Machine Learning in the Analysis of the Mechanical Shredding Process of Polymer Recyclates

Abstract

Artificial intelligence methods and techniques creatively support the processes of developing and improving methods for selecting shredders for the processing of polymer materials. This allows to optimize the fulfillment of selection criteria, which may include not only indicators related to shredding efficiency and recyclate quality but also energy consumption. The aim of this paper is to select methods of analysis based on artificial intelligence (AI) with independent rule extraction, i.e., data-based methods (machine learning—ML). This study took into account real data sets (feature matrix 1982 rows × 40 columns) describing the shredding process, including energy consumption used to optimize the parameters for the energy efficiency of the shredder. Each of the 1982 records in a .csv file (feature vector) has 40 numbers divided by commas. The data were divided into a learning set (70% of the data), a testing set (20% of the data), and a validation set (10% of the data). Cross-validation showed that the best model was LbfgsLogisticRegressionOva (0.9333). This promotes the development of the basis for an intelligent shredding methodology with a high level of innovation in the processing and recycling of polymer materials within the Industry 4.0 paradigm.