Studying the influence of Tri-calcium phosphate in multi stage production of expandable polystyrene with applications of artificial neural networks

Expandable polystyrene is one of the most commonly used polymers. The production of this polymer with the conventional method has some problems which make the process of production to be hard and also reduce the quality of the produced polymer. In this research, besides of implementation of the initiator injection method, adding Tri-calcium Phosphate (TCP) with different percentages ( 3, 6, and 9%) in different states (polymerization of the first stage in 2.5, 3, 3.5, and 4 hours and the amount of used initiator in 70, 75, 80 and 100% of the conventional method and numbers of injection in 6, 8, 10 and 12 times) has been tested and different tests has been conducted on the produced polymer. These tests are: Polymer Dispersion Index, the amount of the tension in yield point, the amount of absorbed pentane and the amount of residual monomer. The results of obtained data from experiments have been simulated with Radial Basis Function (RBF) and Multi-Layer Perceptron (MLP) methods of artificial neural networks and the results of the RBF network had better prediction comparing with MLP network due to having more scientific foundations and ability to filter noises, therefore it can be used to predict the points which have not been experimented. Investigating the experimental data show that in a constant percentage of TCP, by changing the amount of initiator and dosing times and increasing the time of polymerization, the PDI, absorbed pentane, and residual monomer amount and tension in yield point changes. TCP change in different laboratory conditions changes the quality of the polymer with attention to the need of the market and the importance of each item, the information of this research can be used accordingly.