基于多目标参考信号优选的车辆路噪主动控制

在车辆路噪主动控制(Road Noise Cancellation, RNC)过程中，参考信号的选择会直接影响系统的降噪效果。针对传统参考信号选择方法中存在的目标区域单一、参考信号间独立性不足等问题，提出一种改进的多目标参考信号优选方法。首先推导参考信号选择所依据的相干性和条件数评价指标，使用折衷规划法将上述指标构造成综合目标函数，并通过准则间相关性方法确定各项的权重系数；为降低运算量，利用遗传算法建立以该综合目标函数作为适应度函数的多目标参考信号寻优模型，并通过该模型获得一组优选的参考信号，从相干性分析及条件数计算结果易知，优选的参考信号组合在100～500 Hz频段内与目标噪声的相干性较好，且彼此独立；为验证该方法的有效性，以某款燃油车为研究对象，采用多通道NFxLMS算法对该组参考信号进行RNC仿真和实车道路验证。研究结果表明：在降噪效果方面，通过传统多重相干法选择的参考信号在100～500 Hz频带内的平均降噪量为4 dB(A),而通过多目标方法优选出的参考信号在100～500 Hz频带内的平均降噪量达到6 dB(A),RNC系统的控制效果得到增强；在计算效率方面，所提方法的计...

Road Noise Cancellation Based on Multi-objective Optimization of Reference Signals

When road noise cancellation (RNC) on a system is processed, the selection of reference signals directly affects the noise reduction effect. However, the traditional method used to select reference signals is implemented in a single target area and insufficient independence exists among the signals. To address these issues, an improved multi-target reference signal selection method is proposed here. First, the coherence and condition number for reference signal selection were derived and constructed into a comprehensive objective function by using the compromise planning method. The weight coefficients of each item were determined by using the criteria importance through intercriteria correlation method. To reduce computational time, a multi-target reference signal optimization model, applying the comprehensive objective function as a fitness function, was established by using a genetic algorithm. A set of excellent reference signals was obtained from this model. Results from the coherence analysis and condition number calculation show that the chosen reference signals have good coherence with the target noise in the 100-500 Hz frequency band and are independent of each other. To verify the effectiveness of the proposed method, RNC simulations and vehicle experiments were conducted with multi-channel NFxLMS algorithm by employing a fuel vehicle as the research object. The results show that the average noise reduction of the reference signals increases from 4 dB(A), when selected by using the traditional multiple coherence method, to 6 dB(A), when selected by using the proposed multi-objective method, in the 100-500 Hz band, thus enhancing the control effect of the RNC system. Moreover, the computational efficiency of the proposed method is high, with its volume being 0.01% of that of the traditional multiple coherence method. Thus, this method significantly reduces the computational burden and shortens the time required for enhanced reference signal selection when compared with the traditional method.