Optimization of One-dimensional Convolution Time Series Model for Computer Numerical Control Machine Tool Fault Diagnosis

The work efficiency of computer numerical control machine tool is very high. However, the factory will lose huge when the computer numerical control machine tool has some serious faults. To quickly diagnose what kind of fault has occurred in the computer numerical control machine tool, this paper proposes an optimized one-dimensional convolutional time series model. It takes the cutting tool and rolling bearing on the spindle of the computer numerical control machine tool as the research object. The structure of the model is optimized based on the WDCNN model. In addition, Long short-term memory networks and adaptive cross-entropy sensitive learning are integrated. Problems such as too much computation, inactivation of some neurons, overfitting of models, weak correlation of long-time series, and data imbalance have been overcome. The experimental results show that the accuracy of this model can reach 95.11%, which shows the advantages of fault diagnosis ability.