Abstract:Abstract Using artificial intelligence-based tools, this research aims to establish a direct correlation between the alkali-activated concrete (AAC) mix design factors and their performances. More specifically, the machine learning system was fed new property data obtained from AAC mixes used in laboratory experiments. The rheological parameters (yield stress [static/dynamic] and plastic viscosity) of AAC were predicted using the multilayer perceptron neural network (MLPNN) and bagging ensemble (BE) models. In addition, the R 2 values, k-fold analyses, statistical checks, and the dissimilarity between the experimental and predicted compressive strength were employed to assess the performance of the created models. Also, the SHapley additive exPlanation (SHAP) approach was used for examining the relevance of influencing parameters. The BE approach was found to be significantly accurate in all prediction models, with R 2 greater than 0.90, and MLPNN models were found to be moderately precise, with R 2 slightly below 0.90. However, the error assessment through statistical checks and k-fold analysis also validated the higher precision of BE models over the MLPNN models. Building models that can calculate rheological properties of AAC for different values of input parameters could save a lot of time and money compared to doing the tests in a laboratory. In order to ascertain the required amounts of raw materials of AAC, investigators, as well as businesses, may find the SHAP study helpful.

What problem does this paper attempt to address?

The paper attempts to address the problem of establishing a direct correlation between the mix design factors of Alkali-Activated Concrete (AAC) and its performance. Specifically, the researchers used a machine learning system to predict the rheological parameters (static/dynamic yield stress and plastic viscosity) of AAC using new attribute data obtained from laboratory experiments. Additionally, the study evaluated the performance of the created models and explored the correlation of influencing parameters. ### Main Research Questions: 1. **Establishing Correlation**: Establish a direct correlation between the mix design factors and performance of alkali-activated concrete. 2. **Predicting Rheological Parameters**: Use Multilayer Perceptron Neural Network (MLPNN) and Bagging Ensemble (BE) models to predict the rheological parameters (static/dynamic yield stress and plastic viscosity) of alkali-activated concrete. 3. **Model Performance Evaluation**: Evaluate model performance through R² values, k-fold analysis, statistical tests, and the difference between experimental and predicted compressive strength. 4. **Influencing Parameter Analysis**: Analyze the correlation of influencing parameters using SHapley Additive exPlanations (SHAP) method. ### Research Background: - Concrete is the most commonly used construction material, but its production has a significant environmental impact, especially due to the large CO₂ emissions generated during the production of Ordinary Portland Cement (OPC). - Alkali-Activated Materials (AAM) are a promising low-carbon binder alternative to OPC, but their design rules are not yet mature, limiting practical applications. - The development of machine learning algorithms provides effective tools for predicting the properties of complex systems, enabling faster and more accurate predictions of alkali-activated concrete performance. ### Research Methods: - **Data Acquisition and Analysis**: Collect 145 data points from the literature and prepare the data for data mining. - **Machine Learning Modeling**: Use Python and Spyder (5.1.5) scripts through Anaconda Navigator for modeling. Use MLPNN and BE models to predict the static yield stress (SYS), dynamic yield stress (DYS), and plastic viscosity (PV) of alkali-activated concrete. - **Model Validation**: Validate model performance through k-fold cross-validation, statistical tests, and error evaluation methods. - **SHAP Analysis**: Use the SHAP method to analyze the impact of input parameters on the prediction results. ### Research Significance: - **Time and Cost Savings**: By establishing models to predict the rheological properties of alkali-activated concrete, laboratory testing time and costs can be reduced. - **Optimized Design**: The models can help researchers and companies determine the required quantities of raw materials, optimizing the mix design of alkali-activated concrete. - **Improved Accuracy**: The study found that the BE model performed more accurately among all prediction models, with R² values greater than 0.90, while the MLPNN model had slightly lower accuracy, with R² values slightly below 0.90. In summary, this study aims to improve the accuracy and efficiency of predicting the performance of alkali-activated concrete through modern artificial intelligence technology, thereby promoting its application in practical engineering.

Investigating the rheological characteristics of alkali-activated concrete using contemporary artificial intelligence approaches