Machinery Fault Detection using Advanced Machine Learning Techniques

Manufacturing industries are expanding rapidly, making it essential to detect early signs of machine faults for safety and productivity. With the extension of machines' runtime due to industrial automation, breakdown risks have increased, leading to economic and productivity consequences and sometimes even causalities. The surge in industrial big data from low-cost sensing technologies has enabled the development of intelligent data-driven Machinery Fault Detection (MFD) systems based on machine learning techniques in recent years. However, most existing methods are based on supervised pattern classification techniques to detect previously known fault types, which have limitations such as lack of generalization across different operational settings, focusing only on specific machinery and/or data types, and considering the identical and independent distribution of training and testing data. Therefore, my PhD research aims to develop a robust MFD framework for practical use by addressing these limitations.I will explore the potential of ensemble learning, unsupervised and semi-supervised anomaly detection, reinforcement learning, transfer learning, and cross-domain adaptation approaches in MFD. My PhD research will contribute to the field of data-driven MFD by proposing novel, effective solutions that can be applied across various manufacturing applications.