Exploratory data analysis and defects prediction in a glass manufacturing company
Files
Farias_75462000_2023.pdf
Closed access - Adobe PDF
- 8 MB
Farias_75462000_2023.pdf
Closed access - Adobe PDF
- 8 MB
Details
- Supervisors
- Degree label
- Abstract
- This thesis focuses on data exploration and prediction using models within the context of a glass manufacturing company. Among the various defects that can occur in glass, one with a significant impact is the presence of seeds, which adversely affect glass quality and lead to financial losses. Understanding the formation of seeds is crucial for minimizing their occurrence during the manufacturing process. This thesis aims to develop a model capable of predicting the number of seeds, employing two independent approaches: time series modeling and machine learning modeling. The project follows two previous stages: data preprocessing and exploratory data analysis. The time series model is based solely on previous quantities of seeds, while the second approach utilizes a machine-learning model based on other variables. The machine learning model accurately identifies seed quantities surpassing a quality threshold. The findings provide valuable insights for process optimization and quality control in glass manufacturing, offering potential solutions to reduce seed-related defects.