Analysis of Compressive Strength of Anhydrite Binder Using Full Factorial Design

Abstract

Flue gas desulfurization gypsum (FGD gypsum) is obtained from the desulphurization of combustion gases in fossil fuel power plants. FGD gypsum can be used to produce anhydrite binder. This research is devoted to the investigation of the influence of the calcination temperature of FGD gypsum, the activators K2SO4 and Na2SO4, and their amount on the compressive strength of anhydrite binder during hydration. The obtained results showed that as the calcination temperature increased, the compressive strength of anhydrite binder decreased at its early age (up to 3 days) and increased after 28 days. The compressive strength of the anhydrite binder produced at 800 °C and 500 °C differed more than five times after 28 days. The activators K2SO4 and Na2SO4 had a large effect on the hydration of anhydrite binder at its early age (up to 3 days) in comparison with the anhydrite binder without activators. The presence of the activators of either K2SO4 or K2SO4 almost had no influence on the compressive strength after 28 days. To determine which factor, the calcination temperature of FGD gypsum (500–800 °C), the hydration time (3–28 days) or the amount (0–2%) of the activators K2SO4 and Na2SO4, has the greatest influence on the compressive strength, a 23 full factorial design was applied. Multiple linear regression was used to develop a mathematical model and predict the compressive strength of the anhydrite binder. The statistical analysis showed that the hydration time had the strongest impact on the compressive strength of the anhydrite binder using activators K2SO4 and Na2SO4. The activator K2SO4 had a greater influence on the compressive strength than the activator Na2SO4. The obtained mathematical model can be used to forecast the compressive strength of the anhydrite binder produced from FGD gypsum if the considered factors are within the same limiting values as in the suggested model since the coefficient of determination (R2) was close to 1, and the mean absolute percentage error (MAPE) was less than 10%.