Classification of causes for plastic product defects by machine learning and application for the training of workers
In recent years, with the maturing of society and advances in technology, consumers' demand for manufacturing has been increasing. Of particular note in demand is the appearance quality of the product. Under visual inspection at the time of manufacture, skilled workers have classified the c...
Guardado en:
| Autores principales: | , |
|---|---|
| Formato: | article |
| Lenguaje: | EN |
| Publicado: |
The Japan Society of Mechanical Engineers
2021
|
| Materias: | |
| Acceso en línea: | https://doaj.org/article/e396d4e2049745f6a5ee4fd2ba972446 |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| Sumario: | In recent years, with the maturing of society and advances in technology, consumers' demand for manufacturing has been increasing. Of particular note in demand is the appearance quality of the product. Under visual inspection at the time of manufacture, skilled workers have classified the causes of defects in the product's appearance according to their experience and have dealt with them quickly. Due to a serious shortage of workers and the aging of skilled workers, there are many opportunities for inexperienced workers, such as foreign technical interns, to take charge of the work at manufacturing sites. While introducing "automatic visual inspection by a camera," the authors have developed a system that can automatically classify the causes of defects. Also, when the classification work was carried out for the inexperienced workers, the application to the education was seen. The authors propose a classification method based on machine learning with skilled workers' knowledge. This paper analyzes the process of classifying production data into the causes of contaminated products (CP) by skilled workers. First, the occurrence interval of CP was divided into sparse or dense groups. Second, a decision tree learned from causes' labels with a skilled worker was developed as a group label classifier (GLC). When production data were used to validate the prediction capability, high accurate predictions were obtained. This indicates that even inexperienced workers can take measures according to the cause of product defects during production, which is useful for the education of field workers by the GLC. |
|---|