Neural-network-based observer for turbine engine parameter estimation

Abstract

Accurate estimation of unmeasurable engine parameters such as thrust and turbine inlet temperatures constitutes a significant challenge for the aircraft community. A solution to this problem is to estimate these parameters from the measured outputs using an observer. Currently existing technologies rely on Kalman and extended Kalman filters to achieve this estimation. This paper presents a neural-network-based observer that augments the linear Kalman filter with a neural network to compensate for any non-linearity that is not handled by the linear filter. The implemented neural network is a radial basis function network that is trained offline using a growing and pruning algorithm. The neural-network-based observer is trained and simulated to estimate the high-pressure turbine inlet temperature, thrust, and stall margins at different levels of engine degradation for a two-spool turbofan engine. Simulation results show the ability of the observer to accurately estimate the performance parameters. The advantage of this observer is that it does not need explicit estimation of health parameters to accurately estimate the performance parameters.

State estimation problem for fractional-order neural networks using event-triggered state observers

“…Therefore, fractional calculus is extensively applicable in many areas, such as electrical circuits, 3 Lu systems, 4 quantum mechanics, 5 economic systems, 6 neural networks, 7 positive systems, 8 multiagent systems, 9 adsorption and desorption processes with power-law kinetics, 10 and lithium-ion cell. 11 As we all know, the information on the state vectors of dynamical systems is essential for monitoring, stabilizing, [12][13][14] fault diagnosis, fault detection, and isolation. 15 Nevertheless, in many control processes, the information on the state vectors of the systems is often unavailable due to technical or economic reasons.…”

“…As we all know, the information on the state vectors of dynamical systems is essential for monitoring, stabilizing, 12–14 fault diagnosis, fault detection, and isolation. 15 Nevertheless, in many control processes, the information on the state vectors of the systems is often unavailable due to technical or economic reasons.…”

A study on global optimization and deep neural network modeling method in performance-seeking control

“…The increase in the hidden layers would greatly enhance the nonlinear fitting capacity and improve the model precision. Therefore, a new aeroengine model modeling method, deep neural network (DNN), [24][25][26][27] which has deeper network structure and stronger expressive ability than the conventional NN, is proposed to improve model precision.…”

Adaptive recurrent Chebyshev neural network control for permanent magnet synchronous motor servo-drive electric scooter

“…Artificial neural networks (NNs) have emerged as a powerful learning technique to perform complex tasks in highly nonlinear dynamic systems and controls. [4][5][6][7][8][9][10][11] Some of the prime advantages of using NN are their ability to learn based on optimization of an appropriate error function and their excellent performance for approximation of nonlinear functions. There are different paradigms of NNs proposed by different researchers for the task of system identification and control.…”

“…Presently, most of the NN-based system identification and control techniques are based on multilayer feedforward NNs or more efficient variation in this algorithm. [8][9][10][11] This is due to the fact that these networks are robust and effective in modeling and control of complex dynamic plants. [8][9][10][11] Pattern classification using Chebyshev NN was first introduced in Namatame and Ueda.…”