Recent Developments in Structural Health Monitoring and Assessment — Opportunities and Challenges

The following sections are included:

• Preface
• About the Editors
• Contents

Opportunities and challenges of structural health assessment are briefly discussed in this introductory chapter. After giving the historical background, the current state of the art is conceptually discussed. The authors and their team members have developed several sophisticated procedures in the related topics and have been published extensively. They developed papers emphasizing technical aspects of the concepts, which can sometimes be difficult for readers to follow. The team had to face numerous challenges, both conceptual and numerical, during the development of the analytical part of the study. They also had to overcome several challenges during the experimental verification phase of the study. Some of these issues were unknown at that time or overlooked by scholars. They were not emphasized in the technical papers. Some of these challenges are elaborated in this chapter. It is expected that the discussions will help the interested readers develop their own inspection procedures addressing or incorporating some of the factors overlooked in the past. In spite of significant developments in areas related to structural health assessment, their applications in inspecting infrastructures are limited. The authors believe that the discussions included in this chapter will advance the research areas on structural health assessment in a comprehensive way and make inspection methods more implementable.

To promote practical structural health monitoring, techniques for real time accurate and efficient parameter identification of linear and nonlinear systems must be available, one class of which is the Bayesian filter technique discussed in this chapter. The general framework in this chapter unifies the different types of Bayesian filters, and their properties are presented. First, the concept and mathematical foundation of Bayesian filter technique and how it is applied in parameter identification are briefly introduced. The special case of linear system with parameters and states approximated by Gaussian distribution functions can be efficiently solved via the Kalman filter. Its extension to nonlinear system identification can be realized through one of four typical Bayesian filters, namely, the extended Kalman, unscented Kalman, cubature Kalman, and Particle filters. Numerical simulation using a three degrees of freedom system is employed to illustrate the suitability of each filter for the different types of parameter identification problem.

A combination of finite element models for structural systems, Bayesian methods for assimilating measurements into these models, and Monte Carlo simulations for estimating multi-fold integrals offers a powerful framework to tackle problems of structural system identification and prognosis. This framework can consider system nonlinearities, time-varying system matrices, uncertainties associated with mathematical modeling and measurements, data from multiple tests and diverse sensors, and multiple mathematical models. This chapter discusses recent developments related to: (a) difficulties and remedies in implementing Markov chain Monte Carlo (MCMC) samplers when one has a large amount of low noise measurement data, (b) sampling variance reduction in dynamic state estimation by applying the Rao–Blackwell theorem, (c) combining different philosophies of substructuring, namely, component mode synthesis, implicit–explicit mesh partitioning, and Rao–Blackwellization in problems of state and parameter estimation, (d) identification of input and combined input and state estimation in vibrating structures, (e) problems of updating time-invariant/variant reliability models of instrumented structures by combining methods of state estimation with Gisranov’s transformation and subset simulation-based variance reduction schemes, and (f) treatment of computational issues while treating instrumented systems which display numerical stiff behavior. A summary of our contributions made and a few suggestions for future research conclude the chapter.

Significant disparities among the error variances (referred to as heteroscedasticity) for the modal variables are observed in structures, which is conventionally overruled in Bayesian updating methodology by pooling into a single error variance. A recent study proposed a heteroscedastic Bayesian model updating implemented via Gibbs sampling, thereby considerably restricting the choice of priors and also requiring the evaluation of the conditional posteriors in standard form. This study extends an existing Bayesian hierarchical model that allows a conditionally heteroscedastic error model (for the modal variables) to marginally follow a Student’s t distribution, contrasting the commonly adopted Gaussian error model. Due to the analytically intractable nature of the resulting joint posterior, the model is rendered through a Metropolis–Hastings (MH) sampling, a well-acclaimed Markov Chain Monte Carlo (MCMC) algorithm. The proposed modification leads to improved accuracy in the updating variables along with a lesser degree of uncertainty and improved noise immunity. The efficiency of the algorithm is demonstrated numerically using simulated set of partial measurements of a multi-degrees of freedom (MDOF) shear building.

In structural health monitoring and damage detection, particle filtering can potentially play a crucial role. The problems may be solved by posing them within the framework of revealing accurate statistical information from the imprecisely known model parameters and dynamical systems fed with experimentally observed noisy data. While it appears to be lucrative, weight-based particle filters typically experience severe performance failure. The major numerical bottleneck for such underperformance especially for higher dimensional systems happens to be the progressive particle impoverishment owing to weight collapse. In this chapter, we show that such difficulty can be conveniently bypassed by applying additive gain-type nonlinear particle filters. Specifically, we discuss ensemble Kalman filter (EnKF) and the ensemble Kushner–Stratonovich filter (EnKS). The discussion on EnKS encompasses both its non-iterative and iterative forms. We also numerically show that, in the identification of nonlinear and large-dimensional dynamical systems, a substantively superior performance of the non-iterative version of the EnKS may be observed vis-à-vis most existing filters. The costlier iterative version, though conceptually elegant, mostly appears to incorporate a marginal improvement in the reconstruction accuracy over its non-iterative counterpart.

This chapter presents the application of structural health monitoring (SHM) methods based on recursive Bayesian estimation. In this class of SHM methods, an assessment of the state of structural integrity is performed online as information related to structural behavior in the form of response measurements becomes available. SHM methods of this type can be broadly classified as methods that rely on estimated mechanical properties and methods that rely on estimated mechanistic damage models. In the former class of methods, model parameters (typically stiffness and damping related) are assigned to various spatial locations and used as a damage-sensitive feature; the model parameters are recursively estimated and damage is assessed by detecting variations in the parameters. In the latter class of methods, a mechanistic damage model is used for various spatial locations and used as a damage-sensitive feature; the damage index is recursively estimated providing a quantitative damage measure. The application of both types of methods is presented in the context of a large-scale shear wall structure tested on a shake table at the Englekirk Center of the University of California at San Diego. It is shown that the methods provide damage-sensitive features that can be employed for structural damage identification and quantification in the context of large-scale structures.

Recent developments in the unscented Kalman filtering areas when used for health assessment of large structural systems considering only a limited number of noise-contaminated responses measured at a small part of a structure are presented. The structure is represented by finite elements so that the health of each element (beam, column, and/or truss) can be assessed. To implement the concept, the unknown excitation information and the initial state vector must be known. Since they are unknown at the beginning of the inspection, a substructure concept is introduced to generate the required information. This requires a two-stage concept. In the first stage, a small part of the structure where responses will be measured will be considered as the substructure. Using the responses measured in the substructure, the information on the unknown excitation and stiffness parameters of all the elements is identified using an iterative least-squares technique developed by the team earlier. Judiciously using the extracted information, the whole structure can be identified using the unscented Kalman filter (UKF) concept. To improve the efficiency, accuracy, and convergence capability, the algorithm is integrated with the weighted global iteration technique. To avoid excitations caused by multiple sources during field inspections, only a fraction of a second of responses is used for the identification process. Since the information may not be sufficient to identify a system, multiple global iterations are used extending the boundaries of UKF. Its capabilities are demonstrated with the help of a three-dimensional truss-frame structural system in the presence of single and multiple defects. Its superiority over the extended Kalman filter method is established with the help of the same example.

Bayesian filtering-based Structural Health Monitoring (SHM) approaches predominantly employed Extended and Unscented Kalman filter variants (EKF and UKF) for joint estimation of states and parameters. In these approaches, a set of parameters denoting location-based system health indices are either appended in the response state vector of the system and estimated jointly as augmented states or estimated in parallel to the response states employing separate filters through dual estimation approach. This chapter discusses the relative benefits and constraints of the joint and dual estimation approaches specific for SHM problems when dealt with EKF and UKF environment. A comparative study is undertaken in this regard to arrive at an understanding about the relative benefits and applicability of the approaches for SHM. All the approaches are investigated on a numerical eight degrees of freedom mass-spring-damper system in which damage is simulated by numerically reducing stiffness value of a linear spring. The vibration response under Gaussian white noise excitation has been adopted as the measurement. The pros and cons of all the considered approaches are discussed in terms of accuracy, precision, promptness, and computation time.

During the initial construction period, early-age masonry walls are susceptible to lateral loads induced by wind, which may result in cracks or surface-level damages. Failure to accurately identify these localized damages result in local and global failure of masonry walls under construction. In this book chapter, a novel hybrid image processing and deep learning algorithm is proposed for the autonomous crack detection in masonry structures at the construction site. At one end, a convolutional neural network is proposed to detect the presence of cracks in masonry. On the other hand, a matrix density thresholding-based image processing technique is adopted to localize the cracks and perform pixel-level damage identification. The proposed hybrid method is validated using a database of the uncracked and cracked masonry wall.

One of the significant concerns for maintaining railway network reliability is the structural health of aged steel riveted railway bridges, some of which in the United States were built at the beginning of the twentieth century. A reliable structural health monitoring (SHM) network that would anticipate riveted steel bridge deficiencies, optimize maintenance, and, ultimately, extend bridge service life to reduce railway network interruption would be a key component to assuring that the network remains viable. A class of SHM methods that can also potentially predict the remaining useful life of a structure to take advantage of a physics-based model would provide a cost-effective option to current methods that commonly rely on extensive instrumentation. In this regard, the validity of the physical model, its parameterization, and accuracy in emulating actual response is of utmost importance for SHM prognosis of remaining useful life. In the context of Industry 4.0, those physics-based numerical replicas of the physical asset are called digital twins of the structure. This chapter presents and examines the efficacy of a framework for automated model calibration using measured operational and ambient structural response for the development of a precise digital twin of a physical structure. The guidelines provided in this chapter will assist in choosing the right model class for accurate response prediction. The study used an in-service, double-track, riveted steel plate girder railway bridge as a testbed for the proposed framework.

Structural health monitoring is one of the modern technologies to evaluate the safety of civil engineering structures. Modified Social Group Optimization (MSGO) is a human-based meta-heuristic optimization technique which has recently been successfully implemented for structural health monitoring of different modeled civil engineering structures using eigenvalue-based objective function. However, the algorithm has not been explored using other objective functions. This chapter deals with the use of MSGO for damage analysis of a truss bridge considering real-time class AA loading. A Pratt truss bridge has been designed following the Indian Standard Codal provisions using TEKLA software. The damage analysis has been performed in the MATLAB platform and the objective function used for the analysis is based on the vibration data, natural frequencies, of the structure. The objective function has been validated for damage analysis of the benchmark structure proposed by American Society of Civil Engineers (ASCE). It has been observed that the MSGO is effective for damage analysis as it can localize and quantify the damages in the bridge truss accurately.

This chapter introduces smartphone-based methods for civil infrastructure health monitoring and representative case studies. The main focus of such smartphone approaches is on utilizing their embedded sensors, mobile communication functions, and onboard computing capabilities as low-cost alternative instruments and potential data crowdsourcing devices. Recent evolutions of various smartphones’ sensing capabilities are discussed, as such capabilities define the limitations and possibilities of the smartphone in civil engineering applications. Various case studies, particularly utilizing camera and inertial measurement unit (IMU) sensor of the smartphones, are introduced in this chapter, discussing how they benefit from the smartphones and overcome the limitations in their monitoring applications.

Condition monitoring of civil infrastructures involves the application of many contact-based vibration sensors such as accelerometers or strain gauges that require considerable installation effort and involve very high maintenance costs due to harsh environments. To address these drawbacks, vision-based noncontact and targetless vibration sensors are proposed, but the methods yield displacement response at few discrete locations of the structure. With this context, the authors present a novel method of harnessing dense quantitative continuous spatial information from the video of a vibrating structural component to obtain its full-field displacement response. The method applies d’Alembertian of Gaussian filter on the spatiotemporal video signal that is tracked over time using the optical flow of edge. Subsequently, the spatially dense modal properties of the structure are identified from its video using the acquired full-field spatiotemporal displacement response using the Hankel dynamic mode decomposition (Hankel-DMD) method. Experimental validation of the proposed method is presented for two kinds of structures, a three-story steel frame, and an aluminum cantilever beam. The proposed method successfully extracts the dominating vibrational modes of the cables of the Tacoma Narrows bridge from its video, moments before its collapse, thereby validating the efficacy of the proposed method for real-life infrastructure systems.

Emerging augmented reality tools can be used to greatly enhance our ability to capture comprehensive, high-resolution, 3D measurements of critical infrastructure as well as provide a means to interface humans to the data generated by sensor networks deployed on the structures over the cradle-to-grave lifetime of the structure. A framework is presented to address the considerations associated with monitoring and inspecting high-value, high-consequence infrastructure such as is used in chemical, explosive, and nuclear facilities using emerging augmented reality technology.

The following section is included:

• Index