Analyzing the uncertainty of simulation results in accident reconstruction with Response Surface Methodology

Abstract This paper is focused on the uncertainty of simulation results in accident reconstruction. The Upper and Lower Bound Method (ULM) and the Finite Difference Method (FDM), which can be easily applied in this field, are introduced firstly; the Response Surface Methodology (RSM) is then introduced into this field as an alternative methodology. In RSM, a sample set is firstly generated via uniform design; secondly, experiments are conducted according to the sample set with the help of simulation methods; thirdly, a response surface model is determined through regression analysis; finally, the uncertainty of simulation results can be analyzed using a combination of the response surface model and existing uncertainty analysis methods. It is later discussed in detail how to generate a sample set, how to calculate the range of simulation results and how to analyze the parameter sensitivity in RSM. Finally, the feasibility of RSM is validated by five cases. Moreover, the applicability of RSM, ULM and FDM in analyzing the uncertainty of simulation results is studied; the phenomena that ULM and FDM can hardly work while RSM can is found in the latter two cases. After an analysis of these five cases and the number of simulation runs required for each method, both advantages and disadvantages of these uncertainty analysis methods are indicated.