Development of Advanced Technologies for Structural Damage Identification and Prediction

高效的损伤识别方法对于确保各种工程结构（例如车辆框架和民用基础设施中的梁、板或壳元件）的操作效率和可靠性至关重要。这些方法必须灵敏、准确、快速和易于操作，并且必须提供损伤的检测、评估和定位。开发有效的识别技术，特别是针对不同载荷条件下初始阶段存在损伤的结构，还面临许多挑战。一个有效的机械结构状态监测系统需要先进的技术和方法，可以识别和预测结构损伤的发生和发展的早期阶段。监测系统的实时操作对于其作为实际应用的采用也是必不可少的。本论文将在结构静动力分析、先进信号处理算法、新型传感网络和机器学习技术等方面探索和发展新的方法。首先，将开发一种新的荷载传递分析方法，并将其与应力计算相结合，以便在早期阶段提供可能损坏的明确迹象。其次，将创建一种新的可调熵算法，该算法具有适当的噪声滤波器，可以在实际噪声工作条件下可靠地识别损伤。一种柔性压电纳米复合涂料将被引入，以创建一个新的传感器网络设计，可以覆盖大面积的监测对象。利用来自压电传感器的时域和空域信号，将可能定位损伤。为了实现结构损伤识别的真实的实时分析，将引入和开发适当的机器学习算法，以快速准确地学习和估计载荷传递路径和熵结果。最后，将上述所有组成部分结合起来，形成一个全面的损伤识别系统。这项研究的成果将导致全面的损伤识别平台和实用（易于操作和维护）的工程解决方案，以警告即将发生的结构损伤和故障，检测他们在初始阶段。建议的研究将通过实际应用确保不同类型工程结构的可靠性、完整性和安全性作为长期目标。拟议的损害评估方法将是廉价的，易于在工业环境中应用，直接造福于加拿大工业。根据该计划培训的高素质人员将获得必要的理论和实践专业知识，以促进和推进加拿大的上级机械结构状态监测。

Highly efficient damage identification methodologies are essential to ensuring the operational efficiency and reliability of a wide range of engineering structures (e.g. beam, plate, or shell elements in vehicle frames and civil infrastructure). These methods must be sensitive, accurate, fast, and easy to operate and must provide detection, evaluation, and localization of damage. Many challenges have yet to be addressed to develop efficient identification techniques especially for structures with damages at initial stages under varying load conditions. An efficient condition monitoring system for mechanical structures will require advanced technologies and methods, which can identify and predict the occurrence and propagation of structural damage at its earliest stages. Real-time operation of the monitoring system is also essential for its adoption as a practical application. The proposed work will explore and develop new methods in structural static and dynamic analysis, advanced signal processing algorithm, new sensing network, and machine learning technique. First, a new load transfer analysis method will be developed and combined with stress calculations to provide a clear sign of possible damage at an early stage. Second, a novel tunable Entropy algorithm with a proper noise filter will be created that can reliably identify damage under practical noisy working conditions. A flexible piezoelectric nanocomposite paint will be introduced to create a new sensor network design that can cover a large area of monitored objects. With both time and space domain signals from the piezoelectric sensor, it will be possible to localize the damages. To attain real time structural analysis for damage identification, proper machine learning algorithms will be introduced and developed to learn and estimate the load transfer path and Entropy results quickly and accurately. Finally, all of the components listed above will be combined to produce a comprehensive damage identification system. Outcomes of this research will lead to comprehensive damage identification platforms and practical (easy to operate and maintain) engineering solutions to warn impending structural damage and failure by detecting them in the initial stages. The proposed research will ensure the reliability, integrity and safety of different types of engineering structures through practical applications as a long-term goal. The proposed damage assessment methods will be inexpensive and easy to apply in industrial settings for the direct benefit of Canadian industries. Highly qualified personnel trained under this program will gain the theoretical and practical expertise necessary to promote and advance superior mechanical structural condition monitoring in Canada.