Bayesian Regularized Artificial Neural Network Model to Predict Strength Characteristics of Fly-Ash and Bottom-Ash...

Area of Science:

• Materials Science
• Civil Engineering
• Computational Science

Background:

• Geopolymer concrete (GPC) is a sustainable construction material utilizing waste products like fly ash.
• GPC mix design and characterization are complex due to numerous influencing factors.
• Efficient prediction models are needed to accelerate GPC development and adoption.

Purpose of the Study:

• To predict the compressive strength of fly-ash and bottom-ash based GPC using Artificial Neural Networks (ANN).
• To identify the most effective ANN model for accurate GPC strength prediction.
• To optimize GPC mix design and reduce development time and cost.

Main Methods:

• Developed and compared fourteen ANN models with varying training algorithms, hidden layers, and neurons.
• Utilized literature data and in-house experimental results as input features.
• Employed machine learning techniques to predict compressive strength, minimizing mean squared error (MSE) and maximizing correlation coefficient (R).

Main Results:

• The Bayesian regularized ANN (BRANN) model demonstrated superior performance in predicting GPC compressive strength.
• ANN models effectively correlated GPC specimen specifications with compressive strength outcomes.
• Achieved minimized error (MSE) and high correlation (R) for the selected BRANN model.

Conclusions:

• ANN, particularly BRANN, is a powerful and cost-effective tool for predicting geopolymer concrete compressive strength.
• This approach facilitates faster mix design optimization for sustainable GPC.
• The study provides a validated method for forecasting the performance of fly-ash and bottom-ash based GPC.