基于声发射和长短时记忆神经网络的端对端燃气高压调压器故障诊断

在燃气输配系统中燃气调压器处于关键位置，及时有效地诊断其故障类型具有重要意义。文章提出了一种基于声发射信号和长短时记忆神经网络的端到端故障诊断方法，直接利用时域声发射信号对调压器的运行状态进行诊断。首先，设计一个二阶巴特沃斯高通滤波器对采集的声发射信号进行预处理；其次，利用长短时记忆网络（LongShort Term Memory networks，LSTMs）的记忆特性，构建长短时记忆网络端到端（e2e-LSTM）故障诊断模型。实验结果表明，该模型根据输入的高压调压站采集的声发射信号，能够以端到端模式一次性地诊断五种高压调压器故障。

Acoustic emission and LSTM based end-to-end fault diagnosis of gas high-pressure regulator

In the gas transmission and distribution system, the gas pressure regulator is in a key position for the timely and effective diagnosis of its fault types. In this paper, a long short term memory network (LSTM) based end-to-end fault diagnosis method is proposed, in which the acoustic emission signals is used to diagnose the operation condition of the gas pressure regulator. First, a second-order Butterworth high-pass filter is designed to preprocess the acquired acoustic emission signals. Then, by using the memory characteristics of long short term memory networks, an e2e-LSTM fault diagnosis model is established. The experimental results show that, with the input acoustic emission signals collected from the high-pressure regulator station, the model can diagnose five kinds of high-pressure regulator faults at a time in the end-to-end mode.