基于可靠性的多学科设计优化及其在机构设计中的应用

发布时间：2018-04-25 03:41

  本文选题：多学科设计优化 + 序列优化与可靠性评估　； 参考：《电子科技大学》2014年博士论文

【摘要】：现代科学技术的日新月异,使得工程系统愈发复杂化,其具体设计过程往往涉及诸多学科,且学科之间的联系耦合紧密。为了解决传统设计优化方法的局限性,多学科设计优化(Multidisciplinary Design Optimization,MDO)应运而生。MDO方法在充分考虑耦合学科之间协同效应的同时,从工程系统全局的角度进行设计优化,实现提高系统综合性能,缩短研发周期并降低生产成本的目的。不确定性因素广泛地存在于实际工程系统中。特别是在复杂耦合的工程系统中,不确定性因素会随着耦合信息的传播而累积,最终对工程系统的综合性能产生影响,给工程系统的可靠性、稳定性以及安全性带来隐患。为了在设计优化过程中有效考虑这些不确定性因素的影响,基于可靠性的多学科设计优化(Reliability based Multidisciplinary Design Optimization,RBMDO)已成为现代工程系统设计的研究热点之一。迄今为止,对于考虑随机不确定性的RBMDO方法,在结合经典概率论等可靠性分析方法后日趋成熟。同时,由于序列优化与可靠性评估(Sequential Optimization and Reliability Assessment,SORA)策略的采用,使得可靠性分析过程与设计优化过程相互解耦,整个RBMDO过程分解为一系列相互交替进行的确定性MDO与可靠性分析过程,运算效率得以进一步提升。基于SORA策略,本文分别从“RBMDO问题中的确定性MDO方法创新”和“不同可靠性分析方法在RBMDO问题中的引入与应用”两个方面展开研究。具体地,利用大系统递阶控制理论与方法在处理复杂系统协调问题中的策略,针对SORA下RBMDO中的确定性MDO方法进行创新研究;利用鞍点近似方法(Saddlepoint Approximation,SPA)对一阶可靠性方法评估精度的改进,以及利用子集模拟可靠性分析方法(Subset Simulation Reliability Analysis,SSRA)对小概率失效事件的可靠性分析评估效率的改进,分别将上述两种新的可靠性分析方法引入并应用到SORA下RBMDO中的可靠性分析环节。两个方面的理论研究工作围绕SORA策略下RBMDO整体的计算效率提高与优化精度改进展开,拟拓展和完善现有RBMDO的理论体系。最后,在利用已有和上述理论研究成果的基础上,以某型号变体飞行器的翻转驱动机构为例,展开RBMDO方法在机构设计优化中的工程应用研究。系统地分析该机构的多学科耦合特性以及不确定性因素影响。完成该机构的RBMDO同时,与原始设计方案进行对比分析,对优化后的设计方案进一步地进行研究。本文的研究成果主要体现在如下四个方面:(1)关联预测优化(interactionpredictionoptimization,ipo)与关联平衡优化(interactionbalanceoptimization,ibo)。利用大系统递阶控制理论中的关联预测控制与关联平衡控制两种方法,分别结合协同优化(collaborativeoptimization,co)方法的子学科分布式并行优化和协调策略,提出两种新的mdo方法用于sora策略下rbmdo中的确定性优化,避免了co方法中的相容性约束给原rbmdo问题增加非线性程度的缺陷。所提出方法与co方法相比,学科间的协调策略更加简单,对于优化问题的求解具有更高的运算效率和精度。通过算例验证了所提出方法的有效性。(2)基于一阶鞍点近似方法的多学科设计优化(firstordersaddlepointapproximationbasedmdo,fospa-mdo)。当优化问题中存在随机不确定性时,采用一阶鞍点近似(firstordersaddlepointapproximation,fospa)方法对sora策略下rbmdo中的确定性mdo优化结果进行可靠性分析与评估,不需要将随机变量转化成标准正态空间中的标准正态分布随机变量,避免了由于采用传统一阶或二阶可靠性方法所带来的增加原优化问题非线性程度的缺陷。在搜索获得似然设计验算点(mostlikelihoodpoint,mlp)后,利用随机变量与参数的移动向量建立新一轮运算中的确定性优化模型。所提出方法与基于一阶可靠性方法(firstorderreliabilitymethod,form)的rbmdo相比,在保证计算效率的同时,拥有更加准确的可靠性评估精度。通过算例验证了所提出方法的有效性。(3)基于子集模拟可靠性分析方法的多学科设计优化(subsetsimulationreliabilityanalysisbasedmdo,ssra-mdo)。当优化设计对象系统拥有极高的可靠性,需要对其发生小概率失效事件进行可靠性评估时,为了能准确且高效地完成rbmdo,采用基于子集模拟可靠性分析方法(subsetsimulationreliabilityanalysis,ssra)对sora策略下rbmdo中的确定性mdo优化结果进行可靠性分析与评估。该方法将原失效概率计算问题转变为一系列发生概率较高的条件失效概率计算问题,通过马尔可夫链蒙特卡罗仿真(markovchainmontecarlosimulation)方法分别计算各个中间条件失效概率。在搜索获得模拟设计验算点(simulationmostprobablepoint,smpp)后,利用随机变量与参数的移动向量建立新一轮运算中的确定性优化模型。所提出方法与基于传统蒙特卡洛仿真(montecarlosimulation,mcs)的rbmdo相比,在保证计算精度的同时,使用更少数量的仿真样本点,拥有更加高效的计算效率。通过算例验证了所提出方法的有效性。(4)某翻转驱动机构的考虑随机不确定性的多学科设计优化。根据某翻转驱动机构的组成与运动原理,将该机构划分为动力输入学科与动力传递学科。在分别对各个构件以及机构整体进行有限元分析与动力学分析的基础上,构造相应的性能函数响应面。分析实际工程中的不确定性来源,建立该翻转驱动机构的RBMDO模型并利用本文所提出的SSRA-MDO方法在SORA策略下进行优化求解。将优化后的设计方案与原始设计方案进行比较分析,进一步阐述了所得结果的合理性。
[Abstract]:With the rapid development of modern science and technology, the engineering system is becoming more and more complex, and its specific design process often involves many disciplines, and the connection between the disciplines is closely coupled. In order to solve the limitations of the traditional design optimization method, Multidisciplinary Design Optimization (MDO) has emerged as the times require the.MDO method. Considering the synergistic effect among the coupling subjects, the design optimization is carried out from the point of view of the overall engineering system to improve the comprehensive performance of the system, shorten the R & D cycle and reduce the cost of production. The uncertain factors exist widely in the actual engineering system. Especially in the complex coupled Engineering system, the uncertainty factors will follow The accumulation of coupling information will eventually influence the comprehensive performance of the engineering system and bring hidden dangers to the reliability, stability and security of the engineering system. In order to consider the influence of these uncertainties effectively in the process of design optimization, Reliability based Multidisciplinary based on Reliability Design Optimization, RBMDO (RBMDO) has become one of the hot topics in modern engineering system design. To date, the RBMDO method considering random uncertainty is becoming more and more mature after the combination of the classical probability theory and other reliability analysis methods. At the same time, because of the sequence optimization and reliability evaluation (Sequential Optimization and Reliability Assessment, SORA) the adoption of the strategy makes the reliability analysis process and the design optimization process decouple each other. The whole RBMDO process is decomposed into a series of alternative deterministic MDO and reliability analysis processes, and the operational efficiency is further improved. Based on the SORA strategy, this paper separately from "the deterministic MDO method in the RBMDO problem" and "no". The introduction and application of reliability analysis method in the RBMDO problem is studied in two aspects. Specifically, the strategy of the large system hierarchical control theory and method in the coordination of complex systems is used to study the deterministic MDO method in RBMDO under SORA; and the saddle point approximation (Saddlepoint Approximation, SP) is used. A) improvement on the evaluation accuracy of first order reliability method and the improvement of reliability analysis and evaluation efficiency of small probability failure events by using the Subset Simulation Reliability Analysis (SSRA) method (Analysis, SSRA). The two new reliability analysis methods are introduced and applied to the reliability of RBMDO in SORA. The two aspects of the theoretical research work around the overall calculation efficiency of RBMDO and the improvement of the optimization precision under the SORA strategy. The theoretical system of the existing RBMDO is expanded and perfected. Finally, on the basis of the existing and above theoretical research results, the RBMDO square is carried out with the overturning drive mechanism of a certain type of variant aircraft. The study of engineering application in the optimization of mechanism design. A systematic analysis of the multidisciplinary coupling characteristics and uncertainty factors of the mechanism is made. The RBMDO of the organization is completed and the original design scheme is compared and analyzed. The optimized design scheme is further studied. The results of this paper are mainly reflected in the following four Aspects: (1) association prediction optimization (interactionpredictionoptimization, IPO) and association equilibrium optimization (interactionbalanceoptimization, IBO). Using the two methods of associated predictive control and association balance control in the hierarchical control theory of large systems, the sub Discipline Distribution of the cooperative optimization (collaborativeoptimization, CO) method is combined respectively. In parallel optimization and coordination strategy, two new MDO methods are proposed for the deterministic optimization of RBMDO under the Sora strategy, which avoids the defect that the compatibility constraint in the co method increases the nonlinearity of the original RBMDO problem. Compared with the co method, the coordination strategy between the subjects is more simple, and the solution of the optimization problem is higher. Efficiency and accuracy. The effectiveness of the proposed method is verified by an example. (2) the multidisciplinary design optimization (firstordersaddlepointapproximationbasedmdo, fospa-mdo) based on the first order saddle point approximation (fospa-mdo). When the stochastic uncertainty exists in the optimization problem, the first order saddle point approximation (firstordersaddlepointapproximation, fospa) square is used. The reliability analysis and evaluation of the deterministic MDO optimization results in RBMDO under the Sora strategy do not need to convert random variables into standard normal distribution random variables in standard normal space, and avoid the defects caused by the use of traditional first or two order reliability methods to increase the nonlinear degree of the original optimization problem. After obtaining the likelihood design check point (mostlikelihoodpoint, MLP), the deterministic optimization model in a new round of operation is established by using the random variable and the moving vector of the parameter. The proposed method is more accurate than the RBMDO based on the first order reliability method (firstorderreliabilitymethod, form). The effectiveness of the proposed method is verified by an example. (3) the multidisciplinary design optimization (subsetsimulationreliabilityanalysisbasedmdo, ssra-mdo) based on the subset simulation reliability analysis method. The reliability evaluation of the small probability failure event is required when the optimal design object system has high reliability. In order to complete the RBMDO accurately and efficiently, the reliability analysis and evaluation of the deterministic MDO optimization results in RBMDO under the Sora strategy are analyzed and evaluated using the subset simulation reliability analysis method (subsetsimulationreliabilityanalysis, SsrA). This method transforms the original failure probability calculation problem into a series of conditions with higher probability of loss. The failure probability of each intermediate condition is calculated by the Markov Monte Carlo simulation (markovchainmontecarlosimulation) method. After searching for the simulated design checking point (simulationmostprobablepoint, SMPP), the determinacy of the new round operation is established by using the random variable and the moving vector of the parameter. The proposed method is compared with the RBMDO based on the traditional Monte Carlo simulation (montecarlosimulation, MCS). Compared with the RBMDO based on the traditional Monte Carlo simulation (MCS), a less number of simulation sample points are used while the computational efficiency is more efficient. The effectiveness of the proposed method is verified by an example. (4) the random uncertainty of a flipped drive mechanism is considered. According to the composition and movement principle of a overturned driving mechanism, the mechanism is divided into dynamic input subject and dynamic transmission subject. On the basis of finite element analysis and dynamic analysis of each component and mechanism, the corresponding response surface of performance function is constructed. The RBMDO model of the overturned driving mechanism is established by the deterministic source, and the SSRA-MDO method proposed in this paper is used to optimize the solution under the SORA strategy. The optimized design scheme is compared with the original design scheme, and the rationality of the results is further elaborated.

【学位授予单位】：电子科技大学
【学位级别】：博士
【学位授予年份】：2014
【分类号】：TH112
，

本文编号：1799626