UC Merced

In classical control engineering, optimality and robustness have been the main concerns of the control design and maintaining good performance. On the other hand, the third main concern can be considered as smartness with the inevitable grow of Digital Transformation and Industry 4.0 together with the influence of key enabling technologies like Artificial Intelligence (AI), Machine Learning (ML), Big Data (BD) and Edge Computing (EC). These core technologies enable users to increase capabilities of the systems not only for the design of the complex structures with smart control applications but also for maintaining a successful operation afterwards. For this reason, smartness can be considered as one of the most important requirements of maintenance strategies. Many engineering applications require a proper maintenance strategy to address the degradation and failure in the machines, processes and complex systems. In this context, maintenance methodologies play a key role depending on the application type and complexity of the requirements. Reactive and preventive maintenance strategies lead high downtime or waste useful life where they are not handy for a proper maintenance of complex systems. On the other hand, predictive maintenance strategy enables users to find optimal time and part selection to reduce downtime and maximize equipment lifetime. With the introduction of smartness to the predictive maintenance, a new frontier of Smart Predictive Maintenance (SPM) is aimed in this thesis to address main obstacles of traditional predictive maintenance workflow. To introduce smartness into the predictive maintenance framework, key enabling technologies of Digital Twins (DT) and physics-informed Smart Big Data (SBD) is utilized. To enhance the framework, development of the Digital Twin with behavioral matching process and utilization of existing knowledge in the Smart Big Data is demonstrated. The argument of the SPM is supported by a set of case studies including physics-informed transfer learning for fault classification, smart selection of control elements and error recovery for the Radio Frequency Impedance Matching (RFIM) system. Results of the example studies show that SPM is a new and effective systematic approach that can improve maintenance strategies, health monitoring and fault diagnosis applications.