A Notional Framework and Model to Improve Monitoring of Structural Health Systems

Abstract

Structural health monitoring (SHM) systems have rapidly advanced, embracing both technological advances and real-time, in situ monitoring. Such advances are paramount to maintaining structural integrity. Capitalizing on continual data collection, however, SHM systems can reach beyond detection of structural damage to include the development and prediction of time to critical failure. Due to this increased capability, it is inherent that SHM systems capture information differently than current inspection methods. We advocate the increased capability of the SHM system as a difference between assessing a structure and monitoring a structure. This work introduces a framework with which the data captured in the monitoring of structural health is leveraged in order to improve not only the prediction of the state of the structure (presence of damage and extent of damage), but also the estimated time to failure. A statistical model is presented to represent the generalized framework. This model estimates the time to critical crack length in a hierarchal model such that the time until critical crack length occurs is modeled as a function of current crack status, that is, crack length and location. Crack length and location are themselves a joint response modeled from available sensor output and potentially other various environmental or sensor-related measurements. The necessary joint modeling of length (horizontal and vertical) and location allows us to incorporate the inherent correlation between these as part of the estimation process. These estimates are then used to predict the time until failure (critical crack length is achieved). By modeling in this hierarchal fashion, the current and past state of the structure, as well as changes in the structure during the course of monitoring, may be used to more precisely estimate the remaining life, that is, the time until structural failure may occur. Simulated and preliminary data from current specimens will be used to demonstrate these methods.