Study On Fault Identification Of Planetary Gearbox Based On Acoustic Emission

Planetary gear transmission system is the core transmission system of various mechanical equipment because of its excellent characteristics.Its internal structure is complex,so internal parts are easy to be damaged due to the interaction between parts or external environment when working.If these equipments run under the fault state for a long time,they will break down at least,or the production and even personal safety will be affected.Therefore,it is important to carry out real-time monitoring,early fault prevention and diagnosis for planetary gear transmission system.For the diagnosis of mechanical equipment,acoustic emission testing technology is a nondestructive testing technology with higher sensitivity,stronger anti-interference ability and wider frequency response range compared with the traditional vibration detection technology.Acoustic emission(AE)signal is a kind of UHF stress wave pulse signal which can characterize the defects of material structure.A large number of AE signals will be generated during the working process of planetary gearbox.It is research focus and difficulty of planetary gearbox fault diagnosis technology that exploring the deep information of fault signal for fault identification and realizing early fault intelligent diagnosis.So the topic discussed in this paper is based on the real application requirements of planet gear box fault diagnosis,through the acoustic emission testing technology collecting fault signals occurring in the process of planetary gear box,then based on the method of deep learning to explore the data information to complete the fault identification of AE signal for fault diagnosis of mechanical equipment to provide a new feasible way of thinking,subject research work is mainly as follows:(1)According to the operation principle and mechanism of AE signal properties of planetary gear train,this subject designed a planetary gearbox fault detection system with six channels based on STM32.The hardware work mainly includes signal adjusting circuit,data acquisition and transmission circuit,power circuit schematic and PCB design and implementation.The software work includes data acquisition and transmission,upper computer waveform display and storage.This system has realized the error in microsecond multichannel parallel sampling,and realized the working condition of equipment of multipoint real-time monitoring and data storage function.At the same time,it has improved the anti-interference from the sensor level and the system hardware circuit level.The signal acquisition and evaluation experiments show that this system can be applied to different working conditions,the different fault types of equipment fault diagnosis field.Then the signal data is de-noised to provide a real and reliable data set for the fault signal identification experiment.(2)In order to solve the identification problem of a large number of AE fault signals and achieve intelligent diagnosis of equipment.In this paper,a Temporal convolutional network(TCN)based on deep learning method for fault diagnosis of planetary gearbox is proposed,and AEG-TCN,a model for gearbox fault diagnosis,has been established,which avoids the phenomenon of gradient dispersion and weight matrix degradation in the existing fault diagnosis models based on RNN and CNN with the deepening of network layers.And then the author further puts forward,a planetary gear box fault recognition model with parallel TCN-CNN,which could explore the acoustic emission signals from global and local ways.In this paper,we discuss the TCN network architecture and its advantages and disadvantages from its architecture perspective.At the same time,the contrast experiment verifies the effectiveness of ARG-TCN and TCN-CNN model in gearbox fault acoustic emission signal recognition tasks,such as the accuracy is better than traditional algorithms of CNN and RNN,and verifies the feasibility and superiority of TCN application in fault diagnosis field.To sum up,it has a certain application research value and provides a new idea for gearbox fault diagnosis.