Investigation of preparation of Ca-α-Sialon powders via artificial neural networks analysis

Artificial neural networks (ANNs) have been applied in target forecasting since their promising performances in the areas of control and pattern recognition. This paper aims to develop a methodological approach for estimating the relative contents of Ca-α-Sialon by using ANNs. Ca-α-Sialon powders were prepared by the reaction nitridation method using Si, Al, CaCO3 and Al2O3 as main starting materials. The effects of processing parameters such as reaction temperature, contents of Y2O3, TiO2, and Fe2O3 additives, and α-Si3N4 seeds on the synthesis of Ca-α-Sialon were investigated in detail. The results showed that Ca-α-Sialon can be prepared after 8–20 h at 1550 °C, and that the addition of Y2O3 is more effective for promoting the Ca-α-Sialon formation than TiO2 and Fe2O3. The addition of α-Si3N4 seeds plays a positive role in promoting the Ca-α-Sialon formation. A back propagation ANN was used to establish a model to predict the reaction extents (relative contents of Ca-α-Sialon) corresponding to various processing conditions. The results from the reliability verification tests matched well with those predicted by the ANNs model, which proves the effectiveness of the proposed methodology.