Nowadays injection moulding processes can be designed and optimized using both deterministic (Finite Element Method) and heuristic (Response Surface Methodology)techniques. Both techniques can be utilized before real production begins: in the design stage numerical simulations allow the optimal design of the dies and the determination of a suitable combination of “nominal” process parameters for the production. However, during production process conditions can present some deviations from nominal process settings obtained from FEM simulations; for this reason the final geometry varies and possible scrap products can be obtained. This misalignment between numerical results and production results is intrinsically connected to the fact that the numerical simulation of the process is deterministic in nature while the production process is stochastic. To overcome this limit, in the process optimization stage, RSM techniques can be further applied, but they require experience in process variables selection (factors and responses) and in the choice of the range for each factor. Moreover, experimentation is time consuming and the delay of the production, required to perform the whole experimental campaign, is not always convenient.
In order to reduce the required time, moving most of the trial and error activities from the shop floor into the design phase, a new approach for robust design of injection moulding process has been proposed and described in this paper. The approach is based on the integration of numerical modelling, response surface methodology and Monte Carlo stochastic simulations. To clarify the proposed methodology, the paper
details its application to a tub rear cover of a washing machine, made in glass fibre reinforced polypropylene, with six injection points. The controlled output is the length of a radius after warping.
