Predicting the Sintering Curve of Porcelain by Support Vector Regression

Abstract

The densification and deformation of porcelain samples during firing with different compression load were investigated by loading dilatometry. Porcelain has a viscous behavior above 1000°C and the global densification is due to the applied stress and due to the behavior of the material itself. Moreover, the density of the powder compact is very influent in thermal transformations and in the global densification rate. Consequently, dilatometry curves are difficult to explain by analytical methods due to the complex response of porcelain and due to the cost and time required to conduct many experiments, which lead to sparse data sets. In this article, Support Vector Regression (SVR) was applied to predict the behavior of porcelain during sintering from a small set of experiments of densification against time, green density, and applied pressure. The SVR method was used to reconstruct any two‐dimensional densification–temperature curves at any applied pressure, from the experimental data. It is shown that SVR is able to calculate a unique optimum from a very limited number of experiments. Moreover, only two main parameters are required for the calculation. Results provide more information about sintering and creep phenomena as extrapolation at very low applied pressure of densification can be performed.