基于自编码器和支持向量机的气压传动系统气缸泄漏故障诊断

气压传动系统在制造领域应用广泛，对智能化故障诊断与节能有较大需求。泄漏是气动系统最常见的故障类型及能量浪费的最主要因素之一。以最具代表性的执行元件气缸为研究对象，通过对其上游压力与流量信号进行处理分析，实现对下游气缸常见的内外泄漏故障的有效诊断。信号特征提取通过栈式自编码器完成，提取的特征进行聚类处理评估后送入支持向量机(Support Vector Machine, SVM)分类器进行分类，从而对气缸泄漏故障进行分类和定位。结果表明：通过分析上游信号来确定下游元器件故障状态是可行的；且对于泄漏故障实验，在同等条件下，基于流量信号的平均分类准确率可达到96%,基于压力信号的平均分类准确率为87%。

Fault Diagnosis of Cylinder Leakage in Pneumatic System Based on Auto-encoder and Support Vector Machine

Pneumatic system is widely used in manufacturing industries, which has great demand for intelligent fault diagnosis and energy saving. Leakage is one of the most common failure modes and the most important factors of energy waste in pneumatic system. In this study, the most representative actuator pneumatic cylinder is taken as the research object. The effective diagnoses of the internal and external leakage faults of the cylinder are achieved by processing and analyzing the upstream pressure and flow rate signals. The signal feature extraction is completed by AE (Auto-encoder). The extracted features are evaluated by clustering and then sent to SVM (Support Vector Machine) classifier for classification, so as to classify and locate cylinder leakage faults. The experimental results show that it is feasible to determine the fault mode of downstream components by analyzing upstream signals in pneumatic systems. For leakage faults, under the same conditions investigated in this study, the average classification accuracies based on flow rate signal and pressure signal can reach 96% and 87%, respectively.