Research On Fault Diagnosis Of Gearbox Gear Of New Energy Vehicle Based On Deep Neural Network

The gearbox is a key transmission mechanism for automobile power,in which gears are the core components.The gears in the gearbox are affected by changing working conditions and are prone to failure.The vibration signal in the fault state is non-stationary and contains a lot of noise,which greatly increases the difficulty of gearbox fault diagnosis.In order to solve this problem,this paper proposes a fault diagnosis model based on deep neural network,and also proposes a training method for the hybrid model,The reliability and accuracy of the model were verified through experiments.Finally,a gearbox fault diagnosis system based on the model was constructed.The main contents are as follows:This paper proposes a new energy vehicle gearbox gear fault diagnosis method based on deep neural networks.The model is trained and intelligently classified to achieve the diagnosis effect.First of all,for the collected signal contains a lot of noise interference,Stacked Denoising auto-encoders is proposed for processing.SDAE can automatically extract robust feature representations from noisy data through its unique stacking structure and noise reduction training.Subsequently,a Bidirectional Gated Recurrent Unit Network was introduced to process time series data.The Bi GRU network can weaken the gradient and disappear.Its unique gated structure can effectively solve the problem of long-and short-term sequence changes.At the same time,two-way training can make full use of past and future information.In summary,the two neural networks are integrated to build a hybrid neural network,and Adam's optimization algorithm is used to optimize the network parameters.Dropout technology prevents overfitting and combines the two methods to propose a hybrid neural network training method.The vibration signals of gear failures in the gearbox were collected through simulation experiments,and the types of the failures were determined using hybrid neural networks.The average accuracy rate of each failure type was above 99%.The fault diagnosis effect of the method under different signal-to-noise ratio and timevarying speed was explored.The results show that the method has good anti-noise performance.At the same time,the effects of different training set proportions and training imbalances on the diagnostic effect were explored.The results show that the method in this paper has a certain stability and provides a certain guiding role for the proportion allocation of sample sets.Based on the hybrid neural network,a gearbox fault diagnosis system is developed through MATLAB GUI,and the vibration signal collected through the developed experimental platform is taken as an example analysis to prove that the system can effectively diagnose gear fault types.