Publicação
An intelligent decision support system for the textile industry
| Resumo: | In the Textile industry there is a need to design new fabrics more rapidly to meet the demands of consumers. With the fourth industrial revolution bringing the digitalization and automation of manufacturing processes to a new level, the volume of data being stored has increased, enabling the use of Artificial Intelligence tools to enhance textile production processes. This PhD work was executed as part of a Research and Development (R&D) project. The main objective is the development of an Intelligent Decision Support System (IDSS), based on the Adaptive Business Intelligence concept, that will assist in the design of new textile fabrics through predictive and prescriptive analytics. To address this goal several experiments were performed. The first set of experiments was related to the prediction of two laboratory quality tests using an Automated Machine Learning (AutoML) tool. In the second predictive experiment, nine quality tests were selected, using an AutoML tool to compare the value of proposed input fabric yarn and finishing feature representations. Then, an initial prescriptive study was performed within the context of a different R&D project, aiming to develop an IDSS for production planning in the Textile industry. Using an AutoML tool for the predictive module, and NSGA-II to minimize both the cost and production time, the IDSS searched for the best subcontractor allocation plan, achieving interesting results. Based on the previously gained experience, the PhD main IDSS was developed, aiming to support the creation of new textile fabrics. For the predictive module, we compared a Single-Target Regression approach, obtained during the second predictive work, and a Multi-Target Regression, via a deep learning approach, for the prediction of four fabric physical properties and the final textile composition. For the prescriptive module, we compared two methods (NSGA-II and R-NSGA-II), aiming to simultaneously minimize the physical property predictive error and the distance of the optimized values to the learned input space. The full IDSS was evaluated using 100 new fabrics and then presented to the textile company experts, which provided positive feedback. |
|---|---|
| Autores principais: | Ribeiro, Rui Cândido Azevedo |
| Assunto: | Machine Learning Modern Optimization Predictive Analytics Prescriptive Analytics Evolutionary Computation Textile Industry |
| Ano: | 2023 |
| País: | Portugal |
| Tipo de documento: | tese de doutoramento |
| Tipo de acesso: | acesso aberto |
| Instituição associada: | Universidade do Minho |
| Idioma: | inglês |
| Origem: | RepositóriUM - Universidade do Minho |
| Resumo: | In the Textile industry there is a need to design new fabrics more rapidly to meet the demands of consumers. With the fourth industrial revolution bringing the digitalization and automation of manufacturing processes to a new level, the volume of data being stored has increased, enabling the use of Artificial Intelligence tools to enhance textile production processes. This PhD work was executed as part of a Research and Development (R&D) project. The main objective is the development of an Intelligent Decision Support System (IDSS), based on the Adaptive Business Intelligence concept, that will assist in the design of new textile fabrics through predictive and prescriptive analytics. To address this goal several experiments were performed. The first set of experiments was related to the prediction of two laboratory quality tests using an Automated Machine Learning (AutoML) tool. In the second predictive experiment, nine quality tests were selected, using an AutoML tool to compare the value of proposed input fabric yarn and finishing feature representations. Then, an initial prescriptive study was performed within the context of a different R&D project, aiming to develop an IDSS for production planning in the Textile industry. Using an AutoML tool for the predictive module, and NSGA-II to minimize both the cost and production time, the IDSS searched for the best subcontractor allocation plan, achieving interesting results. Based on the previously gained experience, the PhD main IDSS was developed, aiming to support the creation of new textile fabrics. For the predictive module, we compared a Single-Target Regression approach, obtained during the second predictive work, and a Multi-Target Regression, via a deep learning approach, for the prediction of four fabric physical properties and the final textile composition. For the prescriptive module, we compared two methods (NSGA-II and R-NSGA-II), aiming to simultaneously minimize the physical property predictive error and the distance of the optimized values to the learned input space. The full IDSS was evaluated using 100 new fabrics and then presented to the textile company experts, which provided positive feedback. |
|---|