基于可行方向序列无约束极小化技术外点法的改进协同优化策略

Enhanced Collaborative Optimization Strategy Based on Feasible Direction SUMT Exterior Method

指出准协同优化策略(Collaborative optimization,CO)存在的数值缺陷及其原因。针对系统级优化不满足Kuhn-Tucker条件所导致的计算困难,提出一种基于可行方向序列无约束极小化技术(Feasible direction sequential unconstrainedminimization technology,FD-SUMT)外点法的改进协同优化策略(Enhanced collaborative optimization with FD-SUMT method,ECO-FSM)。在系统级优化中使用FD-SUMT外点法,该方法不依赖Lagrange乘子并且能够将系统级设计变量限定在设计变量可行域内,避免传统SUMT外点法设计变量越界所导致的异常。利用学科间动态不一致信息更新系统级优化中的罚因子以加速学科间的协调。利用测试问题检验ECO-FSM的性能,并与其他的CO进行比较研究。研究结果表明ECO-FSM消除了系统级优化中设计变量越界的现象,收敛性、数值稳定性以及收敛速度得以显著提高。将ECO-FSM用于亚声速喷气式客机总体方案优化设计,优化结果表明ECO-FSM具有工程实用性。

The numerical drawbacks of standard collaborative optimization strategy （CO） and their reasons are pointed out. As to resolve the computationally difficulty caused by breaking Kutm-Tucker condition in system level optimization, an enhanced collaborative optimization strategy with feasible direction sequential unconstrained minimization technology （FD-SUMT） exterior method （ECO-FSM） is proposed. FD-SUMT exterior method independent of Lagrange multiplier is employed in system level optimization, which restricts the system level design variables in their feasible region and eliminates the singularity phenomenon resulted from the boundaries violation of design variables in traditional SUMT exterior method. The penalty factor is updated in terms of the dynamical inconsistence information among different disciplines to accelerate the inter-disciplinary coordination. Several test problems are used to validate the performance of ECO-FSM, moreover, comparative study with other enhance CO is also carried out. Study results reveal that boundaries violation of design variables in system level optimization is eliminatedisFD-SUMTirection during ECO-FSM optimization process, and convergence performance, numerical robustness and convergence speed are evidently upgraded as well. Eventually, ECO-FSM is verified to be practical for engineering problems through a conceptual optimization design of subsonic jet transport using ECO-FSM.