Predictive Frameworks for Smart Material Response in Structural Health Monitoring

Abstract

Smart material integration into structural health monitoring (SHM) systems is a revolutionary development towards the ascertainment of safety, reliability, and durability of vital infrastructures. In contrast to traditional sensors, smart materials including piezoelectric ceramics, shape-memory alloys (SMAs) and magnetostrictive composites offer intrinsic sensing and actuation, and facilitate self-adaptive monitoring functions in dynamic and uncertain operating conditions. Nonlinear, hysteretic and environment-dependent behavior is however a basic challenge when it comes to predicting their responses. In this paper, a predictive framework is provided that forecasts and models the behavior of smart materials under varying loading conditions through a flexible and synergistic approach in which physics-informed mathematical models and data-driven machine learning algorithms have been combined. The framework includes constitutive models of piezoelectric, SMA, and magnetostrictive material, and augments it with hybrid learning methods including physics-informed neural networks (PINNs), reinforcement learning based on adaptive predictions, and Bayesian learning based on the quantification of uncertainty. Simulation experiments based on finite element modeling and experimental confirmations of the proposed predictive method on representative testbeds, bridge girders, aircraft wing panels, and long-span cables, indicate the proposed predictive approach is much more accurate in detecting damage, has superior sensitivity to sensor noise and generates less computational load than traditional SHM approaches. Findings show that the hybrid framework not only manages to predict strain responses with over 95 percent accuracy, but also allows proactive maintenance strategies to prolong structural life cycles and save costs of operation. Also, the framework can be applied to various civil, aerospace and energy infrastructure as it is scaled and facilitates real time decision making in safety critical situations. The suggested predictive framework can be used to bridge the gap between system-level intelligence and material-level physics that will give the route to the next-generation SHM systems that will be more resilient, adaptive and have the ability to operate in more complex and changing environments and ultimately result in safer and sustainable built environments.