面向复杂机械系统安全的多学科模型不确定性分析理论与模型验证方法研究

Multidisciplinary modeling and simulation have been playing an increasingly important role in assessing the performance, robustness and reliability of complex mechanical systems in aerospace, maritime, transportation and etc., as they are often much less expensive than physical experiments in analyzing such systems. However, safety design and evaluation of interdisciplinary mechanical systems comprises multiple levels of modeling and simulations with interacting models and codes from multiple coupled disciplines, which all involve a vast quantity of inevitable uncertainties from multiple sources. As multidisciplinary predictive models are used to support decisions and policies made by engineering designers, managers, and public officials in safety engineering, it becomes crucial to build credibility and confidence in these models for enhancing the reliability and robustness of simulation based system safety design. .Considering both the conventional aleatory uncertainty that inherited from system variability and the epistemic uncertainty due to lack of knowledge, this research will be carried out from the following four aspects: 1) Development of concurrent propagation and sensitivity analysis approaches for the two categories of uncertainty from sub-disciplines to the system quantities of interest; 2) Investigation of methods for quantifying the multidisciplinary model uncertainty by comparing data from physical experiments and computer simulations; 3) Development of new validation metrics for multidisciplinary computational models under multiple sources of uncertainty; 4) Exploration of comprehensive validation framework to enhance the confidence of multidisciplinary models through uncertainty quantification, model validation, uncertainty reduction, as well as sensitivity based risk management and cost control. .The methodology developed in this research will be a key enabler to improve multidisciplinary model credibility, reduce the amount of expensive physical testing, and achieve safety design and evaluation for complex mechanical systems with high confidence.

航空航天、海洋工程、交通运输等领域复杂机械系统的安全设计和评估越来越依赖于多学科的建模和仿真，以减少物理实验的次数。然而，多学科建模的复杂性和耦合性，为基于仿真的可靠性或稳健性的分析与设计带来了各种各样的不确定性，从而对多学科系统的安全设计和决策造成非常不利的影响。本项目首次针对多学科模型的不确定性传播、不确定性量化和模型验证等集中需求的综合问题，同时考虑偶然不确定性和认知不确定性，在更深层次上研究不确定性分析与模型验证的理论与算法。重点研究：1）偶然和认知不确定性在在多学科模型中的传播方法和灵敏度分析方法；2）多学科系统模型不确定性的量化和认知不确定性的缩减；3）不确定性多学科模型验证指标的设计；4）集不确定性量化、模型验证、实验与仿真资源的优化配置等于一体的综合模型验证策略。本项目的研究将提高多学科模型对真实机械系统的预测能力，减少昂贵的物理实验，加强多学科系统的可靠性和稳健性设计。

本项目针对多学科模型的不确定性传播、不确定性量化和模型验证等集中需求的综合问题，同时考虑偶然不确定性和认知不确定性，在更深层次上研究不确定性分析与模型验证的理论与算法。主要完成了以下四个方面的研究工作。.1）在偶然和认知不确定性的模型灵敏度指标的研究中，采用随机变量描述偶然不确定性，模糊数描述认知不确定性，并基于模型输出统计矩的条件隶属函数和无条件隶属函数之间的区别，提出了衡量两类不确定性对模型输出的总体不确定性的影响和贡献的灵敏度指标及其稳定性指标，全局而量化地衡量两类不确定性对系统输出的影响，为不确定性的量化和缩减起到铺垫作用。.2）在基于Bayesian推理理论的多学科模型验证框架的构建和系统安全性预测研究方面，充分考虑复杂系统所涉及的多源不确定性、多尺度的试验和仿真数据、充分缩减认知不确定性，从而给出可信度高的系统安全性评估。将所提出的模型验证和预测方案应用于液体加压容器的层次型多学科系统，最终给出其安全操作区间的预测。.3）在针对一般复杂机械结构的模型验证策略研究方面，首先通过实验数据验证结构的有限元模型，然后建立结构系统的高保真可靠性模型和低保真可靠性模型，并验证低保真可靠性模型的可行性，最后采用低保真可靠性模型进行最终可靠性的预测。通过该验证框架，不仅可以高效利用宝贵的实验数据，也可以节省长时间的仿真成本。.4）在电动汽车制动系统安全与稳定性研究的研究中，针对电动汽车制动系统同时存在摩擦制动和电机回馈制动系统，展开电动汽车制动系统分层控制研究，提高电动汽车安全性与稳定性。在上层控制器中，基于车辆动力学模型和轮胎动力学模型，并充分考虑轮胎-路面的接触特性，设计自适应控制器，得到每个车轮的理想制动力矩。在下层控制器中，考虑摩擦制动系统和电机回馈制动系统动态响应特性的差异，基于上层制动力矩信号变化频率对制动力矩进行分配。进一步对信号通信质量不确定性对制动控制性能的影响进行了分析。