【作者】 张玉璘；

【导师】 黄尚廉；

【作者基本信息】 重庆大学 ， 仪器科学与技术， 2002， 博士

【摘要】 随着现代社会的发展，噪声控制问题日益引起了人们的高度重视和极大关注。无论在民用领域还是军事领域，噪声不但污染环境，损害健康，致使设备疲劳损伤，而且可能危及生命，噪声控制已经成为了一个迫切需要解决的世界性难题，世界各国都投入了大量的人力和物力对此展开了深入的研究。就噪声主动控制系统而言，目前普遍使用的是自适应控制策略，而且已经取得了丰硕的成果。但自适应控制尚不完善，其中的许多算法存在着明显的不足和特定性。因此运用新的信息分析手段对控制算法进行研究是一个非常有潜力的、具有重要意义的工作。针对这一问题，本文选择了以控制算法和控制方法为切入点，从理论上和实验上进行了研究和探讨。从总体上讲，本文的主要研究工作如下：1.本文首先分析了几种常用的最小均方(Least Mean Square, LMS)算法，利用计算机仿真对每一种算法的性能进行比较，总结了影响LMS算法的重要因素，这些都为算法的改进工作提供了重要依据。2. 本文把小波分析理论和LMS算法结合起来提出了两种结构形式的小波自适应算法，即：分解LMS算法（Decomposition LMS, D-LMS）和分解重构LMS算法（Decomposition and Reconstruction LMS, DR-LMS），对每一种算法进行了深入的理论分析。为了提高小波自适应算法的实时性，推导出了MALLAT算法的一种高效的实现方法。为了从仿真的角度考察小波自适应算法的性能，建立了系统辨识的仿真模型，获得了分别采用小波自适应算法和LMS算法的系统辨识结果。结果表明，与LMS算法相比，小波自适应算法的性能得到了很大的提高，使得控制系统的设计具备了强有力的工具。3. 根据Filtered-X LMS算法的思想和随机自适应控制理论，建立了小波自适应控制系统。根据D-LMS算法和DR-LMS算法，本文设计了两种形式的小波自适应控制器，在解决小波自适应控制的系统辨识问题时，引入了并行在线系统辨识方法，把D-LMS算法和DR-LMS算法应用了到该辨识方法中，提高了系统辨识的精度。4. 在以上研究的基础上，把小波自适应控制系统应用到基于智能结构思想的噪声主动控制中，完成了小波自适应控制系统的硬件和软件设计，建立了实验系统，并进行了大量的实时控制实验。实验结果表明，无论对<WP=5>于频率简单的噪声信号还是频率成分复杂的噪声信号，小波自适应控制系统都能对其进行有效的抑制，从而从实验上证明了小波理论的引入改善了自适应算法的性能，提高了自适应控制系统的稳定性和适应性。更多还原

【Abstract】 With the development of society, the topic of noise control has been received considerable attention. Either in civil or military domain, noise is polluting the environment, doing harm to our health. Even our lives are badly endangered, so the noise control is becoming important and urgent. The works in this domain have been widely investigated all over the world.As far as the active noise control system, the adaptive control strategy is generally adopted, and that the results are comparatively satisfying. However, the adaptive control also is not perfect in many aspects, for instance, the deficiencies in many adaptive algorithms are extraordinarily distinct. Consequently it is promising and significant to improve the performance of adaptive algorithm by means of new methods. Aimed at this matter, the control algorithm and control way are investigated in this paper.On the whole, the work in the paper is made up of the following part:1. Firstly, the several LMS (Least Mean Square, LMS) algorithms are analyzed and their performance is compared by the simulation results, which offer gist for the improvement of algorithm.2. Combining the Wavelets and LMS algorithm, the paper brings forward two classes of Wavelets Adaptive Algorithm (WAF) with different structure, viz., D-LMS algorithm and DR-LMS algorithm. Theoretically investigation on each algorithm is made in details. In order to achieve the real-time of WAF, the fast use of WALLAT algorithm is brought out in the paper. To study on the performance of WAF, the simulations of system identification with D-LMS algorithm and DR-LMS algorithm are made. The simulations indicate that the performance of WAF is greatly improved, which offer the powerful and efficient method for design of control system.Based on Filtered-X LMS algorithm and random adaptive control, the Wavelet Adaptive Control System (WACS) is constructed in the paper. The wavelet adaptive controller is developed; the adaptive equation of the parameter matrix of each controller is presented in details. Because, in adaptive control, the precision and real-time of system identification can have direct effect on control quality, the method of system identification is<WP=7>3. vital. In the paper, the parallel on-line system identification is applied in control system, which effectively solves the problem in system identification.4. Finally, active noise control with the wavelet adaptive control system (WACS) is explored. Corresponding hardware and software of wavelet adaptive controller are developed. The real-time experiment with WACS shows that noise with simple or complex frequency components can be greatly and rapidly reduced, which indicates that introducing Wavelet to LMS algorithm has improved the performance of LMS and WACS.更多还原