A Unified Machine Learning and Simulation Approach for Enhanced Concrete Mix Optimization and Structural Behavior Prediction using Novel PhysiCon-XCNN Model

Abstract

The structure of Concrete is explicit in the quality and life of the building. The mixture of the materials in a correct proportion results in long-lastingness. Though the strength of the concrete is examined by the civil engineers in a manual methodology, there are some limitations and deviations in concluding high-quality results. While approaching with this empirical method, it leads to manpower, time-consuming, costly, and most importantly, material wastage, where a lot of materials are used for each trial. So, determining the quality of life of concrete and classifying it from highest to lowest order is always a demanding task. Existing ways in AI systems face a few challenges in predicting the quality of materialistic and structural properties. When looking forward to the strength through classification, it is resource-intensive. The Novel PhysiCon-XCNN Model, integrating civil with machine learning with simulation methods, gives new possibilities in the development of technology. To resolve the challenges in classifying the quality of concrete, a machine learning framework is being proposed with the help of proportions of the materials for compressive strength in an efficient way. An open-source Concrete Compressive Strength dataset is used for the evaluation process for the proposed method. The proposed method achieves a significant performance in the prediction with a low MSE of 22.63, improving the accuracy to 92.4. The 34.98 MPa of stress distribution corresponding of 29,572.62 MPa of elastic modulus. Additionally, obtained 6.42 MPa of reduction in effective strength under different environmental situations. This helps in the growth of the building industry, focusing on sustainability. The classification is processed with technology and reduces the traditional applicability.