基于贝叶斯推理的乘员约束系统参数识别

为了克服测量响应的不确定性给乘员约束系统参数识别带来的困难,利用马尔科夫链蒙特卡洛(Markov Chain Monte Carlo,MCMC)采样和近似模型构造技术,提出一种基于贝叶斯推理的乘员约束系统不确定性参数识别方法.该方法结合约束系统参数的先验分布和测量响应,通过马尔科夫链在未知参数联合概率密度空间进行抽样,从而获得了织带刚度缩放系数和质量流率缩放系数的后验边缘概率密度函数.识别结果表明,相比于传统确定性识别方法,基于贝叶斯推理的不确定性参数识别方法不仅能有效给出乘员约束系统参数的概率分布,而且能够保证参数寻优的全局收敛性.

Parameters Identification for Occupant Restraint System Based on Bayesian Inference

In order to overcome the difficulty for the parameters identification of occupant restraint system caused by the measured uncertainty, this paper proposed an uncertain identification method for the parameters of occupant restraint system based on Bayesian inference, which combined Markov Chain Monte Carlo (MCMC) sample and surrogate model. This method firstly obtains the prior distributions of identified parameters and measured responses, and then the MCMC sampling is applied to the joint probability density of unknown parameter. Then, the marginal posterior probability distributions of scale factor of webbing and rate of flow can be calculated. Compared with the traditional method of determined identification, the identified results show that the Bayes inference method for uncertain parameter identification not only obtains the probability distributions effectively, but also ensures the global convergence of identified parameter.