Research On Fault Algorithm Of Rotating Machinery Based On Convolutional Neural Network

With the development of industrial level,people have been inseparable from mechanical equipment.Fault diagnosis of mechanical equipment is a key technology to ensure people’s production and life stably and efficiency.For the mechanical equipment is usually complicated and refined,it is time-consuming and inefficient to carry out manual fault diagnosis.With the development of deep learning and sampling technology,the health detection of mechanical equipment has entered the intelligent era.Aircraft hydraulic pump is an important mechanical component of aircraft.The work condition of hydraulic pump will seriously affect the flight safety of the aircraft.Therefore,it is necessary and important to carry out fault diagnosis of aircraft hydraulic pump.Convolutional neural network has powerful classification and recognition ability,and is widely used in fault diagnosis of rotating mechanical equipment.For the aircraft hydraulic pump is a kind of rotating mechanical equipment,convolutional neural network has a good application prospect in aircraft hydraulic pump fault diagnosis.This paper is based on the collected axial and radial vibration data of aircraft hydraulic pump.In order to enhance the classification ability of convolutional neural network,the axial and radial vibration data are sent into the neural network at the same time.By increasing the number of input channel,the convolutional neural network can enhance the feature extraction ability of the sample time-domain signal.In this paper,an independent input multi-input convolutional neural network(IIMI-CNN)is design.In the IIMI-CNN model,multiple inputs are sent to the convolution and pooling layers twice individually,and then mix them through the concatenate connection layer.The output of the concatenate connection layer is sent to the convolution and pooling layers twice,and the classification result will output through the fully connected layer.The recognition accuracy of IIMI-CNN on aircraft hydraulic pump data can reach 98%.However,in the training process,sometimes the fitting is unstable.In order to improve the fitting stability IIMI-CNN,a combined input multi input convolutional neural network(CIMI-CNN)is proposed in this paper.Multiple inputs are vertically combined at the beginning of the model to form two-dimensional inputs,and then send them to the convolution neural network.The fitting condition of the network is stable,and the recognition accuracy can reach 99%.By adding noise to the input signal,the anti-noise test of the above two models is carried out.The results show that CIMI-CNN has strong anti-noise performance.By using two motor current data from the bearing data set of Paderbom University,the generalization performance of the two multi input convolutional neural networks is verified.The results show that both of them have certain generalization,and the generalization of CIMI-CNN is stronger.