Project/Area Number |
01850105
|
Research Category |
Grant-in-Aid for Developmental Scientific Research (B).
|
Allocation Type | Single-year Grants |
Research Field |
土木構造
|
Research Institution | Tokai University |
Principal Investigator |
KITAHARA Michihiro Tokai University, School of Marine Science and Technology, Associate Professor, 海洋学部, 助教授 (60135522)
|
Co-Investigator(Kenkyū-buntansha) |
NOTAKE Masayoshi Mitsubishi Research Institute, Mathematical Engineering Dept., Manager, 数理工学部, 室長
SUZUKI Hideo Ono Sokki, Acoustics Laboratory, Manager, 音響技術研究所, 部長
HAMADA Masanori Tokai University, School of Marine Science and Technology, Professor, 海洋学部, 教授 (30164916)
SAKODA Shigemi Tokai University, School of Marine Science and Technology. Lecturer Professor, 海洋学部, 講師 (50056230)
|
Project Period (FY) |
1989 – 1990
|
Project Status |
Completed (Fiscal Year 1990)
|
Budget Amount *help |
¥5,300,000 (Direct Cost: ¥5,300,000)
Fiscal Year 1990: ¥1,000,000 (Direct Cost: ¥1,000,000)
Fiscal Year 1989: ¥4,300,000 (Direct Cost: ¥4,300,000)
|
Keywords | Nondestructive evaluation / Elastic wave theory / Knowledge base / Ultrasonics / Artificial intelligence / Neural network / AI / 超音波散乱法 |
Research Abstract |
A system for the ultrasonic nondestructive evaluation was proposed in this study. The system is based on the ultrasonic knowledge base by the theoretical and experimental characterizations of waveforms from defects in a structural component. The way to construct the knowledge base has the following three steps : (1) Numerical simulation of theoretical scattering waveforms from defects. (2) Calibration of theoretical waveforms by experimental waveforms. (3) Developments of knowledge base by the use of theoretical waveforms. The performance of the proposed nondestructive evaluation system was confirmed in the following process : (1) Developments of a neural network system which can extract features of waveforms. (2) Learning of network by the use of knowledge base based on theoretical waveforms. (3) Performance check of the system by the experimental data input. It was confirmed that the proposed neural network system has the versatility to evaluate the features of waveforms from defects and to detect the defects. The theoretical knowledge base calibrated by the experimental data provides wide range of synthetic data for the neural network evaluation system.
|