Experimental evaluation of constrained model predictive control for active noise control in ducts

This paper presents an experimental evaluation of a constrained Model Predictive Control (MPC) strategy for tonal Active Noise Control (ANC) in acoustic ducts. The controller is implemented in real time using Simulink® at a sampling frequency of 2.4 kHz, explicitly enforcing constraints on the control signal, its rate of change, and the anti-noise amplitude. Experimental comparisons with classical adaptive controllers, including FxLMS and MOV-FxLMS, show that while adaptive methods can achieve comparable steady-state attenuation when unconstrained, the MPC approach provides faster convergence, smoother transients, and reliable operation under actuator limitations. Time- and frequency-domain analyses confirm that the constrained MPC avoids windup effects and preserves spectral purity under bounded operation. The results demonstrate that constrained MPC is a practical and computationally feasible solution for real-time ANC applications subject to physical constraints.