【作者】 石逸林； 李金策； 李宏光；

【Author】 Yilin Shi;Jince Li;Hongguang Li;College of Information Science & Technology,Beijing University of Chemical Technology;

【机构】 北京化工大学信息科学与技术学院；

【摘要】 控制回路的阀门粘滞故障诊断是过程工业安全生产的重要问题。本文提出了一种基于深度学习的故障诊断模型,将其用于解决过程控制回路不同类型阀门粘滞故障的分类问题。该方法框架包括LSTM(Long Short-Term Memory)的稀疏编码-解码模式以及Softmax分类器,能够处理复杂系统中高维非线性且具有时间依赖性的序列数据。借助于LSTM-SAE网络对故障时间序列的数据特征进行有效提取,Softmax分类器能够实现对故障类型的精确识别。本文将所提出方法与其它深度学习模型在工业阀门粘滞故障特性仿真模型平台上进行了有效对比验证,结果表明LSTM-SAE故障诊断模型具有较好的快速性和准确性,并且具有较强的实际应用性。更多还原

【Abstract】 The valve stiction fault diagnosis of the control loop is an important problem in process industry safety production.In this paper,a fault diagnosis model based on deep learning is proposed to solve the problem of classification of different types of valve stiction faults in the process control loop.The framework of this method includes the sparse encoding-decoding mode of long-short-term memory neural network(LSTM) and the Softmax classifier,which can process high-dimensional nonlinear and time-dependent sequence data in complex systems.With the help of LSTM-SAE network to effectively extract the data characteristics of the fault time series,the Softmax classifier can realize the accurate identification of the fault type.In this paper,the proposed method is compared with other deep learning models on the simulation model platform of industrial valve stiction fault characteristics.The results show that the LSTM-SAE fault diagnosis model has better rapidity and accuracy,and has strong practical applicability.更多还原