Intelligent Quality Control System for Polymeric Films Based on Big Data Processing and Mathematical Modeling of Technological Processes

An intelligent system is described for the formation of advice to managerial production staff on quality control of polymeric films of a wide range in a routine mode and in case of emergency situations that are associated with defective films. The system is designed to control extrusion calendering production of films. The production is distinguished by its multi-stage nature, flexibility of the hardware design of the stages, the presence of a recycle for returning waste for processing, a large number and complexity of connections between the parameters of raw materials, equipment, technological mode and product quality indicators, the lack of quality control of the intermediate (extrudate) from which the film is made. Production data is characterized by a large volume, high generation velocity and a variety of sources. The complexity of the control object has required the creation of algorithms for processing its big data to predict product quality based on machine learning methods and the construction of multivariate models for describing an industrial facility. Regression analysis, artificial neural networks and boosting decision trees are used to build predictive models. The choice of method depends on the type of distribution law of production data, their volume and requirements for accuracy and processing speed. The models for describing an industrial facility are deterministic mathematical models of technological processes at the production stages for calculating unmonitored quality indicators of extrudate and film, a base of expert knowledge about emergency situations, their reasons and recommendations for elimination for forming advice to operators, as well as a data bank of film types, production line configurations, monitored and calculated production parameters for adjusting the system to hardware and technological design of production and the range of its products. Testing using data from industrial production of polyvinyl chloride-based packaging films for pharmaceuticals and food products has confirmed the operability of the system. The implementation of the system at the factory in Saint Petersburg has showed an increase in control efficiency by reducing the mass of defective film and increasing the yield of conditioned products.