DEVELOPMENT OF PREDICTION MODELS FOR COMPRESSIVE STRENGTH IN CEMENT MORTAR WITH BENTONITE USING MACHINE LEARNING TECHNIQUES

Abstract

In this study, the effects of bentonite-substituted cement mortar, cement compressive strength, cement quantity, spread values, water absorption percentages by weight, and porosity values on the 28-day compressive strength were investigated using Multiple Regression, Adaptive Neuro-Fuzzy Inference System and the intuitive optimization method known as Particle Swarm Optimization. Based on the results obtained from 18 data points, with 4 of them used for testing and 14 for training, effective and ineffective input parameters were identified in comparison to Multiple Regression. Subsequently, Particle Swarm Optimization and Adaptive Neuro-Fuzzy Inference System main models were designed according to the obtained results. As a result of the study, it was determined that cement compressive strength, cement quantity and water absorption parameters have a higher impact on compressive strength compared to other parameters. It was found that the best accuracy model was achieved with the Particle Swarm Optimization model, and the results of the Multiple Regression model can also be used in predicting outcomes.