神经网络分布式多通道协同控制声屏障原理研究  

Basic Study on Neural Active Noise Barrier of Distributed & Coordinated Multi Channels.

作　　者：谭成翔[1] 赵松龄[1] 

机构地区：[1]同济大学声学研究所

出　　处：《环境科学》1997年第3期47-49,共3页Environmental Science

基　　金：中国博士后科学基金资助项目

摘　　要：声屏障的高度、材料、结构、重量和风载荷等要素受到景观要求和承载结构等条件的制约，进一步劣化了常规声屏障在低频宽带噪声衍射和透射控制方面的缺陷．本文为此提出神经网络分布式多通道智能化协同控制声屏障方法．神经网络自组织形成时空二维非线性控制矢量，以形成优化的分布式消声器阵列．采用多目标优化．神经网络通过自学习．首先优化各传感器和次声源的空间配置，以形成适当的消声空间分布并降低硬件开销；其次简化控制器结构；最后精细优化各联结权．研究表明，该方法对宽带。The height,material,architecture,weight of the noise barrier are constrained by the landscape requirement and load bearing frame,which deepens the deficiency of the conventional noise barrier in low frequency wide band noise diffraction and transmission. To improve the performance of the conventional noise barrier,neural active noise barrier of distributed and intelligently coordinated multichannels is presented. The adaptive nonlinear control vector in time and space domains is self organized by the proposed compound neural network to drive the distributed secondary sources array optimized. Multiobjectives optimization is adopted. Through training, at first,the neural network optimizes the geometrical distribution of the sensors and secondary sources, in order to synthesize the suitable noise reducing space distribution and reduce the hardware cost; secondly, the structure of the neural controller is simplified. At last, the reserved connection weights get precisely optimized. The scheme is demonstrated to be applicable especially for wideband and large area noise barrier.

关 键 词：噪声控制 声屏障 神经网络 

分 类 号：TB535[理学—物理]