1998 Fiscal Year Final Research Report Summary
LEGO-type Structural Design and Binocular Visual Interface
| Project/Area Number |
09651022
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (C)
|
| Allocation Type | Single-year Grants |
|---|
| Section | 一般 |
|---|
| Research Field |
海洋工学
|
|---|
| Research Institution | Osaka Prefecture University |
|---|
Principal Investigator |
KISHI Mitsuo College of Eng., Osaka Prefecture University, Assistant Prof., 工学部, 助教授 (00145814)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
YAMADA Yomoki College of Eng., Osaka Prefecture University, Assistant, 工学部, 助手 (90240027)
IZUMI Masao College of Eng., Osaka Prefecture University, Assistant Prof., 工学部, 助教授 (60223046)
|
|---|
| Project Period (FY) |
1997 – 1998
|
| Keywords | Structural design / Optimum design / Neural networks / Boltzmann machine / Visualization / Stereoscopy / Head mounted display / LEGO-type structure |
|---|
| Research Abstract |
LEGO is a toy block of North Europe origin. LEGO blocks play has some analogy with practical structure designs. Such structures are composed of the elements from finite sets on account of the industrial standard and for reasons of economical/building efficiencies. We call them by the name of 'LEGO-type structures.' Scaffolding structures are typical examples of the LEGO-type structure, and very large space structures may be also.
We classify LEGO-type structural designs into three modes : system design, connection design, and elements design. Considered in this study is the connection design mode for LEGO-type truss structures. LEGO-type blocks of the same kind are used, and the loading conditions are assumed to be given. Assembling the blocks we try to find the optimal configuration. The design variables to represent the structural configuration are discrete ; so that the design optimization is a discrete optimization problem.
Neural networks consist of mutually interconnected neurons. Some discrete optimization problems can be programmed and solved on artificial neural networks that minimize their own energy. The discrete variables are represented numerically by neurons. The neuron state takes binary values of one or zero. To escape from the local minimum we introduce the Boltzmann machine, which is a probabilistic neural networks model combined with the simulated annealing algorithm.
Numerical examples are provided to verify the applicability of the proposed method. In order to visualize the optimal structure, we introduce a virtual reality system using a head mounted display. Despite binocular stereopsis, it is not easy to recognize the configuration of a complicated truss structure.
|
