复杂结构多学科多目标概率设计的混合代理模型方法

Around the engineering background of integrated probabilistic design of complex structure, this project proposes hybrid surrogate model methodology for multi- discipline multi-object dynamic probabilistic analysis to address the issues of many design variables and high nonlinearity, which fuses multiply existing surrogate models, combined with intelligent algorithm, mixed sampling technology and decomposition coordination strategy. And then multi-level multi-layer optimization design method is explored to deal with the problems of large-scale optimization parameters and massive loop iterations on the basis of hybrid surrogate model methodology and multilevel decomposition hierarchical coordination strategy. The two technical bottlenecks, namely correlation processing between multi-disciplinary and multi-objective and analytical accuracy and computational efficiency, are broken though. The main contents include the studies on the single objective hybrid surrogate model method of complex structural probabilistic analysis, the multiple hybrid surrogate model method and decomposed coordinated hybrid surrogate model method of complex structural probabilistic analysis, the complex structural optimization design approach based on the single objective hybrid surrogate model, the complex structural optimization design approaches based on the multiple hybrid surrogate model and decomposed coordinated hybrid surrogate model, the studies on the experimental verification for typical rotor structures of gas turbine engine core machine. The developed methodologies are validated by using the typical rotor structures of gas turbine engine core machine. The efforts of this project will have direct application value for probabilistic analysis and optimization design of aeroengine typical rotor structure, and have important academic significance for developing probabilistic design theory of complex structures.

围绕复杂结构一体化概率设计工程背景，本项目融合多种现有代理模型方法，结合智能算法、混联抽样技术、分解协调策略等理论，研究多学科多目标动态概率分析的混合代理模型方法，解决多设计参数和高度非线性问题，在此基础上结合多级分解层次协调策略，探索多级别多层次优化设计方法，处理大规模优化参数和大量循环迭代问题，进而突破“多学科和多目标之间的相关性处理”和“分析精度与计算效率”这两方面技术瓶颈。主要内容包括：复杂结构概率分析的单目标混合代理模型方法、多重和分解协调混合代理模型方法研究、基于单目标混合代理模型的复杂结构优化设计方法研究、基于多重和分解协调代理模型的复杂结构优化设计方法研究、燃气涡轮发动机核心机典型转子结构的试验验证研究，并以燃气涡轮发动机核心机典型转子结构为例，加以验证。研究成果将对航空发动机典型转子结构概率分析和优化设计具有直接应用价值，对发展复杂结构概率设计理论具有重要的学术意义。

围绕复杂结构多学科多目标概率设计问题，以航空燃气涡轮发动机核心机的压气机和涡轮等典型转子结构为研究对象，开展了复杂结构多学科多目标动态概率分析和多级别多层次优化设计方法研究。针对涉及多个学科载荷交互作用且载荷具有时变性的复杂结构概率分析问题，提出了3种高效率高精度单目标混合代理模型方法；其次，对多学科耦合相关性和多目标失效相关性问题进行了探究，提出了7种多重混合代理模型和分解协调混合代理模型方法；之后，为解决涉及多个学科的复杂结构设计优化分析问题，给出了5种多级别协调优化设计方法；接着，针对复杂结构复杂工况作用的单一结构多失效模式和多构件单一失效模式的多目标设计优化分析问题，给出了2种多级别多层次整体优化设计方法和多级别多层次协同优化设计方法；其中，将航空燃气涡轮发动机核心机的压气机和涡轮等典型转子结构作为试验研究对象，获取了大量的试验数据和运行数据，通过与有限元分析结果以及所建立的代理模型计算结果进行对比，改进、完善本项目所研究提出的方法；最后，总结和回顾了复杂结构动态概率设计的理论和方法，剖析了现有多学科多目标动态概率分析及优化设计的混合代理模型方法，展望了复杂结构动态概率设计的发展方向。本项目最终得到了复杂结构单目标混合代理模型、多重混合和分解协调混合代理模型、多级别协调优化设计、多层次整体优化设计和多级别多层次协同优化设计方法，这些成果在复杂结构概率设计及优化工程中将有很好的应用前景。

内容获取失败，请点击重试