The thesis examines the possibilities of improving PVC window production from the perspective of lean manufacturing and reducing environmental impacts. An analysis of the current state was conducted using Value Stream Mapping (VSM), time analysis, simulation and EHS analysis. The environmental impact was assessed by calculating the CO₂ equivalent, while production waste was analyzed according to its causes. With proposed improvements, simulated production increased from 130 to 143 windows per day. A YOLO model was developed and evaluated for detecting scratches on PVC profiles. On the easier validation set, the model detected approximately 85 % of scratches, while recall dropped to 46 % on the more difficult set. Early detection enables a reduction in CO₂ emissions associated with production rejects.
|
