Experimental Investigation of Microjet Engine Anomaly Detection Based on ANN Data Clustering Analysis

Please enable javascript in your browser.

Khodadadi, Mahyar | 2017

387 Viewed

Type of Document: M.Sc. Thesis
Language: Farsi
Document No: 49616 (45)
University: Sharif University of Technology
Department: Aerospace Engineering
Advisor(s): Saghafi, Fariborz
Abstract:
In this study, the possibility of detecting known and unknown anomalies in a micro jet engine is evaluated. the proposed method is neural network and clustering analysis. The process consists of two main parts, first part is experimental abnormal behavior simulation and second part is validation of proposed detection method. In the first part a micro jet engine (Olympuse HP) were used and we created pre-designed errors in engine normal behavior to record significant changes in its performance. Recorded outputs consist of six different variables: RPM, Exhaust gas temperature, Throttle command, Fuel pump voltage, Thrust and Fuel weight. In the second part, for analyzing data, we used clustering method using neural networks, which is a non-regulatory methods, This method will detect unusual behavior without any prior knowledge. After applying the data to the network, we tried to change network parameters and data configuration to achieve optimal lassification. Finally at the first step we succeeded to detect pre-defined and undefined abnormal behavior and in the next step we achieved a proper classification between our unusual behaviors
Keywords:
Clustering ; Anomaly Detection ; Neural Network ; Data Clustering ; Microjet Engine ; Experimental Simulation

No TOC