Some Research On Engineering Structure Optimization Algorithm And On Its Application

Traditional structure design is based on the experiences of designer and completed by experiments. Then a long design period would be taken and lots of materials would be wasted. The structure designs or structure modifications, combined with the structure optimization, can solve this problem. The structure optimization, based on the structure analysis, using certain criteria or optimization method, modifies the structure shape, geometry dimension and material layout, in the purpose of satisfying certain performance and reducing the costs. Nowadays, the structure optimization, with the help of the large commercial finite element software, achieves the virtual design of the products, which is a great achievement. However, the commercial finite element software is composed of the common program, which some times does not have the ability to analyze for specific problem, facing many complicated engineering problems. And the present computation method has a low efficiency, which cannot satisfying the engineering practical needs completely.There are three ways available to solve the problems: firstly, make the second development on the software that has been widely used, secondly, use the own programs, thirdly, make the full use of the software in existence.There are lots of problems when the second development is made. But experience is got during the study process. MSC.Nastran software is a widely used one in the world which is based on CAE. This software can solve almost every aspect of linear problems. The input and output format have become the surveyor's pole in the structure analysis field. DMAP as the advanced computer language of MSC.NASTRAN has been made all the solving sequences. To be familiar with or to mast this language can help us use the software better and when modification is made on the exiting sequences more functions will be available. In this paper the grammars and usage of the language are described to present the whole view to the readers. At last examples are given.Design of experiment method (DOE)which is based on statistics can get a particular information of the system via a series of simulation of the physics system or computer system and create a model to describe the characters of the system. Hyperworks is a famous engineering software system in the world and integrate of the CAE preceding process, analysis process and post process. The software of Hyperstudy consists of DOE study, optimization study and stochastic study. In this paper the relationship between design variables with structure response is got by DOE study. The global optimization is solved by Sequential Response Surface Method in section of optimization study. All the results are used to make the preceding process of the size optimization solution, then the optimal result will be got.The optimization method, with another name of the mathematical planning method, is a subject of strong application. It is classified into group optimization, linear planning, non-linear planning and so on. Many problems in the engineering are the constraint nonlinear problems. There are three methods for resolving the nonlinear problems: the first method is to directly analyze the constraint optimization problems without the transformation; the second method is to transform the nonlinear constraint optimization to the non-constraint optimization; the third method is to modify the current iteration point by resolving the simple approximate optimization problems, which is constructed by utilizing the relating information of the original nonlinear constraint problems. When the scale of the problems is not very big, the outside penalty function method has the advantages of simple idea, good convergence and is easy to apply. The outside penalty function method is chosen in this paper, combined with the finite element method. It is programmed and is simply applied to optimize the structure.Design of experiment (DOE) which is mainly based on statistics can describes the objective system by a series of experiments or simulation about the model. HyperWorks is famous software in engineering field in the world which gathers the CAE pre-processing, solving and post-processing. Hyperstudy which is one part of HyperWorks is mainly in DOE method. There are three modules in Hyperstudy including DOE study, Optimization study and Stochastic study. In the paper the design variables and the constraints are got by DOE study module firstly, and secondly the global convergent approximation is generated by Sequential Response Surface Method in Optimization study module, and thirdly the constraints are get from the Stochastic study module. Then the optimal result is got by using Optistruct. Finally the door of a car with an objective that minimizes the weight takes the length, wide and highness with lower and upper boundaries for the design parameters. The weight of the door is reduced without any reduce of the landscape orientation stiffness. The vertical stiffness is improved, and a lot of material can be saved when volume-produce.