堆栈自编码器下机电故障信号多尺度滤波方法研究  

Research on Multi-Scale Filtering Method of Electromechanical Fault Signals under Stack Self-Encoder

作　　者：宁永安 NING Yongan(Guoneng Shendong Coal Group Co.,Ltd.,Yulin 719315,China)

机构地区：[1]国能神东煤炭集团有限责任公司,陕西榆林719315

出　　处：《自动化仪表》2024年第8期42-46,51,共6页Process Automation Instrumentation

摘　　要：煤矿机电故障信号组成成分较复杂,会导致煤矿机电故障诊断质量下降。对此,提出基于堆栈自编码器的煤矿机电故障信号多尺度滤波方法。首先,采用非线性时间序列以及相空间重构的方式构建煤矿机电故障函数并采集故障信号。其次,通过成本函数与稀疏约束相结合的方式设计堆栈自编码器,将采集到的故障信号输入堆栈自编码器,以实现故障信号分类。最后,以分类结果为基础,通过集合经验模态分解(EEMD)将故障非平稳信号转变为平稳信号,并采用多尺度滤波对故障信号展开寻优处理,以获取煤矿机电故障的多尺度调解信号,从而实现煤矿机电故障信号多尺度滤波。经试验验证,所提方法对故障信号进行多尺度滤波处理后,大多数频率成分得到了保留,同时滤除了不需要的频率成分。煤矿机电故障诊断准确率平均值为99.71%、诊断时间平均值为0.24 s。该方法能够实现准确、快速的煤矿机电故障诊断。Coal mine electromechanical fault signals are composed more complex,which leads the reduction of the quality of coal mine electromechanical fault diagnosis.In this regard,a multi-scale filtering method of coal mine electromechanical fault signals based on stack self-encoder is proposed.Firstly,the nonlinear time series and phase space reconstruction are used to construct the coal mine electromechanical fault function and collect the fault signal.Secondly,the stack self-encoder is designed by combining the cost function and sparse constraints,and the collected fault signals are input into the stack self-encoder to realize the classification of the fault signals.Finally,based on the classification results,the ensemble empirical modal decomposition(EEMD)is used to transform the fault non-smooth signals into smooth signals,and the multi-scale filtering is used to optimize the fault signals,in order to obtain the multi-scale mediation signals of the coal mine electromechanical failures,so as to realize the multi-scale filtering of coal mine electromechanical fault signals.After experimental verification,most frequency components are retained and unwanted frequency components are filtered out at the same time after multi-scale filtering of the fault signal by the proposed method.The average value of coal mine electromechanical fault diagnosis accuracy is 99.71%,and the average value of diagnosis time is 0.24 s.This method can realize accurate and fast fault diagnosis of coal mine electromechanical.

关 键 词：堆栈自编码器 机电故障 故障信号 多尺度滤波 信号去噪 集合经验模态分解 

分 类 号：TH-69[机械工程]