| Project/Area Number |
13650491
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (C)
|
| Allocation Type | Single-year Grants |
|---|
| Section | 一般 |
|---|
| Research Field |
Control engineering
|
|---|
| Research Institution | KYUSHU UNIVERSITY |
|---|
Principal Investigator |
FURUZUKI Takayuki KYUSHU UNIVERSITY, Graduate School of Information Science and Electrical Engineering, Research Associate, 大学院・システム情報科学研究院, 助手 (50294905)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
MURATA Junichi KYUSHU UNIVERSITY, Graduate School of Information Science and Electrical Engineering, Associate Professor, 大学院・システム情報科学研究院, 助教授 (60190914)
HIRASAWA Kotaro Waseda University, Graduate School of Information, Production and Systems, Professor, 大学院・情報生産システム研究科準備室, 教授 (70253474)
|
|---|
| Project Period (FY) |
2001 – 2002
|
| Project Status |
Completed (Fiscal Year 2002)
|
| Budget Amount *help |
¥2,400,000 (Direct Cost: ¥2,400,000)
Fiscal Year 2002: ¥1,100,000 (Direct Cost: ¥1,100,000)
Fiscal Year 2001: ¥1,300,000 (Direct Cost: ¥1,300,000)
|
|---|
| Keywords | Nonlinear model / Neural networks / Neuro-fuzzy networks / Learning / Prior knowledge / Nonlinear control / Robust control / Fault diagnosis / 線形構造 / 線形制御技術 |
|---|
| Research Abstract |
Neural networks have recently attracted much interest in system control community because they learn any nonlinear mapping. However, from a user's point of view, neural networks are not user friendly, That is, they are not easy-to-use ; more specifically they do not have structures favorable to the applications of system control and fault diagnosis. To solve these problems, the following studies have been carried out.
1. A modeling scheme has been developed, which consists of two parts : a macro-net part and kernel-net part. The macro-net part is a user-friendly interface constructed using application specific knowledge and the nature of network structure. The kernel-net part is a flexible multi-input-multi-output (HIMO) nonlinear model such as neural networks and neurofuzzy networks.
2. An optimization scheme has been developed. The scheme consists of two learning loops. It has been studied to increase robustness of the algorithm to local minima and over-fitting, by using such as homotopy and hierarchical techniques.
3. Applications of the proposed modeling scheme to controller design and fault detection of nonlinear systems have been studied. Some new approaches are proposed and confirmed through numerical simulations.
|
