基于粒子群优化的可拓神经网络故障诊断方法研究  

Study on Extension Neural Network Optimized by PSO Algorithm for Fault Diagnosis

作　　者：文天柱[1] 许爱强[2] 吴明辉[1] 

机构地区：[1]海军航空工程学院研究生管理大队,山东烟台264001 [2]海军航空工程学院科研部,山东烟台264001

出　　处：《计算机测量与控制》2014年第2期364-366,369,共4页Computer Measurement &Control

基　　金：总装武器装备预研基金项目(9140A27020212JB14311)

摘　　要：基于可拓学的故障诊断方法是智能故障诊断领域中较为新颖的研究方向;首先介绍了可拓故障诊断领域的研究现状;然后重点分析了可拓神经网络模型,包括它的结构和故障诊断原理,由于该模型存在参数设置主观、易早熟等问题,进而提出了基于粒子群优化的可拓神经网络模型,该模型以关联度作为测度工具物理意义明确,通过粒子群算法进行参数优化避免算法早熟;最后采用汽轮发电机组振动信号频谱数据进行算法验证,结果表明该算法能够正确诊断出全部故障,且诊断精度高。extension fault diagnosis is a novel research direction in the field of intelligent fault diagnosis. First the research status of ex- tension fault diagnosis is introduced. Second the extension neural network model is analyzed including its structure and diagnostic principle. But as the problems of parameters setting subjective and algorithm precocious, the extension neural network model based on PSO algorithm is proposed, which take dependent degree as the tool of measurement and optimize the parameters by using PSO algorithm. At last, the algo- rithm is verified with frequency data of vibration signals which is come from turbine--generator set. The results show that using this algo- rithm the entire fault can be correctly detected and the precision is high.

关 键 词：故障诊断 可拓神经网络 粒子群 物元 关联函数 

分 类 号：TP277[自动化与计算机技术—检测技术与自动化装置]