中文摘要：

为了使CSSO方法在系统级优化时能用于自主筛选来自各子学科间的设计变量的最优值以提高系统级优化效率和收敛速度,提出了一种改进型的并行子空间优化方法（improved concurrent subspace optimization,I-CSSO）。该方法在传统CSSO方法基础上通过增加一个自动筛选程序,即将各子学科优化方案及系统级优化方案进行了对比,从中优选出最优方案作为下一轮迭代的初始方案进行了优化,同时应用拉丁超立方试验设计方法采取样本点并构建了数据库,近似模型基于径向基神经网络模型。以某一机床主轴设计优化为例,基于Isight平台搭建了所设计的I-CSSO方法框架,并与传统的CSSO方法进行了对比。研究结果表明：所提出的I-CSSO方法有效提高了算法的计算效率。

英文摘要：

Aiming at selecting optimal design variables of each sub discipline and then passed them to system level when it was optimizing in CSSO method, improving the optimization efficiency and convergence speed of system level, an improved concurrent subspace optimization （I-CSSO） was proposed. An automatic screening process was added for the proposed method based on the traditional CSSO, that is, by comparing schemes from sub-disciplines and system level, an optimal scheme was determined and taken as the initial scheme for next iteration, meanwhile Latin hypercube of DOE method was used to set up a database, an approximate model was constructed by using radial basis func- tion neural Network. I-CSSO was incorporated into Isight platform and applied to the design of a machine tool spindle. The results indicate that the efficiency of the proposed I-CSSO gains over the conventional CSSO.