Answers are grounded in this paper and cite their sources.
This analysis demonstrates improved quality control using big data algorithms, highlighting predictive models in polymeric films.
Algorithms for primary processing of big data on polymeric film production have been developed. They allow forming complete data sets (quality indicators of polymeric material at the main stages of production and control actions affecting them) and selecting methods for data mining to build predictive models to control consumer characteristics of films (number of surface defects of various types, thickness, thickness variation, color coordinates, shrinkage degree). The algorithms take into account the multi-assortment nature of production and the flexibility of its hardware design, large volume and high speed of data accumulation, the presence of incorrect values of production parameters in the data due to malfunctions of automation equipment and recording errors, unstructured data collected from various sources, incomplete data due to the lack of monitoring of some material quality indicators at stages. Primary processing of these data read from the industrial database includes decoding (conversion into real values), filtering, structuring by production stages and types of parameters of the control object, calculation of unmonitored quality indicators to ensure data completeness, statistical analysis and selection of a machine learning method for predicting consumer characteristics of the film. To calculate unmonitored quality indicators, deterministic mathematical models of technological processes at the main stages (extrudate preparation and forming, drawing and cooling of the film) are used. Statistical analysis allows you to check the data distribution law and determine significant control actions for the analyzed quality indicators. As a result of primary processing, a machine learning method is selected depending on the type of data distribution law, data volume and requirements for prediction accuracy and speed. With a normal data distribution, multifactor regression models are used to predict consumer characteristics, on the basis of which control advice is formed. The developed algorithms are implemented as a software package. Its testing on data from industrial extrusion-calendering production of uncolored and colored rigid packaging films based on polyvinyl chloride has confirmed the operability of the algorithms. Their application ensures completeness and increases the speed of formation of large-volume structured data sets required for subsequent processing for the purpose of controlling multi-assortment production
No takes yet. Share an insight, caveat, or question.
Damrin et al. (2025) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: