本文根据不同级别的故障诊断技术要求和液压系统特点,从压力脉动信号中提取故障信息,采用自回归模型参数作故障特征、用模式识别技术对系统各种状态进行分类,形成一种液压系统故障识别与诊断的方法。在泵源系统故障诊断试验研究中,采用三种分类方法,对不同传感器位置、不同模型阶次和样本容量时的确诊率、漏警率和虚警率作了比较。诊断方法原理正确,效果良好,具有很好的应用前景。 In this paper, fault information is extracted from the pressure pulsation signal according to different levels of fault diagnosis technical requirements and characteristics of the hydraulic system. Autoregressive model parameters are used as fault features, and the system status is classified by pattern recognition technology to form a hydraulic system fault Methods of identification and diagnosis. In the study of pump system fault diagnosis, three classification methods were used to compare the diagnostic accuracy, missing alarm rate and false alarm rate at different sensor positions, different model order and sample size. Diagnosis principle is correct, the effect is good, with good application prospects.