基于网格实现的汽轮机基础优化设计

工程优化设计往往需要进行大规模的数值计算,拥有大量闲置资源的网格环境为建立这种高性能计算平台提供了可能.但是网格资源的动态性、异构性和分布性的本质特征,阻碍了网格技术在工程应用上的普及.为了利用网格环境中大量的闲置资源来协同解决实际工程中复杂的优化设计问题,建立了一个4层结构的高性能网格计算平台,并利用Kriging近似模型,在该平台上开发了以减轻基础重量和降低基础振幅为目的的多目标汽轮机优化设计的网格算法.使用该算法,在网格平台上对两个汽轮机基础进行了优化设计,与序列线性规划方法的结果比较表明所开发的优化算法有较高的计算精度.还分析了当使用不同数量的计算节点时网格的加速情况,说明所发展的优化方法能够在网格环境中高效地运行,搭建的网格平台也适合于工程优化设计.

Optimization Design of Turbine Engine Foundation on Grid

Optimization design plays a very important role in the engineering applications,which often involves huge computational effort and requires powerful computing environment.However,the distributed,dynamic and heterogeneous characteristics make it more difficult to make good use of the abundant idle resources.In order to integrate the massive idle resources to super-powerful environment by resources sharing and cooperative working on Internet,a four-layer high-performance grid platform is constructed for solving the complex engineering optimization problems.The dynamic analysis program developed is sealed to the nodes of the grid as black-box for computing dynamic response.An effective optimization method using Kriging model is proposed to do dynamic optimization design of the turbine engine foundation on the grid platform.The optimization problem is to find the design variables such that both the maximum dynamic displacement and structural weight are minimum and certain side constraints are satisfied.The cross sections of the beams and columns are considered as design variables.Kriging model is used to build the approximate mapping relationship between the forced vibration amplitude and design variables,reducing expensive dynamic reanalysis.Two engineering examples are carried out successfully on the platform using the proposed method,and the computing efficiency is compared with different number grid nodes.In comparison with the result of sequential linear programming,the optimization results show that the method has good accuracy.The speed-up is also analyzed when the nodes with different number are used showing that the method has very high efficiency for grid computing,and the grid platform constructed is suitable for the engineering optimization design.