Neural network- driven optimization of injection moulding parameters for enhanced recycling

Abstract

Recycling practices in injection molding can reduce costs and enhance sustainability. This research especially addresses small- to medium-sized companies. The use of scrap material is often related to unknown material behavior. The use of neural networks can assist in the optimization of processing conditions to ensure processibility and product quality. The focus is set on two main process parameters: the smallest cushion volume and the specific switching pressure, as a basis for ensuring optimal material flow and pressure distribution. Virgin and recycled glass fibre reinforced Styrene maleic anhydride (SMA), as well as a mixture thereof was mechanically and Theologically characterized to collect relevant ground truth data. A model was developed using artificial neural networks (ANN), based on 6650 data points covering various process conditions and the different material compositions. This ANN-based model shows potential for improved material utilization and waste reduction, laying the foundation for future AI deployment in sustainable manufacturing.