基于声场分解和稀疏正则化的二维空间次级声源布局优化

Optimization of Secondary Sources Configuration in Two-Dimensional Space Based on Sound Field Decomposition and Sparsity-Inducing Regularization

在有源噪声控制系统电声器件布局设计中,目前的优化方法需要预先确定误差传感器的布局,这样大大增加了后续次级声源布局优化的工作量。利用高阶传声器拾取声场信息,在波域构造有源控制代价函数,从而解除了对误差传感器布局信息的需求。在此基础上,根据初级声场的稀疏特性,引入稀疏正则化方法,通过求解欠定方程的稀疏近似解,实现了次级声源布局(个数和空间位置)优化。讨论了备选次级声源个数对优化结果的影响,并与均匀布局和遗传算法优化布局结果进行了比较。结果表明,所提方法在不依赖于误差传感器布局信息的情况下可以有效地优化次级声源布局,降噪效果与遗传算法布局结果相近,其他系统评价指标明显更优,有利于有源系统的稳定运行。

During the design of transducers configuration for an active noise control system, current optimization methods need to predetermine the error sensors configuration, which significantly increases the workload of later optimization of the secondary sources configuration. In this study, a new method free from specific error sensors configuration information is presented that higher order microphones are used to capture the sound field so as to formulate the cost function in wave domain. In addition, according to sparsity characteristics of the primary sound field, sparsity-inducing regularization is introduced to optimize the secondary sources configuration, including the number and positions, by calculating a sparse approximate solution to underdetermined equations. Effects of the number of candidate secondary sources are discussed, and the comparison with the uniform configuration and the optimized configuration using the genetic algorithm is performed. Results show that the proposed method can optimize the secondary sources configuration effectively independent of the error sensors configuration information. The noise reduction of the proposed method is close to that by the genetic algorithm, while other evaluation metrics for the system are much better, which would benefit the stability of active noise control system.