Research On Crack Identification Of High-speed Train Gearbox Based On Surrogate Model

With the rapid development of high-speed train in China,the proportion of high speed and heavy haul trains is increasing gradually.The interaction between the various structures of the train are becoming more and more serious due to the more complex operating environment,which lead s to the increasing problems of the vehicle operation safety.Therefore,it can effectively avoid the burst fault and ensure the operation safety of the train through the safety monitoring of the train key parts.Gearbox is a key component in the running departments of high-speed train.Its structural performance not only affects the speed of train operation,bu t also has a great effect on the structural performance of bogie,which may cause the train instability in the serious situation.As an important form of structural damages,crack has an important impact on the normal operation of the gearbox.Therefore,how to judge and identify the crack in the gearbox timely and effectively is an urgent problem to be solved.Because there are many parameters involved in the gearbox system in the train running condition,it takes long time and low efficiency to establish the crack database by calling the finite element model.Using agent model technology to replace the traditional method of using finite element model to modify and establish crack database can effectivel y simplify the reconstruction of a large number of sam ple models and the number of times of finite element calculation in the construction of crack database.High accuracy of crack identification can be achieved through the surrogate model correction technique,and the recognition error is kept within 3%.This paper aims to synthetically consider the working condition of the gearbox in the train running state,start with the identification of the crack fault in the gearbox,and construct the Kriging surrogate model between the crack parameters of the gearbox and the dynamic response of the gearbox structure.The particle swarm optimization(PSO)algorithm is used to identify the target crack parameters in the surrogate model.The main research contents are as follows:(1)The natural vibration characteristics o f box structure are analyzed by modal analysis.Because the local stiffness of the structure can be reduced when damage occurred in the gearbox,and its dynamic characteristics can be reflected by modal parameters such as natural frequency,so as to evalua te the structural parameters of the established gearbox model.Where the damage area is likely occurred has been found based on the inherent properties of the gearbox,which provides guidance for the setting of the crack position in the construction of the crack model.(2)A method of gearbox crack identification based on Kriging surrogate model is proposed.The Kriging surrogate model between the crack model parameters and its structural dynamic responses is constructed by using the crack samples that selected by the Latin hypercube sampling and its corresponding modal frequency responses.It can replace the relationship between the original crack parameters and its structural dynamic responses,so as to effectively reduce the reconstruction times of the cr ack model and the finite element calculation process.(3)The effect of internal and external excitations on the dynamic performance of the gearbox in the running condition of high-speed train is studied.The dynamic model of high-speed train that based on the internal and external excitation is established to simulate the running state of the gearbox,and the vibration signal of the gearbox is collected by simulation.The dynamic response of the gearbox under the dynamic load is studied through the transie nt dynamic analysis and the harmonic response analysis,which can be used to analyze the structure characteristics of the gearbox in the train running state.(4)The Kriging surrogate model is constructed to replace the corresponding relationship between the original design variables and the structural dynamic responses by using the train parameters,operation conditions,crack structure and other samples and its corresponding dynamic responses of the gearbox under the dynamic load,which can effectively reduce the amount of calculation.T he identification of crack can be achieved high accuracy by using the particle swarm optimization algorithm to optimum identification and using the criteria of increasing sample to iterative optimization.The validity of cr ack identification method based on Kriging surrogate model is verified by experiments.