Injection molding process modeling using back propagation neural network method

Polymer material is now widely used to replace metal materials. One of the most common processes used to form polymers is the injection molding machine process. Unfortunately, the relationship between parameters and the quality of the results of this process is quite complex and not yet known certainly that to predict the quality of the results which is still based on established parameters is difficult to do. Back propagation neural network (BPNN) is an algorithm in artificial neural network proposed in this research used to predict the quality of the result of tensile strength and impact strength of biocomposite material on injection molding machine process based on some process parameters such as barrel temperature, injection pressure, holding pressure and injection velocity. To obtain good BPNN network structures, several combinations of the number of neurons in the hidden layer and activation function have been attempted where the mean square error (MSE) is used as a reference. The best BPNN network is the network that has the smallest MSE value. The results showed that the network BPNN network model 2 hidden layers has the number of neurons in each hidden layer 9, with tansig activation and trainrp training function in the smallest MSE value that is 0.0467.