How Crash Reconstruction Works
A reconstruction turns marks, damage and stored data into speeds, positions and timing. Every method rests on assumptions, and the best work says plainly what each number depends on.
A crash lasts a fraction of a second, but it leaves records: marks on the pavement, bent metal, where the vehicles came to rest and, increasingly, data stored in the vehicles. Crash reconstruction reads those records back into a sequence of events. How fast was each vehicle going, where did they meet, and who could have done what, and when?
The answers are estimates. A good reconstruction reports ranges, names its assumptions and tests one method against another.
The physics
Conservation of momentum
In a collision, the vehicles push on each other with equal and opposite force for the same brief moment, so their combined momentum (mass times velocity, with direction) is about the same just after impact as just before. Knowing each vehicle’s mass, its direction of travel and its speed and direction just after separation, a reconstructionist can solve for the speeds before impact.
Post-impact speeds usually come from how far each vehicle traveled to rest and the friction it met on the way. Directions come from the scene diagram. Small errors in angles can produce large errors in the result, especially in shallow-angle collisions.
Crush energy
Damaged metal absorbed energy. Reconstructionists measure crush depth at several points across the damage and combine those measurements with stiffness values for the vehicle type, derived from crash tests, to estimate the energy absorbed and the vehicle’s change in velocity. Crush analysis is often run as a cross-check on momentum.
Skid to stop
When a vehicle slides to a stop, its kinetic energy is spent as friction over the slide. Setting the two equal and converting to miles per hour and feet gives the familiar formula:
S = √(30 × d × f)
- S is the speed at the start of the slide, in miles per hour.
- d is the slide distance, in feet.
- f is the drag factor: the deceleration the tires and surface produced, as a fraction of gravity. It is measured at the scene with test skids or a drag sled, or taken from published ranges, and adjusted for grade.
- 30 is a rounded constant from the unit conversion (2 × 32.2 ft/s², divided by the square of 1.467 ft/s per mph, is about 29.9).
A car that left 100 feet of marks on a surface tested at 0.70 gives √(30 × 100 × 0.70), or √2,100: about 46 mph.
That figure is a minimum. It counts only energy spent along the measured marks, not braking before the marks began or any speed left over when the vehicle struck something. Antilock brakes keep the wheels turning and may leave faint marks or none. Heavy trucks need their own analysis: air brakes take a moment to apply and axles do not all brake alike, so a drag factor measured with a car does not transfer to a loaded tractor-trailer.
The data and the measurements
Event data recorders
Federal rules do not require passenger vehicles to have an event data recorder. Under 49 CFR Part 563, recorders that are installed in passenger cars and in light trucks, vans and buses rated at 8,500 pounds or less, built on or after September 1, 2012, must meet uniform standards1.
Part 563 defines an event as a crash that produces a change in velocity of at least 8 km/h (about 5 mph) within 150 milliseconds, or that deploys a non-reversible restraint such as an air bag1. The recorder must keep the change in velocity during the crash; vehicle speed, accelerator position and service brake status in the seconds before; and driver seat belt status, air bag warning lamp status, deployment times and ignition cycle counts1. The pre-crash window is now 5 seconds at 2 samples per second. A 2024 amendment extends it to 20 seconds at 10 samples per second2, phasing in, as revised in 2026, from vehicles built on or after September 1, 2028 to all covered vehicles built on or after September 1, 2031, with later dates for small-volume and multi-stage manufacturers31.
Air bag deployment records must be locked against overwriting; other event records may be replaced by later events. Manufacturers must make a retrieval tool commercially available1.
Heavy trucks fall outside Part 563. Their engine control modules often keep event and last stop records under rules each engine maker sets. See What Data Is Stored in a Commercial Truck’s Electronic Control Module?
Mapping the scene
Calculations are only as good as the measurements behind them.
- Total stations. A surveying instrument measures angle and distance to a prism held on each point of interest: the ends of a tire mark, a gouge, a rest position. It is precise, but it records only the points someone chose.
- Laser scanning. A 3D scanner records millions of points, capturing surfaces no one thought to measure at the time. It sees only what is in its line of sight, so a scene usually needs several setups.
- Photogrammetry. Measurements are extracted from photographs, including drone imagery. It can recover positions from police or bystander photos taken before the scene was cleared, with accuracy that depends on image quality and known reference distances in the frame.
Simulation
Simulation software models vehicle motion from mass, tire properties, steering, braking and initial speeds, and shows whether a proposed sequence produces the evidence that was found. It tests a hypothesis rather than supplying one, and more than one set of inputs can often match the same marks and rest positions.
Where each method runs out
Momentum weakens when one vehicle far outweighs the other. A loaded tractor-trailer’s velocity barely changes when it strikes a car, so small errors in the car’s measurements swamp any estimate of the truck’s speed.
Crush analysis depends on stiffness data that exists mainly for passenger vehicles in tested configurations. Underride crashes, side impacts and truck structures are poorly covered.
Skid formulas give a floor, not a figure, and lean on a drag factor that can vary along the path.
Vehicle data reports what the vehicle’s own designated speed source indicated1. Tests on heavy trucks found recorded speed runs low during hard braking, when the wheels slip4. Clocks drift, and records can be overwritten.
Scene mapping cannot recover marks swept away before anyone measured or photographed them.
Simulation inherits every error in its inputs.
The strongest reconstructions use several methods and explain where they agree and where they do not.
