Research on application of DL model parameter tuning based on optimization algorithm in fault identification

초록

Unlike traditional fault diagnosis methods that rely on expert experience for manual feature extraction or model-based techniques with high data quality requirements, deep learning models leverage convolutional neural networks and Transformers to automatically extract features. This capability reduces the influence of noise and enhances generalization, significantly improving diagnostic accuracy and efficiency. Optimized parameter configurations can enhance model accuracy, expedite the training process, and mitigate overfitting issues. The deep learning model parameter tuning method based on optimization algorithms circumvents the tedious manual parameter adjustment process, improving both model accuracy and convergence speed. For instance, model accuracy can increase by 5% to 15% after tuning. Different optimization algorithms exhibit varying performances and functions in fault identification tasks, making the selection of an appropriate algorithm vital for enhancing training efficiency and accuracy.