Research On Nonlinear Active Noise Control Based On Feedforward And Feedback Hybrid Structure

With the wide application of rotating machinery in daily life,the noise problem caused by it has aroused widespread concern in society,and various noise control technologies have emerged.Among them,active noise control(ANC)technology is the most representative.ANC technology has achieved rich theoretical results and practical experience in linear control field,but the actual noise environment generally has nonlinear problems,such as the reference signal is chaotic noise and the acoustic paths are the nonlinear system.These nonlinear problems severely limit the application of the conventional linear ANC technology,which makes the nonlinear ANC technology attract much more attention.According to the physical structure,the control system can be divided into feedforward control,feedback control and feedforward and feedback hybrid control.The feedforward control structure is stable but not suitable for time-varying and nonlinear objects.The feedback control structure is robust and suitable for time-varying and nonlinear objects,but the system stability and control accuracy are restricted.By combining the advantages of the former two,the feedforward and feedback hybrid control structure has a wider application environment,and higher control accuracy and system stability.Rotating device noise can be classified into narrowband noise,broadband noise,and broadband and narrowband hybrid noise according to their frequency band characteristics.Therefore,the paper intends to study novel and efficient nonlinear ANC methods based on the feedforward and feedback hybrid structure for narrowband noise,broadband noise,and broadband and narrowband hybrid noise,respectively.For the narrowband noise,the narrowband ANC system based on Filtered-Error Least Mean Square algorithm(FE-NANC)is studied,and its dynamic and steady-state characteristics and step-size boundary are analyzed in the mean and mean-squared sense.Under the condition of nonlinear primary path,the mathematical model between reference signal frequencies and primary noise frequencies is established by analyzing the primary noise frequency characteristics.The feedforward frequency expansion and feedback spectrum estimation techniques are used to quickly and accurately obtain the primary noise frequencies as new reference frequencies.Based on the FE-NANC system,the low complexity narrowband nonlinear ANC method is studied.In addition,the de-noising performance and convergence rate of the method are further improved by introducing an error separation technique.For the broadband and narrowband hybrid noise,the frequency characteristics of the primary noise are analyzed firstly,and the de-noising mechanisms of feedforward Function Link Artificial Neural Network(FLANN)structure and feedback Finite Impulse Response(FIR)structure are studied,and their respective de-noising advantages and shortcomings are analyzed in detail.Then ANC method with FLANN-FIR hybrid structure is proposed in this paper.This method is composed of the feedforward FLANN structure and feedback FIR structure,which can not only improve the system convergence rate,but also have a significant de-noising performance for both the linear part and the nonlinear part of the primary noise.For the broadband chaotic noise and random noise,the feedback nonlinear ANC method is studied for noise reduction.The feedback ANC system is the feedback prediction process for primary noise,such that it is not affected by acoustic feedback.However,broadband chaotic noise and random noise cannot be linearly predicted,and the nonlinear prediction accuracy is not good.In order to solve this problem,the wavelet packet is used to decompose the broadband chaotic noise and random noise firstly,and then the linear independent prediction is performed for each decomposed part to build a new feedback nonlinear ANC method.Compared with the traditional feedback nonlinear ANC methods,the proposed method is more effective in reducing the broadband chaotic noise and random noise,especially the broadband random noise.For the acoustic feedback problem in nonlinear ANC,a new nonlinear ANC method using bilinear FLANN(BFLANN)filter is proposed in this paper.This method includes FLANN structure,feedback output term,and the product term of input and output.Because FLANN and the product term of input and output all have nonlinear mapping capabilities,such that BFLANN can handle more complex nonlinear ANC problems than FLANN.Moreover,in BFLANN,the feedback output and the product term of input and output can compensate linear acoustic feedback and nonlinear acoustic feedback respectively,thus it can successfully solve the acoustic feedback problem in nonlinear ANC.In addition,the detailed theoretical derivation fully proves that when the input of BFLANN is bounded,its output is also bounded.In order to verify the effectiveness of the proposed nonlinear ANC method based on the feedforward and feedback hybrid structure in this paper,a closed space ANC experimental platform with a size of 2.2m×1.1m×1.2m is designed.The controller of ANC experimental platform consists of real-time controller dSPACE DS1104 and MATLAB/Simulink software.Through the experimental platform,the proposed methods are performed in different nonlinear acoustic environment.The experimental results fully demonstrate that:(1)The FE-NANC system not only saves computational cost but also reduces the multi-frequency narrowband noise well;(2)The FLANN-FIR method improves the convergence rate and de-noising performance of conventional nonlinear ANC methods;(3)The BFLANN method successfully reduces the effects of linear acoustic feedback and nonlinear acoustic feedback.