一种改进的CPSO-BP神经网络故障诊断技术研究  

An Improved Technology on CPSO-BP Neural Network Fault Diagnosis

作　　者：王慧[1] 董增寿[1] 张春梅 

机构地区：[1]太原科技大学 [2]新疆69028部队

出　　处：《太原科技大学学报》2013年第5期342-347,共6页Journal of Taiyuan University of Science and Technology

基　　金：山西省自然科学基金(2012011015-4)

摘　　要：针对BP神经网络对液压泵进行故障诊断时存在收敛速度慢、诊断精度不高的问题,文中提出了一种CPSO-BP诊断网络。该网络用混沌运动对PSO算法进行改进,构成CPSO算法,克服了PSO算法早熟的缺点。然后采用CPSO算法对BP网络的权值及阈值进行优化,用该网络对柱塞泵的常见故障进行诊断,并与PSO-BP网络的诊断结果进行比较。实验表明该网络的故障诊断能力及诊断精度都得到了有效的提高。A CPSO-BP neural network is proposed due to the slow convergence speed and low diagnostic accuracy of the BP neural network for fault diagnosis of hydraulic pump. The CPSO-BP neural network with chaotic motion to improve PSO algorithm can overcome the shortcomings of precocious PSO algorithm. CPSO algorithm was used to optimize the weights and thresholds of BP network for fault diagnosis of hydraulic pump, then was compared with the PSO-BP network diagnostic results. Experiments show that the fault diagnosis capacity and diagnostic accuracy of the network have been effectively improved.

关 键 词：液压泵 故障诊断 混沌 粒子群算法 BP神经网络 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]