Control of Powertrain Noise Using a Frequency Domain Filtered-x LMS Algorithm

An enhanced, frequency domain filtered-x least mean square (LMS) algorithm is proposed as the basis for an active control system for treating powertrain noise. There are primarily three advantages of this approach: (i) saving of computing time especially for long controller’s filter length; (ii) more accurate estimation of the gradient due to the sample averaging of the whole data block; and (iii) capacity for rapid convergence when the adaptation parameter is correctly adjusted for each frequency bin. Unlike traditional active noise control techniques for suppressing response, the proposed frequency domain FXLMS algorithm is targeted at tuning vehicle interior response in order to achieve a desirable sound quality. The proposed control algorithm is studied numerically by applying the analysis to treat vehicle interior noise represented by either measured or predicted cavity acoustic transfer functions. The simulation results show that the proposed algorithm, by increasing block size and utilizing proper step size, can yield a more precise targeted reduction and at the same time suppress unintended overshoot compared to the traditional time domain algorithm.