Investigations on machine learning and deep learning techniques for machinery fault diagnosis and tile defect detection in cyber physical systems

Abstract

In recent years many industries are moving towards the vision of newlineIndustry 4.0 which includes autonomous, adaptive self-diagnosing machines newlinereferred to as Cyber Physical Systems (CPS). CPS is an integrated machinery newlinethat constitutes physical, networking, and computational components of newlineIndustry 4.0 to provide cutting-edge functionalities. The widespread adoption newlineof CPS has transformed several industrial sectors by enabling them to newlineoptimize their manufacturing processes, save costs, and boost production. newlineThe capacity to accurately identify faults in end-products and automated newlinemachinery is one of the major challenges in CPS. This is crucial for industrial newlinequality control applications since regular maintenance downtime and newlinedamaged end-products can cause considerable losses in production and newlinerevenue. newlineThis research thesis explores the potential of machine learning and newlinedeep learning techniques in the cognition domain of CPS for industrial newlineappliances such as machinery fault diagnosis and tile defect detection. The newlineinvestigation delves into diverse algorithms and architectures, evaluating newlinetheir performance using actual datasets from real-world applications. Firstly, newlinean enhanced deep learning methodology that can identify various induction newlinemotor faults, including bearing faults, motor imbalances, and misalignments newlinewas developed. Furthermore, different machine learning techniques were newlineanalysed to forecast equipment breakdowns and enhance industry newlineperformance indicators through preventive maintenance. Finally, an newlineoptimized deep learning technique is proposed for the identification and newlineclassification of defective tiles in an assembly line for the production of tiles newline