# Chapter 12 | Staging a Car Crash | RICSAC 1

## Introduction

In Virtual CRASH 3 one can setup and refine a collision sequence with incredible speed. Virtual CRASH automatically updates simulations in real time as input parameters are tuned and adjusted. In this write-up we will explore how to simulate a collision between two automobiles. The exact same techniques will be applicable to collisions involving pickup trucks, commercial vehicles, or any other vehicle type. We will attempt to reproduce results from a staged collision test from the Research Input for Computer Simulation of Automobile Collisions (RICSAC) series. In this reconstruction, we will attempt to solve for the pre-impact speeds given knowledge of the pre-impact orientations and post-impact rest positions and orientations. The purpose of this write-up is to show the Virtual CRASH user the typical workflow used during an accident reconstruction analysis; it is not intended to be a refined study of RICSAC collisions, as such studies have been performed by other researchers. We recommend reading references (1), (2), and (3) for more information on the RICSAC tests and test conditions.

A video of the resulting simulation can be found below:

## RICSAC 1

In the RICSAC 1 collision, a 1974 Chevrolet Chevelle Malibu impacted a 1974 Ford Pinto. The figure below depicts the impact configuration, as well as pre and post-impact trajectories.

Import Scale Diagram

Using the procedure outlined in reference (4), import the diagram shown in the previous figure into the Virtual CRASH environment (see below). You can copy and paste the diagram directly from this pdf file by simply copying a screen capture, pasting into Window’s Paint application, and using the crop tool. After you have imported, scale, and oriented your diagram, it is recommended you freeze the image object to prevent accidental modification as you build your simulation. Note, some minor adjustments will be made to the image scale due to slight scale-bar inaccuracies in the Smith and Noga diagrams.

With the scale diagram as your environment’s backdrop, you can easily place additional graphical elements into your Virtual CRASH scene to suit your needs. In this example, the scale diagram was rotated and positioned so that it properly aligned with the impact and rest positions described in the “Summary Form” of page 2-8 of reference (2); this form is shown in Table 1 below. Note the data shown in this table is assuming SAE conventions, with the $$z$$-axis running parallel with gravitational acceleration, whereas in Virtual CRASH, the z-axis is aligned anti-parallel with gravity. This implies one must take $$y \rightarrow -y$$ and $$yaw \rightarrow -yaw$$ when reading data from this table for use in Virtual CRASH. The diagram was further refined using the “Axes” tool, which allows one to place a coordinate axis system within the environment. This tool also allows the user to place points within the $$x-y$$ plane. This feature was used in order to place crosshairs at the impact and rest positions of each vehicle (see below). The “Text” tool was used to annotate the diagram.

Table 1

Place Vehicles into Scene

Now that we’ve finished building our simulation environment, we are ready to place our vehicle models into the scene. Looking through reference (5), one can look for reasonable exemplar vehicle shells for the subject vehicles. In this case, a Skoda Felicia seems to reasonably match the body style of the Pinto. To place a vehicle into the scene, left-click on the “+” symbol next to “cars” in the left side control panel to reveal the car library. Next left-click on the “+” symbol next to “Skoda,” and then left-click and hold “Felicia.” Drag the mouse into the scene and release the left mouse button. You should see the vehicle model in the scene.

Using this same procedure, we will next place an AMC Matador into the scene as our exemplar for the Malibu.

Modify Vehicle Data

Using the methods described in reference (6), modify your vehicle properties to match the geometric size of each vehicle, center-of-gravity positions, and wheel placement (see below). In this case, the weights were taken from the “Crash Test Summary” on page 7-4 of reference (3) (shown as Table 2 below), whereas the wheel position (Table 3 and Table 4) and size data (Table 5 and Table 6) were taken from Expert Autostats. On page 7-20 of reference (3), the front/rear weight distributions are given for both vehicles. Using this distribution, we can solve for a more precision value for the distances between the centers-of-gravity and the front-axles. Doing this, one obtains 3.81 ft between the cg and front axle for the Pinto, and 4.24 ft for the Malibu. Note, when the cg location is adjusted after a vehicle has already been placed in the scene, take care to reposition the initial start positioning to the desired location as it will shift when the cg location is modified.

Table 2

Table 3

Table 4

Table 5

Table 6

Touch-up the Vehicle Mesh

Now that the vehicle wheels have been repositioned to match the Autostats data, the Pinto’s vehicle mesh needs to be touched up to correct the awkward relative position of the rear wheels and the rear wheel wells.

Using the method described in reference (7), the vertices can be easily adjusted to correct this problem.

Place Vehicles at Pre-Impact Positions

Using the data in Table 1, we can place the vehicles at their pre-impact positions. The Malibu should be given coordinates $$(x,y,yaw) =$$ (-10.8 ft, y=-1.0 ft, -30 deg) and the Pinto should be placed at (0.0, 5.5, 90 deg). The vehicle positions can be precisely controlled using the “position-local” menu in the left side control panel, and the orientations can be controlled using the “rotation-local” menu.

Place Clone Vehicles at Post-Impact Positions

Select both vehicles, and using the clone tool, make a clone of both vehicles (see below). We’re going to use the cloned vehicles to mark the rest positions of the vehicles.

Next using the rest positions indicated in Table 1, move the clone vehicles to their proper points of rest, and give them both the proper final orientations.

Now with the clone vehicles selected, select Create > Physics > Remove Physics From Selection. This will ensure our clone vehicle shapes are simply used for rest position markers, and won’t accidentally conflict with the physics simulation.

It is known, and has been reported elsewhere, that the diagrams included in Reference (1) have minor scaling errors that can cause slight inaccuracies in our final results. To correct for this, we will use the reported points of impact and rest, as well as pre- and post-impact orientations to adjust the diagram from Reference (1). Select the diagram image and adjust the scale length down iteratively until good agreement is observed between the pre and post-impact vehicle objects and pre- and post-impact locations indicated in the diagram. In this diagram, it was sufficient to lower the scale length to 9.843 ft (see below). Our intention is the use the diagram to give us an indication of post-impact trajectories via tire marks or paint marks, but we adjust our simulation inputs until we have good agreement between the simulated and actual rest positions and orientations indicated in Table 1.

Go to the left side control panel and click on the name of your Malibu vehicle. Hold the left mouse button and drag it down so that you also select the name of your Pinto vehicle; this is another way to select multiple objects at once. You can use the left control panel menus to modify values that are common to simultaneously selected objects. You should notice both vehicles have red boxes in the simulation environment indicating they are both currently active selections.

Next, go to the “contact” menu and change the locked-wheel drag factor “adhesion” value to $$\mu=$$ 0.87 as indicated by the tire pavement drag factor in Table 1.

Enter Rolling Resistance Values

With both vehicles still selected, go to the left side control panel and left-click on “sequences” to access the braking data. Next, left-click on the empty box to the left of “wheels separately” to adjust the braking at each wheel as a percentage of the adhesion value. Table 1 indicates that the rear wheels of both vehicles had a drivetrain resistance of 20% $$\mu$$ whereas the front wheels had a rolling resistance of 1% $$\mu$$. Set the “brake lag” value to 0 seconds so that the braking is enabled at time = 0 seconds.

Establish Reasonable Search Window

Though it is not completely necessary, to help speed up our search for a simulation solution to our case, we’ll first study the post-impact motion of both vehicles, and look for post-impact speeds that allow the vehicles to traverse the path from the point of impact to point of rest. We will then use conservation of linear momentum (COLM) to estimate the implied pre-impact speeds for both vehicles, which will then give us a better idea of what pre-impact speeds to give our vehicles within the full simulation.

Set the Pinto aside in your simulation environment. Now with your Malibu rest position maker object visible, adjust the initial speed, velocity vector orientation (vni), and yaw rate (omega-z) until you get the Malibu to come to rest in the area indicated by the marker. In this case, the Malibu’s impact position was adjusted forward so that the rear tires began making marks in approximately the same area depicted in the diagram. In this case, our results indicate a post-impact speed of about 4.2 mph at an angle of 10.39 degrees in the Earth frame (yaw + vni) will allow the Malibu to traverse the required distance to go from impact to rest (see below).

Next, repeat this process for the Pinto.  Here we see an initial speed of 15.723 mph with vni of 67.424 degrees and yaw rate of -0.379 rad/s is sufficient to project our Pinto to the point of rest. This implies the velocity vector is directed along -22.576 degrees in the Earth frame.

With the final velocity data, we can use momentum conservation to estimate the pre-impact velocity. Here we simply use a spreadsheet tool to perform the calculations. Without loss of generality, we can simplify our problem by first rotating our x-axis by 30 degrees (our "Adjustment Angle" in the spreadsheet below) such vehicle 1 is traveling along the x-axis. This implies that vehicle 2 is traveling along the pre-impact heading given by yaw = -120 degrees. Our estimates for pre-impact speeds are given by:

These estimates are shown in yellow below:

These result indicate reasonable first estimates for both vehicles would be about 16 to 17 mph.

Note, we can take this preliminary analysis one step further if we wish, because while searching for potential post-impact velocities during this preliminary search phase, there will typically be a large range of possible values for (vi, vni, omega-z) that will yield the correct final vehicle positions to within a reasonable tolerance. We know that variations in final velocity vector orientation and magnitude imply a range of possible pre-impact velocities are possible from COLM. To explore a reasonable window of pre-impact velocities to start our full simulation within, we can use a Monte Carlo analysis to determine these pre-impact velocity windows (see Reference (9)). The Excel application RC Monte Carlo (see Reference (10)) provides a very simple and powerful user interface that is ideal for this task. In the case of RICSAC 1, a very simple scan of values for (vi, vni, omega-z) for each vehicle suggested the full range of possible final velocity vector values:

Assuming equal chances (i.e., Uniform “Probability Density Function”) for these inputs, 15,000 simulations yielded the following results from RC Monte Carlo. For the Malibu, we have:

Here we see for the 15,000 monte carlo experiments, a pre-impact speed window of 13 to 18 mph is possible (min to max) for the Malibu. For the Pinto we have from the RC Monte Carlo tool:

These results suggest the Pinto may have a pre-impact speed within the window of 13 to 21 mph (min to max).

Therefore, using these results from momentum conservation, we have established reasonable pre-impact speed windows within which to search for solutions in our full simulation. With this information, we can proceed with our full simulation.

Begin Iterating the Pre-Impact Speeds

With the pre-impact configuration set, we are now ready to optimize our simulation. Here we assume no initial knowledge of the pre-impact speeds. In order to converge upon a final solution for our reconstruction, we must iterate across our physics model’s input parameters, which must be kept within a physically meaningful range. It is possible that given a final post-impact state which we are trying to match, there can be multiple solutions which do equally as well to match our intended goal; in this case our goal is have the vehicles come to rest as close as they can to the measured rest positions.

Let’s start by increasing the pre-impact speed of the Malibu. Go to the “dynamics” menu and slide the speed up until you see the Malibu stop near the final position marked on the diagram (see below). You can use your mouse’s scroll wheel to move the “v” slider. Note, at about 15 mph, the Malibu comes close to the final position.

Now adjust the Pinto’s initial speed. You’ll notice that the Virtual CRASH system automatically updates and plays the simulation based on your current inputs (see below). At 15 mph pre-impact speed for both vehicles, the Pinto is about a car-width short of the final target, implying there is not enough initial kinetic energy for the Malibu in the current simulation. Iteratively increase the speeds of each vehicle until a better match is found.

We will stop this first round of iterating with v[Malibu] = 17.485 mph and v[Pinto] = 16.864 mph. In this case, both vehicles are much closer to their rest points. Here the rest points are shown with the cloned vehicles models we created earlier.

Open the Report

To see how well we’re doing so far with our simulation, go to the left side control panel and left click on “report dynamics” in the “tools” menu (see below). We only need to look at the final positions of the vehicles, so set the “time” increment setting to 1 second. This will display the vehicle data every second in the report table. Now we can read off the final positions and orientations of the vehicles. The final $$(x,y,yaw)$$ configuration for the Malibu is (-2.294 ft, 1.351 ft, -7.893 deg) and is (6.690 ft, -1.824 ft, 115.760 deg) for the Pinto.

Collision Data

Using the methods shown in reference (8), access the first impact collision data by left-clicking on “auto-ees” and then selecting “previous contact” in the left control panel. Note, you can access data in subsequent collisions by pressing “next contact.” In the “defaults” menu, we can see the collision model parameters used. Under “object 1” we see that the Pinto’s $$\Delta v =$$ 15.062 mph and the Malibu’s $$\Delta v =$$ 10.077 mph (see below). These values could also have been obtained in the report, by setting the output time increment value to a much smaller value and reading off and analyzing the velocity vector data just before and just after impact. The results for the current state of our simulation are shown in Figure 172 above. The results are summarized in Table 7 (Note, here we use adjusted $$\Delta v$$ values from Reference (11), where the authors corrected for accelerometer displacement away from the vehicles’ centers-of-gravity). Thus far, the largest discrepancy appears to be the final position and orientation of the Pinto. Our initial velocity estimates are known already to within 16% of the measured values, $$\Delta v$$ to within 18%, and PDOF values to within 11 degrees. We were able to determine these in just a few minutes of work. Notice the initial velocity values shown in Table 7 are slightly less than the initial velocity settings used for the simulation. This is because the first impact occurs at time = 0.065 seconds. Prior to the moment of first impact, both vehicles are decelerating according to the rolling resistance drag factors entered into the sequences menus for both vehicles.

Table 7

Create User Contact

Depending on the needs of your case, you may already wish to end the reconstruction project here; however, we will try to do some further optimization to better match the final positions and orientations of our vehicles. To do this, we need to adjust the parameters of the impulse vector for the first impact. With the initial contact already selected from the previous step, left-click on “create user contact” under the “selection” menu. Again, to ensure the first impulse is selected, press “previous contact” until you see you have clearly selected the impulse corresponding to first contact. Once you press “create user contact” a new ees object will be created and will be visible in the left side control panel.

This new ees object will have its own collision properties which you can now modify.

Next, go to the objects window in the left side control panel, and hide the “auto-ees” objects to make things easier to see in the diagram. Left click on the new ees object in the left side control panel. Deselect “auto-position” in the contact menu, so that we may move the centroid of the impulse vectors. Note you can enable deformation of the vehicle meshes by selecting “deform,” though we will not do that in this simulation. Make sure your mouse cursor is on “Select And Move” or “Select, Move And Manipulate.”

Virtual CRASH will initially automatically determine the centroid position for the impulse vectors based on the intersection of the vehicle bounding boxes at the collision time, which controlled by the “depth of penetration” parameter. This centroid position can be adjusted based on your accident reconstruction analysis to yield more consistent results for your case. You can reposition the centroid either by using the “position-local” menu or by using left-click-hold-and-drag. If you interactively drag the impulse centroid around the scene to adjust its position, you may want to freeze underlying objects to prevent accidental selection. Though you can place the impulse centroid anywhere in the scene, obvious care must be taken to ensure it is being correctly positioned with respect to both vehicles such that it is consistent with the physical evidence.

You’ll note in this case, dragging the centroid further up on the y-axis will tend to decrease the torque on the Malibu, thus causing less rotation in its final state.

Now that you are comfortable controlling the impulse model parameters, with your first collision object selected, increase the coefficient of restitution in order to increase the separation velocity between the two vehicles after the first impact. You’ll notice this gives greater distance for the Pinto to decelerate before the secondary impacts. Recall, you can determine the instant the impulses are exchanged between vehicles by going to the timing menu and noting the time when bounding boxes begin overlapping. The impulses will be exchanged at this time plus the depth of penetration, which by default is set to 0.03 seconds. However, in our case we set the depth of penetration to 0 seconds in our ees object so that the impact occurs as soon as the vehicles are in contact. By doing this, we can adjust the initial positions of the vehicles near the area impact, which is somewhat indicated by the tire marks on the diagram. Using this technique, one can quickly set up a collision between vehicles, where the simulation starts at the moment of impact.

Continue fine-tuning the initial positions, impulse centroid, pre-impact speeds, and restitution values until you are satisfied with the final position and orientations of the vehicles. In this case, we were able to converge on a reasonable solution after a few more minutes, where v[Pinto] = 19.57 mph, v[Malibu] = 17.9 mph, $$\varepsilon =$$ 0.18, and the impulse was located at $$(x,y,z) =$$ (-2.584 ft, 2.683 ft, 1.411 ft).  The initial positions were set to $$(x,y) =$$ = (-9.257 ft, -0.119 ft) for the Malibu and $$(x,y) =$$ (-0.021 ft, 5.443 ft)  for the Pinto.

The final results can be seen in Table 8. Our final rest positions are simulated to better than 0.42 ft for the Pinto and 0.36 ft for the Malibu. The final heading angles are simulated to better than 0.26% for the Pinto and 0.12% for the Malibu. The pre-impact speeds estimates are accurate to better than 1 mph (1.4%) for the Pinto and 2 mph (9.8%) for Malibu. The $$\Delta v$$ value for the Pinto is simulated to within better than 2 mph (10.8%) and for the Malibu to within 1 mph (1.8%). The PDOF is simulated to within 2.2% for the Pinto and 0.94% for the Malibu. We note that better accuracy ranges, both in position and in pre-impact velocities, were achieved above compared to those shown by similar software models (see SAE 960885). In our study, we were quickly and easily able to optimize our simulation by hand because of the Virtual CRASH system’s ability to visualize the simulation trajectories in real time as inputs are updated by the user. In this exercise, our optimization was based on matching the final rest positions and orientations.

Table 8

References

(1) “Examples of Staged Collisions in Accident Reconstruction,” R. Smith and J. Noga, NHTSA, US DOT.

(2) “Research Input for Computer Simulation of Automobile Collisions, Volume IV. Staged Collision Reconstructions,” NHTSA, US DOT, DOT HS 805 040.

(3) Research Input for Computer Simulation of Automobile Collisions, Volume II. Staged Collision Reconstructions,” NHTSA, US DOT, DOT HS 805 040.

(4) Chapter 9 | Scaling Images

(5) “Current Vehicle & Object Library,” http://www.vcrashusa.com/s/VirtualCRASH_ObjectLibrary.pdf

(6) Chapter 6 | Modifying Vehicle Properties

(7) Chapter 11 | Touching Up the Polygon Mesh

(8) Chapter 10 | Reading Collision Data

(9) Fundamentals of Statistics for Traffic Crash Reconstruction, Andrew Rich and Michelle Fish-Rich, IPTM Press, ISBN: 978-1-934807-13-2. https://store.iptm.org/products/fundamentals-of-statistics-for-traffic-crash-reconstruction.