Frequency-domain fuzzy-FxLMS algorithm for active sound profiling in hybrid powertrains in various speeds

Abstract

This paper proposes a Fuzzy-FxLMS algorithm for improving sound quality (SQ) across varying speeds in hybrid electric powertrains. The algorithm integrates a frequency-domain FxLMS framework with a fuzzy system, forming the core of the Active Sound Profiling (ASP) system. ASP utilizes this combined approach to modulate the amplitude and relative phase of primary noise, achieving SQ improvement. First, primary noise signals are acquired from the hybrid electric powertrain at specific speeds. A sensitivity analysis identifies key design frequencies impacting SQ attributes like Loudness, Sharpness, Roughness, and Tonality. For these frequencies, a multi-objective optimization determines the optimal amplitude and relative phase of the design frequencies. For speeds between the optimized ones, the fuzzy system directly obtains the desired amplitude and relative phase, ensuring low computational cost and control over sound characteristics. This Fuzzy-FxLMS approach, implemented in the frequency domain, achieves SQ optimization across varying speeds. Results demonstrate improvements in Loudness and Roughness, even for speeds between those with optimized profiles. This work paves the way for enhanced sound perception in non-stationary noises, including coast-up/down scenarios in hybrid electric vehicles, while maintaining low computational cost.