An Extended Multi-Model Regression Approach for Compressive Strength Prediction and Optimization of a Concrete Mixture

TL;DR

This paper introduces a multi-model regression framework combined with genetic algorithm-based optimization to improve concrete compressive strength prediction and mixture optimization, outperforming existing single-model approaches.

Contribution

It presents a novel weighted multi-regression approach and a multi-objective optimization method for concrete mixture design, enhancing prediction accuracy and optimization efficiency.

Findings

Multi-model regression outperforms single models in accuracy.

Genetic algorithm-based optimization yields a Pareto front for cost-strength trade-off.

The proposed method achieves significant improvement over existing models.

Abstract

Due to the significant delay and cost associated with experimental tests, a model based evaluation of concrete compressive strength is of high value, both for the purpose of strength prediction as well as the mixture optimization. In this regard, several recent studies have employed state-of-the-art regression models in order to achieve a good prediction model, employing available experimental data sets. Nevertheless, while each of the employed models can better adapt to a specific nature of the input data, the accuracy of each individual model is limited due to the sensitivity to the choice of hyperparameters and the learning strategy. In the present work, we take a further step towards improving the accuracy of the prediction model via the weighted combination of multiple regression methods. Moreover, a (GA)-based multi-objective mixture optimization is proposed, building on the obtained multi-regression model. In particular, we present a data aided framework where the regression methods based on artificial neural network, random forest regression, and polynomial regression are jointly implemented to predict the compressive strength of concrete. The outcome of the individual regression models are then combined via a linear weighting strategy and optimized over the training data set as a quadratic convex optimization problem. It is worth mentioning that due to the convexity of the formulated problem, the globally optimum weighting strategy is obtained via standard numerical solvers. Employing the proposed GA-based optimization, a Pareto front of the cost-CS trade-of has been obtained employing the available data set. Moreover, the resulting accuracy of the proposed multi-model prediction method is shown to outperform the available single-model regression methods in the literature by a valuable margin, via numerical simulations.