الفهرس | Only 14 pages are availabe for public view |
Abstract Fault detection and diagnosis (FDD) have become critical issues, particularly for industrial systems that rely on sophisticated control capabilities. FDD ensures project performance requirements such as reliability, survivability, maintainability, availability, cost efficiency, safety, and quality. This thesis is concerned with FDD monitoring framework execution, detecting system faults, isolating the broken components (faults) within the system, and then identifying faults. Two different AI methodologies were proposed and applied in two different case studies for FDD. Those methodologies are FDD based on Recurrent Neural Networks (RNNs) and Distributed Neural Networks (DNNs). Overall, both methodologies demonstrated the potential of neural networks, specifically RNNs and DNNs, in fault detection and diagnosis in dynamic systems. They show the importance of preprocessing data and training neural networks with labelled data for accurate fault detection. These methodologies could be applied to enhance system reliability and performance in various engineering systems such as level control systems and network of distributed motors. |