Proposal for the implementation of a predictive quality system for injection moulding

The manufacturing industry is rapidly evolving with the adoption of innovative technologies, particularly in data-driven quality control and process optimization. Injection moulding is a widely used polymer processing method, whose quality is influenced by the interplay of multiple process parameters including temperature, pressure, injection speed, and cooling rates. Traditional quality control approaches in injection moulding primarily rely on post-production inspection, which often results in late defect detection, leading to increased material waste, higher production costs, and inefficiencies in manufacturing. These limitations highlight the need for a more proactive and intelligent approach to quality assurance. Thus, the injection moulding process can benefit significantly from the introduction of a predictive quality system that leverages machine learning algorithms and advanced sensor-based monitoring for real-time quality prediction. By continuously analysing process data, such a system can predict and prevent defects before they occur, enabling adaptive process control, minimized waste, and optimized production performance. The study proposes a methodology that includes the identification of relevant process data and production setpoints, real-time data acquisition and the elaboration of a control system based on the implementation of both supervised and unsupervised machine learning models to support decision-making and process optimization. To validate the proposed approach, the predictive quality system was tested using a real industrial case study, regarding an injection moulding machine for the production of automotive polymer components. The results demonstrated the practical applicability of the system in a real-world manufacturing environment and provided valuable insights for the implementation of predictive quality applications in manufacturing.