A digital twin approach for sustainable construction: predictive optimization of concrete strength using industry 4.0 principles

Description

Peipei Xin, Haytham Isleem, and Mohammad Khishe build a digital twin for concrete and use it to hunt for stronger, greener mixes without pouring a test batch, in a 2025 study. Construction talks about digital twins often but rarely validates one end to end. The team treated a public dataset as the digital feed. It trained a deep neural network as the virtual twin of the material. Four classical machine learning models served as benchmarks for predicting compressive strength.

The twin predicted held-out results with an R-squared of about 0.90 and an average error near five megapascals. A plain linear model managed only 0.63, showing that the material's nonlinear interactions demand the heavier machinery. The authors present the workflow as a reproducible blueprint: digitalize the data, build the twin, and optimize mixes by simulation. They argue the approach can cut the physical experiments behind low-carbon concrete design dramatically.

Catalog record

Source: https://europepmc.org/articles/PMC12820324