Real-time prediction of noise signals for active control based on Bayesian forecasting and time series analysis

Abstract

Rapid urbanization has intensified environmental noise pollution due to hectic construction activity. Generally, conventional noise barriers/enclosures can provide passive noise control, but do little to mitigate low-frequency noise, as the acoustic wavelength is significantly longer than the dimension of barrier-type structures, leading to inefficient coupling. In this study, we consider a smart barrier/enclosure featuring a combination of both passive and active noise control functions for dealing with construction noise, in which the passive approach is regarded as a low-pass filter to remove the high-frequency components of noise signals. The filtered signals are mainly focused on a relatively narrow bandwidth that can be treated by active noise control. An active control technique based on the superposition principle is a promising alternative approach, in which anti-noise speakers emit antiphase acoustic waves to reduce undesirable noise. The effectiveness of active control systems is mainly governed by adaptive algorithms. However, a time delay issue is still inevitable for most adaptive systems. To reduce the time delay and enhance the accuracy of adaptive systems, real-time estimation of noise signals based on a Bayesian inference-based dynamic linear model is proposed in this work. Bayesian forecasting is a learning approach that enables the computation of conditional probabilities, and a dynamic linear model is a mathematical tool for times series analysis. On-site measured data were used as observation values for time series analysis with the dynamic linear model. When the Bayesian forecasting approach is combined with times series analysis, prior information and measurement data are naturally integrated for better condition identification and assessment. Based on the Bayesian statistical framework, a “forecast-observation-analysis” cycle can be formulated to mitigate time-delay issues in adaptive noise control.