Superior Performance of Artificial Neural Networks for Optimizing Concrete Mix Design Using Local Materials in Ayacucho, Peru (2024)

Ramos J.; Lizarbe H.; Estrada J.; Ayala R.; Tenorio M.; Bravo R.; Bellido V.; Ircañaupa A. · Mathematical Modelling of Engineering Problems (2025) · DOI: 10.18280/mmep.120802

Resumen

The traditional concrete mix design in Peru presents significant limitations due to the lack of consideration of the nonlinear behavior of materials. This study developed a predictive model based on Artificial Neural Networks (ANNs) to optimize material dosage in concrete mixes, using the physical and mechanical properties of locally sourced aggregates from Ayacucho during 2024. The city’s high-altitude conditions (2,761 m a.s.l.), unique aggregate mineralogy, and economic constraints make conventional methods less accurate and less efficient, leading to excess cement use and material waste. A database of 806 experimental records from local research, including granulometric, absorption, compressive strength, and proportion variables, was statistically cleaned and validated. Multilayer Perceptrons (MLPs) with a feedforward architecture were trained in MATLAB using the Levenberg-Marquardt algorithm with backpropagation. The optimal architecture (18-11-12-4, representing 18 input variables, two hidden layers with 11 and 12 neurons, and 4 output variables) achieved a correlation coefficient (R) of 0.9947, coefficient of determination (R2) of 0.9895, and a Nash-Sutcliffe Efficiency (NSE) of 0.9784, with a global Mean Absolute Percentage Error (MAPE) of only 0.9175%, outperforming a multivariate multiple linear regression model (R2= 0.9507; NSE = 0.9413; MAPE = 1.4130%). Practical evaluation indicates that the ANN can reduce cement overuse by up to 3%, lowering production costs by approximately USD 2.12 per cubic meter and decreasing waste generation. Bland-Altman analysis confirmed acceptable agreement with the ACI 211.1 method, validating its technical applicability. In conclusion, ANN-based modeling represents a robust, adaptable, and resource-efficient tool for concrete mix design in high-altitude regions such as Ayacucho, offering both economic and environmental benefits. © (2025), (International Information and Engineering Technology Association). All rights reserved.

Instituciones: Universidad Nacional de San Cristóbal de Huamanga