located in the viewof the optical system. Since digitizing these points requiredmoving the CMM around the exterior of the car, a series ofoverlapping points were taken in each measurement location,and each point cloud captured in each measurement locationwas aligned in a single coordinate system by singular valuedecomposition [3].Data Processing

Kinematics Data ProcessingAll sensor data were filtered and debiased. Road load cells,vehicle accelerometers, and string potentiometers, werefiltered to CFC60 and vehicle angular rate sensors werefiltered to CFC180 [6]. For the purpose of calculating theangular accelerations necessary for determining vehicle CGkinematics from kinematics measured at the sensor cube, the

Downloaded from SAE International by Guangxi University of Tech, Friday, December 06, 2013 09:53:55 PM

angular velocity data were filtered with a Butterworth 4-polezero-phase low-pass digital filter with a 25 Hz cutoff. Roadloads were summed to determine total vertical reaction force.The time of initial vehicle to road contact was determined towithin 1 ms from high speed video, and all time history datawere time shifted so that t=0 corresponded with the time ofinitial vehicle contact with the road surface.

Once the CMM data point clouds were unified, vehicle localand global (inertial) coordinate systems were defined. Astandardized SAE definition for the vehicle coordinatesystem was used [6]. The vehicle's positive X axis wasdefined by a vector directed from the rear roll axis point tothe front roll axis point; the positive Y axis was defined by anaverage of vectors connecting similar points from the driver'sside to the passenger's side door latch plates; and positive Zwas determined by the cross product of X and Y vectors.These vectors were calculated in the CMM coordinate systemand were used to define the rotation matrix Rvehcmm.The point cloud was then rotated from the CMM coordinatesystem to the vehicle local coordinate system by Rvehcmm andits origin was translated to the center of the line connectingthe front and rear roll axis marker points. The front/rear right/left weight distribution was used, with the CMM-determinedlocations of the center of the wheels, to determine thelocation of the vehicle CG on the XY plane. Then the originof the vehicle coordinate system was translated to the CGpoint.

The global (inertial) coordinate system was defined based onthe sled system rails used to propel the roadbed. The roadbedmotion was defined to travel in the positive X′ (global)direction and the Z′ axis was defined to be perpendicular tothe test facility floor, with positive pointing downward. Theglobal Y′ axis was determined by the cross product of Z′ withX′.

Figure 8. Vehicle local and global (inertial) coordinate

systems

To transform the sensor cube kinematics measurements to thevehicle local and global (inertial) coordinate systems rotationmatrices defining the relationships between these frames hadto be identified. The transformation between the sensor cubeand vehicle local coordinate system Rvehcube, which remainedconstant throughout the test, was determined from the set ofsensor cube orientation points captured with the CMM.Sensor accelerations and angular rates were transformed tothe vehicle local coordinate system by Rvehcube.

To calculate vehicle kinematics in the global reference frame,the time history of the transformation Rglobalveh relating thevehicle local system to the global system was determined.Beard and Schlick formulated a method to update the matrixRglobalveh at each time step using only local frame angularvelocity measurements [1]. The details of this method arepresented concurrently with this study (Kerrigan et al. 2013,SAE).

Before transforming the acceleration data from the sensorcube to the local coordinate system using Rvehcube, the sensoraccelerations were corrected for the effect of gravity. Globalacceleration time histories were then calculated bytransformation of vehicle frame acceleration (at the vehicleCG) time histories by the Rglobalveh time history. Globalvelocities and displacements were determined by numericalintegration of the global acceleration and global velocities,respectively.

String Potentiometer TrilaterationData were sampled from the potentiometers at 10 kHz duringthe test. At each time step, 0.1 ms increments, the three-dimensional location of the A-pillar and B-pillar attachmentpoints were solved for using a trilateration technique. For thisprocess it is assumed that the intersection of the three pointsexisted on the surface of three spheres with centers fixed atthe locations where the strings come out of thepotentiometers. It is also assumed that those string orientationpoints do not move relative to the vehicle CG and coordinatesystem during the test. For this analysis, a potentiometercoordinate system is defined to have its origin at the locationof one of the points where the string exits the potentiometersensor. The potentiometer x axis is defined to point in thepositive direction such that a second string base is at alocation (d, 0,0) and the third potentiometer lies in the +xyplane at a location (i,j,0). From this the distance from each ofthe three pots (r1, r2, and r3) to the location of intersection ofthe three strings can be written as

Downloaded from SAE International by Guangxi University of Tech, Friday, December 06, 2013 09:53:55 PM

Rearranging the equations to solve for the unknowncoordinates x, y, and z of the intersection point yields

The z coordinate, expressed as the positive or negative squareroot, can have zero, one, or two possible locations. Thepositive z location was chosen to locate the point above thevehicle floor. The resulting coordinate location time historiesfor the A-/B-pillar points were rotated into the standardvehicle coordinate system using the CCM data points.

Optical System ProcessingBefore installation into the vehicle, the optical system wascalibrated in the ARAMIS software using a specializedcalibration object from the manufacturer (GOM OpticalMeasuring Techniques). A measurement volume of 1280 mmby 1280 mm by 1175 mm was produced for the test. Imageswere captured at 915 by 915 pixel resolution, with an averageinitial focal length to the structure of 915 mm. Each pixel inthe image corresponds to a 1.5 by 1.5 mm square on thevehicle surface. The black dots in the paint patterncorresponded to between 4 to 8 pixels in diameter on average.The imagers recorded synchronized frames at 3000 Hz for800ms, 100 ms before the event and 700 ms after roof-to-ground contact.

The images were analyzed in ARAMIS at 1000 Hz, usingevery third frame in the series. ARAMIS recognizes thesurface structure of the measured object in an image, andassigns 2D coordinates to the image pixels. For 3Dmeasurements, two cameras are calibrated with the softwareto record synchronized images. ARAMIS usesphotogrammetric methods to combine the 2D coordinates foreach pixel, as observed from the left and right camera images,into a common 3D coordinate for the analysis. The softwareobserves the deformation of the object through a series ofimages by discretizing each image into facets, unique squaregroupings of pixels, similar to finite elements used incomputational analyses. The facets are identified by the graylevel structure of the individual pixels within the facet. Thesoftware tracks the changing location of each facetthroughout the image series to calculate the displacementsand strains of the object's surface. The pixel size of each facetand the pixel step, overlapping area of adjacent facets, can beadjusted in the software and manually optimized for theanalysis conditions including image resolution, geometriccomplexity of measured object, and desired resolution of theresulting strain and displacement fields. In order for thesoftware to divide the images into recognizable facets, theremust be sufficient variation of pixel gray scale on the object's

surface. Surfaces that are heterogeneous in color require theapplication of a stochastic paint spray or dot pattern toproduce pixel variation.

A facet size of 20 pixels and a facet step of 10 pixels wereused in the analysis. This provides an output of individuallycalculated points in a 10 by 10 pixel grid, equivalent to every15 mm on the physical surface. Using the physical coordinatelocations captured by the CMM and their correspondinglocations in the ARAMIS software coordinate system, atransform was applied to the optical output to align it with theSAE vehicle coordinate system.

RESULTSForce and Kinematics

The peak sum total vertical force recorded by the roadbedwas 98,754 N at 40.5 ms after touchdown (Figure 9). Afterthe initial peak, the force time history shows oscillatorybehavior of approximately 3 Hz, centered around 19,000 Nwhich corresponds to the weight of the vehicle.

Figure 9. Roadbed sum total vertical force

The peak resultant global vehicle CG acceleration was 7.8 gat 10.9 ms (Figure 10). The peak for global X′ accelerationwas 7.2 g at 11.0 ms, the global Y′ was 4.1 g at 14.9 ms, andthe global Z′ acceleration was −6.9 g at 24.8 ms. Globalvelocity resultant reached a maximum of 2.87 m/s at 15.8 ms.The peak global Z′ velocity was 2.83 m/s at 9.4 ms (Figure11). Global resultant displacement had a maximum of 0.566m at 121.4 ms (Figure 12). The peak global Z′ displacementwas 0.563 m at 121.7 ms.

Downloaded from SAE International by Guangxi University of Tech, Friday, December 06, 2013 09:53:55 PM

Figure 10. Vehicle CG global accelerations

Figure 11. Vehicle CG global velocities

Figure 12. Vehicle CG global displacements

The maximum acceleration measured at the B-pillar was 31.1g at 5.5 ms after impact, and at the A-pillar was 15.7 g at 7.4ms (Figure 13). The accelerometers measure approximately 1g before t=0, as the vehicle is falling under gravity.

Figure 13. Resultant Acceleration at A-/B-Pillars

Roof Deformation Measurement

The peak resultant deformation reported by the optical systemwas 146.5 mm at the top of the B-pillar 116 ms after contact(Figure 14). The structure subsequently unloaded to aminimum resultant deformation value of 77.1 mm at 291 ms(Figure 15). This accounted for a 47.3% rebound, normalizedby the peak deformation.

Figure 14. ARAMIS resultant displacement overlaidonto left camera image at t=116ms, maximum 146.2 mm

Downloaded from SAE International by Guangxi University of Tech, Friday, December 06, 2013 09:53:55 PM

Figure 15. ARAMIS resultant displacement overlaidonto left camera image at t=291ms, maximum 77.1 mmThe maximum resultant deformation at the A-pillar was 87.9mm at 123 ms reported by the optical system, and 88.6 mm at121 ms reported by the string potentiometers (Figure 16). Atthe time of peak resultant, the component deformationsreported by the optical system were −20.9 mm in the X,−81.4 mm in the Y, and 25.3 mm in the Z directions. Thestring potentiometers reported −48.6 mm in the X, −73.5 mmin the Y, and 9.5 mm in the Z directions.

Figure 16. A-Pillar displacement comparison between

String Potentiometers and ARAMIS output

The maximum resultant deformation at the B-pillar was 146.2mm at 112 ms reported by the optical system, and 145.0 mmat 114 ms reported by the string potentiometers (Figure 17).At the time of peak resultant, the component deformationsreported by the optical system were −24.1 mm in the X,−134.2 mm in the Y, and 52.8 mm in the Z directions. Thestring potentiometers reported −27.7 mm in the X, −132.3mm in the Y, and 52.4 mm in the Z directions. The 3 Hzoscillation is also present in the roof deformationmeasurements at the A-/B-pillars.

Figure 17. B-Pillar displacement comparison between

String Potentiometers and ARAMIS outputThe vehicle exhibited elastic rebound effects captured by theoptical system (Table 1). The unloaded displacement valuesare taken from t=291ms when the first deformation minimumoccurs after contact. The normalized rebound values arecalculated by dividing the amount of deformation rebound(peak minus minimum) by the peak deformation.

Table 1. Optical displacements at the A-/B-pillars fortimes of maximum deformation and maximum

unloading

Downloaded from SAE International by Guangxi University of Tech, Friday, December 06, 2013 09:53:55 PM

Force-Deflection

The maximum total sum force measured during the test at theroadbed of 98,754 N was reached at 48 mm of resultantdeformation for the A-pillar, 67.5 mm of resultantdeformation for the B-pillar, and at 96 mm of global verticaldisplacement of the vehicle's CG following initial roof-to-ground contact (Figure 18). The maximum force reached inthe IIHS roof crush test for this same vehicle was 100,617 Nat 85 mm of platen deformation.

Figure 18. Force vs Deflection curves for the A-/B-Pillarresultants, CG global vertical Z displacement, and IIHS

roof crush test

DISCUSSIONOptical System

To successfully implement the optical deformationmeasurement technique on-board a dynamic test, thecalibration and focal intersection of the stereo imagers mustbe maintained throughout image capture. With peakaccelerations of around 8 g, it was imperative that thecameras be rigidly fixed with respect to each other as anydivergent motion between the two cameras can cause error ortotal loss of results. A rigid connection to stiff structuralmembers in the vehicle frame was designed to prevent rigidbody motion of the camera beam with respect to themeasured surfaces. In this test, the optical system maintainedcalibration between the two imagers during the vehicle-to-ground impact acceleration spike and subsequent vibrations.The nominal intersection deviation between the two imagerscalculated by ARAMIS was 0.8 pixels, and did not changesignificantly during the impact. Some areas of complexgeometry on the hat section roof stiffeners, and areas of

significant buckling on the roof sheet were not trackedthroughout the entire deformation.

Although the images were recorded at 3000 Hz during thetest, they were processed at 1000 Hz in ARAMSIS to reducecomputational expense and post processing effort. Due to thelimited size of the ruggedized lenses used there wassignificant vignetteing of the images (Figures 7, 14, 15). Thepartially blocked view interfered with the software's ability toautomatically match facets between the left and right images,requiring manual pairing for each time increment. Futuretests should utilize the full image sensor chip to increaseresolution and accuracy, while avoiding manual postprocessing. This was a limitation of the current study;however, it appears that 1000 Hz is sufficient to capture thefast rate deformation of a vehicle roof during an invertedground contact.

Deformation Measurement Comparison

The string potentiometer attachment locations on the A-/B-pillars were chosen to be the existing curtain airbag tappedholes to minimize the installation's impact on vehiclestructure. Due to geometric constraints on the vehicle interior,the string potentiometers were not mounted orthogonally inline with the vehicle coordinate axes, but rather to maximizethe angle between the three strings to improve the accuracy ofthe trilateration technique (Figure 5). Orthogonal alignment isnot necessary when utilizing trilateration; however it isimperative that the potentiometer bases are rigidly mountedto the vehicle structure so that they remain fixed in thevehicle coordinate system, and are assumed to not moverelative to the vehicle CG. High speed video of the stringpotentiometers refuted concerns of string oscillationsimpairing the calculation, as only slight vibrations areapparent in the video and do not translate into the collecteddata signal. The oscillations in the string potentiometer signal(Figures 16, 17) are present before the drop initiates when thestrings are not moving, and are attributed to signal noise.The deformation time histories of the optical system andstring potentiometer trilateration correlated well (Figures 16,17). The resultant displacements at the A-/B-pillars for thetwo techniques matched almost exactly, reporting adifference in peak deformation of 0.7 mm and 2 ms at the A-pillar, and a 1.2 mm and 2 ms at the B-pillar. The B-pillarcomponent displacements were in good agreement betweenthe two methods with the maximum difference of 13%occurring in the X direction.

Component Disagreement and OngoingTesting

The lack of A-pillar component agreement was at firsthypothesized to be related to a coordinate transformationerror in the string potentiometer processing, as the resultantmatched closely and the X and Y directions were in phase but

Downloaded from SAE International by Guangxi University of Tech, Friday, December 06, 2013 09:53:55 PM

differed in magnitude; however, no processing error wasfound. The same transformation that was applied to the A-pillar was applied to the B-pillar data, which matchedbetween the two measurement methods. A second test on thesame vehicle, a full rollover test with vehicle roll and roadbedtranslational velocity, was performed with the stringpotentiometers and optical system mounted in the samepositions and orientations as the drop test. The output fromthe two systems produced matching resultants at the A-/B-pillars and components at the B-pillar, but again thecomponent deformations at the A-pillar were indisagreement.

Subsequent full rollover testing on a different vehicle,utilizing the same data collection and processing methodsdetailed in this paper but with different mounting locationsand orientations for the string potentiometers and the opticalsystem, produced matching resultant and component outputsat both of the two compared locations. The discrepancybetween the test described in this study and these other testshas led to the hypothesis that there is some geometric effectspecific to the orientation of the string potentiometers in thisvehicle during the drop and rollover tests that caused the A-pillar component displacement errors. The exact source ofthese errors remains unknown at this time, and furtherinvestigation will be performed through subsequent testing.In this case of the drop test's A-pillar componentdisagreement between the two measurement methods, theoptical output is hypothesized to be the more accurate of thetwo methods for several reasons. Firstly, the opticalcomponent deformation curves along the roof rail betweenthe A-pillar location (where a discrepancy exists) and the B-pillar (where the two measures agree) exhibited smoothcontinuity. The string potentiometer component contours didnot match the deformation contours at any point on thevehicle structure. Secondly, the string potentiometerspredicted over 50 mm of displacement in the X direction atthe A-pillar, a value roughly double that of the highestpredicted X component anywhere on the structure by theoptical system. And thirdly, the string potentiometerspredicted complete unloading back to 0 mm in the Z directionwithin 400 ms of contact in an area that retained measurableplastic deformation post-test.

As observed through the subsequent full rollover tests, theoptical system was able to accurately track roof deformationwhen vehicle roll and sled translational velocities were addedto the test conditions. The resulting multi-plane kinematicsdid not have a noticeable effect on optical recording andprocessing. It is hypothesized that the optical system wouldalso perform accurately in a multiple roll event, such as theJ2114 dolly rollover test, as long as airborne debris did notdamage the imagers or obstruct the view of the measuredsurfaces. The imagers would need to have sufficient on-boardmemory available for recording the entire multi-roll event at

high speed, exceeding the memory requirements for a singleroll on the DRoTS.

Validation

Since the two independent techniques were in agreement forthe A-/B-pillar resultants and B-pillar componentdeformations, and they utilized two very differentdeformation measurement strategies, the optical and stringpotentiometer systems were validated for the on-board use inhigh-rate vehicle deformation tests. While the componentdistribution of measurements was validated at the B-pillar,small differences at the A-pillar suggest that further analysismay provide greater certainty in the componentmeasurements of the string potentiometers. The opticalsystem produced true multi-point component displacementfields for the vehicle structure, which is ideal for use invalidating finite element models.

Component Deformations

Due to the +155 degree initial roll angle and −5 degreespitch, the Z component of the force vector acting on thevehicle roof was 90.6% of the total force, whereas the Ycomponent was only 42.3% of the total force, roughly half ofthe Z component. The majority of the loading vector is in thevertical direction; however, the component displacementresults show the majority of deformation in the Y direction(Table 1). At the time of peak deformation, the Y componentwas 92.7% and the Z component was 28.8% of the resultantat the A-pillar. Similar results were recorded at the B-pillarwith the Y component 91.8% and the Z component 36.1% ofthe resultant.

Quasi-Static and Dynamic LoadingComparison

The IIHS roof strength to weight ratio is calculated bydividing the maximum force prior to 127 mm of platendeformation during the quasi-static test by the measured curbweight of the vehicle. This ratio was reported as 5.15 for thetested vehicle. The dynamic strength to weight ratio wascalculated from the drop test by using the peak sum totalroadbed force (98,754 N) divided by the as tested weight ofthe vehicle (1911.9 kg * 9.81 m/s2). The dynamic ratio was5.27 for the tested vehicle. The force-deflection behaviorsbetween the dynamic and quasi-static roof crush tests werealso similar, with the peak force of the quasi-static test 1863N larger, and the displacement at the time of peak force was11 mm less (Figure 18). The global CG vertical displacementwas used as the deformation measure in the dynamic test, asthis was the most accurate analog for total platen deformationfrom the IIHS test. The vehicle exhibited a stiffness, the slopeof the force-deflection curve, under dynamic deformation thatis similar to that under quasi-static deformation. Nosignificant rate effects were apparent in the comparison.

Downloaded from SAE International by Guangxi University of Tech, Friday, December 06, 2013 09:53:55 PM

Future Research

Unfortunately, neither the optical or string potentiometerdeformation measurement approach can be utilized in avehicle with seated and belted ATD's, roof liners, and curtainairbags due to line of sight and physical string obstructions.This limits direct correlation of roof deformation timehistories and occupant injury metrics. Ideally, a method todetermine deformation time history that is compatible withsimultaneous ATD injury testing could be developed,possibly using strain gauges installed on the vehicle structurebehind the roof liner. Another potential approach, measuringroof deformation in one test and occupant injury risk inseparate replicate test, is highly dependent on a preciselyrepeatable rollover test fixture.

compared to the Z direction; however the vehicle structurehad a majority of deformation in the Y direction.

The dynamic stiffness of the vehicle to roof loading wassimilar to the quasi-static stiffness, with limited ratedependence shown.

REFERENCES1. Beard, D. A., & Schlick, T. (2003). Unbiased RotationalMoves for Rigid-Body Dynamics. Biophysical Journal, 85,2973-2976.

2. Croteau, J., Zolock, J., Larson, R., Bare, C. et al.,“Dynamic Response of Vehicle Roof Structure and ATDNeck Loading During Dolly Rollover Tests,” SAE Int. J.Passeng. Cars - Mech. Syst. 3(1):407-449, 2010, doi:10.4271/2010-01-0515.3. Hanson, R. J., & Norris, M. J. (1981). Analysis of

Measurements Based on the Singular Value Decomposition.Siam J. Sci. Stat. Comput., 2(3), 363-373.

4. Insurance Institute for Highway Safety, “CrashworthinessEvaluation Roof Strength Test Protocol (Version II).”October 2012.

5. Kerrigan, J., Jordan, A., Parent, D., Zhang, Q. et al.,“Design of a Dynamic Rollover Test System,” SAE Int. J.Passeng. Cars - Mech. Syst. 4(1):870-903, 2011, doi:10.4271/2011-01-1116.6. SAE International Surface Vehicle RecommendedPractice, “Instrumentation for Impact Test - Part 1 -Electronic Instrumentation,” SAE Standard J211/1, Rev. July2007.

SUMMARY/CONCLUSIONSThis study evaluated the dynamic stiffness (slope of theforce-deflection curve) of the tested vehicle as compared tothe quasi-static stiffness under similar roof loading. Theaccuracy of deformation measurements made by an opticalsystem mounted inside a vehicle subjected to dynamicloading was also evaluated against independent deformationmeasurements by string potentiometers.

The result of this study, validation of the optical system foruse in recording multi-point high-rate vehicle deformations ina crash test, provided a methodology for multi-point motiontracking that may be a useful data collection tool in rollovercrash testing. The optical system produced deformation datafor a wide area of vehicle structure, which is well suited forvehicle finite element model validation. Use of opticalmeasurements shows promise as a data collection method forresearchers studying the complex relationship between roofdeformation timing, vehicle to occupant interactions, andinjury mechanisms.

The string potentiometer trilateration approach was validatedfor providing resultant displacement time history of high ratedeformation. High speed video of the string potentiometersrefuted concerns of string oscillations impairing results.The roof-to-ground loading vector for +155 degrees rollproduced half the component force in the Y direction

The Engineering Meetings Board has approved this paper for publication. It has

successfully completed SAE's peer review process under the supervision of the sessionorganizer. This process requires a minimum of three (3) reviews by industry experts.All rights reserved. No part of this publication may be reproduced, stored in a

retrieval system, or transmitted, in any form or by any means, electronic, mechanical,photocopying, recording, or otherwise, without the prior written permission of SAE.ISSN 0148-7191

Positions and opinions advanced in this paper are those of the author(s) and not

necessarily those of SAE. The author is solely responsible for the content of the paper.SAE Customer Service:

Tel: 877-606-7323 (inside USA and Canada)Tel: 724-776-4970 (outside USA)Fax: 724-776-0790

Email: CustomerService@sae.org

SAE Web Address: http://www.sae.orgPrinted in USA