Structural health monitoring of non-linear hysteretic structures.
Thesis DisciplineMechanical Engineering
Degree GrantorUniversity of Canterbury
Degree NameDoctor of Philosophy
Ensuring life safety is the primary design and maintenance requirement for civil structures designed to be serviceable for a specific lifetime. However, structures subjected to a number of factors may experience quicker or more gradual deterioration than anticipated, or even a premature loss of function. Scheduled visual inspection is the oldest and most commonly used damage monitoring technique, but has significant disadvantages when quick damage assessment and certainty are needed, such as after a major seismic event. Quick, accurate and quantitative determination of the damage state is vital following an earthquake, to estimate damage, remaining life-time, and ensure safe re-occupancy, if possible.
Rapid development of sensor technology and increasing computing power has enabled continuous structural monitoring using various sensing techniques. The measured data can be analyzed using structural health monitoring (SHM) methods. SHM refers to all elements of the process of identifying mechanical properties of a structural system, comparing it with previous states, detecting changes/abnormalities, and relating these to damage. A successful SHM method should be able to automatically identify and locate damage after large, non-linear response events.
The majority of existing, primarily vibration based, SHM techniques have serious limitations in situations where a quick, accurate, and quantitative assessment is needed. More specifically, many SHM techniques perform well when structures behave linearly and are subjected to ambient loads, but this does not apply to earthquake events. Moreover, some methods can only work off-line, involve significant computational effort and/or human input, and/or do not provide any indication of damage location and/or severity.
To address these limitations, this thesis explores the application of a novel SHM implementation strategy composed of a novel modal parameter identification and its subsequent application to a proven hysteresis loop analysis (HLA) method. The study demonstrates the proposed strategy can be readily used to track the performance of non-linear degrading structures subjected to strong ground motion, essentially in real-time and without human input. Thus, the proposed tools can be used to support/replace visual inspection results, reduce downtime, minimize business disruptions and, most importantly, maximize life safety.
More specifically, this thesis proposes and analyses the application of a novel modal parameter identification technique, which performs in near real-time and, most importantly, is efficient when approximating non-linear structures subjected to relatively short duration ground motion inputs. The technique operates in modal space and is based on a pre-defined optimization process, which decouples frequency response spectra of interfering, generally higher frequency, modes. Optimization can be realized over relatively short time windows to provide continuous monitoring of highly non-linear, degrading structures.
In particular, identified modal parameters can be readily used to identify damage. However, modal parameters can have very poor sensitivity to damage and are often difficult to interpret. Thus, it is challenging to infer the location and severity of damage based on detected changes/variation in modal parameters alone. In this research, the identified time-varying modal parameters are used to decompose the structural response and reconstruct single mode dominant restoring force-deformation hysteresis loops, which can be readily analyzed using recently developed hysteresis loops analysis (HLA). The versatility and robustness of HLA has been explored in a number of studies. However, the analyzed case structures employed in these validation cases exhibited very small contribution from the higher modes, which typically can cause significant irregularities, and make effective implementation of HLA more problematic. Hence, this thesis aims to improve the robustness of HLA, using mode segregation and reconstruction of single mode dominant, regular shape hysteresis loops from non-linear structural response.
First, this research develops a modal parameter output-only identification technique, which is validated for a simple time-invariant linear structure. Second, the output-only method is extended to an input-output method enabling operators to carry out near-real time identification of non-linear structures, which is validated for a simple time-varying non-linear structure. Third, the input-output method is validated using the simulation results of a more complex non-linear multi-degree-of-freedom structure, formulated using fiber elements. Finally, the proposed SHM strategy, consisting of continuous modal parameter identification and subsequent application of HLA is validated for two experimental non-linear structures.
Overall, this thesis proposes a novel system identification technique, which performs outputonly identification of linear structures and, more importantly, provides input-output real-time modal parameter tracking of highly non-linear structures. Thus, the method extends the application of modal SHM methods to non-linear cases. The proposed technique performs successfully without operator input and can be easily automated to provide continuous modal tracking and damage detection. The technique performs both as stand-alone for damage detection and in combination with HLA for damage quantification as demonstrated for highly non-linear cases.