Predição do índice de irregularidade longitudinal com acelerações suspensas: integração de sensores de smartphones, simulações e inteligência artificial

Abstract

This study presents an innovative approach to predicting the International Roughness Index (IRI) using acceleration data collected by smartphone sensors mounted on vehicles. The research integrates computational simulations, real and hypothetical vehicle models, and practical analyses on tangible and non-tangible pavement profiles. A quantitative and exploratory methodology was employed, combining field data collection, compliance with ISO 2631-1 comfort standards, and the application of artificial intelligence models such as neural networks and advanced regressions. The results demonstrate that smartphones, when properly calibrated with reference equipment, can serve as effective response-type instruments for evaluating pavement irregularities. Artificial intelligence models showed high performance in IRI prediction, achieving correlation factors ranging from 77% to 99%. Additionally, comparisons between real and simulated vehicles revealed statistically significant differences in measured accelerations, highlighting the influence of vehicle dynamics on analysis outcomes. This study offers a cost-effective and accessible solution for monitoring road infrastructure, enhancing resource allocation and maintenance strategies.