Testing and Model Identification of a Turbojet Engine Using Neural Networks | ||||
| International Conference on Aerospace Sciences and Aviation Technology | ||||
| Article 58, Volume 14, AEROSPACE SCIENCES & AVIATION TECHNOLOGY, ASAT - 14 – May 24 - 26, 2011, May 2011, Page 1-18 PDF (822.04 K) | ||||
| Document Type: Original Article | ||||
| DOI: 10.21608/asat.2011.23294 | ||||
 View on SCiNiTO | ||||
| Authors | ||||
| I. M. Atia; A. M. Bayoumy | ||||
| Egyptian Armed Forces, Egypt. | ||||
| Abstract | ||||
| Artificial Neural Networks (NN) are a well-known tool among artificial intelligence techniques that are able to reproduce arbitrary relationships existing between input and output variables of even highly non-linear systems. In this paper, a small turbojet engine SR-30 is tested on a minilab test-rig. Then linear ARX (AutoRegressive with eXternal input) structure and nonlinear neural network representations are used for modeling the dynamics of this small turbojet engine. This modeling is based on real engine data obtained from testing of the SR-30 engine. In order to build a feed forward NN model, one could identify the nature and characteristics of its dynamics and the order of the system to be modeled by using conventional linear system identification. This step is used to obtain a linear ARX model. Using the input/output relationship of this model, a neural model is trained for the SR-30 turbojet engine that represents the nonlinearity of the engine throughout its full operating range. Validation of this neural model is performed using another set of the experimental data. The work shows that neural model could capture system nonlinearity and represent the real engine dynamics better than the linear ARX model. | ||||
| Keywords | ||||
| Small turbojet engines; artificial intelligence; Neural Networks; system identification; Modeling and simulation; engine testing | ||||
|
Statistics Article View: 157 PDF Download: 215 |
||||
