Research On Improvement Of Ant Lion Optimizer And Its Application On Mechanical Fault Diagnosis

With the continuous improvement of national mechanization development,the relationship between various equipment becomes complex,and the consequences of equipment failure will be unpredictable.Therefore,the mechanical fault diagnosis technology has been highly valued by scholars all over the world.However,how to adopt accurate and efficient fault diagnosis methods for mechanical equipment has always been a difficulty with first consideration in the field of mechanical engineering.In view of the shortcomings of traditional feature selection and pattern recognition methods in the field of mechanical fault diagnosis,in this paper we applies ant lion optimizer(ALO)and its improved version to the mechanical fault diagnosis,completes the intelligent optimization selection of the mechanical fault feature set and parameter optimization of the extreme learning machine(ELM).The main work of this paper can be summarized as follows:(1)Firstly,the basic idea and mathematical model of ALO are introduced,and the effectiveness of ALO is tested by benchmark functions.Secondly,the simplified evolution trajectory model of ALO is transformed into the difference equation of discrete sequence and solved.The local convergence of ALO is proved successfully.Then,the theory of two sufficient and necessary conditions for the global convergence of stochastic optimization algorithm prove that ALO belongs to the local search optimization algorithm.Finally,one-way analysis of variance experiment is designed to analyze the parameter effectiveness of ALO,and the influence rule of the initial parameters on the solution performance of ALO is obtained,which can solve the blindness problem of ALO parameter setting effectively.(2)In this paper,an improved binary ant lion optimizer(IBALO)is proposed,the mechanism of population protection set is introduced,and it takes ants with optimized potential into protection set,group in the set iterates parallelly with the main group in order to enhance the global optimization capability.The above of improvements are integrated into the hybrid feature selection model,applied to UCI standard test databases and rolling bearing fault databases respectively.The experimental results show that the recognition accuracy and feature reduction ability of the hybrid feature selection model based on the IBALO are improved significantly.(3)In this paper,an improved continuous ant lion optimizer mixed with estimation of distribution algorithm(EDA)and variable step Levy disturbance strategy(ELALO)is proposed,during the initialization of population,the individuals with poorly fitness will redistribute by the Gaussian probability model,in view of the stagnation of iteration,the Levy disturbance strategy is introduced and the adaptive change of disturbance step length is controlled,the experimental results show that the above strategy of improvements significantly increase the convergence ability and adaptability of ALO.To avoid the problem of local optimum and instability during ELM training process,ELALO is used to optimize the initial parameters of ELM,the results show that the recognition rate of ELALO-ELM is improved significantly through the case of rolling bearing fault pattern recognition.