A Parallel Ensemble Learning Model for Fault Detection and Diagnosis of Industrial Machinery

Accurate fault detection and diagnosis (FDD) is critical to ensure the safe and reliable operation of industrial machines. Deep learning has recently emerged as effective methods for machine FDD applications. However, the gradient descent optimization method that is commonly used in deep learning suffers from several limitations, such as high computational cost and local sub-optimal solutions. Accordingly, this paper proposes a new parallel ensemble model comprising hybrid machine and deep learning for undertaking FDD tasks. Composed of three levels of learning, the proposed ensemble model employs two base learners and a meta-learner, and is executed in parallel processing platform to achieve efficient computation. The base learners adopt a hybrid Back-Propagation (BP) and Particle Swarm Optimization (PSO) algorithms to exploit the corresponding local and global optimization capabilities for identifying optimal features and improving FDD performance. The proposed model is validated through a series of experiments using two benchmark data sets, i.e., CWRU and MAFaulD. The results demonstrate a high performance with accuracy rates of 98.45% and 99.79% for CWRU and MAFaulD, respectively. Its parallel implementation is able to reduce the computation time, resulting in a speed-up of 5.9 time and 7.17 time, respectively. These findings indicate that the proposed model is effective and efficient for FDD of industrial machinery, making it a promising solution for implementation in real-world environments.

Automated summary

This paper proposes a new parallel ensemble model for fault detection and diagnosis (FDD) tasks using hybrid machine learning and deep learning methods. The model achieves efficient computation through parallel processing, and is validated to have high performance and efficiency in FDD of industrial machinery.