| Project/Area Number |
07555511
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (A)
|
| Allocation Type | Single-year Grants |
|---|
| Section | 試験 |
|---|
| Research Field |
Material processing/treatments
|
|---|
| Research Institution | Saitama University |
|---|
Principal Investigator |
OSHIMA Kenji Saitama University, Graduate School of Science and Engineering, Professor, 大学院・理工学研究科, 教授 (70026087)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
KANJYOU Yoshihiro NKK Engineering Research Center, Engineer, 応用技術研究所, 技師
SUGITANI Yuji NKK Engineering Research Center, Chairman Engineer, 応用技術研究所, 主幹
KUBOTA Takefumi Himeji institute of Technology, Faculty of Engineering, Professor, 工学部, 教授 (40047585)
KANEKO Yasuyoshi Saitama University, Information Processing Center, Lecturer, 総合情報処理センター, 講師 (10233892)
YAMANE Satoshi Saitama University, Graduate School of Science and Engineering, Associate Profes, 大学院・理工学研究科, 助教授 (10191363)
|
|---|
| Project Period (FY) |
1995 – 1996
|
| Keywords | Intelligent welding robot / knowledge based information and control / Weld pool control / Cooling time control / Fuzzy inference / Neural network / Multi-layred fuzzy control / switch back welding |
|---|
| Research Abstract |
Knowledge based information processing and control of welding robots by using fuzzy neural network is discussed. The authors apply the fuzzy-neural network to estimate weld pool penetration depth and cooling time, and control welding current and cooling time by fuzzy control.
First, images of weld pool surface and gap are taken with CCD cameras, simultaneously. By using image processing, pool surface shape and gap are obtained. Next, the authors made neural networks to estimate the weld pool shape (the penetration depth and the back bead width) and heat input to base metals, which corresponds to the welding current. The penetration depth is detected by using the fuzzy neural network and the information such as the weld pool surface, the gap and the current in case of the variation of the gap. By using multi-layred fuzzy control, the current is controlled to keep the estimated penetration depth the desired value.
Neural networks are constructed to estimate the cooling time by using gap, current and welding speed. Training data of neural network are mede by carrying out numerical simulations concerning heat transfer in the weld pool. Current and welding speed are determined by using neural networks to get the desired cooling rate, and are applied to the welding. A good quality of the welding is obtained.
The authors consider the simulation result and propose switch back welding to get the stable back bead and desired cooling rate. The validity of the one-side back bead welding without backing is verified to carry out simulations and experiments.
|
