Optimization of Sample Size, Data Points, and Data Augmentation Stride in Vibration Signal Analysis for Deep Learning-Based Fault Diagnosis of Rotating Machines

In recent years, deep learning models have increasingly been employed for fault diagnosis in rotating machines, with remarkable results. However, the accuracy and reliability of these models in fault diagnosis tasks can be significantly influenced by critical input parameters, such as the sample size, the number of data points within each sample, and the augmentation stride in vibration signal analysis. To address this challenge, this paper proposes a new adaptive method based on Bayesian optimization to determine the optimal combination of these input parameters from raw vibration signals and enhance the diagnostic performance of deep learning models. This study utilizes a one-dimensional convolutional neural network (1-D CNN) as the deep learning model for fault classification. The proposed adaptive 1-D CNN-based fault diagnosis method is validated via vibration signals collected from motor rolling bearings and achieves a fault diagnosis accuracy of 100%. Compared with existing CNN-based diagnosis methods, this adaptive approach not only achieves the highest accuracy on the testing set but also demonstrates stable performance during training, even under varying operating conditions. These results indicate the importance of optimizing the input parameters of deep learning models employed in fault diagnosis tasks.