Research on numerical control machine fault diagnosis based on distribution adaptive one-dimensional convolutional neural network

Abstract

Numerical control machine is a high-precision and high-automation equipment, if the problem occurred in operation, it can affect the processing conditions of the machinery parts first. If the fault is aggravated, it can eventually cause the numerical control machine to stop. Spindle bearings and tools are the most vulnerable parts of numerical control machine. Previously, resonance demodulation technique was used for bearing fault diagnosis. Empirical analysis or neural network was used for tool fault diagnosis. However, the numerical control machine is an entirety, the fault is usually caused by multi-dimensional factors, the above method doesn’t work when two types of faults occur at the same time. To diagnose faults of numerical control machine, a fault diagnosis model named distribution adaptive deep convolutional neural networks (DADCNN) was proposed. This model was based on One-dimensional convolution algorithm. The Batch Normalization algorithm was involved to overcome the problem of changing data distribution in the middle layer. The t-SNE algorithm was used to visualize and view the feature classification results. Experiments show that the accuracy of this model can reach 90.29%, and it has good fault diagnosis capabilities.