Research on the optimal selection method of sensors/actuators in active structural acoustic control for helicopter based on machine learning

• An active structural acoustic control strategy based on radiation panel elements is proposed. • An intelligent sensors/actuators selection algorithm based on machine learning is proposed. • A simulation study is conducted using a helicopter cabin model as the object. • The method has the advantages of good control effects and low computing cost. • The method provides application potential in active structural acoustic control. The reasonable placement of sensors/actuators is crucial in helicopter cabin noise control technology based on smart active panels. Addressing the issues of inadequate noise reduction in structural modal control and the high computational complexity in acoustic radiation modal control, this paper proposes a control strategy based on radiation panel elements, aiming to identify and prioritize the control of the dominant acoustic radiation panel elements contributing significantly to cabin noise. To achieve this goal, in combination with machine learning technology, an intelligent optimal selection algorithm is proposed. The algorithm initially establishes a multi-line spectrum complex neural network model to accurately fit the relationship between panel vibration and radiation noise, thereby identifying the dominant acoustic radiation panel elements to guide the optimal placement of sensors. Building on this foundation, genetic algorithms are employed to optimize the selection of actuators. Ultimately, an efficient and collaborative sensors/actuators placement scheme is formed. To validate the feasibility and effectiveness of the algorithm, a simulation study is conducted using a helicopter cabin model as the object. The results indicate that the optimized sensors/actuators scheme can achieve good control effects in the active control process. For 2 line spectra, the average sound pressure levels at the monitoring points are reduced by 18.0 dB and 22.3 dB, respectively. Meanwhile, the scheme exhibits notable effects on global vibration and noise reduction. The simulation results also demonstrate that under the same control scale, the algorithm yields significantly better control effects than the empirical selection; compared to the traditional genetic algorithm, the algorithm ensures the control effects while reducing the calculation time by 50 %, significantly improving the calculation efficiency.