Neural network based multi-objective active vibration optimization method for shell structure

Active control method mainly focuses on vibration and noise suspension at present thus can't satisfy the requirement of frequency characteristics control. Therefore, based on the multi-objective parallel processing ability of neural network, the multi-objective vibration optimization method is proposed to deal with this problem. First, the frequency-domain control frame is constructed based on neural network algorithm. Compared with traditional time-domain methods, the proposed control frame just require once FFT in each iteration and no IFFT needed, so the control efficiency can be guaranteed. Second, hybrid error criterion is constructed by combining global frequency error and frequency node error together to improve the adaptability, reliability and anti-interference ability. Third, the controllability problem of the multi-objective method in implementation is studied through mathematical analysis. At last, the effectiveness of the proposed multi-objective method is verified through vibration optimization on eight points of shell structure.