| Large column structure, which occupies an important proportion in weight, is a majorpart and the supporting part of a lifting system. The current structure is so heavy that they arehard to be assembled and transported. Therefore, It is necessary to optimization this structureon the basis of meeting the requirements of strength and rigidity. Because of the large numberof design parameters of the large column, it will take long time to optimize them when use thefinite element model directly, and the efficiency of optimization is too low. In this paper thisproblem can be solved when use surrogate model instead of the traditional Finite elementanalysis. This paper mainly carried out from the following aspects to study the issue:Firstly, in this paper， the development status of optimization algorithm and the theoryof surrogate model strategy were introduced. A comparison study was conducted for the basicprinciple and application environment of the common used design of experiment andsurrogate model.Secondly, according to relevant data provided by the enterprise, the large columnstructure parametrical finite element model was established by using SOLIDWORKS andWorkbench. According to the practical working condition, the static analysis of the largecolumn model under four different working conditions were researched by Workbench. In thestatic analysis used the remote force to represent the computational load. After static analysisthe most dangerous working conditions was obtained, at the same time it verify the existingstructure had large optimization space.Afterwards, through the sensitivity analysis of big column structure parameters selectedseven design factors to construct surrogate model.729groups of samples were obtained byusing the orthogonal test design. The total displacement and equivalent stress value of thecorresponding sample point were obtained by combining the SOLIDWORKS and Workbench.Evaluate the total displacement and equivalent stress surrogate model by using40groups oftest samples which were generated randomly. After all,8groups of multiple correlationcoefficient and root mean square error were obtained. Selected the higher precision polynomial response surface equivalent stress surrogate model and polynomial responsesurface total displacement surrogate model to replace the large column structure finiteelement model for the optimization calculation.Finally, in the optimization of large column structure pursing minimum weight as theoptimization goal, regard the polynomial response surface equivalent stress surrogate modeland polynomial response surface total displacement surrogate model may not exceed theallowable value as constraint condition in genetic algorithm.The results show that, the optimized model meets the design strength and stiffnessrequirement and its weight reduced27%, and the operation time was greatly reduced too. Thecontent of dissertation provides the optimization principle and method of the structure for theenterprise and has certain engineering application value. |
