Abstract To address the issues of hyperparameter optimization and, consequently, the limitation in the diagnostic accuracy of a convolutional neural network (CNN) model in the domain of computer numerical control (CNC) machine tool bearing fault diagnosis, the paper suggests an artificial-intelligence-based diagnostic approach, which combines an improved Exponential Triangular Optimization algorithm (IETO) with a CNN model. First, a better IETO is created. The algorithm brings out a phase-switching mechanism (feedback-based). The switching threshold is dynamically updated in relation to the search status to optimize exploration and exploitation. In the meantime, a joint hyperparameter optimization plan is developed around covariance learning and joint evaluation. It is a strategy that characterizes the coupling among hyperparameters and steers them to be updated in coordination. Second, IETO algorithm is combined with a CNN model in order to develop an end-to-end bearing fault diagnosis model. It has been experimentally shown that, at a self-constructed five-axis CNC machine tool spindle-bearing fault dataset, the presented approach has a 99.65% diagnostic accuracy. This is quite high compared to a number of mainstream baselines, such as the original ETO-CNN, Transformer, CNN-BiLSTM, and YOLOv8 models. Moreover, the results of the ablation experiments confirm the efficiency and need of every improvement strategy. On the whole, the offered approach offers a valid way to perform high-precision and self-adjusting fault detection of the CNC machine tools bearings.
Gang et al. (Wed,) studied this question.