Opportunities and challenges in health monitoring of constructed systems by modal analysis

Dynamic testing of constructed systems was initiated in the 1960’s by civil engineers interested in earthquake hazards mitigation research. During the 1970’s, mechanical engineers interested in experimental structural dynamics developed the art of modal analysis. More recently in the 1990’s, engineers from different disciplines have embarked on an exploration of health monitoring as a research area. The senior writer started research on dynamic testing of buildings and bridges during the 1970’s, and in the 1980’s collaborated with colleagues in mechanical engineering who were leading modal analysis research to transform and adapt modal analysis tools for structural identification of constructed systems. In the 1990’s the writer and his associates participated in the applications of the health monitoring concept to constructed systems. In this paper, the writers are interested in sharing their experiences in dynamic testing of large constructed systems, namely, MIMO impact testing as well as output-only modal analysis, in conjunction with associated laboratory studies. The writers will try to contribute to answering some questions that have been discussed in the modal analysis and health monitoring community for more than a decade: (a) What is the reliability of results from dynamic testing of constructed systems, (b) Can these tests serve for health monitoring of constructed systems? (c) Are there any additional benefits that may be expected from dynamic testing of constructed systems? (d) Best practices, constraints and future developments needed for a reliable implementation of MIMO testing and output-only modal analysis of constructed systems for health monitoring and other reasons. decades following IMAC I, the modal analysis community propelled this specialty into a significant research and application area. Today, mechanical and aerospace engineers take advantage of modal analysis for supporting mechanical systems design, quality control during manufacturing, control of operational vibration, acoustics, and, damage diagnosis applications. These applications have fostered the development of an industry for producing excitation devices, sensors, data acquisition hardware and data processing software. Although the fundamental structural dynamics theory for constructed and mechanical systems is the same (introduced to civil engineers by aerospace engineers in the 1950’s), it is the writers’ opinion that there has not been sufficient cross-fertilization of the seismic monitoring and dynamic testing applications in earthquake engineering by transforming the recent advances made by the modal analysis community related to sensing, data acquisition and data post-processing. Also, the importance of taking advantage of pattern recognition and other signal processing advances made in the telecommunications field for advancing modal analysis applications on constructed systems are becoming quite clear. A third and more recent research area termed “health monitoring” has been initiated in the 1990’s. Since 1997, the International Workshop on Structural Health Monitoring (IWSHM) that has been organized at Stanford University, has brought together biannually a large community of researchers from aerospace, space, automotive, earthquake and infrastructures fields. In addition to IWSHM, there have been many additional international workshops and conferences held since the 1990’s related to post-earthquake damage evaluation, structural control during earthquakes, condition assessment of constructed systems by nondestructive testing, etc. that included discussions of health monitoring. Today, the health monitoring community has significantly broadened, such that, in addition to aerospace, space and automotive fields, the terms health monitoring and health assessment are widely used by many academic and practicing civil engineers specializing in the design, condition assessment, evaluation, maintenance and retrofit of all types of constructed systems. There is consensus amongst the civil engineering research community that health monitoring promises to be a critical enabler for performance-based civil engineering applications, lifecycle cost-based maintenance management of constructed systems, and asset-management applications to entire infrastructures. The senior writer has participated in each of the earthquake engineering, modal analysis and health monitoring research and application fields since he received his PhD in 1973. During 1990-1997, writers had the opportunity of testing a mid-rise building (Hosahalli & Aktan 1994, Aktan et al 1995, Miller et al. 1993, Aktan et al. 1997, and Catbas et al. 1998), and three different types of highway bridges to controlled damage and failure. After the test bridges were loaded to various levels of damage, they were unloaded and their modal analysis was conducted. By transforming the modal vectors resulting from multi-input, multi-output (MIMO) testing by impact to modal flexibility, and by virtually loading the modal flexibility by various load patterns, the writer and his associates showed that the virtual deflection patterns of a bridge may be used as a conceptual and sensitive structural condition and damage indicator (Toksoy & Aktan 1994, Aksel et al 1994, Catbas et al 2005). The virtual deflection patterns were validated by loading the test bridges with trucks, and measuring and correlating the deflection patterns with the virtual deflections obtained from modal flexibility. Since 1992, the writers have been investigating continuous monitoring applications on constructed systems and how these may be used for long-term monitoring of highway bridges. After 1997, they started exploring the challenges in structural identification and lifecycle health monitoring of large-scale, long-span bridges, representing a special class of constructed system in terms of very large (kilometers) scale and complexity. These systems proved to be excellent field laboratories not just as constructed systems but in fact by representing infrastructure systems that cannot be fully observed and conceptualized without recognizing and incorporating every one of the interacting engineered, human and natural sub-systems and elements necessary for understanding and defining their performance (Aktan & Faust 2003). 1.2 Objective and scope The writers’ objective in writing this paper is to take advantage of their experiences in dynamic testing of a number of actual building and bridge structures in the field and formulate an overview of the state-of-art in dynamic testing of constructed systems, especially for health monitoring applications. They are motivated to discuss the most critical prerequisites for reliable applications of MIMO and output-only dynamic testing that may lead to a meaningful understanding of the global mechanical characteristics of large constructed systems and their subassemblies. The writers are motivated to contribute to answering some questions that have been discussed in the modal analysis and health monitoring community for more than a decade: (a) What is the range of reliability of results from dynamic testing of constructed systems; (b) Can dynamic tests serve for health monitoring of constructed systems? (c) Are there any additional benefits that may be expected from dynamic testing of constructed systems? (d) What are the best practices, constraints and future developments needed for a reliable implementation of MIMO testing and output-only modal analysis of constructed systems for health monitoring and other reasons? Measuring the global mechanical characteristics of large constructed systems by dynamic testing requires the assumptions of their observability, time-stationarity (at least during the observation period), and linearity (of the structure and damping mechanisms). None of these three principal assumptions can be strictly valid for structures which are very large and complex, or subject to daily temperature fluctuations of 10 degrees Celsius or higher. Given that there have been very few examples of dynamic testing of constructed systems or their subassemblies in the field, with adequate instrumentation designed in a scientific context, any new application becomes an exploration into the unknown and is governed by significant epistemic uncertainty. This greatly impacts the reliability of the data and the results that are extracted from the data, and points to the importance of experience and taking advantage of any heuristics as well as the necessity of concerted efforts for mitigating many common human/application errors that impact the reliability of the test results. Hence, to make field testing of a constructed system in the field produce reliable and sufficiently complete results, it is necessary to approach this as a scientific exploration and not as a routine, process-oriented engineering application. The system-identification concept becomes a necessary and fundamental approach to designing such an exploration. The writers would like to defend this viewpoint by providing examples from the laboratory and from the field. To illustrate the relationship between MIMO and output-only testing, the results from a laboratory specimen are presented. To illustrate the best recommended practices for MIMO and outputonly testing of systems in the field, applications to actual bridge structures will be discussed. 1.3 Brief review of recent reported applications of dynamic testing of constructed systems Since 1980, dynamic testing and modal analysis have been explored in integrated analytical and experimental research on: condition assessment of offshore platforms (Rubin & Coppolino 1986), large space structures (Stubbs et al 1990, Kim & Bartkowicz 1993), seismic vulnerability evaluation of buildings (Hosahalli & Aktan 1994) and condition assessment of highway bridges (Brownjohn et al 1986, Mazurek & DeWolf 1990, Salane & Baldwin 1990, Nigbor et al 1992, Raghavendracher & Akta