RSAS-FxLMS: A Robust Active Noise Control Method for Vibrator Operation Noise in Precast Beam Plants
Year: 2026
Type: article
Abstract: To improve active noise control of 120–180 Hz vibrator noise in precast beam plants under strong impulsive disturbances, this paper proposes a robust scaled adaptive step-size filtered-x least mean square (RSAS-FxLMS) algorithm. The conventional scaled adaptive step-size filtered-x least mean square (SAS-FxLMS) algorithm is prone to weight instability and divergence under strong impacts. The proposed method embeds a dual stabilization mechanism: an adaptive divergence detector based on short-term error energy triggers weight vector reset and step-size rollback, while a norm constraint is imposed on the weight vector after each iteration. The signal preprocessing chain is optimized using a high-order bandpass filter and an offline secondary path model identified with swept-frequency excitation and a weighted-error criterion. Simulations using field-measured vibrator noise show that the proposed algorithm achieves a noise reduction of 33.3 dB at 150 Hz, which is 27.63 dB higher than that of the conventional SAS-FxLMS algorithm, and an average noise reduction of 29.10 dB across the 120–180 Hz target band, which is 20.91 dB higher than that of the conventional SAS-FxLMS algorithm. Under continuous high-intensity impacts, the weight vector remains bounded, and the error signal re-converges quickly without divergence. The algorithm also achieves faster convergence in more than 80% of the convergence period. These results demonstrate that the proposed method effectively suppresses strong impulsive narrowband vibrator noise with high reduction depth and strong robustness. The evaluation uses field-recorded vibrator noise as the input, whereas the primary and secondary acoustic paths are modelled; it is therefore an open-loop, offline assessment and does not constitute a closed-loop or real-time experimental validation of an ANC system, so that the results provide an algorithmic reference for industrial ANC rather than a demonstration of field performance.
Source: Applied Sciences
DOI: 10.3390/app162211165
Institution: Hubei University of Technology
Language: English
Subfield: Signal Processing
Field: Computer Science
Domain: Physical Sciences
Open Access status: gold
Estimated APC paid: $2,896
OpenAlex ID: W7225800925