Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
share announcement

23 March 2026
Materials | New Study Reveals Most Accurate AI Model for Predicting Concrete Strength

A study published in the international journal Materials (ISSN: 1996-1944) shows that an Artificial Neural Network (ANN) model outperforms other common machine learning models—including Support Vector Machines (SVMs), Multiple Linear Regression (MLR), and Regression Trees (RT)—in predicting the compressive strength of concrete.

Traditionally, determining concrete compressive strength requires a time-consuming 28-day process involving the casting and testing of samples. This study compared the predictive performance of four models using a dataset of 1,030 samples containing eight key ingredients. The results demonstrate that the Artificial Neural Network, capable of learning complex non-linear relationships, performed best, achieving the lowest error across key metrics such as Root Mean Square Error (RMSE) and Mean Absolute Percentage Error (MAPE).

Summary of key performance results:

Model Root Mean Square Error (RMSE)
Artificial Neural Network (ANN) 6.09 MPa
Support Vector Machine (SVM) 6.31 MPa
Regression Tree (RT) 6.60 MPa
Multiple Linear Regression (MLR) 10.45 MPa

Watch the video abstract

For more information about this innovative comparison and its implications for the construction industry, readers can access the article page to watch the video abstract prepared by the authors.

This video was produced with the support of MDPI’s Academic Video Service, which helps researchers bring their work to a wider audience through professional visual summaries and animations.

Read the full article:

“Performance Comparison of Machine Learning Models for Concrete Compressive Strength Prediction”
by Amit Kumar Sah and Yao-Ming Hong
Materials 2024, 17(9), 2075; https://doi.org/10.3390/ma17092075

More News & Announcements...
Back to TopTop