基于声发射特征参数的阀门内漏神经网络识别

为解决阀门内漏在线快速检测问题,基于声发射检测技术,采集了DN150闸阀在上下游压差0.8 MPa典型工况下的声发射信号,提取振幅、振铃计数、能量、平均信号电平、信号均方根值5个特征参数作为输入矢量,内漏结果作为输出矢量,建立了3层BP神经网络识别阀门内漏的模型,利用实验数据样本对该模型进行了验证。结果表明,阀门内漏识别准确率超过90%,有利于提高阀门安全运行在线监测水平。

Valve Internal Leakage Neural Network Identify Based on Acoustic Emission Parameters

In order to solve the rapid detection of valve internal leakage, using acoustic emission detect technology, it is collected the acoustic emission signals of DN150 gate valve while the difference pressure between upstream and downstream is 0.8 MPa and extracted five parameters of amplitude, ringing count, energy, average signal level and signal root mean square as the input vector while internal leakage results as an output vector. Finally, a 3 level BP neural network model is established to identify valve leakage by experimental data to verify the model. The results show that the accuracy of valve internal leakage identification is more than 90%, which is beneficial to improve the level of online monitoring of valve safe operation.