Research On The Gearbox Fault Detection Method Based On Particle Swarm Optimization And LSTM

Gear box as a very core part of many components in the mechanical transmission system,because of its internal structure is very complex,complex and changeable working environment and other special reasons,it often fails,especially the gear failure in the gear box.Gear box has been used in many fields,such as: chemical,aerospace,navigation,electrical and other fields.The failure of the gear box may cause the failure of the mechanical system and threaten the safe operation of the mechanical equipment.Therefore,it is of practical significance to carry out research on gear box fault.This article first analyzes the causes of gearbox faults,and on this basis,determines that gear faults are the most important part of gearbox faults,and provides a comprehensive introduction to the types of faults in the gears inside the gearbox.Furthermore,the mechanism of gear vibration was further explored,and it was clarified that there is a certain connection between the vibration signal during gear meshing and gear faults.The principle of particle swarm optimization algorithm and short-and long-term memory neural network is also analyzed.Then,in response to the problem that the vibration signals of gear faults often have non-stationary characteristics and contain deep feature information that is difficult to fully utilize,time-domain analysis was conducted on the vibration signals,and 15 time-domain indicators were extracted.The Long Short Term Memory(LSTM)neural network was applied to gear fault detection.The extracted 15 time-domain indicators were sent to the LSTM network for detection,but the detection effect was not satisfactory.After analysis,it was believed that the time-domain analysis made the gear fault information more comprehensive,but some time-domain features had repetitive representation information and certain connections between features.In order to reduce information redundancy,improve the independence between features,and reduce operational costs,Principal Component Analysis(PCA)was used to successfully reduce the dimensionality of features..Secondly,in the process of using LSTM to detect gear fault categories,it was found that the values of learning rate and the number of hidden layer neurons directly affect the accuracy of LSTM detection,resulting in LSTM model detection accuracy not meeting the requirements and the model not operating at its optimal state.To solve this problem,the Particle Swarm Optimization(PSO)algorithm in swarm intelligence optimization algorithm is introduced.At the same time,the PSO is prone to be trapped in the local optimization problem.The PSO is improved from the change mode of inertia weight and learning factor,and is compared with the target function,The test results can verify that the improved Particle Swarm Optimization(IPSO)algorithm has better optimization ability.The improved particle swarm optimization algorithm was used to optimize the learning rate and number of hidden layer neurons of LSTM,and the most suitable values were selected to assign to the LSTM model.Finally,a gear fault detection model based on IPSO-LSTM was established.Finally,in order to verify the excellent detection ability of the gear fault detection model based on IPSO-LSTM,experiments were conducted on Convolutional Neural Networks(CNN),Classical LSTM,and PSO-LSTM models using the same gear fault data.The experimental results were analyzed using multiple evaluation indicators.After analysis,it was found that the gear fault detection model based on IPSO-LSTM had the highest detection accuracy,The classification of gear fault types is more accurate and has better performance,providing a good guarantee for the safe operation and maintenance of the gearbox.