Publicação
Use of machine learning to improve the lot release process and enhance Quality 4.0 adoption
| Resumo: | In modern manufacturing, digital transformation and the integration of advanced analytics are reshaping production processes to meet stricter quality and efficiency demands. The final stage of the lot release process is critical in manufacturing, ensuring that products meet quality standards before shipment. With Industry 4.0, Quality 4.0 has emerged, integrating advanced technologies to enhance quality management systems. This thesis, carried out as part of a project with BOSCH Car Multimedia, explores the application of machine learning to improve the lot release process, contributing to the advancement of Quality 4.0 in manufacturing environments. The research aims to develop a machine learning framework to predict customer complaints using automated test and repair data from automotive production lines. Using the Action Research methodology in conjunction with the Cross-Industry Standard Process for Data Mining, a comprehensive dataset of production line data was analyzed. Four machine learning models were selected and compared using cost-sensitive learning and threshold-moving techniques to address class imbalance. In addition, a non-sampling threshold-moving approach was proposed to adapt predictive capabilities to specific product requirements. Model performance was assessed using the F1-Score and the Matthews correlation coefficient. The results demonstrate the effectiveness of the machine learning framework. Specifically, when comparing the different machine learning models analyzed, XGBoost achieves the best result with an F1 score of 72.4% and a Matthews correlation coefficient of 75%. This data-driven approach improves the lot release decision process over heuristic methods and supports Quality 4.0. The proposed framework was tested on real-world data and offers promising results in improving lot release decisions and reducing customer complaints. By also enabling threshold adjustments without retraining, it provides flexibility and efficiency, advances predictive capabilities, and enhances Quality 4.0 adoption. |
|---|---|
| Autores principais: | Lobo, Armindo Augusto Gonçalves |
| Assunto: | Imbalanced Datasets Lot Release Machine Learning Quality 4.0 Aprendizagem Automática Dados Desiquilibrados Libertação de Lotes Qualidade 4.0 Engenharia e Tecnologia::Engenharia Eletrotécnica, Eletrónica e Informática |
| Ano: | 2025 |
| 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 modern manufacturing, digital transformation and the integration of advanced analytics are reshaping production processes to meet stricter quality and efficiency demands. The final stage of the lot release process is critical in manufacturing, ensuring that products meet quality standards before shipment. With Industry 4.0, Quality 4.0 has emerged, integrating advanced technologies to enhance quality management systems. This thesis, carried out as part of a project with BOSCH Car Multimedia, explores the application of machine learning to improve the lot release process, contributing to the advancement of Quality 4.0 in manufacturing environments. The research aims to develop a machine learning framework to predict customer complaints using automated test and repair data from automotive production lines. Using the Action Research methodology in conjunction with the Cross-Industry Standard Process for Data Mining, a comprehensive dataset of production line data was analyzed. Four machine learning models were selected and compared using cost-sensitive learning and threshold-moving techniques to address class imbalance. In addition, a non-sampling threshold-moving approach was proposed to adapt predictive capabilities to specific product requirements. Model performance was assessed using the F1-Score and the Matthews correlation coefficient. The results demonstrate the effectiveness of the machine learning framework. Specifically, when comparing the different machine learning models analyzed, XGBoost achieves the best result with an F1 score of 72.4% and a Matthews correlation coefficient of 75%. This data-driven approach improves the lot release decision process over heuristic methods and supports Quality 4.0. The proposed framework was tested on real-world data and offers promising results in improving lot release decisions and reducing customer complaints. By also enabling threshold adjustments without retraining, it provides flexibility and efficiency, advances predictive capabilities, and enhances Quality 4.0 adoption. |
|---|
Atividades financiadas
Carregando projetos financiados...