Work

A computational framework for optimizing LECA content in concrete using ANN and RSM

Year: 2025

Type: article

Abstract: The incorporation of lightweight expanded clay aggregate (LECA) into concrete mixtures offers significant reductions in density but is often accompanied by undesirable losses in compressive strength, presenting a critical design trade-off. To address this trade-off, the present study develops a computational framework for optimizing LECA content, aiming to minimize compressive strength loss (CSL) while maximizing density loss (DL). A dataset of 129 experimental records was analyzed. An exploratory data analysis (EDA) was performed to investigate distributional properties and inter-variable relationships. Several artificial neural network (ANN) architectures, varying in hidden layers, neurons, and training algorithms, were systematically evaluated using k-fold cross-validation to identify the most accurate predictive model. Performance of the selected ANN was compared with response surface methodology (RSM) to establish a computational optimization framework. The results demonstrated superior predictive performance for ANN, achieving coefficients of determination (R2) of 0.983 for CSL and 0.993 for DL. For instance, using the ANN approach, a multi-objective optimization in which CSL was weighted twice as heavily as DL identified an optimal LECA content of 10.37%. This proportion minimized CSL while simultaneously maximizing DL. By balancing CSL with DL, the framework offers practical guidance for the design of lightweight concretes in construction applications.

Source: Results in Engineering

DOI: 10.1016/j.rineng.2025.108805

Authors: Seyed Hossein Mousavinia, Mohammad Ali Dastan Diznab

Institution: Arak University

Language: English

Topic: Concrete and Cement Materials Research

Subfield: Civil and Structural Engineering

Field: Engineering

Domain: Physical Sciences

SDG: Industry, innovation and infrastructure

Open Access status: gold

Estimated APC paid: $1,940