DDDAS-TMRP: A Dynamic Data Driven System for Structural Health Monitoring and Critical Event Prediction

在过去的二十年里，计算科学和工程的许多方面都取得了长足的进步。更高保真的数学模型、更高阶的近似方法和更快的求解算法已经被开发用于许多应用。硬件制造商也打破了计算速度的障碍。因此，建模和仿真在缩短设计周期和提高系统性能方面的潜力在当今几乎每个工程领域都得到了认可。然而，对于许多复杂的结构系统，即使是最复杂的计算模型也必然会因为各种原因而不完美。动态数据驱动应用系统(DDDAS)的概念是一个独特的范例，它利用成熟的计算和传感器网络技术来弥补模型缺陷和不可预见的系统演化和刺激条件，减轻设计缺陷对长期和短期系统安全的影响，并使维护计划和危机管理能够做出明智的决策。本项目将通过模拟和实验室实验设计、实施、建立和演示一个DDDAS，用于在不同时间尺度上运行的复杂结构系统的健康监测、故障预测和危机管理。该项目将在以下方面取得进展：(1)应用：开发用于生成混合分析、计算和数据更新模型的逆算法，该模型代表各种时间尺度上退化系统的行为；(2)应用测量系统和方法：开发在高度动态环境中保持自主传感器网络基础设施的方法，在存在系统损坏或退化的情况下保持路由能力和时间同步，传感器评分和自我一致性检查，以及减少模拟器-传感器节点通信的动态模型；(3)数学和统计算法：开发用于在线识别退化系统、损害跟踪、稳定的自适应降阶建模和近实时计算的创新算法，以及(4)系统软件基础设施：开发概念验证软件架构，该架构概括了上述领域的所有进展，并进一步解决了在运行时动态选择和优化跨不同计算资源的不同功能的组件的问题。开发实时的、数据驱动的冗余利用方案，以解决嵌入在多学科系统中的有线和无线传感器网络的服务质量下降，作为操作条件的函数。拟议研究的一个独特特点是一套试验台和相关的实验室实验，其设想不仅是为了证明设想的DDDAS，而且也是为了支持该项目的强大教育组成部分。

During the last two decades, great strides have been achieved in many aspects of computational sciences and engineering. Higher-fidelity mathematical models, higher-order approximation methods, and faster solution algorithms have been developed for many applications. Computing speed barriers have also been shattered by hardware manufacturers. As a result, the potential of modeling and simulation for reducing design-cycle time and enhancing system performance is recognized today in almost every field of engineering. However, for many complex structural systems, even the most elaborate computational models remain bound to be imperfect for numerous reasons. The Dynamic Data Driven Application Systems (DDDAS) concept is a unique paradigm for exploiting maturing computational and sensor networking technologies to compensate for model deficiencies and unforeseen system evolution and stimulus conditions, mitigate the effect of design imperfections on long-term as well as short-term system safety, and enable informed decision for maintenance planning and crisis management.This project will design, implement, build, and demonstrate by simulations as well as by laboratory experiments a DDDAS for health monitoring, failure prediction, and crisis management of complex structural systems operating in various time-scales. The project will enable advances in: (i) Applications: Development of inverse algorithms for generating hybrid analytical, computational, and data-updated models representing the behavior of degrading systems at various time-scales; (ii) Application Measurement Systems and Methods: Development of methods for preserving autonomous sensor network infrastructure in highly dynamic environments, preserving routing capabilities and time synchronization in the presence of system damage or degradation, sensor scoring and self-consistency checks, and dynamic models for reducing simulator-sensor node communication; (iii) Mathematical and Statistical Algorithms: Development of innovative algorithms for on-line identification of degrading systems, damage tracking, stable adaptive reduced-order modeling, and near real-time computing, and (iv) Systems Software Infrastructure: Development of a proof of concept software architecture that encapsulates all the advances in the above areas and furthermore addresses the issue of dynamically selecting and optimizing at run-time components of heterogeneous functionality over heterogeneous computing resources. Development of real-time, data driven, redundancy exploitation schemes for addressing the service quality degradation of wired and wireless sensor networks embedded in multidisciplinary systems, as a function of operating conditions. A unique feature of the proposed research is a set of test-beds and associated laboratory experiments conceived not only to demonstrate the envisioned DDDAS, but also to support a strong educational component of the project.