贝叶斯网络数据细化的柴油发电机故障诊断

Data Refinement Approach to Diesel Engine Fault Diagnosis Based on Bayesian Networks

针对传统故障诊断方法在不确定问题诊断方面的不足,提出了基于贝叶斯网络的数据细化的柴油发电机故障诊断法。对柴油发电机转子的某些特定故障,结合专家知识确定转子特定状态下故障与振动频率、幅值及相关描述的依存关系,将获取的观测数据细化处理,利用结构学习,构建了基于贝叶斯网络的柴油发电机故障诊断模型,通过参数学习确定各节点的条件概率。实验结果表明,在已知信息具有模糊性和不完备性时,基于贝叶斯网络数据细化的故障诊断技术可明显提高诊断正确率。

Since the restriction of traditional method in uncertain problem fault diagnosis, the data refinement approach in diesel engine fault di- agnosis was proposed based on Bayesian networks. This approach used expert knowledge to determine the structure of Bayesian networks, and then it es- tablished the Bayesian networks model of vibration fault diagnosis with the refined data from data nodes which were relevant each other, at last, it got the conditional probability of data nodes through parameter learning. Experimental results indicate that the diagnosis system based on Bayesian net- works has preferable diagnosis efficiency and satis- factory accuracy.