Computerized Automation and Machine Learning for Prediction of Melt Mass-Flow Rate Using Custom-Built Machine in Industrial IoT

In the polymer industry, where polymers are transformed into a wide range of plastic products, accurate quality assessment through melt flow testing remains critical but challenging. Predicting polymer melt characteristics such as flowability with machine learning (ML) algorithms offers opportunities to optimize processing, reduce raw material and labor costs, and ensure stable, high-quality production. Two key metrics, the melt mass-flow rate (MFR) and the melt volume-flow rate (MVR), serve as industry standards for material quality control and new mixture development, both defined by strict specifications for testing machine geometry and standardized protocols. This work presents a custom-built melt flow testing machine designed to meet these specifications, integrated with computerized automation for consistent execution of the test procedure. Using a representative dataset of materials, testing temperatures, and applied weights, multiple ML models were trained to predict MFR quickly and accurately. Comparative evaluation with cross-validation demonstrated that neural network regression achieved superior results, with a root mean square error (RMSE) of 0.3 g/10 min and a mean absolute percentage error (MAPE) of 3.2%. The results underscore the promise of combining tailored ML algorithms with custom-built testing equipment and automation to deliver faster and more precise melt flow characterization. This proof-of-concept not only advances polymer processing but also paves the way for smarter, more sustainable manufacturing. By enabling consistent quality in recycled polymers, optimizing resource consumption, and integrating ML within the industrial Internet of Things (IoT), this approach points toward a future of intelligent, efficient, and reliable polymer production.