Koopman-Based Nonlinear Identification and Adaptive Control of a Turbofan Engine
arXiv:2604.01730v1 Announce Type: new Abstract: This paper investigates Koopman operator-based approaches for multivariable control of a two-spool turbofan engine. A physics-based component-level model is developed to generate training data and validate the controllers. A meta-heuristic extended dynamic mode decomposition is developed, with a cost function designed to accurately capture both spool-speed dynamics and the engine pressure ratio (EPR), enabling the construction of a single Koopman model suitable for multiple control objectives. Using the identified time-varying Koopman model, two controllers are developed: an adaptive Koopman-based model predictive controller (AKMPC) with a disturbance observer and a Koopman-based feedback linearization controller (K-FBLC), which serves as a benchmark. The controllers are evaluated for two control strategies, namely configurations of spool speeds and EPR, under both sea-level and varying flight conditions. The results demonstrate that the
arXiv:2604.01730v1 Announce Type: new Abstract: This paper investigates Koopman operator-based approaches for multivariable control of a two-spool turbofan engine. A physics-based component-level model is developed to generate training data and validate the controllers. A meta-heuristic extended dynamic mode decomposition is developed, with a cost function designed to accurately capture both spool-speed dynamics and the engine pressure ratio (EPR), enabling the construction of a single Koopman model suitable for multiple control objectives. Using the identified time-varying Koopman model, two controllers are developed: an adaptive Koopman-based model predictive controller (AKMPC) with a disturbance observer and a Koopman-based feedback linearization controller (K-FBLC), which serves as a benchmark. The controllers are evaluated for two control strategies, namely configurations of spool speeds and EPR, under both sea-level and varying flight conditions. The results demonstrate that the proposed identification approach enables accurate predictions of both spool speeds and EPR, allowing the Koopman model to be reused flexibly across different control formulations. While both control strategies achieve comparable performance in steady conditions, the AKMPC exhibits superior robustness compared with the K-FBLC under varying flight conditions due to its ability to compensate for model mismatch. Moreover, the EPR control strategy improves the thrust response. The study highlights the applicability of Koopman-based control and demonstrates the advantages of the AKMPC-based framework for robust turbofan engine control.
Executive Summary
This article presents a novel application of Koopman operator theory to the multivariable control of a two-spool turbofan engine. The study integrates a physics-based model with an extended dynamic mode decomposition algorithm tailored to capture spool-speed and engine pressure ratio dynamics, enabling the construction of a unified Koopman model for diverse control objectives. Two adaptive control frameworks are developed: an AKMPC incorporating a disturbance observer and a K-FBLC as a benchmark. Evaluated under sea-level and varying flight conditions, the results indicate that the Koopman-based identification allows for flexible model reuse and accurate prediction. While both controllers perform similarly in steady states, the AKMPC demonstrates greater robustness under dynamic conditions due to its disturbance-compensating capabilities. The EPR control strategy further enhances thrust response performance, showcasing the viability of Koopman-based methodologies in aerospace control applications.
Key Points
- ▸ Development of a physics-based model for training data generation
- ▸ Use of extended dynamic mode decomposition tailored for spool-speed and EPR dynamics
- ▸ Creation of two adaptive control frameworks—AKMPC and K-FBLC—for comparative evaluation
Merits
Innovative Application of Koopman Theory
The integration of Koopman operator concepts with aerospace control systems represents a significant theoretical advancement, offering a unified predictive framework for multivariable dynamics.
Flexible Model Reusability
The ability to apply a single Koopman model across multiple control formulations enhances modularity and reduces development overhead.
Robustness Validation Under Realistic Conditions
Evaluation under varying flight conditions validates the practical applicability and resilience of the proposed control strategies.
Demerits
Model Assumption Dependency
The effectiveness of the Koopman model may be limited by the accuracy of the underlying physics-based assumptions, particularly under extreme or unmodeled conditions.
Computational Complexity
Extended dynamic mode decomposition and real-time adaptation of Koopman models may impose higher computational burdens compared to conventional control methods.
Expert Commentary
The article represents a compelling synthesis of modern control theory and practical aerospace engineering. The use of Koopman operator theory to unify disparate control objectives into a single predictive model is both elegant and pragmatic. The distinction between AKMPC and K-FBLC—particularly the AKMPC’s enhanced robustness due to disturbance compensation—demonstrates a nuanced understanding of real-world operational variability. Moreover, the fact that EPR control improves thrust response underscores the value of targeting specific performance metrics in engine design. While the computational demands remain a practical hurdle, the methodological rigor and empirical validation elevate this work beyond theoretical speculation into the realm of viable engineering solutions. This is a significant contribution to the field of adaptive aerospace control.
Recommendations
- ✓ Extend the study to include validation on actual hardware or hybrid simulation-hardware platforms to further substantiate real-time applicability.
- ✓ Explore the scalability of the Koopman framework to other propulsion systems or multi-domain aerospace platforms.
Sources
Original: arXiv - cs.LG
