Aprendizagem de máquina aplicada na previsão de demanda por setup do processo de tratamento químico de tecido na indústria têxtil

Abstract

The increasing use of Machine Learning (ML) has led many processes to depend less on human action. In this article, the fabric dipping process in a textile industry was selected, where the decision to stop production, evaluating the need for adjustments in the parameters (setup) of the equipment, depends mainly on human action. This is a decision that impacts the indicators of production capacity and product quality. Based on these data, an attempt was made to create and test 7 ML models with different techniques to decide on the application of the setup and evaluate performance metrics. Historical data were collected, the basis for training the models, then followed by a process of preparation, organization, structuring, and exploratory analysis, producing a data set that was divided into 2 groups for training and testing. In the training phase, simulations were performed with different combinations of hyperparameters to identify the best estimators. Finally, the models were tested with unknown data and KNN was the winning model, showing the best results.