Machine Learning Algorithms For Optimizing Mix Design Of Concrete

Abstract

The procedure of concrete mix design is a multifaceted and intricate procedure aimed at determining the optimal combination of materials to produce high-quality concrete with desirable performance characteristics. In the realm of current literature and modern business practice, several approaches to concrete mix design have emerged, with the methodologies evolved from The Three Equation Method gaining significant popularity. The determination of the concrete class is contingent upon the compressive strength, which is often regarded as a pivotal characteristic of concrete. The predictability of concrete's compressive strength is crucial for the effective utilization of concrete structures since it directly influences their safety and durability. In recent times, there has been a notable surge in interest in machine learning, with projections for its future development becoming more optimistic. The field of data mining has garnered significant interest due to the advancements in machine learning algorithms, which have shown the ability to identify patterns that are challenging for human cognitive abilities to discern. In this study, we aim to use cutting-edge advancements in machine learning methodologies to optimize concrete mix design. In the course of our study, we compiled a comprehensive database consisting of several realistic recipes together with corresponding damaging laboratory experiments. This collection was then used to train the chosen optimum architecture of an Artificial Neural Network (ANN). The architectural representation of the ANN has been successfully transformed into a mathematical equation, hence enabling its practical implementation in many applications.