机械系统贝叶斯网络参数估计

Parameter estimation of Bayesian networks of mechanical system

由于数据缺乏、单元分布不同等,机械系统贝叶斯网络模型的参数估计面临很多困难。为此,采用贝叶斯方法分别对根节点服从多项分布、指数分布、正态分布、Weibull分布的概率估计和多项分布情况下的条件概率估计进行了研究。所提方法解决了建模过程中的小子样、多分布等难题,在机械系统贝叶斯网络建模中有很大优势。

Due to lack data and different distributions in components, there are some problems in pa-rameter estimation of the mechanical systems Baysian network model. This paper uses Bayes method to assess the probabilities of polynomial distributions, exponential distributions, normal distribu-tions, Weibull distribution of root nodes and conditional probabilities of polynomial distributions. The proposed method addresses the small sample and multi-distribution problems and has good advantage in BN construction of mechanical systems.