Advanced Failure Classification in Rotating Machinery Using Dense Convolutional Neural Networks

It is essential to ensure the reliability and effectiveness of rotating machinery in modern industry, particularly when applying Condition-Based Maintenance (CBM). Conventional fault diagnosis techniques are inefficient due to the large volume and complexity of data generated by signals in industrial parts such as bearings. Purpose: This research addresses the challenge of accurately and efficiently analyzing complex vibration data produced by rotating machinery. A Dense Convolutional Neural Network (DCNN) model for improved failure classification of rolling element bearings is proposed. Methodology: In this research a methodology is proposed based on deep learning techniques, and particularly one dimensional (1D) convolutional layer to classify bearing conditions from normal operation to critical failure with high accuracy. The system learns patterns in vibration signals and provides a more automated and effective solution than the current methods. Findings: The performance of the Dense Convolutional Neural Networks (DCNNs) was ensured by a large and diverse dataset, achieving higher accuracy, precision, and recall than classical approaches. These results indicate that the DCNNs can substantially enhance the current predictive maintenance approaches by a significant margin, which will in turn decrease downtime, reduce maintenance costs, and increase productivity. This research demonstrates how Artificial Intelligence (AI) enhances industrial maintenance by enabling more accurate and efficient diagnostic systems. It also introduces a scalable model applicable to diverse industrial contexts, including both developed and emerging economies.