基于近似模型的螺旋轴式移箱机构多学科设计优化

Multidisciplinary design optimization of screw axis type moving-box mechanism based on surrogate models

针对移箱机构优化设计中存在多目标、强耦合及非线性问题,以螺旋轴式移箱机构为研究对象,开展了基于近似模型的移箱机构多学科设计优化研究。将移箱机构多学科优化设计分解为运动学、动力学和结构力学3个学科,进行了学科分析;采用拉丁超立方试验设计方法,产生初始样本点,构建了系统级变量与学科级优化之间的Kriging近似模型;采用协同优化方法,搭建了基于iSIGHT软件的螺旋轴式移箱机构多学科设计优化集成平台,采用多岛遗传算法和序列二次规划法相结合的混合算法优化求解,并进行了验证。结果表明,优化后的移箱机构滑套最大加速度由1.286e5 mm·s^(-2)减至9.253e4 mm·s^(-2),降低了28.05%;螺旋轴一阶固有频率由238.265 Hz增至264.538Hz,增加了11.03%;机构质量由2265.318g减小至1848.310g,降低了18.41%。采用近似技术进行移箱机构多学科设计优化,在保证计算精度的前提下,降低了求解时间,而且优化后的移箱机构整体性能最优。

In view of multi-target, strong coupling, nonlinear problems in optimization design of moving-box mechanism, based on the surrogate models, the research on multidisciplinary design optimization of moving-box mechanism was executed with the screw axis type moving-box mechanism as the research object. The multidisciplinary design optimization of moving-box mechanism, decomposed into the 3 disciplines of kinematics, dynamics and structural mechanics were analyzed. The Latin hypercube design method was used to generate the initial sample points and the Kriging model between the system level variables and the discipline level optimization was established. Based on iSIGHT software, the multidisciplinary design optimization platform on the moving-box mechanism was set up by using collaborative optimization method and calculated by using the hybrid algorithm of the multi-island genetic algorithm method and the sequence two quadratic programming method. The optimization results showed that the maximum acceleration of sliding sleeve was reduced to 9.253 e4 mm·s^-2 by 1.286 e5 mm·s^-2, which was decreased by 28.05%;the first natural frequency was increased from 238.265 Hz to 264.538 Hz, which was increased by 11.03%;the weight of the mechanism was reduced to 2 265.318 g by 1 848.310 g, which was reduced by 18.41%. Thus, by using the approximation technique and multidisciplinary design optimization, the overall performance of moving-box mechanism was most optimal under the higher accuracy of computation and the lower computing time.