The Prediction and Analysis of Acoustic Metamaterial Based on Machine Learning

Abstract

Low-frequency noise is regarded as special environmental noise problem which has a serious impact on people’s lives. The traditional technology cannot effectively insulate the low-frequency noise, while the emergence of acoustic metamaterial makes it come true. This paper studies the acoustic insulation of membrane-type acoustic metamaterial. Due to different weights of the factor, an analytical method based on analytic hierarchy process (AHP) is proposed to analyze the affecting factors of acoustic insulation performance. The orthogonal array results show that the main factors affecting acoustic insulation are membrane thickness, membrane preload force and attached mass. A factor-weighted k-Nearest Neighbor (kNN) classification approach is proposed to predict different levels of acoustic insulation, which also provides a reference for the analysis of acoustic insulation. The experimental results demonstrate that when k = 3, the maximum classification accuracy of acoustic insulation is 98.2% by using AHP-kNN approach, which makes the accuracy for acoustic insulation is higher than other three baselines.