PREDICTION OF HARDNESS, FLEXURAL STRENGTH, AND FRACTURE TOUGHNESS OF ZrO2 BASED CERAMICS USING ENSEMBLE LEARNING ALGORITHMS

Abstract

Flexural strength, hardness, and fracture toughness are the basic mechanical properties of ceramic materials. Manufacturers widely use this set of properties to ensure the durability of ceramic products. However, many factors, such as chemical and phase compositions, sintering temperature, average grain size, density, and others, affect these properties, making it challenging to estimate corresponding reliability parameters correctly. Experimental examination of the impact of these factors on the mechanical properties of ceramics is a rather time-consuming and resource-consuming procedure. This work aims to predict the mechanical properties of zirconia ceramics using machine learning tools. The authors have created an experimental database for predicting the hardness, flexural strength, and fracture toughness of ZrO2-based ceramics based on chemical composition, phase composition, microstructural features, and sintering temperature on the mechanical properties of zirconia ceramics. To solve this problem, we compared the effectiveness of using five single machine learning algorithms and five ensemble methods of different classes. We found a high accuracy of the predicted values of each of the three mechanical properties using ensemble methods from the boosting class (CatBoost, Ada-Boost, and XGBoost). The authors developed a stacked ensemble of machine learning methods to improve the accuracy of determining the hardness property prediction task. The effectiveness of linear and nonlinear meta-regressors in the scheme of the developed ensemble is investigated. We obtained an increase in accuracy of more than 10% (R2) using our approach.