Prediction of the physical properties of barium titanates using an artificial neural network

• Published in: Applied physics. A, Materials science & processing Materials science & processing, 2017-04, Vol.123 (4), p.1-12, Article 274
• Main Authors: Al-Jabar, Ahmed Jaafar Abed, Al-dujaili, Mohammed Assi Ahmed, Al-hydary, Imad Ali Disher
• Format: Article
• Language: English
• Subjects: Applied physics; Artificial neural networks; Barium; Barium titanates; Ceramics industry; Characterization and Evaluation of Materials; Condensed Matter Physics; Dielectric properties; Heating rate; Machines; Manufacturing; Materials science; Nanotechnology; Neural networks; Optical and Electronic Materials; Physical properties; Physics; Physics and Astronomy; Porosity; Pressing; Process parameters; Processes; Shrinkage; Sintering (powder metallurgy); Surfaces and Interfaces; Thin Films
• ISSN: 0947-8396; 1432-0630
• DOI: 10.1007/s00339-017-0885-6

Barium titanate is one of the most important ceramics amongst those that are widely used in the electronic industry because of their dielectric properties. These properties are related to the physical properties of the material, namely, the density and the porosity. Thus, the prediction of these properties is highly desirable. The aim of the current work is to develop models that can predict the density, porosity, firing shrinkage, and the green density of barium titanate BaTiO 3 . An artificial neural network was used to fulfill this aim. The modified pechini method was used to prepare barium titanate powders with five different particle size distributions. Eighty samples were prepared using different processing parameters including the pressing rate, pressing pressure, heating rate, sintering temperature, and soaking time. In the artificial neural network (ANN) model, the experimental data set consisted of these 80 samples, 70 samples were used for training the network and 10 samples were employed for testing. A comparison was made between the experimental and the predicted data. Good performance of the ANN model was achieved, in which the results showed that the mean error for the density, porosity, shrinkage, and green density are 0.02, 0.06, 0.04, and 0.002, respectively.