Calibration of the Filling Simulation of an Injection Moulding Process by Artificial Neural Networks

The reliability of numerical simulation of the thermoplastics injection moulding process by means of finite elements and finite differences methods strongly depends on the accuracy of the parameters that describe the material from both a rheological and a thermal point of view. Simulations are often un-calibrated due to the scarcity and inaccuracy of available data. The main objective of the paper is to calibrate the filling phase of numerical simulation of an injection moulding process through the use of pressure transducers placed in the mould cavity. The software calibration was formulated as an inverse problem, considering the numerical code as a direct model. The inverse problem aims at finding a set of viscosity parameters which minimises an objective function representing, in the least square sense, the difference between experimental and numerical pressure data. In order to minimise this objective function a feed-forward back-propagation artificial neural network (ANN) approach was used. The melt pressure developments estimated by the simulation using calibrated-viscosity data was very close to the measured ones.