| Traffic accident simulation is a very powerful method of supporting the conclusions of an accident reconstruction. However, two technical problems exist in the process of conducting a traffic accident simulation. The first problem is the excessive amount of time required to conduct the simulation. The second problem is that more than one set of input parameter combinations can produce a satisfactory result. Thus, to improve the quality of accident simulation, it is necessary to have a systematic and efficient approach to conduct the trial and error process to reproduce a simulated scenario that more accurately reflects the physical evidence.;A blocked coordinate descending (BCD) converging algorithm is proposed in this dissertation as a means of improving the convergence of traffic accident simulation results. This algorithm was modified from a hill-climbing algorithm, and divides a large converging task into several smaller converging tasks. BCD improved the converging efficiency while maintaining adequate accuracy.;The most significant contributing factors (i.e. speed, steering angle, braking or accelerating wheel forces, etc.) that affect the simulation results of cases where a single vehicle loses control and the factors (i.e. impact speed, impact configuration, wheel forces, post-impact steering angle, etc.) that affect the simulation of cases where two cars collide were analyzed. An automatic coordinate converging system (ACCS) was developed to conduct the convergence.;The impact configurations and final rest positions of ten staged collision cases was used to evaluate the performance of the ACCS system. The results show that the performance of converging simulation results with staged collisions are effectively and efficiently improved by using a BCD converging algorithm. |
