Yazar "Al-Zurfi, Rafah Hussein Jumaah" seçeneğine göre listele

Listeleniyor 1 - 1 / 1
  • Küçük Resim
    Öğe
    POWER SYSTEM FAULT IDENTIFICATION AND CLASSIFICATION IN FUEL CELLS VIA ARTIFICIAL NEURAL NETWORK
    (2024-01) Al-Zurfi, Rafah Hussein Jumaah
    The research delineated in this thesis is poised to contribute significantly to the Domain of fault diagnosis in industrial processes, with a specific emphasis on employing sophisticated processing and pattern recognition methodologies for bearing analysis. The primary thrust of the investigation is centered on the application of vibration analysis to discern and diagnose issues in bearings. To this end, an Artificial Neural Network (ANN) is deployed for the analysis of input-output datasets extracted from a Matlab-Simulink-based Proton Exchange Membrane Fuel Cell (PEMFC) model, specifically the 6kw-45Vdc model. The articulated ANN is designed to furnish steady-state predictions predicated on the provided input. Subsequently, the output of the PEMFC is scrutinized vis-a-vis The model's output, particularly in response to emergent events inducing alterations in the plant's output voltage or current. A residual signal is systematically monitored and employed as a diagnostic tool to identify and characterize defects within the system. The empirical phase of data collection entails meticulous acquisition from a system or test rig, with due consideration accorded to diverse fault typologies, encompassing Abrupt, Incipient, and Intermittent faults. The steady-state simulation is built around three inputs: heat, fuel pressure, in addition air pressure, as well as two outputs: voltage and current. Matlab's Simulink platform serves as the instrumental medium for comprehensive system modeling. The subsequent research phase pivots towards the utilization of an Artificial Neural Network for condition categorization. A nuanced exploration and juxtaposition of various supervised learning algorithms, inclusive of support vector machines, random forests, and extreme learning machines, is undertaken to discern the optimal method for effecting bearing fault classification. In summation, this research orchestrates a methodically comprehensive approach to fault diagnosis, encompassing meticulous data collection, exacting system modeling via Simulink, and the judicious application of advanced machine learning paradigms through an Artificial Neural Network. The overarching objective is the discernment and diagnosis of bearing faults within the context of industrial processes.