随机与区间不确定下基于近似灵敏度的序列

为解决复杂系统多学科可靠性设计优化过程中由于存在多源不确定性和多层嵌套而导致的计算效率低的问题，将近似灵敏度技术与两级集成系统综合策略(Bi-level integrated system synthesis，BLISS)和功能测度法集成，提出一种能同时处理随机和区间不确定性的序列化多学科可靠性设计优化方法。基于概率论和凸模型对混合不确定性进行量化，提出一种随机和区间不确定性下的混合可靠性评价指标，并基于功能测度法建立多学科可靠性设计优化模型。采用近似灵敏度信息替代实际灵敏度值，将近似灵敏度技术同时嵌入多级多学科设计优化策略和多学科可靠性分析方法中，避免每轮循环都进行全局灵敏度信息的分析与迭代，提高了计算效率。基于序列化思想同时将四层嵌套的多学科可靠性设计优化循环和三层嵌套的多学科可靠性分析过程进行解耦，形成一个单循环顺序执行的多学科可靠性设计优化过程，避免了每轮循环对整个可靠性分析模型进行迭代分析的过程，减少灵敏度分析和多学科分析次数。以汽车侧撞工程设计为例，验证了该法具有同时处理随机和区间不确定性的能力，并且计算效率较传统方法分别提高了10.98%和23.63%，表明该法具有一定工程实用价值。

Sequential Multidisciplinary Reliability Design and Optimization Based on Approximate Sensitivity Method under Random and Interval Uncertainties

To resolve the low computational efficiency problem of conventional reliability-based multidisciplinary design optimization (RBMDO), which is caused by the multi-source uncertainties and multi-nested loops during the design process of RBMDO. A new mixed uncertainties multidisciplinary design optimization (MUMDO) method is proposed by integrating the approximate sensitivity technique, bi-level integrated system synthesis strategy and performance measure approach, which can deal with both random and interval uncertainties simultaneously. The MUMDO model is formulated based on a new proposed reliability evaluation index and the mixed uncertainties quantification method using the probability theory and convex model. The approximate sensitivity information is employed to replace the real sensitivity value during the MUMDO, avoiding a large number of iterations of sensitivity calculations in each cycle. In addition, the four-layered nested MUMDO flowchart has been decoupled into a series of sequential execution of multidisciplinary design optimization and multidisciplinary reliability analysis based on the sequential optimization and reliability assessment (SORA). As a result, it is not necessary to evaluate the whole reliability model in each cycle, and a great number of sensitivity analysis and multidisciplinary analysis iterations are eliminated. Taking a vehicle side impact design as an example, the results show that the proposed method can deal with the random and interval uncertainties simultaneously. And also, the efficiency of the proposed method has been improved by 10.98% and 23.63% respectively compared to that of conventional methods. Therefore, it is valuable in engineering design and optimization.