一种用于故障诊断的演化小波神经网络

Genetic Optimization of a Wavelet Neural Network Used in Fault Diagnosis

针对小波神经网络实现故障模式识别时存在的“维数灾”问题 ,提出了利用遗传算法在小波网络的学习过程中优化网络结构的方法 ,可有效减少小波基元 ,加速收敛。同时为提高遗传优化的收敛速度和精度 ,避免“早熟”现象 ,采用基于实数编码的遗传算法。给出了各个控制算子的自适应调整策略 ,并设计了增加和删除操作对遗传算法进行改进。仿真结果证明了该算法的有效性。

To solve the problem of 'dimension disaster' for the wavelet neural network in fault mode recognition, a genetic optimization algorithm that can search for the optimum wavelet parameter and network neurons adaptively in the learning process is proposed At the same time, a new genetic algorithm based on real number code with self adaptive strategy for the control operators and delete and addition operation is improved in order to overcome premature The simple structure and high convergence rate of the new algorithm are demonstrated by simulation results \;