Bearing Fault Detection Based on Audio Signal Using Pre Trained Deep Neural Networks

In the current study, we delve into advanced deep learning techniques, focusing on Convolutional Neural Network (CNN) and deep Multi-Layer Perceptron (MLP) architectures to enhance fault detection in crucial machine components such as rolling bearings. The main idea is to utilize a Stacked Auto-Encoder (SAE) to initialize the model and improve its feature extraction capability. Moreover, departing from traditional vibration-based analyses, we pioneer the use of audio signals for fault detection. These ideas are investigated for CNN and MLP architectures, and the performance of the pre-trained models is compared with that of two other models, namely models with the same architectures trained from scratch and the SAE encoder equipped with a softmax classifier. Comprehensive testing and comparison reveal that integrating a pre-trained SAE model into the Deep Neural Network (DNN) can result in remarkable error detection through prior feature learning

In the current study, we delve into advanced deep learning techniques, focusing on Convolutional Neural Network (CNN) and deep Multi-Layer Perceptron (MLP) architectures to enhance fault detection in crucial machine components such as rolling bearings. The main idea is to utilize a Stacked Auto-Encoder (SAE) to initialize the model and improve its feature extraction capability. Moreover, departing from traditional vibration-based analyses, we pioneer the use of audio signals for fault detection. These ideas are investigated for CNN and MLP architectures, and the performance of the pre-trained models is compared with that of two other models, namely models with the same architectures trained from scratch and the SAE encoder equipped with a softmax classifier. Comprehensive testing and comparison reveal that integrating a pre-trained SAE model into the Deep Neural Network (DNN) can result in remarkable error detection through prior feature learning.