Determining Mechanical and Physical Properties of Phospho-Gypsum and Perlite-Admixtured Plaster Using an Artificial Neural Network and Regression Models

Abstract

This research investigates the utilization of artificial neural networks for improving the mechanical and physical properties of phospho-gypsum and perlite-admixtured plaster. The values obtained were modeled using an artificial neural network. Phospho-gypsum (CaSO<inf>4</inf>.2H<inf>2</inf>O) is known as a by-product of waste material of the phosphoric acid production process. Perlite is an amorphous volcanic glass. This study examined the effects of perlite and phospho-gypsum additives on fresh and hardened properties of plaster putty and also the feasibility of a plaster with these additives and heat insulation properties. Mixture and physico-mechanical properties after mixture conforming to standards have been provided. The values obtained were modeled with both multiple regression analysis and an artificial neural network. The R² values for multiple regression analysis with test data were between 0.5264 and 0.9883. R2 value of the artificial neural network was found to be 0.9907. The test results of these mixtures have been compared and the plaster mixture with best values was obtained. © 2017, HARD Publishing Company. All rights reserved.