Research On Narrowband Active Noise Control Algorithms Based On Momentum Term

Noise pollution has emerged as the third biggest environmental problem only to theair pollution and water pollution in the world. The traditional passive noise control(PNC)techniques can effectively cancel high-frequency acoustic noises. However, they are in-effective in attenuating low-frequency noises. The active noise control(ANC) techniquescan solve this problem. Compared with PNC techniques,ANC systems usually are lessbulk, lighter weight, easy to control, and e?cient at low-frequency noises. In all kindsof low-frequency noise source, the narrowband noises which are generated by rotatingmachines and devices with reciprocating motion, such as engines, cutting machines, fans,etc. become more evident by the spread use of industrial equipments. In ANC researcharea, the narrowband active noise control(NANC) systems can suppress the narrowbandnoises more e?ciently in various engineering and environmental systems. Therefore,the NANC system have become an important research topic in ANC area, and garneredsignificant attention.In practical NANC system application, primary noise always is nostationary, thesecondary path may be variational, and the uncorrelated disturbance noise is present inthe system. For these cases, the control algorithm must be cope with these variations andensure enhanced stability of the system. Meanwhile, the convergence speed, steady-stateerror and tracking ability are also the key characteristics that reflect the system perfor-mance. How to reduce the influence of these factors on the performance of the system,and improve the overrall performance of the system, is the key problem need to be solvedfor the control algorithm of the system. The momentum technique has small amount ofcalculation, parameter boundary determination, and can effectively accelerate the con-vergence and the tracking ability of the system. Therefore, a new research thinking isprovided to solve the above problems by introduing the momentum technique into NANCsystem. This dissertation conside the linear NANC system as the research object, andcarry on research on NANC system based on momentum from the statistical theory andsignal processing point of view. The main contents of this dissertation are as follow:Firstly, the momentum NANC algorithm introduce an additional term and a key pa-rameter into the system. However, the mechanism that the influence of the parameter onthe transient and steady-state behaviors of the system is not clear, and the correspondinganalytical and theoretical explanations are still missing. To solve this problem, the statis-tical performance of NANC system based on such algorithm is investigated in both thetransient and steady-state behaviors. Different equations which describe the convergencebehaviors of the mean and mean squared estimation errors are derived and discussed indetail. Using related difference equations, the steady-state expressions of the system aredeveloped in closed forms. Meanwhile, the influence of introduced parameter on the step-size, steady-state error and steady-state boundary of the system is obtained from theseexpressions. We finally provided theoretical proofs and explanations for the system’s per-formance improvement and the existing problem caused by momentum technique. Therelated conclusions provide a theoretic guide for the improvement of the algorithm, pa-rameter setting and practical application. In addition, some conclusions could providereference for the application of the momentum technique in other fields. Extensive simu-lations are conducted to prove the validity of the analytical findings.Secondly, a novel adaptive filtered-x weighted accumulated least-mean-square al-gorithm termed variable forgetting factor and variable step-size(VFVS-FXWALMS) isproposed to deal with the nostationary primary noise according to the influence of thestep-size and forgetting factor on the performance of the system obtained by the analysis.At first, using the error autocorrelation to construct the the scheme of variable forgettingfactor for solving the downside of the fixed forgetting factor, and enhance the stability ofthe system. Furthermore, the scheme of variable step-size is proposed by using variableforgetting factor, which not only inherit the merits of the scheme of variable forgettingfactor but also further improve the steady-state performance of the system. Thus, thebetter step-size and forgetting factor values are obtained by using the proposed VFVS-FXWALMS algorithm, and improve considerable the overall performance of the system.Extensive simulations are conducted to demonstrate the superior performance of the pro-posed algorithm in various scenarios.Thirdly, this dissertation focus on the NANC system with online secondary pathmodeling(SPM) system based on auxiliary noise. The auxiliary noise not only affect-s the convergence of the NANC subsystem, but also increases the energy of the errornoise. Moreover, the most of auxiliary noise power strategies have all been made for theBANC system, which could not be directly applied to NANC system because the differentstructure between BANC and NANC system. On the other hand, the convergence speed,tracking ability and modeling accuracy of SPM restricts the performance of NANC sys-tem. Aiming at these issues, we proposed a new online SPM method with auxiliary noisepower scheduling strategy based forgetting factor. According to the range of forgettingfactor, the new auxiliary noise power scheduling strategy is designed for the online SPMsubsystem based on the convergence status of forgetting factor which could accuratelytrack the change of secondary path. Consequently, the proposed method not only obtainsgood modeling accuracy but also increases noise attenuation considerably. Extensive sim-ulations are conducted to demonstrate that the proposed system can effectively improvesthe overall performance of the system in various scenarios.Fourthly, the performance of the conventional NANC system may degrade severelywhen the uncorrelated disturbance noise is present in the system. The hybrid structurecontrol system based on the traditional thinking use common error to update system,which caused the mutual interference between the subsystems, and reduced the conver-gence speed of the system. For the above problem, we proposed a new hybrid structureNANC system on the basis of the traditional hybrid structure control system. In the pro-posed system, a sunusoidal noise cancellation(SNC) subsystem is designed to solve theproblem caused by the common error, and provide suitable error signals for the other sub-systems. Moreover, an optimization strategy is used to further improve the convergenceof the system. Extensive simulations are conducted to demonstrate the convergence of theproposed system is more superior to the traditional hybrid structure control system.