| 作成者 |
|
|
|
|
|
|
|
| 本文言語 |
|
| アクセス権 |
|
| 権利関係 |
|
|
|
| 概要 |
A deep learning-based method is proposed for estimating the leg length of fillet welds produced by CO₂-shielded gas metal arc welding. The method uses welding log data, which correspond to welding cur...rent, welding voltage, and wire feed rate acquired from the welding power source during welding, together with sound generated near the molten pool. The welding current, welding voltage, and wire feed rate were converted into statistical features, while the sound was converted into acoustic features; these features were used as training data. Estimation accuracy was evaluated in two environments: a laboratory environment and an actual fabrication environment. Three input conditions were considered: (1) welding log data only, (2) sound data only, and (3) both welding log and sound data. The results showed that the combined use of welding log data and arc sound achieved the highest estimation accuracy.続きを見る
|
Mendeley出力
