Statistical Structural Health Monitoring and Damage Detection for Highly Variable Environments

结构健康监测的目的是在很长一段时间内检测损坏并评估关键的物理基础设施。评估的基础设施包括管道、桥梁、钢缆、飞机、火车和发电厂。由于结构健康监测数据的复杂性和可变性，目前的技术只能合理地评估有限的信息量。然而，随着结构健康监测扩展到更复杂的问题，其数据包括几十年来数百tb的结构数据，大数据战略变得至关重要。该奖项支持将统计、大数据战略与结构健康监测系统相结合的基础研究。新的集成框架将能够在多年的恶劣、高度多变的环境中跟踪和预测结构损坏。该研究将为在灾难性破坏发生之前检测结构损伤提供工具。这项研究的结果将通过提高安全性和降低许多民用和机械系统（从飞机到桥梁到管道）的生命周期成本，使美国经济和社会受益。该研究还促进了高度的跨学科合作，因为该研究采用了电气和计算机工程、计算机科学、机械工程和土木工程等学科的方法和解决问题。这种多样化的主题将促进多学科工程教育，并有助于招募和扩大代表性不足的学生参与研究。该项目将在大型（千兆字节到太字节）结构健康监测数据集中统计识别几乎不可见的、至关重要的趋势。解决了三个主要挑战：(1)扭曲的环境条件；(2)对大型数据集的缩放能力差；(3)没有可靠的标准统计。该项目通过创建基于动态时间扭曲、奇异值分解、因素分析和最大似然统计的模块化大数据结构健康监测框架来解决这三个挑战。大数据结构健康监测框架是通过几个短期（小时到天）和长期（多年）结构健康监测实验来评估的，这些实验将复合材料板引入各种损伤和环境条件。这些实验为研究环境参数对复合结构健康监测数据的影响提供了统计验证和独特的机会。

Structural health monitoring aims to detect damage and assess critical, physical infrastructures over long periods of time. The infrastructures assessed can include pipelines, bridges, steel cables, aircrafts, trains, and power plants. Due to the significant complexity and variability of structural health monitoring data, current technology can only reasonably assess limited amounts of information. Yet, as structural health monitoring expands to more complex problems and its data includes hundreds of terabytes of structural data over several decades, big data strategies become essential. This award supports fundamental research to integrate statistical, big data strategies with structural health monitoring systems. The new, integrated framework will enable the tracking and the anticipation of structural damage in harsh, highly variable environments over many years. This study will provide tools for the detection of structural damage before catastrophic failures occur. Results from this research will benefit the U.S. economy and society by improving safety and reducing life-cycle costs of many civil and mechanical systems, from airplanes to bridges to pipelines. The research also promotes highly interdisciplinary collaboration as the research employs methods and solves problems found across the disciplines of electrical and computer engineering, computer science, mechanical engineering, and civil engineering. This diverse subject matter will promote a multidisciplinary engineering education and help to recruit and broaden the participation of underrepresented students in research.This project will statistically identify barely visible, critically important trends in large (gigabyte to terabyte size) structural health monitoring data sets. Three primary challenges are addressed: (1) distorting environmental conditions, (2) poor scaling to large data sets, and (3) no standard statistic for reliability. This project addresses these three challenges through the creation of a modular big data structural health monitoring framework based on dynamic time warping, singular value decomposition, factor analysis, and maximum likelihood statistics. The big data structural health monitoring framework is assessed through several short-term (hours to days) and long-term (multi-year) structural health monitoring experiments that introduce composite panels to a variety of damage and environmental conditions. These experiments provide statistical validation as well as a unique opportunity to study the effects of environmental parameters on composite structural health monitoring data.