| Project/Area Number |
05402031
|
|---|
| Research Category |
Grant-in-Aid for General Scientific Research (A)
|
| Allocation Type | Single-year Grants |
|---|
| Research Field |
機械工作・生産工学
|
|---|
| Research Institution | KEIO University |
|---|
Principal Investigator |
INASAKI Ichiro Keio University, Faculty of Science and Technology, Professor, 理工学部, 教授 (30051650)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
AOYAMA Tojiro Keio University, Faculty of Science and Technology, Associate Professor, 理工学部, 助教授 (70129302)
|
|---|
| Project Period (FY) |
1993 – 1994
|
| Project Status |
Completed (Fiscal Year 1994)
|
| Budget Amount *help |
¥15,500,000 (Direct Cost: ¥15,500,000)
Fiscal Year 1994: ¥1,400,000 (Direct Cost: ¥1,400,000)
Fiscal Year 1993: ¥14,100,000 (Direct Cost: ¥14,100,000)
|
|---|
| Keywords | Grinding / Artificial intelligence / Genetic algorithm / Fuzzy reasoning / Neural network / Data base / Dressing / 人口知能 |
|---|
| Research Abstract |
The purpose of this study is to establish an intelligent learning model which can imitate the decision making process of skilled operators for setting up the grinding parameters. The function of the system is to provide both the dressing condition and the grinding condition which can achieve the required surface roughness and the specific grinding energy. The specific grinding energy has a decisive influence on the occurrence of grinding burn. Therefore, it is an important output to be observed.
In order to establish such intelligent learning model, we employs the neural network to imitate the associative memory of operators. The operator must adopt the grinding conditions with which he could achieve successful results in his experiences. Such process is imitated in our system by applying two different types of neural network i.e., a conventional Feed Forward Network and a Brain-State-in-a-Box Network. These two networks were combined into a hybrid network to imitate the associative memory of operators. The system is able to provide a combination of dressing and grinding condition which can meet the required surface roughness. The effectiveness of the proposed system was confirmed through a series of computer simulations.
The second system proposed has an ability to imitate the learning function of operators. In order to achieve such function, we applied a genetic algorithm and a fuzzy rule. The system can learn causalities between the input and the output in the grinding process through practical operations. The system can, consequently, establish a grinding data base. The availability of the system was confirmed through both the computer simulations and the practical grinding experiments.
|
