柱塞泵滑靴磨损信号随机森林算法故障诊断  

Fault Diagnosis of Slipper Wear Signal of Piston Pump Based on Random Forest Algorithm

作　　者：张月平[1] 田伟华 刘艳红[2] ZHANG Yue-ping;TIAN Wei-hua;LIU Yan-hong(School of Automotive and Mechatronic Engineering,Zhoukou Vocational and Technical College,He’nan Zhoukou 466000,China;School of Electrical Engineering,Zhengzhou University,He’nan Zhengzhou 450001,China)

机构地区：[1]周口职业技术学院汽车与机电工程学院,河南周口466000 [2]郑州大学电气工程学院,河南郑州450001

出　　处：《机械设计与制造》2024年第6期150-153,共4页Machinery Design & Manufacture

基　　金：河南省高等学校重点科研项目(21B460018)。

摘　　要：为了提高低负载下柱塞泵滑靴磨损故障状态诊断精度,提出基于随机森林算法的柱塞泵滑靴磨损故障状态识别方法。重点分析了低负载条件下各类柱塞泵滑靴故障信号的频域特征值,构建了特征数据库。验证了上述方法的适应性,并测试了柱塞泵不同程度松靴故障的诊断情况。研究结果表明:滑靴磨损后频域表现出明显的波动性,袋外错误率和决策树数量呈现反比变化规律,基本都在0.05附近,将决策树最优棵数n设定在400。以随机森林算法诊断特征数据库时,在250组样本中只发生了1组误识别情况,达到了98.75%的识别准确度。随机森林方法训练时间、训练准确度与测试准确度都比其它各算法更优。松靴故障诊断结果获得了高于99.5%的总体诊断准确度。采用随机森林方法柱塞泵磨损故障状态诊断表现出优异适应性,能够对柱塞泵各故障状态进行准确诊断。In order to improve the accuracy of slipper wear fault diagnosis of piston pump under low load,a method for slipper wear fault identification of piston pump based on random forest algorithm was proposed.The frequency domain eigenvalues of the fault signals of the slipper of the piston pump under the condition of low load were analyzed emphically,and the characteristic da-tabase was constructed.The adaptability of above method was verified,and the fault diagnosis of the piston pump with different degrees of loose boots was tested.The results show that the frequency domain has an obvious fluctuation after slipper wear,out-of-Bagerror and the number of decision trees show an inversely proportional change rule,which are basically around 0.05,and the optimal number of decision trees n is set at 400.When the random forest algorithm is used to diagnose the feature database,only one group of misidentification occurs in 250 groups of samples,and the recognition accuracy reaches 98.75%.The training time,training accuracy and testing accuracy of random forest method are better than other algorithms.The overall diagnostic accuracy is higher than 99.5%.The random forest method shows excellent adaptability to the wear fault diagnosis of the piston pump and can accurately diagnose each fault state of the piston pump.

关 键 词：柱塞泵 故障诊断 随机森林算法 滑靴磨损 准确度 

分 类 号：TH16[机械工程—机械制造及自动化]