Data-Driven Design for a Modified Asphalt Binder Based on a Reduced Order Model and Simplicial Homology Global Optimization

Abstract

The orthogonal experimental method is currently the most commonly used tool for determining an appropriate composition of modified asphalt binders. However, it cannot automatically output a reasonable composition when dealing with multi objective optimization and must go through the process of repeated comparison manually. Obviously, the accuracy and efficiency of the orthogonal method are easily affected by subjective factors, which grow tremendously as the sizes of the experiments and data increase. In order to enrich the design method, the objective of this study is to develop a new framework to determine reasonable compositions of modified asphalt binder automatically, accurately, and efficiently. Hence, a data-driven multiobjective optimization method is established based on the experimental data, reduced order model (ROM), and simplicial homology global optimization (SHGO). The ROM is utilized to characterize the complex relationship between material composition and the corresponding performance. The average relative errors between experimental results and prediction values of the ROM are less than 1.5%, and the maximum relative errors are less than 8.7%. The SHGO is selected to determine the reasonable compositions in the global space. Finally, the properties and costs of the SBS-modified asphalt binders, orthogonal-based modified asphalt binder, and ROM-SHGO-based modified asphalt binder are tested to verify the feasibility of the developed method. The ROM-SHGO-based modified asphalt binder has the lowest cost under the premise of similar performance when taking the cost as the constraint. The developed ROM-SHGO method provides feasible, useful, and promising tools for composition design of modified asphalt binder based on laboratory test data.