Authors:
Ibrahem M. A. Ibrahem
1
;
Ouassima Akhrif
2
;
Hany Moustapha
1
and
Martin Staniszewski
3
Affiliations:
1
Department of Mechanical Engineering, Ecole de Technologie Supérieure, 1100 Notre Dame St. W, Montreal and Canada
;
2
Department of Electrical Engineering, Ecole de Technologie Supérieure, 1100 Notre Dame St. W, Montreal and Canada
;
3
Digitalization Project Manager, Siemens Canada, 9545 Cote de Liesse Road, Montreal and Canada
Keyword(s):
Neural Networks, NARX Model, Gas Turbine, Aero Derivative, Modelling, Simulation, Dynamic Neural Networks.
Related
Ontology
Subjects/Areas/Topics:
Informatics in Control, Automation and Robotics
;
Signal Processing, Sensors, Systems Modeling and Control
;
System Identification
;
System Modeling
Abstract:
In this paper, the modelling of aero derivative gas turbine engine with six inputs and five outputs using two types of neural network is presented. Siemens three-spool dry low emission aero derivative gas turbine engine used for power generation (SGT-A65) was used as a case study in this paper. Data sets for training and validation were collected from a high fidelity transient simulation program. These data sets represent the engines operation above its idle status. Different neural network configurations were developed by using of a comprehensive computer code, which changes the neural networks parameters, namely, the number of neurons, the activation function and the training algorithm. Next, a comparative study was done among different neural models to find the most appropriate neural network structure in terms of computation time of neural network training operation and accuracy. The results show that on one hand, the dynamic neural network has a higher capability than the static
neural network in representation of the engine dynamics. On the other hand however, it requires a much longer training time.
(More)