基于贝叶斯网络分类器的产品故障率分类研究

Research of product failure rate based on Bayesian network classifiers

针对装备在不同配置及使用环境的条件下运行的故障率等级差异,详细介绍并分析了现有各贝叶斯分类器的特点和构建算法。在此基础上,提出了基于贝叶斯网络的产品故障分类模型建模方法用于指导实际分类任务的模型建立和应用。通过法国某装备生产企业的实例分析,实验结果证明在所有的贝叶斯网络分类器及传统的决策树C4.5分类器中,树型朴素贝叶斯分类器能够取得最好的分类效果,并为后续的维修资源配置及产品运行能力优化提供有效的理论支持。

For identifying the product failure rate grade with diverse configuration and different operation condition, introduced the useful Bayesian networks classifiers. Also described their algorithms and characters in detail. On the basis of these classifier models, listed the procedure of building product failure rate grade classifier for guiding the modeling and application the actual cases. Carried out the France enterprise case study and the results show that, with the comparison to other Bayesian networks classifiers and traditional decision tree C4.5, the tree augmented natve-Bayes classifier get the best general performance with highest precision, which can build a firm keystone for later maintenance resource distribution and operation optimization.