In modern gas turbine engine control, Direct Thrust Control (DTC) is an effective method for achieving the desired thrust. Model Predictive Control (MPC) has the characteristics of handling constraints while accomplishing command tracking, making it a promising approach for implementing DTC. However, since the performance of DTC is highly sensitive to MPC’s tuning parameters, developing an efficient optimization strategy for these parameters becomes imperative. Therefore, a Model Predictive Direct Thrust Control (MP-DTC) architecture is designed, and its asymptotic stability is proven. Additionally, the influence of the tuning parameters on control performance is analyzed. Then, a tuning strategy for MP-DTC architecture is proposed. This strategy combines the objectives of DTC to design a Multi-Objective Optimization (MOO) index function, and uses Multi-Objective Grey Wolf Optimizer (MOGWO) to solve its Pareto front and obtain the tuning parameters. In the Hardware-in-Loop (HIL) experiments, the proposed MPDTC architecture achieves the shortest settling time and smallest overshoot compared to the latest DTC scheme. Its MOGWO-based tuning strategy provides more Pareto-optimal solutions, ensuring optimal selection based on rapidity and stability, and maintains precise DTC even under component degradation and various operating conditions, thereby providing robustness, optimality, and generalizability.

Direct Thrust Control (DTC) is effective in dealing with the mismatch between thrust and rotor speed in traditional engine control. Among the DTC architecture, model-based thrust estimation method has less arithmetic consumption and better real-time performance. In this paper, a direct thrust controller design approach for gas turbine engine based on parameter dependent model is proposed. In order to ensure the stability of DTC control system based on parameter dependent model, there are usually conservatism detects. For the purpose of reducing the conservatism in the solution process of filter and controller, an Equilibrium Manifold Expansion (EME) model with bounded parameter variation of engine is established. The design conditions of Kalman filter for discrete-time EME system are introduced, and the proposed conditions have a certain suppression effect on the input noise of the system with bounded parameter variation. The engine thrust estimator stability and H∞ filtering problems are solved by the polytopic quadratic Lyapunov function based on the Linear Matrix Inequalities (LMIs). To meet the performance requirements of thrust control, the Grey Wolf Optimization (GWO) algorithm is applied to optimize the PID control parameters. The proposed method is verified on a Hardware-in-Loop (HIL) platform. The simulation results demonstrate that the DTC framework can ensure the stability of engine closed-loop system in large range deviation tests. The filter and controller solution method considering the parameter variation boundary can obtain a solution that makes the system have better performance parameters. Moreover, the proposed filter has better thrust estimation performance than the traditional Kalman filter under the condition of sensor noise. Compared with Augmented Linear Quadratic Regulator (ALQR) controller, the PID controller optimized by GWO has a faster response in simulation.

Fault diagnosis plays a significant role in conducting condition-based maintenance and health management for gas turbines (GTs) to improve reliability and reduce costs. Various diagnosis methods developed by modeling engine systems or certain components implement faults detection and diagnosis based on the measurement of systemic parameters deviations. However, these conventional model-based methods are hindered by limitations of inability to handle the nonlinear nature, measurement uncertainty, fault coupling and other implementing problems. Recently, the development of artificial intelligence algorithms has provided an effective solution to the above problems, triggering broad researches for data-driven fault diagnosis methods with better accuracy, dynamic performance, and universality. This paper presents a systematic review of recently proposed intelligent fault diagnosis methods for GT engines, according to the classification of shallow learning methods, deep learning methods and hybrid intelligent methods. Moreover, the principle of typical algorithms, the evolution of enhanced methods, and the assessment of pros and cons are summarized to conclude the present status and look forward to the future in the field of GT fault diagnosis. Possible directions for development in method validation, information fusion, and interpretability of intelligent diagnosis methods are concluded in the end to provide insightful concepts for scholars in related fields.