基于贝叶斯方法的设备级电磁脉冲效应评估

在电磁脉冲效应试验中,由于所获得的效应数据量一般较少,利用经典的数理统计方法对效应阈值场强进行概率拟合非常困难。基于贝叶斯数理统计方法,对油气管道数据采集与监视控制系统的中心控制系统客户端的电磁脉冲效应数据拟合分析。选择了Weibull分布模型和正态分布模型作为假设模型,并依次求得这两种模型参数的先验分布、似然函数和后验分布。通过拟合优度检验,最后基于贝叶斯信息准则选择形状参数、尺度参数分别为8.87,21.11的Weibull模型作为更合理的阈值场强概率分布模型。

Effect assessment of electromagnetic pulse at equipment level based on bayesian method

Normally, very limited data can be collected during the electromagnetic effect experiments. So it is hard to fit effect threshold curve based on classical mathematical statistics. This paper employs Bayesian mathematical statistics to fit the data of pipeline Supervisory Control and Data Acquisition systems Master Control Set client effect experiment. We choose Weibull model and normal model as the hypothesis models, and calculate the model parameters prior distribution, likelihood function and posteriori distribution orderly. After tests of goodness of fit, we find that the Weibull model with shape parameter equals 8.87 and scale parameter equals 21.11 is more suitable for effects evaluation based on BIC calculated result, providing a more reliable threshold probability distribution for effect assessment of electromagnetic pulse at equipment level.