Sensors: GOALI: Networked Estimation and Decision Computing for Structural Health Monitoring

该项目的目标是为结构健康监测（SHM）系统级应用领域的网络传感器开发决策和控制理论、状态估计算法和数学工具。我们将开发用于时空数据诊断和预测解释的分布式计算算法。最初的重点是航空航天应用，但研究结果将足够普遍，可用于许多其他领域。该算法将使用嵌入式优化来准确估计损伤的演化时空模式。计算将可扩展到非常大的传感器阵列。提出的方法是通过产业合作开发航空航天领域的创新工程应用，然后推广结果。现有的SHM工作旨在开发传感器、数据查询方法和低水平信号处理，以获得单个损伤估计阵列。该项目将基于从分布式网络传感器获得的一系列空间数据阵列，开发系统级统计算法，用于估计结构健康状态、监测和决策支持。该算法将允许可扩展和可重构的分布式并行计算实现。知识价值。我们建议开发结构损伤的时空估计算法，这些算法可以通过网络计算扩展到非常大的空间数据阵列。工程需求和潜在的实际影响来自与霍尼韦尔的合作。与现有工作相关的问题表述有以下特点：(i)我们专注于现有传感器数据的数学处理，而不是开发新的传感系统；（ii）我们关注的是数据的时空处理，而现有的SHM文献关注的是获得损伤的单一空间模式。对损害的时间演变进行建模和趋势分析将有助于预测；（iii）我们通过分布式计算实现的基于约束优化的估计算法来追求系统可扩展性。现有的分布式算法都是针对少量参数进行优化的。我们将估计大的时空阵列；（iv）决策系统设计将包括不确定性、缺失数据、自我修复能力等，而不是考虑理想化和简化的数学问题。建议工作的主要步骤和挑战包括：1。建立了损伤的时空演化模型和最优估计准则。2. 利用同似然指数的嵌入式凸优化，开发最优统计估计和检测问题的解决方案。3. 通过网络分布式计算开发可扩展算法。广泛的影响。本提案中的算法和系统工程方法将通过减少误报和提高可扩展性来加速SHM技术的工业采用。这将使结构能够根据情况进行维护，减少航空公司航班延误，提高军用飞机的战备状态。其他航空航天应用包括下一代航天器和空间栖息地。与霍尼韦尔的合作将促进向航空航天工程实践的过渡。这项研究将有助于在其他领域开发大型网络传感系统。这些包括民用结构（建筑物，桥梁等），船舶和地面车辆以及工业厂房的SHM。这些结果也可以扩展到医学、地球物理（例如地震相关损害）以及生物或生态系统损害的传感和健康评估。我们将采取步骤，通过将研究成果纳入现有课程的课程，发展网络传感器决策技术领域的研究生培训。在计划中的大学访问期间，工业合作项目负责人将教授一门面向工业的研究生课程。参与项目或参加课程的学生将有机会在霍尼韦尔实习，这将为他们提供额外的实践培训

The objective of the proposed project is to develop decision and control theory, state estimationalgorithms, and mathematical tools for networked sensors in a system-level application area ofstructural health monitoring (SHM). We will develop distributed computing algorithms for diagnosticand prognostic interpretation of the spatio-temporal data. The initial focus is on aerospaceapplications but the research results will be suffciently general to be used in many other areas. Thealgorithms will use embedded optimization for accurate estimation of the evolving spatio-temporalpattern of the damage. The computing will be scalable to very large sensor arrays.The proposed approach is to develop an innovative engineering application in aerospaceSHM area through industrial collaboration and then generalizing the results. Existing SHM workis aimed at developing sensors, data interrogation methods, and low-level signal processing toobtain a single damage estimate array. This project will develop system-level statistical algorithmsfor estimation of structural health state, monitoring, and decision support based on series of spatialdata arrays obtained from distributed networked sensors. The algorithms will allow for scalableand reconfigurable distributed parallel computing implementation.Intellectual merit. We propose to develop algorithms for spatio-temporal estimation of structuraldamage that can be scaled to very large spatial data arrays through networked computing.The engineering need and potential for practical impact come from collaboration with Honeywell.The place of the problem formulation with respect to existing work is characterized by thefollowing highlights: (i) We concentrate on mathematical processing of the existing sensor data,rather than developing new sensing systems; (ii) We focus on spatio-temporal processing of thedata, while the existing SHM literature is focused on obtaining a single spatial pattern of thedamage. Modeling and trending the temporal evolution of the damage will enable prognostics;(iii) We pursue system scalability through constrained optimization-based estimation algorithmsimplemented through distributed computing. Existing distributed algorithms are for optimizinga small number of parameters. We will estimate large spatio-temporal arrays; (iv) The decisionsystem design will include uncertainty, missing data, self-healing capability etc, as opposed toconsidering an idealized and simplified mathematical problem.The main steps and challenges in the proposed work include: 1. Formulating models of spatiotemporalevolution of the damage and criteria of estimation optimality. 2. Developing solutions ofoptimal statistical estimation and detection problems through embedded convex optimization of alog-likelihood index. 3. Developing scalable algorithms through networked distributed computing.Broad impact. Algorithmic and system engineering approaches in this proposal would accelerateindustrial adoption of the SHM technology by reducing false alarms and improving scalability.This would enable condition-based maintenance of structures reducing airline flight delays and improvingmilitary aircraft readiness. Other aerospace applications include next generation spacevehicles and space habitats. The transition into aerospace engineering practice will be facilitatedby the collaboration with Honeywell. The research would help in developing large networked sensingsystems in other areas. These include SHM of civil structures (buildings, bridges, etc), marineand ground vehicles, and industrial plants. The results could be also extended to sensing andhealth assessment in medicine, geophysics (e.g., earthquake related damage), and bio or ecosystemdamage.We will take steps towards developing graduate training in the area of decision technology for thenetworked sensors by including the research results in the curriculum of existing courses. Duringthe planned university visit, the industrial co-PI will teach an industrial-oriented graduate-levelcourse. The students involved with the project or attending the courses will have an opportunityto work as interns at Honeywell, which will provide them with an additional practical training.1