Development of an integrated neural network system for prediction of process parameters in metal injection moulding

In this present work attempts have been made to develop an integrated neural network system for prediction of process parameters such as injection pressure and injection time in metal injection moulding (MIM) process. The current system has been developed by integrating the different aspects of MIM process. The aspects that are addressed in this system are the physical model of MIM filling stage based on governing equations of mould filling, and process parameters for debinding and sintering stages generated by experimentation. In this work the feed forward type of neural network has been used, which was initially trained with the analytical data before incorporating as part of an integrated system. In this work Gauss training method has been incorporated for the usage of function approximation. This integrated system has been implemented in MatLAB environment by using neural networks toolbox. This integrated system was successfully tested to solve the real world problems of MIM process. The analytical algorithm based on governing equations of mould filling process first produces a feasible injection time for the MIM process. Injection time data is then used to train the neural network system. In order to validate the results generated by the neural network system are checked with the simulation results of the “Moldflow” software and found that the results generated by integrated neural network system are not different from the simulated results.