Prediction of Biphasic Calcium Phosphate Synthesis Conditions Using Artificial Neural Network

Wet chemical methods are the most widely used routes for biphasic calcium phosphate bioceramics synthesis in which control of synthesis condition and reaction factors are very important. To predict and control the synthesis condition، mathematical models can be used. Artificial neural networks are computational tools inspired by the nervous systems of living organisms that help us to get better understanding about the complicated problems. Powders were synthesized using aqueous solution containing different calcium /phosphorus ratio. HNO3 and NH4OH were used to adjust the pH of the solution mixture during the process. The precipitation was calcined at 1100˚ C for one hour. The chemical composition and Ca/P ratio were determined by inductively coupled plasma atomic emission spectroscopy. Phase identification of powders and evaluation of the functional groups of specimens were carried out by X-ray diffraction and Fourier transform infrared spectroscopy respectively. Four three-layered feed forward networks with ten neurons in the hidden layer، linear sigmoid stimulation and Levenberg-Marquardt learning algorithm were trained using data obtained from the experiments designed in four different patterns. The best result was obtained with the network consists of 80% training process، 15% validation process and 5% testing process samples by changing number of samples in each step. To ensure optimal performance of four networks، each network was studied using four new data. The predicted results show a good comparison with those obtained experimentally.