Multi-Objective Density Diagrams Developed With Machine Learning Models to Optimize Sustainability and Cost-Efficiency of UHPC Mix Design

Abstract

The emergence of ultra-high-performance concrete (UHPC) as an attractive solution for precast and prestressed applications has coincided with global efforts towards sustainable construction. The increasing need for tools capable of intuitively demonstrating the effect of concrete mixture composition on mechanical performance, cost and eco-efficiency concurrently has motivated this work in an effort to promote design of more sustainable solutions to help meet environmental goals. Such tools are needed to effectively evaluate the environmental impact of UHPC given the outstanding mechanical properties of the material coupled with high volumetric embodied CO2. Meanwhile, artificial intelligence (AI) techniques have emerged as a great opportunity for game-changing tools capable of effectively modeling the synergistic relationships between mix proportions and material performance. This work couples machine learning models with orthogonal arrays to generate machine-learning-based tools to evaluate the tradeoffs between emissions, cost and mechanical performance concurrently. Random forest and k-nearest neighbors’ models are ensembled to predict the compressive strength of UHPC mixtures and generate Performance Density Diagrams (PDDs). These predicted strengths are then coupled with volumetric environmental factors and unit costs to generate eco- and cost-efficiency density diagrams. The makeup of these tools facilitates the evaluation of rather complicated trends associated with mix proportions and multi-objective outcomes, allowing AI-based tools to be of easy use by industry personnel on a daily basis, while serving as decision-making aids during mix design stages and provide proof of mixture optimization that could be introduced in Environmental Product Declarations. The PDD developed herein enabled the design of a mix with compressive strength of 155 MPa, while keeping the aggregate-to-cementitious ratio above unit. Other mixtures were developed from these models and compared to several different concretes from the literature. Results show that high paste content, high strength (and ultra-high strength) concrete technologies are not necessarily detrimental to cost or eco efficiencies. For the different indices evaluated, optimum solutions were mostly obtained with these types of concrete, which means that industry trends toward requiring minimization of embodied CO2 in concrete on a per volume basis are misguided and do not minimize the embodied CO2 in concrete structures.