一种结合云模型和QPSO优化的RBFNN及其应用

A RBF NEURAL NETWORK COMBINING CLOUD MODEL AND QPSO AND ITS APPLICATION

提出一种基于云变异操作的量子行为粒子群优化算法(QPSO-CM)的径向基函数神经网络(RBFNN)学习方法。首先QPSO算法利用云模型加入云变异操作,增加算法多样性;然后利用减聚类算法确定RBF神经网络径向基层的单元数;最后用QPSO-CM算法对RBF神经网络的参数(中心与宽度)和连接权重进行优化。将此算法用于齿轮的故障诊断,仿真诊断结果表明此方法是有效的,具有较好的分类效果,诊断精度高、收敛速度快。

In this paper we introduce a learning method of radial basis function （RBF） neural network which is based on quantum-behaved particle swarm optimisation with cloud mutation operation （QPSO-CM）. First, the QPSO adds the cloud mutation operation by using cloud model to increase its diversity; then the method uses the subtractive clustering method to determine the unit number of radial basis layer in RBF neural network; finally, it optimises the parameters （central position and directional width） of RBF neural network and the connection weight by QPSO-CM. Applying the method to gear faults diagnosis, the simulation results show that this method is effective with high diagnosis accuracy and fast convergence.