Learning quality characteristics for plastic injection molding processes using a combination of simulated and measured data

During the initial sampling of injection molds, the determination of suitable process parameter values to achieve a desired quality of the resulting parts, can be a time-consuming and demanding task. This is due to the complex viscoelastic properties of injection molding processes. Conducting technological investigations and using simulation techniques are popular approaches to support the design of the regarded process. However, while the former approach can require extensive research efforts, it can be difficult to design simulations and validate their prediction accuracy, especially when few process measurements are available as a baseline. In addition, the knowledge obtained by both, simulation and technologically based approaches, is only valid for the analyzed process configurations. In contrast, models based on machine learning (ML) approaches can provide forecasts for previously unseen data and can be evaluated quickly. Unfortunately, a high amount of data is required to train such models reasonably. In this contribution, a novel ML-based methodology to predict quality characteristics of an injection molding process for different process parameter values using an intelligent combination of simulation data and measurements, is presented.