Quality prediction in molded door skin using data mining
A door skin is a high-density wooden board and is the main component in the manufacture of doors. To ensure its commercialization, it must comply with demanding quality standards, the main one that measures the force necessary to detach the door skin from the structure of a door. Quality tests are c...
| Autores Principales: | Troncoso-Espinosa, Fredy Humberto, Castro-Albornoz, Karen |
|---|---|
| Formato: | Artículo |
| Idioma: | Español |
| Publicado: |
Editorial Tecnológica de Costa Rica (entidad editora)
2021
|
| Materias: | |
| Acceso en línea: |
https://revistas.tec.ac.cr/index.php/tec_marcha/article/view/5395 https://hdl.handle.net/2238/13584 |
| Sumario: |
A door skin is a high-density wooden board and is the main component in the manufacture of doors. To ensure its commercialization, it must comply with demanding quality standards, the main one that measures the force necessary to detach the door skin from the structure of a door. Quality tests are carried out every two hours and the results are obtained after five hours. If the results show that the door skins are outside the required quality standard financial losses are generated during this waiting time. This research proposes the use of data mining using machine learning techniques to continuously predict this measure of door skin quality and reduce the economic losses associated with waiting for quality tests. For the use of data mining, a database was created using historical record of the variables of the production process and quality tests. The methodology used is the discovery of knowledge in KDD databases (Knowledge Discovery in Databases). The application of this methodology allowed identifying the main variables that affect the quality of the door skin and training four machine learning algorithms to predict the quality. The results show that the algorithm that best performance is Neural Net and allows to show that the implementation of the Neural Net algorithm will reduce the economic losses associated with waiting for the results of the quality tests. |
|---|