| Summary: | 碩士 === 國防大學理工學院 === 機械工程碩士班 === 100 === In engineering optimization, the Design Of Experiment (DOE) method is usually used to reduce the times of simulation. The Response Surface Method (RSM) which is developed by the statistics of mathematics for simplify the relationship between design variables(x) and the response(f (x)) to find the best design in engineering model. Searching the optimization point with this method can save efforts of computer calculating process, but there still exist a difference between the solution of simulation and real reaction. If the size of difference is too large, it means that the optimization has less evidence to convince people. Up to the present, the major search method of the large-scale numerical optimization program always use Sequential Quadratic Programming (SQP), such method is used to analyze a general function with the search direction and step size in a lot of iterations for finding the best parameter solution.
Considering the computer technology, the performance of the hardware or stability of parallel computing technology has grown so much, so it can directly find the optimization point with the concept of the Finite Element Method (FEM) which simulates engineering models with the boundary and initial conditions in analysis software. After looking the entire analysis processes like a black box or a complex function, which can search the optimal design with traditional numerical methods, the final simulation results will get a strong reliability from the optimum design theory.
In this study, a general rail model (UIC 60) structure optimization is used as an example, and it will separate 2 methods, the RSM of DOE and directly numerical search method (SQP), to find the structure optimum design. Finally, compare the result of the 2 optimum methods to get the proof.
|
|---|
