الفهرس 
IntroductionOnly 14 pages are availabe for public view
 |  | from | 148 |  |  |
| | | from | 148 | | |
Abstract
The thesis is structured into five chapters, which develop and evaluate methods for diagnosing and prognosing bearing faults. It reviews existing approaches, covers tradi- tional and deep learning-based techniques, and discusses the development of a neural architecture search method and a genetic algorithm to optimize performance.
Introduction: introduces machinery maintenance, emphasizing its importance and pro- viding a general framework for data-driven bearing fault diagnosis and prognosis. It gives an overview of deep learning algorithms and optimization techniques applied to this domain. The target, objectives, novelty, and contribution of the research are out- lined, followed by a description of the thesis’s organizational structure and the content of subsequent chapters.
Literature Review: provides an overview and contextual information about the crucial importance of bearings in rotating machinery, including details on their construction, components, failure modes, and monitoring strategies. It discusses both conventional and deep learning methodologies employed for diagnosing and predicting bearing faults. Particular emphasis is given to the significance of deep learning techniques in this area. Additionally, the chapter delves into optimization approaches used for enhancing deep learning models. Lastly, it concludes by summarizing the key findings, identifying re- search gaps, and acknowledging any limitations encountered during the investigation.
Design and Development of a Bearing Fault Diagnosis Algorithm Based on Deep Learning: starts with an introduction and then proceeds with data collection and preprocessing, involving dataset description, data preprocessing procedure, and dataset splitting for
the primary and subsequent experiments. The methodology includes using RL-NAS for identifying an accurate CNN architecture, model training, and performance evaluation metrics. The chapter concludes with the results and discussions of three experiments: the primary experiment, the second experiment, and the third experiment.
Deep Neural Network Optimization for Bearing Fault Prognosis: utilizes a genetic algo- rithm to optimize the hyperparameters of a deep neural network for estimating the re- maining useful life of bearings. The PRONOSTIA bearing dataset, a widely used dataset in this area, is employed for this purpose. The chapter covers the data preparation, pro- viding a detailed description of the data and explaining the signal preprocessing steps. Additionally, it describes the architecture of the deep neural network and the training process. The optimization procedure involving the genetic algorithm is discussed, and the chapter concludes with a presentation and discussion of the performance results of the optimized network.
x
Conclusion and Future Work: Summarizes the main findings of the study and suggests directions for future work.