中文摘要：

为克服事故再现结果不确定性分析中试验样本少、模型非线性等问题，基于支持向量回归（SVR）与子区间（SI）技术，提出一种求解事故再现结果取值区间的支持向量回归一子区间（SVR-SI）方法。先用拉丁超立方抽样（LHS）设计仿真试验，获得结果后用SVR方法建立事故再现模型的近似响应面模型；然后用sI算法获得事故再现结果的取值区间。最后给出2个数值算例及1个真实事故案例。计算结果表明，用SVR—SI方法在较少试验样本条件下，仅需较少仿真次数，即可获得事故再现结果的取值区间。

英文摘要：

The small test sample and nonlinear models are two problems in analyzing uncertainty of the accident reconstruction result. In order to overcome such problems, a method for calculating the interval of an accident reconstruction result was worked out based on SVR and techniques of SI. In SVR-SI, the LHS was used for generating sample set firstly, and then SVR was used for obtaining the response surface model, after that, a proposed SI algorithm was used for calculating the interval of the accident reconstruction result with the obtained response surface model. Finally, two numerical cases and a real accident case were given. Results show that ideal results can be obtained by SVR-SI method with a smaller test sample, and fewer simulation times.