Hybrid Artificial Intelligence Model for the Transition from Predictive to Prescriptive Maintenance in Textile Manufacturing

Abstract

This paper presents the development and validation of a hybrid artificial intelligence framework integrating predictive and prescriptive capabilities to optimize maintenance in the textile industry. The research addresses the high failure rate of critical textile machinery components—such as needles, feeders, guides, and tensioners—whose traditionally reactive management leads to unplanned downtime, increased operational costs, and deterioration of final product quality. The methodology comprises: historical data acquisition, preprocessing, feature engineering (MTBF, MTTR), development of predictive models (ARIMA, XGBoost), and a prescriptive module that transforms predictions into optimized decisions through multi-objective optimization subject to operational constraints. The results demonstrate significant improvements: reduced downtime from unexpected breakages, near-total elimination of fabric waste due to defective seams, decreased sudden failures, increased specific MTBF, and reduction of non-conforming products. It is concluded that the value of predictive maintenance is enhanced when complemented by a prescriptive module that translates predictions into contextualized decisions, bridging the gap between theory and its practical application in real production environments. The framework provides a replicable roadmap for extending the benefits of prescriptive maintenance to the textile sector.