Near-Real Time Quality Prediction in a Plastic Injection Molding Process Using Apache Spark

The automotive industry is undergoing wide scope transformation. Industry 4.0 has both expanded the possibilities of digital transformation in automotive, increased its importance to all mobility ecosystem and being driven by continued digitization of the entire value chain. Manufacturing data which is unceasingly flow during serial production is one of the great sources towards Industry 4.0 goal to fully automatizing complex human dependent processes. However, there are few challenges to consider such as collecting and filtering various data from shop floor in given production cycle time range and make them ready for real time analytics as well as constructing efficient data pipeline to reach useful outcomes which is reliable enough to meet customer expectations. In this study, we will extract meaningful relation between injection machine parameters from Farplas Automotive Company's shop floor and describe their effects on the product quality. We will train and test machine learning models with different hyperparameters and test model performance to identify defected products. Finally, we will show implementation of streaming data pipeline using Kafka and Spark to be able to analyze injection machine data and effectively predict plastic injection product's OK-NOK condition real time even before human operator reaches the product itself. Consequently, detecting defected products will be independent from human attention which makes production areas one step closer to dark factory.