自适应变异麻雀算法优化LSTM变速箱故障诊断

Gearbox Fault Diagnosis Based on LSTM Optimized by Adaptive Mutation Sparrow Search Algorithm

针对麻雀搜索算法(Sparrow Search Algorithm,SSA)优化深度长短时记忆网络(Long-short Term Memory,LSTM)模型参数时存在陷入局部最优、后期收敛精度不高的问题,对SSA算法进行改进,提出一种自适应变异麻雀搜索算法(Adaptive Mutation Sparrow Search Algorithm,AMSSA)。AMSSA在SSA基础上,引入发现者和跟随者数量自适应调整策略、发现者和跟随者柯西变异策略,提高算法的寻优能力。以AMSSA为LTSM模型参数优化方法,建立变速箱故障诊断模型,并进行实验验证。结果表明:相比于SSA,AMSSA优化LSTM的诊断精度提升4%;相比于其他3种类型优化算法,在诊断精度提升的同时耗时更短。

In order to solve the problem that the sparrow search algorithm(SSA)fell into the local optimum and the later convergence accuracy was low when optimizing long-short term memory(LSTM)model parameters,an adaptive mutation sparrow search algorithm(AMSSA)was proposed.On the basis of SSA,AMSSA introduced the adaptive adjustment strategy of discoverer and follower numbers and Cauchy mutation strategy of discoverer and follower to improve the optimization ability of the algorithm.Taking AMSSA as the model parameter optimization method of LSTM,a gearbox fault diagnosis model was established and verified by experiments.The results show that compared with SSA,the diagnostic accuracy of LSTM optimized by AMSSA is improved by 4%.Compared with the other three types of optimization algorithms,the diagnosis accuracy is improved with less computer-time consuming.