摘要
针对单一径向基函数(RBF)神经网络在反应釜故障诊断中泛化能力不足的缺点,设计了基于粒子群(PSO)算法优化的RBF神经网络。利用PSO算法操作简单、容易实现等特点及其智能背景,对RBF神经网络的参数、连接权重进行优化,并用经PSO算法优化的RBF神经网络对反应釜故障进行仿真诊断。仿真诊断结果表明,PSO算法优化的RBF神经网络具有较好的分类效果,较RBF诊断模型精度高、收敛快,具有推广应用价值。
A new PSO algorithm with dynamically changing inertia weight and study factors based on improved adaptive PSO was proposed,where the inertia weight of the particle was adjusted adaptively based on fitness of the particle.The diversity of inertia weight made a compromise between the global convergence and local convergence speed,so it can alleviate the problem of premature convergence effectively.The algorithm was applied to train RBF neural network and a model of fault diagnosis for CSTR was established,compared with PSO algorithm,the proposed algorithm can improve the training efficiency of neural network effectively and obtain good diagnosis results.
出处
《中国机械工程》
EI
CAS
CSCD
北大核心
2012年第18期2204-2207,共4页
China Mechanical Engineering
基金
浙江省自然科学基金资助项目(Y1110686)
关键词
RBF神经网络
粒子群优化算法
故障诊断
连续搅拌反应釜
RBF neural networks
particle swarm optimization（PSO）
fault diagnosis
continuous stirred tank reactor（CSTR）
