| Project/Area Number |
12650246
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (C)
|
| Allocation Type | Single-year Grants |
|---|
| Section | 一般 |
|---|
| Research Field |
Intelligent mechanics/Mechanical systems
|
|---|
| Research Institution | Fukui University |
|---|
Principal Investigator |
YAMADA Yasuhiro FUKUI UNIVERSITY DEPT. OF MECHANICAL ENGINEERING, ASSOCIATE PROF., 工学部, 助教授 (40220412)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
MASUDA Masanobu INDUSTRIAL TECHNOLOGY CENTER OF FUKUI PREFECTURE, RESEARCHER, 機械電子部, 主任研究員
KOMURA Yoshiaki Fukui University DEPT. OF MECHANICAL ENGINEERING, PROF., 工学部, 教授 (00020214)
|
|---|
| Project Period (FY) |
2000 – 2001
|
| Project Status |
Completed (Fiscal Year 2001)
|
| Budget Amount *help |
¥2,600,000 (Direct Cost: ¥2,600,000)
Fiscal Year 2001: ¥1,100,000 (Direct Cost: ¥1,100,000)
Fiscal Year 2000: ¥1,500,000 (Direct Cost: ¥1,500,000)
|
|---|
| Keywords | Inspection / Neural Network / Wavelet Transform / Regression analysis / Reaction Force / 計算機援用検査 / 統計手法 / 力覚検査 / ニューラルネット / ウェーブレット |
|---|
| Research Abstract |
This paper looks at a system for inspection of the quality ofa probe's reaction force characteristics. This system, until now considered difficult to realize, automates the inspection method utilizing the touching of an inspector's finger.
1. Wavelet transformation and a neural network (NN) approach is applied to the system for probes to learn an inspector's finger judgment. We provide an input layer of a NN with thirty-three nodes corresponding to a time series of reaction forces of a probe and an output layer with one node corresponding to ajudgment ; being one of non-defective, defective, or unable to judge. From experimental results, the effectiveness of the system has been clarified. The inspection results of skilled inspectors were learned by the proposed wavelet-neural network system.
2. Regression analysis and a NN approach is applied to the system for probes to learn an inspector's finger judgment. We provide an input layer of a NN with forty-three nodes corresponding to a time series difference between reaction forces of the test probe and the best quality probe, and an output layer with three nodes corresponding to a judgment. From experimental results, the effectiveness of the system has been clarified. The inspection results of skilled inspectors were learned by the proposed regression analysis -neural network system.
3. A graphical presentation system is proposed. All performances of the regression analysis-neural network system are displayed in the graphical presentation system. The graphical presentation syste*m supports unskilled inspector' s training in inspection skills.
In time better trained neural networks will further improve the speed and accuracy of probe inspections, thereby reducing inspection costs.
|
