Multi-objective optimization for composition design of civil materials based on data-driven method

The performance of civil materials depends on their compositions. The material design is an important, challenging, and time-consuming work during the construction due to the complex relationship among different compositions and the conflict between properties and economic cost. This study developed a data-driven multiobjective optimization design framework to determine the reasonable material compositions based on laboratory experiment, reduced order model (ROM), and multi-objective optimization (MOO). The ROM was used to capture the complex relationship between the material compositions and the corresponding material performance. The MOO was utilized to determine the pareto optimal solution of compositions in the global space. Taking examples of cement grouts and modified asphalt binders, the develop framework was demonstrated and verified. The developed framework provides an excellent tool for composition design of civil materials.