Semester of Graduation

Summer 2026

Degree

Master of Science (MS)

Department

Department of Civil and Environmental Engineering

Document Type

Thesis

Abstract

Longitudinal rumble strips (RS) are low-cost safety countermeasures that alert drivers through tire-pavement sound and vibration when vehicles drift from the travel lane. This can help in reducing some types of traffic crashes such as lane-departure and run-off-road crashes. However, RS can increase exterior noise and cause complaints from residents. Previous RS noise studies have generally evaluated limited designs and configurations, relied mainly on A-weighted sound levels, and provided limited assessment of interior alerting, vibration, low-frequency annoyance metrics, complaint locations’ performance, and traffic noise predictive modeling. This thesis therefore develops and evaluates a comprehensive field-based framework for assessing and predicting RS acoustic and vibration performance under varying geometric, operational, and site conditions.

This thesis comprehensively evaluates and predicts RS noise and vibration using controlled pass-by measurements from 26 RS installations across 18 Louisiana sites. The tested designs included sinusoidal, raised, and cylindrical milled profiles installed as continuous shoulder RS, shoulder RS with bicycle gaps, and centerline RS. Exterior noise was measured at 25 ft and 50 ft following AASHTO TP-98, interior noise following SAE J1477, and vibration using ISO 2631-1 frequency weightings. Two test vehicles were operated at 45, 55, and 65 mph, resulting in approximately 792 valid runs. Acoustic descriptors included LeqA, LAFmax, SEL, one-third-octave-band spectra, and the low-frequency-corrected sound pressure level LA′. The dataset was used to develop linear mixed-effects models, threshold-compliance models, and machine learning models with leakage-safe grouped validation. SHapley Additive exPlanations (SHAP) were used to interpret selected Random Forest models.

The results indicated that the tested sinusoidal RS produced the lowest exterior above-baseline noise increments, remained below the 10-dBA exterior threshold after LA′ correction, and maintained sufficient interior noise and vibration for driver alerting. In contrast, the complaint-location cylindrical installation exceeded the 10-dBA above-baseline threshold at 480-ft. Statistical modeling identified RS design family and vehicle speed as the dominant predictors of exterior noise increments, with cylindrical geometries producing substantially greater noise impacts than sinusoidal profiles. Relative to sinusoidal profiles, cylindrical RS increased ΔLeqA by 7.09 dBA, while within cylindrical installations, each additional 0.10 in. of vertical depth increased ΔLeqA by approximately 1.94 dBA. Random Forest best predicted exterior ΔLeqA (out-of-fold R² 0.605, RMSE 2.78-dBA), while absolute ON-RS and combined ON/OFF models reached R² of 0.718 and 0.842.

These findings support the development of noise-conscious RS design practices that balance roadway safety and community noise impacts. Sinusoidal profiles similar to those evaluated in this study can provide effective driver alerting while minimizing exterior noise increases, whereas deep cylindrical designs should be avoided near noise-sensitive receptors. The proposed mechanistic-statistical framework can support evidence-based RS selection, complaint-risk screening, and design-stage evaluation while maintaining interior alerting performance consistent with driver-warning requirements.

Date

7-20-2026

Committee Chair

Hassan, Hany

LSU Acknowledgement

1

LSU Accessibility Acknowledgment

1

Download

Available for download on Friday, July 20, 2029

Share

COinS