基于支持向量机回归的材料参数反求方法

高强度钢的应变率效应对汽车件碰撞性能的影响研究是国际上的研究热点,不同类型的高强度钢将呈现出不同的应变率效应。因此,如何获取精确的材料参数是保证汽车碰撞计算机仿真结果可靠性的前提。如果直接通过标准拉伸试验获取相关材料参数,并没有考虑材料在碰撞过程中的特性,会引入较大的误差。为此,采用直接碰撞过程反求材料参数的方法,将参数反求的问题转换为测量值和仿真值最小二乘最小的优化问题。此外,由于参数反求中存在大量不确定性因素,为同时保证反求结果的稳健性和精度,采用基于最小二乘支持向量机回归技术的近似模型算法。近似模型技术保证了反求的效率、最小二乘支持向量机最大限度地保证了反求结果的精度和稳定性。通过对高强度钢的试验试验和反求结果的比对,验证了算法的性能。

Material Parameter Inverse Technique Based on Support Vector Regression

The study on strain rate effect of advanced high stiffness steel(AHSS) for vehicle crashworthiness became the hot spot in this research field recent year.Different strain rate should lead to different influence.The accurate material parameter is the key important issue for reliable simulation.The direct material parameter identification methods commonly don’t consider the crash effect.It might introduce the large errors.Therefore,a parameter inverse method by considering the crash effectis is proposed.Moreover,there are a lot of uncertain factors during inverse identification procedure.In order to enhance the efficiency and reliability of the proposed inverse method,the least square support vector regression(LSSVR)-based metamodeling is implemented for the inverse method.The LSSVR is a modeling algorithm based on structural risk minimal,therefore the reliability of the proposed inverse method can be promise.The metamodeling technique is used to improve the efficiency.According to the comparison between the data from experiments and inverse method,the suggested inverse method is proved to be a feasible technique for AHSS.