Scholarlink

The mechanical characterization of asphalt pavements is carried out by means of performance parameters that are representative of the asphalt mixtures’ behavior when subjected to different external conditions. In this regard, one of the fundamental parameters frequently used during the design processes of flexible pavements is known as stiffness modulus and, in this study, it was obtained by means of 4-point bending tests, carried out at the Department of Road Structures of the Czech Technical University in Prague. The experimental campaign was performed by testing samples of 2 distinct mixtures at 4 different temperatures (0, 10, 20, and 30 ℃) and at 11 different frequencies (0.1, 1, 2, 3, 5, 8, 10, 15, 20, 30, and 50 Hz) so that the mechanical behavior could be deeply investigated for wide variations in external conditions. Laboratory results then allowed data-driven methodologies to be developed in order to model this phenomenon and predict the stiffness modulus by means of different soft-computing techniques, namely support vector machines and artificial neural networks. Both algorithms, properly developed and optimized, provided satisfactory results. However, artificial neural networks, by virtue of a slightly higher complexity, provided more accurate predictions and a higher reliability, evaluated using several error and correlation metrics.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Support Vector Machine and Neural Network Approaches for Stiffness Modulus Prediction of Bituminous Mixtures Under Four-Point Bending Tests

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Nicola Baldo,
  • Fabio Rondinella,
  • Jan Valentin,
  • Marcin D. Gajewski,
  • Jan B. Król

摘要

The mechanical characterization of asphalt pavements is carried out by means of performance parameters that are representative of the asphalt mixtures’ behavior when subjected to different external conditions. In this regard, one of the fundamental parameters frequently used during the design processes of flexible pavements is known as stiffness modulus and, in this study, it was obtained by means of 4-point bending tests, carried out at the Department of Road Structures of the Czech Technical University in Prague. The experimental campaign was performed by testing samples of 2 distinct mixtures at 4 different temperatures (0, 10, 20, and 30 ℃) and at 11 different frequencies (0.1, 1, 2, 3, 5, 8, 10, 15, 20, 30, and 50 Hz) so that the mechanical behavior could be deeply investigated for wide variations in external conditions. Laboratory results then allowed data-driven methodologies to be developed in order to model this phenomenon and predict the stiffness modulus by means of different soft-computing techniques, namely support vector machines and artificial neural networks. Both algorithms, properly developed and optimized, provided satisfactory results. However, artificial neural networks, by virtue of a slightly higher complexity, provided more accurate predictions and a higher reliability, evaluated using several error and correlation metrics.