PCA和GA-PSO-RBF集成的发电机组远程故障诊断

首先基于LPC2290核心芯片的arm嵌入式工控平台设计了远程数据采集系统,使故障诊断系统通过Internet在线获取发电机组状态数据。然后给出了主元分析(PCA)和GA-PSO-RBF神经网络集成的故障诊断方法。故障模式向量先通过PCA降维,降低RBF神经网络的规模和计算时间。针对RBF神经网络参数难以设置、收敛速度慢等不足,介绍了一种具有遗传算法中的选择、交叉、变异操作的遗传-粒子群算法(GA-PSO),用于RBF神经网络的参数优化过程。最后以转子振动试验台仿真发电机组,实现了状态信息的远程获取;通过故障诊断仿真测试验证了PCA和GA-PSO-RBF集成诊断方法的有效性。

Remote integrated fault diagnosis for generator unit using PCA and GA-PSO-RBF

Firstly, the remote data acquisition system based on arm embedded industrial platform with LPC2290 core chip is designed, so that fault diagnosis system access to state data of the generator unit online via the Internet. Then inte- grated fault diagnosis method using principal component analysis （PCA） and GA-PSO-RBF neural network is given. Fault pattern vectors are dimension-reduced using PCA, in order to reduce the scale and computing time of RBF neural network. Since the parameters of RBF neural network is difficult to set, and it has slow convergence rate, GA（genetic algo- rithm）-PSO（particle swarm optimization） algorithm, which has operations of selection, crossover and mutation of GA is in- troduced, and employed to optimize the parameters of RBF neural network. Finally, using the rotor vibration test-bed to simulate the generator unit, remote access to the status information is succeeded. The simulation tests also show that the inte- grated fault diagnosis method using PCA and GA-PSO-RBF is an efficient diagnosis approach.