Koduah Amoako

Date of Award

December 2025

Document Type

Thesis

Degree Name

Master of Science (MS)

Department

Mechanical Engineering

First Advisor

Djedje-kossu Zahui

Abstract

This thesis presents a comprehensive, non-invasive framework for real-time monitoring of road surface and subgrade conditions using acoustic sensing and machine learning. Traditional Tire-Road Coefficient of Friction (TRCF) estimation techniques often rely on intrusive sensors, offline data analysis, or laboratory measurements, making them impractical for real-time applications. The present research introduces an alternative approach that utilizes the acoustic emissions generated by tire-road interaction to infer both surface friction and subsurface integrity dynamically.Measurement microphones were mounted beneath a test vehicle to record tire-road noise signatures across different pavement textures and vehicle speeds. The collected acoustic signals were processed using Fast Fourier Transform (FFT) to convert them into the frequency domain, followed by Principal Component Analysis (PCA) for dimensionality reduction and feature extraction. PCA was employed to identify dominant frequency components and reveal underlying correlations between surface texture, subgrade stiffness, and vehicle speed. Principal Component (PC) vectors were generated and compared across multiple test locations and speeds to evaluate their sensitivity to varying road conditions. The comparison of PC vectors provided a clear distinction between smooth, intermediate, and rough pavement surfaces. Specifically, PC1 and PC2 captured the dominant energy components associated with tire-road contact dynamics, while higher-order components (PC3-PC5) represented minor variations and noise. The relative magnitude and orientation of PC vectors demonstrated consistent patterns that correlated with road texture classification, confirming PCA’s effectiveness in isolating meaningful spectral features from complex acoustic data. Experimental trials conducted on roads around the University of North Dakota including Washington Street, Cherry Street, and University Avenue, validated the proposed method. The PCA-based results revealed that variations in the principal components were consistent with changes in pavement roughness and subgrade response, with low-frequency components indicating subsurface weakening or moisture accumulation. Overall, this study establishes a scalable, data-driven approach for intelligent infrastructure assessment. By integrating acoustic sensing with unsupervised learning, the framework enables continuous estimation of road surface and subgrade conditions without intrusive testing. The findings have significant implications for real-time traction monitoring, autonomous vehicle safety, and predictive maintenance systems, providing a foundation for the next generation of smart transportation infrastructure.

Download

Share

COinS