Research On Optimal Placement Of Secondary Sound Sources And Error Points Based On Neural Network

With the development of marine equipment,the acoustic stealth characteristics of underwater vehicles have been widely concerned.For large-scale underwater vehicles,the radiated noise generated by mechanical vibration has low-frequency line spectrum characteristics,and it is difficult to eliminate this noise by traditional passive control methods.As a control method for low-frequency noise,active noise control has been paid more and more attention.For the underwater active noise control system,it is obviously unrealistic to achieve a better noise reduction effect by deploying a large number of secondary sound sources and error sensors due to the limited space resources in practical engineering applications,so the placement position of secondary sound sources and error sensors determines the actual control effect.Based on the advantage of neural networks in dealing with nonlinear complex mapping problems,an optimal arrangement method of secondary sound sources and error sensors for underwater complex elastic structure models is proposed.Firstly,the dual-channel active control system for pulsating ball source in free field is studied,and the optimization results obtained by analytical method and neural network prediction are compared and analyzed.Secondly,the numerical simulation of active noise control is carried out for the cylindrical shell model,and the neural network model is trained by the numerical simulation results.Then,the active control experiment is carried out in the anechoic pool by using c RIO embedded controller,and the prediction results of the neural network are compared with the measured results of the pool experiment,which verifies the reliability of the prediction results of the neural network model.Thirdly,Then,combined with the underwater double-layer cylindrical shell model,the reliability of the neural network-based optimization method is verified by numerical simulation,and the optimal placement results of secondary sound source and sensor are given.Finally,the vibration radiation sound field of underwater double-layer stiffened cylindrical shell is tested on the outfield lake,and the off-line control simulation analysis is carried out by using the measured data.The neural network model is trained according to the sample data obtained by off-line control,and then the optimized secondary sound source and error sensor placement scheme are given.Numerical simulation and experimental results show that neural network can solve the problem of optimal placement of secondary sound sources and error sensors for complex models,and different optimal placement schemes can be given by selecting different noise reduction criteria,which is of guiding significance to engineering practice.