Machine Learning (ML) Enhances Concrete Strength Prediction

A recent article published in Scientific Reports proposed using machine learning (ML) to enhance concrete compressive strength (CS) prediction. A graphic user interface (GUI) was also designed to bridge the gap between complex computational predictions and real-world applications.

ML Enhances Concrete Strength Prediction
Pearson correlation of input and output variables. Image Credit: https://www.nature.com/articles/s41598-024-66957-3

Background

Portland cement (PC) manufacturing accounts for approximately 7 % of global CO2 emissions. Thus, incorporating waste and recycled materials into concrete is considered a viable and sustainable solution to meet the growing demand for concrete and prevent the degradation of natural resources.

Various industrial by-products such as ground granulated blast furnace slag (GGBS), granite powder, and fly ash can substitute PC in concrete. However, their impact on concrete CS needs careful evaluation. Conventional laboratory tests to evaluate CS are now considered inefficient and not cost-effective.

Recently, the artificial intelligence (AI) sector has advanced significantly through the evolution of various ML models. Predicting concrete CS using ML models allows for optimizing mix designs, ensuring the concrete meets the required performance standards without extensive mixture trials that waste time and resources. Thus, this study explored ML models to predict concrete CS and display the results through a GUI for practical applications.

Methods

In this study, the researchers employed ML models to enhance the prediction of concrete compressive strength (CS) by analyzing 1,030 experimental data points from previous research, with CS values ranging from 2.33 to 82.60 MPa. The main aim was to examine the efficacy of different ML models in predicting concrete CS.

Data from the 1,030 datasets, including components such as cement and aggregates, were collected and analyzed through histograms and heat maps. Subsequently, two types of predictive models were explored: non-ensemble and ensemble models. The non-ensemble models included regression-based, evolutionary, neural network, and fuzzy-inference-system approaches, while the ensemble models consisted of adaptive boosting, random forest, and gradient boosting techniques.

The input parameters for the ML models included cement, blast-furnace slag, aggregates (coarse and fine), fly ash, water, superplasticizer, and curing days, with CS as the output variable. The performance of the ML models was evaluated by comparing the predicted and actual values using seven performance indices: determination coefficient (R²), Willmott index (WI), root mean square error (RMSE), scatter index (SI), mean absolute error (MAE), mean absolute percentage error (MAPE), and mean bias error (MBE).

Additionally, k-fold cross-validation was performed to check the reliability and accuracy of the ML models. Furthermore, a sensitivity analysis using Shapley-Additive-exPlanations (SHAP) was conducted to comprehend the impact of each input variable on CS. Finally, a Python web application was developed to make the ML models accessible and usable for concrete prediction through an intuitive graphical user interface (GUI).

Results and Discussion

The performance of the ensemble and non-ensemble ML models was compared to analyze their prediction capabilities. The evaluation framework included visual methods such as scatter plots, violin boxplots, and Taylor diagrams to ensure the reliability and precision of the developed models. Among all models, CatBoost predicted the concrete CS with maximum accuracy during the testing stage. It exhibited an R² value of 0.966, MAE of 2.27 MPa, and RMSE of 3.06 MPa.

The CatBoost, XGBoost, and Random Forest (RF) ensemble models performed well across low (2.33-25.02 MPa), medium (25.08-55.02 MPa), and high (55.06-82.60 MPa) CS ranges. Notably, CatBoost exhibited consistently superior performance in all ranges. The SHAP analysis using CatBoost helped identify the most dominant parameters influencing CS prediction. The most important parameter was the concrete age, followed by cement content. While water, coarse aggregates, and fly ash parameters exhibited moderate effects, superplasticizers and fine aggregates demonstrated minimal effects.

The user-friendly GUI developed in this study was introduced on an open-source platform like GitHub. This Python-based web application could help designers predict concrete CS rapidly and cost-effectively. Additionally, it supports real-time and precise predictive capabilities, enabling model refinement and improvement. Thus, the proposed GUI improves efficiency and resource management in the construction sector compared to conventional laboratory tests.

Conclusion

Overall, the researchers examined seven ML models from ensemble and non-ensemble categories to predict concrete CS; the CatBoost exhibited the best performance with high accuracy and generalization capability. However, the efficacy of the proposed ML models was limited by the range and quality of the input data. Thus, they might not apply to datasets or conditions extremely different from the training data. Moreover, these models did not account for all factors that could influence concrete CS, such as curing temperature and environment.

The researchers suggest expanding the dataset by considering more varied concrete designs and external factors in the future. In addition, deep learning models can enhance predictive accuracy using this vast and complex dataset. The proposed GUI can also be integrated with greater user guidance features and tutorials to expand usage. Finally, validating the models with real-world data will ensure their robustness and reliability in practical construction applications.

Journal Reference

Elshaarawy, M. K., Alsaadawi, M. M., & Hamed, A. K. (2024). Machine learning and interactive GUI for concrete compressive strength prediction. Scientific Reports, 14(1), 16694. DOI: 10.1038/s41598-024-66957-3, https://www.nature.com/articles/s41598-024-66957-3

Disclaimer: The views expressed here are those of the author expressed in their private capacity and do not necessarily represent the views of AZoM.com Limited T/A AZoNetwork the owner and operator of this website. This disclaimer forms part of the Terms and conditions of use of this website.

Nidhi Dhull

Written by

Nidhi Dhull

Nidhi Dhull is a freelance scientific writer, editor, and reviewer with a PhD in Physics. Nidhi has an extensive research experience in material sciences. Her research has been mainly focused on biosensing applications of thin films. During her Ph.D., she developed a noninvasive immunosensor for cortisol hormone and a paper-based biosensor for E. coli bacteria. Her works have been published in reputed journals of publishers like Elsevier and Taylor & Francis. She has also made a significant contribution to some pending patents.  

Citations

Please use one of the following formats to cite this article in your essay, paper or report:

  • APA

    Dhull, Nidhi. (2024, July 31). Machine Learning (ML) Enhances Concrete Strength Prediction. AZoBuild. Retrieved on December 12, 2024 from https://www.azobuild.com/news.aspx?newsID=23576.

  • MLA

    Dhull, Nidhi. "Machine Learning (ML) Enhances Concrete Strength Prediction". AZoBuild. 12 December 2024. <https://www.azobuild.com/news.aspx?newsID=23576>.

  • Chicago

    Dhull, Nidhi. "Machine Learning (ML) Enhances Concrete Strength Prediction". AZoBuild. https://www.azobuild.com/news.aspx?newsID=23576. (accessed December 12, 2024).

  • Harvard

    Dhull, Nidhi. 2024. Machine Learning (ML) Enhances Concrete Strength Prediction. AZoBuild, viewed 12 December 2024, https://www.azobuild.com/news.aspx?newsID=23576.

Tell Us What You Think

Do you have a review, update or anything you would like to add to this news story?

Leave your feedback
Your comment type
Submit

While we only use edited and approved content for Azthena answers, it may on occasions provide incorrect responses. Please confirm any data provided with the related suppliers or authors. We do not provide medical advice, if you search for medical information you must always consult a medical professional before acting on any information provided.

Your questions, but not your email details will be shared with OpenAI and retained for 30 days in accordance with their privacy principles.

Please do not ask questions that use sensitive or confidential information.

Read the full Terms & Conditions.