From available field data describing the site of the collision, the vehicles involved, and the collision damage, CRASHEX can be used to iteratively develop a time-reversed reconstruction, refined to eliminate inadvertent error and for internal consistency until a "best-case" reconstruction is established. In the alternative, using the reconstructionist’s method of choice a long and careful prior analysis may be performed off-line to develop a preferred solution.

In either case, CRASHEX can then apply the "finishing touch" of Finite Difference Analysis, efficiently determining the reliability, given the uncertainties of input determination, of that best or central case. To most effectively serve this purpose, rather than to require of the reconstructionist the purchase of a long-term license for only occasional use, the use of CRASHEX, once downloaded, is offered on a pay-per-day basis.

The analytical approach is described in A. G. Fonda’s Society of Automobile Engineers paper 2007-01-0713, "The Practical Application of Finite Difference Analysis in Accident Reconstruction." As demonstrated there, upon provision of the array of inputs to a best case previously developed by any means of reconstruction or simulation, variations from that best case can be treated by the method of Finite Differences. Because the results depend on the transfer functions of the particular physical event, independent of the method of analysis, the sensitivities of the outputs to variations in the inputs, the width of the band of probability, are transferable from one means of reconstruction, CRASHEX in this instance, to another means of reconstruction or simulation.

ASSIGN LIKELY VARIATIONS OF MEASUREMENT

Variations statistically representative of measurement error potential may be imposed on every input to the best case, giving an array of the effect of each of those variations on every output. Usually these variations will be twice the standard deviation for that type of measurement, judiciously assuming for each item Casual, Ordinary, or Precise means of measurement as found in the tables of 2007-01-0713 (or, as available, its predecessor papers 2003-01-0469 and 2004-01-0418).

COMPUTE LIKELY VARIATIONS OF APPROACH SPEED

With that accomplished, the root of the sum of the squares of all these deviations can be found for every output of the solution. This result will be the numerical reliability of each parameter of interest, for instance the speed at impact of the bullet car as in 2007-01-0713. These are the bounds not likely to be exceeded due to mis-measurement, given the available evidence. These bounds are usually dominated by just three or four major contributors, all other contributors and their uncertainties being trivial by comparison. The reconstructionist will thus be in a position to say, as to possible errors in each and every one of his or her assumptions, that its likely variations due to inadvertence, given the evidence as reported, have already been considered, while unlikely variations need not be considered.

Within just a few hours of analyst time, and mere seconds of computer running time, the bands of the probable variations from the central case which are likely to have occurred can be determined. CRASHEX is thus a very powerful method applicable as needed, whether to develop a central case given the evidence, or to evaluate the reliability of the results of a prior, non-CRASHEX study of an impact event. NEXT