Health Assessment of Engineering Structures Using Graphical Models

Health Assessment of Engineering Structures Using Graphical Models

Abbas Moustafa, Sankaran Mahadevan
DOI: 10.4018/978-1-4666-1640-0.ch014
(Individual Chapters)
No Current Special Offers


A hybrid qualitative-quantitative health assessment of structures using the bond graph theory is presented in this chapter. Bond graph (BG) is an energy-based graphical-modeling tool for physical dynamic systems, actuators, and sensors. BG provides domain-independent framework for modeling dynamic systems with interacting components from multiple domains. Discrete structures are modeled using one-to-one bond graph elements, while continuous structures are modeled using finite-mode bond graphs. BG facilitates the construction of temporal causal graph (TCG) that links the system response to the damaged component or faulty sensor. TCG provides qualitative damage isolation, which is not possible using most existing system identification techniques. This leads to rapid isolation of damage and significant reduction in computations. Quantitative identification of damage size is performed by analyzing the substructure containing the damaged component, using the nonlinear least-squares optimization technique, thus reducing the computations. The health assessment algorithm developed in this chapter combines the Generic Modeling Environment (GME), the Fault Adaptive Control Technology (FACT) software, and Matlab Simulink®. Numerical illustrations on BG modeling of a hydraulic actuator and system identification of a fifteen-story shear building and a high-rise structure under earthquake loads are provided.
Chapter Preview

1. Introduction

This chapter proposes a graphical, domain-independent, energy-based framework that is capable of modeling multidisciplinary systems with interacting components from structural, mechanical, electrical, and hydraulic domains. This framework is based on the bond graph (BG) theory introduced by Paynter (1961) and developed by Karnopp, Rosenberg and Margolis (Rosenberg & Karnopp 1983, Karnopp & Margolis, 2006). For example, an electrical circuit and a mechanical system can be described with the same bond graph model. The use of bond graphs in electrical and mechanical engineering is well established. This method, however, has not received significant research attention in civil engineering. The BG model of a dynamic system represents the system equations of motion implicitly in a graphical form using bond graph elements. These elements model inertial, stiffness, damping and external forces. BG elements include serial and parallel junctions that govern the dynamic equilibrium of the structure subsystems.

Civil structures deteriorate over time and experience damage due to natural events such as earthquakes and wind. Structural health monitoring (SHM) is a process that aims at providing accurate and in-time information of the structural health condition of existing structures. A comprehensive review on recent advances in health assessment of structures can be found in Doebling et al (1998), Alvin et al (2003), Chang et al (2003), Koh et al (2003), Lui & Ge (2005), Gonzalez & Zapico (2008) and Moustafa et al (2010). System identification (SI) techniques can be grouped into parametric and non-parametric methods. The parametric methods identify changes in the structure global parameters (e.g. natural frequencies, mode shapes and modal damping) or in the local parameters (e.g. members stiffnesses and damping) to characterize the structural damage (Doebling et al 1998, Alvin et al. 2003, Chang et al, 2003, Koh et al, 2003, Lui & Ge, 2005). In general, the parametric methods can detect the damage location but require complete measurements and extensive computations for large structures. The non-parametric methods, on the other hand, require less measurement and have better adaptability to large structures but provide a global assessment on the health status of the structure (Gonzalez & Zapico, 2008).

Sensor performance also degrades with time under varying environmental conditions (Koh et al, 2003, De Oliveira et al, 2004, Elouedi et al, 2004, Blackshire et al, 2006, Glisic & Inaudi 2007). Different degradation mechanisms have been observed in different types of sensors (surface-bonded or fully-embedded) under various environmental effects such as temperature- and moisture-cycling (Elouedi et al, 2004, Blackshire et al, 2006, Glisic & Inaudi 2007). Sensor performance is particularly relevant in the field of road infrastructure where the loading conditions affecting the main structure (traffic loads, temperature cycling, etc.) also affect the sensor measurements. Details on recent sensor technologies can be found in (Ansari 2005, Manders et al, 2006). While sensor faults are difficult to be handled using existing SI techniques, bond graphs are capable of modeling both the system components and the sensors. This enables damage detection in both structural components and sensors. Bond graphs facilitate also the extraction of damage signatures off-line before sensor data collection thus providing rapid identification of the damage location through qualitative comparison of predicted and observed signatures. The quantification of the damage size is performed by analyzing the substructure containing the damaged component only, thus, reducing the computational costs. The idea of using BG in system identification of frame structures was introduced by these authors (Moustafa et al, 2010).

Complete Chapter List

Search this Book: