一种改善鲁棒性的噪声有源控制自适应神经网络方法  

A method of active noise control using adaptive neural networks with improved robustness

作　　者：曾成[1] 张奇志 伍萍辉[2] 

机构地区：[1]北京机械工业学院,北京100085 [2]湖南工程学院

出　　处：《声学学报》2003年第1期79-85,共7页Acta Acustica

基　　金：北京市教委资助科技项目(99KJ44)

摘　　要：控制对象参数的时变是噪声有源控制付诸实际应用所面临的主要问题之,传统的控制方法通常不考虑对象参数时变。本文首先引入一个能方便进行在线自适应的扩展控制对象自适应神经网络模型,在此基础上提出一种噪声有源控制的自适应神经网络方法。通过在控制过程中分别对控制网络和模型网络进行自适应,解决了控制对象参数的时变问题,显著改善了整个系统的鲁棒性。实验结果表明,对于控制对象参数的突变扰动,该方法具有良好的鲁棒稳定性。The time-variance of plant parameters, which is not considered usually by traditional control methods, is one of the main problems in the application of active noise control. This paper firstly introduces an extended plant adaptive neural network model which can be on-line adapted conveniently, then presents a method of active noise control using adaptive neural networks. With separately adapting of the control network and the model network in control, the problem of time-variance of plant parameters is solved, and the robustness of whole system is improved obviously. It is showed by experimental results that this method has a good robust stability for abrupt disturbances of the plant.

关 键 词：鲁棒性 噪声有源控制 自适应神经网络 控制网络 模型网络 控制对象参数 时变问题 ANC 

分 类 号：TP273.22[自动化与计算机技术—检测技术与自动化装置]