基于改进型Hopfield神经网络的潜污泵故障诊断方法

FAULT DIAGNOSIS METHOD OF SUBMERSIBLE SEWAGE PUMP BASED ON IMPROVED HOPFIELD NEURAL NETWORK

为了实现对潜污泵运行时的故障问题进行精准诊断,提出一种改进型Hopfield神经网络(HNN)故障诊断方法。利用BP神经网络进行编码操作,克服HNN神经网络的编码缺陷,并通过粒子群优化算法(PSO)对HNN神经网络连接权值进行优化,提高改进型神经网络的全局收敛能力,得到改进型HNN神经网络模型。基于现场实验,获得潜污泵故障运行时的振动信号特征向量,将特征向量作为样本数据对改进型神经网络进行训练,并对潜污泵的故障类型进行诊断。研究结果表明:改进型HNN神经网络全局收敛能力较好,对潜污泵典型故障的诊断准确率达到90%以上,可以实现对潜污泵运行时的故障进行精确诊断。

In order to accurately diagnose the fault of submersible sewage pump,an improved Hopfield neural network(HNN)fault diagnosis method was proposed.BP neural network was used for coding operation to overcome the coding defects of HNN neural network.The connection weights of HNN neural network were optimized by particle swarm optimization(PSO)algorithm to improve the global convergence ability of the improved neural network,and the improved HNN neural network model was obtained.Based on the field experiments,the vibration signal feature vector of the submersible sewage pump under fault operation was obtained.Then the feature vector was used as sample data to train the improved neural network,and the fault types of the submersible sewage pump were diagnosed.The results show that the improved HNN neural network has better global convergence ability,and the typical fault diagnostic accuracy of the submersible sewage pump is more than 90%,which can realize the accurate diagnosis of the fault during the operation of the submersible sewage pump.