摘要
基于粒子群优化算法(PSO)的神经网络具有良好的训练性能,输出的整体误差小于BP算法。用PSO作为一种粗优化或离线学习过程,用神经网络学习作为一种细优化或在线学习过程。这两种方法综合使用可以大大提高传动箱故障诊断性能。诊断系统把常用的7个时频动态特征参量作为BP网络的输入层的输入,把传动箱中常见的6种特征作为网络的输出。诊断系统拓扑结构为7126的3层BP网络,规定系统误差0·001。结果表明,粒子群优化算法对多故障征兆有较好的故障识别率,作为一种有效优化方法,在机械故障诊断领域具有良好的应用前景。
Neural network based on particle swarm optimization (PSO) algorithm has good training characteristics, its integrated output deviation is less than BP algorithm. Taking PSO as a study process of coarse optimization or off-line, neural network as a study process of fine optimization or on-line. Integrated two methods, characteristics of gearbox fault diagnosis can be improved greatly. The result showed that PSO algorithm has better fault diagnosis recognition resolution for multi-fault sign. As an effective optimization method, PSO algorithm has a good application prospect in the field of mechanical fault diagnosis.
出处
《火炮发射与控制学报》
北大核心
2006年第4期54-58,共5页
Journal of Gun Launch & Control
基金
国家自然科学基金项目资助(50575214)
关键词
人工智能
粒子群优化
群体智能
神经网络
故障诊断
artificial intelligence
particle swarm optimization
swarm intelligence
neural network
fault diagnosis