Hybrid Optimisation of the Sensing Matrix for a Multi-Sensor Active-Passive Sound Absorption System
일시11/27 금 · 10:00~10:40
발표장컨퍼런스홀
- 저자
- 신유진, Jordan Cheer, Felix Langfeldt(사우스햄튼대)
- 소속
- 사우스햄튼대
- 좌장
- 김성현(현대차), 박춘수(표준과학연), 박현우(동아대), 손기성(세안기술)
키워드
확장요약문
A multi-sensor feedback control framework for active-passive sound absorption has been established to achieve near-perfect absorption over a broad frequency range. The system employs seven feedback sensors, whose signals are combined through a sensing matrix with 21 weighting coefficients for the PID controller. Since the proposed framework contains a large number of weighting coefficients, manually tuning these coefficients is impractical.
An optimisation problem was defined to maximise sound absorption, with constraints imposed to maintain feedback system stability. Three optimisation strategies were investigated: a genetic algorithm (GA), sequential quadratic programming (SQP), and a hybrid strategy. In the hybrid approach, GA is first employed for global exploration, followed by SQP for local refinement. Their characteristics are summarised in Table 1.
The numerical results in Fig. 1 show that all three strategies achieved high broadband absorption, with mean absorption coefficients of 0.98, 1.00, and 1.00 for GA, SQP, and the hybrid strategy, respectively. Both SQP and the hybrid strategy achieved near-perfect broadband absorption, while the hybrid strategy required only 10 min compared with 112 min for GA and 46 min for SQP. Overall, the hybrid strategy achieved comparable absorption performance to SQP with reduced computational time, demonstrating its effectiveness for optimising the multi-sensor feedback control system.