Artificial neural network modeling of ceramics powder preparation: Application to NiNb2O6

This paper studied the modeling of the synthesis process of NiNb2O6 (NN) powder using an artificial neural network (ANN). The characteristic of interest was the amount of NN phase percentage produced from the synthesis process. Three controlling factors affecting the mentioned characteristic were dwell time, calcined temperature and heating/cooling rate. Design of experiments (DoE) technique was used to analyze the relationship of controlling factors to the amount of NN phase. The results show that calcined temperature is the most important factor affecting the amount of NN phase. The dwell time and heating/cooling rate are less significant on the phase but longer dwell time and higher heating/cooling rate are appreciable for the slightly higher purity. Multiple regression was also used to compare the results and the ANN was found to significantly outperform the regression analysis.