Optimization of Concrete Structure Design Using Machine Learning and Multi-Objective Optimization Algorithms: An Analytical Review

Optimization of Concrete Structure Design Using Machine Learning and Multi-Objective Optimization Algorithms: An Analytical Review

bahar salari1

1) Civil Engineering Expert

IOI Code : ABGG-ABKKGZ
Publication : 11th.International Conference on Researches in Science & Engineering & 8th.International Congress on Civil, Architecture and Urbanism in Asia(11icrsie.ir)
Abstract :
    The design, mixing, and performance optimization of concrete structures, particularly Ultra-High-Performance Concrete (UHPC) and Recycled Aggregate Concrete (RAC), have traditionally relied on trial-and-error methods and simple empirical models. These approaches are costly, time-consuming, and inefficient in balancing conflicting criteria such as strength, durability, cost, and environmental impacts. In recent years, the integration of Machine Learning (ML) algorithms for predicting mechanical and durability properties, together with Multi-Objective Metaheuristic algorithms for identifying optimal mix designs, has become a dominant approach in the design of concrete materials and structures. This paper provides an analytical review of four recent studies in the fields of UHPC optimization, RAC optimization, and seismic beam-column connections in reinforced concrete structures. It explains the common framework underlying these approaches, including data collection and augmentation, the development of learning-based surrogate models such as XGBoost, LightGBM, Artificial Neural Networks (ANNs), and the Sparse Identification of Nonlinear Dynamics (SINDy) method, as well as the application of multi-objective optimization algorithms such as NSGA-II, AGE-MOEA-II, and Cuckoo Search. The results indicate that these integrated approaches can improve the coefficient of determination (R²) for compressive strength prediction to more than 95%, while simultaneously reducing material costs by up to 52% and carbon emissions by up to 45%, without compromising mechanical performance or code requirements. Finally, the existing challenges, including the need for more extensive experimental datasets, the necessity of incorporating uncertainty, and the requirement for experimental validation of optimized designs, are discussed, and potential directions for future research are proposed.

Keywords : ultra high performance concrete (uhpc) recycled aggregate concrete (rac) machine learning multi objective optimization nsga ii cuckoo search carbon emissions mix design