1992 Fiscal Year Final Research Report Summary
Development of Optimization Method for Redundant Structure Based on Reliability Assessment Considering Damage of Collapse Mode
| Project/Area Number |
03650364
|
|---|
| Research Category |
Grant-in-Aid for General Scientific Research (C)
|
| Allocation Type | Single-year Grants |
|---|
| Research Field |
船舶構造・建造
|
|---|
| Research Institution | Osaka University |
|---|
Principal Investigator |
MURAKAWA Hidekazu Osaka University, Welding Research Institute, Associate Professor, 溶接工学研究所, 助教授 (60166270)
|
|---|
| Project Period (FY) |
1991 – 1992
|
| Keywords | Reliability Assessment / Optimum Design / Damage of Collapse Mode / Significance of Optimization / Large Scale Structure / MonteCarlo Simulation / Learning Algorithm / Hierarchical Optimization |
|---|
| Research Abstract |
The fundamental philosophy of structural design can be divided into two according to the criterion which defines the failure of a structure. One is the allowable stress design based on the initial failure of individual member. The other is the design based on the ultimate strength of a structure. The design or the assessment of the safety of the structure can be further divided into deterministic and probabilistic approaches. However, regardless of these differences, the general goal of the design is an optimization of the structure with maintaining sufficient reliability.
In this research, optimum designs of truss structures based on the deterministic and the probabilistic estimation of the strength for both the initial failure and the ultimate failure criteria are studied. Especially, a new probabilistic optimization method based on the ultimate strength is proposed. It employs the idea of optimum distribution of safety indices for each failure mode. Further, the significance of the optimum design based on the reliability is discussed through the comparison with deterministic approaches. Also, to deal with large scale structures, a hierarchical optimization method using MonteCarlo simulation with learning algorithm is proposed.
|
Research Products
(3 results)
