柴油机供油系统状态的朴素贝叶斯诊断方法研究

Research of Fault Diagnose Technology for Fuel Injection System of Diesel Engine Based on Nave Bayesian Classifier Method

针对朴素贝叶斯方法的缺点,提出了基于主分量分析方法的柴油机供油系统朴素贝叶斯诊断技术;该方法利用历史诊断记录,通过主分量分析方法对训练样本的输入维数进行约简,将高维相关的特征信号转换为低维相互独立的特征信号,并在此基础上进行贝叶斯诊断分析,从而改善了贝叶斯方法中要求的属性信息之间的独立性限制,实验结果表明,基于主成分分析方法的贝叶斯故障诊断技术对于简化诊断模型,减少算法执行时间,提高诊断速度具有重要作用。

Since the restriction of attribute dependency in nave Bayesian method, PCA based Bayesian diagnose method for fuel injection system of diesel engine is proposed. With this method, attributes of fault symptom is reduced based on PCA using history record of diagnose. Because of transforming multi--dimensional correlated characteristics signal into low dimensional independent characteristics signal, The limitation on the independence among characteristics signal is relaxed via PCA. Experiment result indicates that PCA based Bayesian diagnose model is effective to simple diagnose model, reduce time of arithmetic executing and enhances diagnose speed.