DIAGNOSTICS OF THE TURBINE ENGINE CONTROL SYSTEMS BY TRAINED NEURAL NETWORKS
The problems of control systems stability margins diagnostics by trained neural networks are considered. Within the fault-tolerant control system framework, a sequence of data processing on the formation of diagnostic attributes, identification of symptoms of malfunctions and diagnosis of the control systems technical state was proposed. Test studies of diagnostic procedures illustrate the ability of trained neural networks to recognize the symptoms of parametric and topological system failures in terms of the diagnostic attributes of various categories: measurement samples, coefficients of the difference equations and eigenvalue moduli. Computer simulations of diagnostic procedures for the purpose of assessing the stability margin of the turbo-generator control system shows the possibility of solving the problems of neural network diagnostics and, consequently, ensuring an active fault-tolerance of control systems.
Authors: E. V. Guzaev, Yu. A. Korablev, M. Yu. Shestopalov, D. H. Imaev
Direction: Automation and Control
Keywords: Control system, model, fault, symptom, diagnosis, fault tolerance, neural network, learning, turbine engine
View full article