Deep CNN and Adaptive Habitat Biogeography-Based Optimizer Algorithms for Rotating Machine Fault Detection and Classification

With the abundance of condition monitoring data, research on data-driven methods for rotating machine fault diagnosis has garnered increased attention. However, existing approaches often struggle with extracting representative features from raw data and assume feature distribution consistency between training and testing domains, which does not hold in many real-world scenarios. This study proposes a novel methodology for diagnosing faults in rotating machines by utilizing the automatic feature extraction capabilities of deep learning and the improved performance of ensemble learning, specifically through the integration of deep convolutional neural networks (DCNN) and an adaptive habitat biogeography-based optimizer for hyperparameter optimization of the deep CNN. Experimental validation conducted on a dataset collected from URMM/CRTI test bench demonstrates the effectiveness of the proposed methodology.