D-SSAE与PSO-SVM在转子-滑动轴承系统故障诊断中的应用

Application of D-SSAE and PSO-SVM in Fault Diagnosis of Rotor-plain Bearing System

旋转机械广泛应用于石油化工、航空航天等行业,而转子-滑动轴承系统是旋转机械的核心部件,对其进行故障诊断和状态监测具有重要意义。针对转子-滑动轴承的故障诊断问题,提出一种改进的深度堆叠稀疏自编码器(D-SSAE)和粒子群算法(PSO)优化后的支持向量机(SVM)相结合的故障诊断算法。首先,对振动信号进行快速傅里叶变换(FFT),以稀疏自编码器(SAE)组合构造堆叠稀疏自编码器(SSAE),以提出的函数动态改进dropout方法和自适应时刻估计法(Adam)进行迭代训练网络,实现特征的自动提取;其次,将频域信号输入诊断网络进行测试,以PSO-SVM为分类器得到诊断结果;最后,利用滑动轴承故障模拟实验台对算法进行了实验验证,结果表明,提出的方法具有故障识别率高、泛化能力强、训练时间合理等优点。

Rotating machinery is widely used in petrochemical,aerospace and other industries,and rotor-plain bearing system is the core part of rotating machinery,so it is of great significance to fault diagnosis and condition monitoring.Aiming at the fault diagnosis problem of rotor-plain bearings,in this paper,a fault diagnosis algorithm which combines the improved deep-stacked sparse autoencoder(D-SSAE)and support vector machine(SVM)optimized by particle swarm optimization(PSO)was proposed.First,the vibration signals were transformed by fast Fourier transform(FFT),and the stacked sparse autoencoder(SSAE)was constructed by combining the sparse autoencoder(SAE).The proposed function dynamically improved dropout method and adaptive moment estimation(Adam)were used to perform an iterative training network to achieve automatic feature extraction.Secondly,the frequency domain signals were input into the diagnostic network for testing,and the diagnosis results were obtained by using PSO-SVM as the classifier.Finally,the sliding bearing fault simulation test platform was used to verify the algorithm.The results show that the proposed method has the advantages of high fault recognition rate,strong generalization ability and reasonable training time.