Modal parameter identification of a multiple-span post-tensioned concrete bridge using hybrid vibration testing data
Introduction
In recent decades, the evaluation of the actual behaviour of in-service bridges and their safety throughout life cycle via measurements of structural responses has been attracting increasing research efforts as an alternative to visual inspections or localised non-destructive tests [1], [2], [3], [4], [5], [6], [7], [8]. Particularly attractive are vibration measurements that can be processed for the estimation of the modal characteristics of bridges, as well as for the calibration of the finite element models (FEMs) used to simulate their behaviour. The information about the identified modal parameters and the updated FEMs can be used for design validation [9], [10], [11], [12], damage detection [13], [14], [15], [16] and structural health monitoring [17], [18], [19], [20], [21], [22].
The estimation of the modal characteristics often requires the application of system identification methods that process output-only, or ambient, vibrations, since for large-scale civil structures strong enough controlled input excitation cannot be directly imparted or measured in most practical situations. A variety of system identification techniques have been developed using only structural response data, which are typically referred to as operational modal analysis (OMA) or output-only modal analysis. The principal assumption made in the development of output-only identification techniques is that the input can be represented by a stationary white noise process. Stochastic techniques in the frequency domain, such as peak-picking [23], frequency domain decomposition [24] and maximum likelihood identification [25], have then been developed based on response auto and cross-spectral densities. However, due to limitations in frequency resolution as well as leakage errors, time domain methods are continuously being developed and enhanced. They arguably tend to provide better results, especially for large number of closely spaced modes hidden in the data. They usually can be subdivided into one-stage and two-stage methods. In the two stage approaches, firstly free vibration response is extracted from measurements based on random decrement functions or response correlation functions. Subsequently, modal parameters are identified using the classical system identification algorithms, e.g. the Ibrahim time domain method [26] or the least-squares complex exponential method [27], based on impulse or free response function estimates. In contrast to the two-stage approaches, one-stage system identification methods, such as the notable data-driven stochastic subspace identification (SSI-data) algorithm [28], [29], [30], can be used to identify modal parameters directly from the output-only measurements. Due to the advantage of state space representation of system dynamics in which the observation vector is separated from the system state vector, SSI-data has shown robustness in noise-contaminated testing environments and proved capable of yielding high quality modal identification results [31], [32], [33], [34], [35], [36]. However, its drawbacks are the computational effort and time required.
Amongst other state-of-the-art OMA techniques are the auto-regressive (AR) time series method [37], [38] and the eigensystem realization algorithm with observer/Kalman identification (ERA-OKID) [39], [40], which were originally developed for the identification of lightly damped linear structures in aerospace applications. They have shown good computational efficiency while preserving high identification accuracy. The parametric AR representations of measured responses capture the temporal covariances of the observed responses and can be converted into estimates of the natural frequencies, damping ratios and mode shapes of the structure. Recently, multivariate AR models have started to be applied to model the dynamic behaviour of vibrating structures and to infer their modal parameters when subjected to ambient vibration, such as industrial structures excited by hydro-electric generators or hydraulic pumps [41], [42], buildings [43] or cable stayed bridges [44].
In the ERA-OKID method, the observer/Kalman filter is used to remedy the lack of knowledge of initial conditions, extend the applicability of the ERA to general input-output systems, and create unique system Markov parameters. The ERA-OKID algorithm has proven to be very successful in several input–output cases [45], [46], [47]. In recent years, the ERA-OKID method has been extended to output-only measurements by splitting the state-space model into the deterministic and stochastic subsystems [48], [49], [50]. The strength of the ERA-OKID algorithm lies in the fact that the original state space equations are augmented with the observer gain for deriving the transformed state space model, which can eliminate the input and feedthrough matrices for large model orders, stabilize the system and increase its identification accuracy.
Although these two system identification algorithms can be found in civil engineering and SHM applications, their reliability and computational efficiency when confronted with full-scale, multiple span bridges have been poorly investigated, especially for comprehensive comparisons to the classical SSI-data method, which is better understood in these contexts. Such an application and comparisons are thus part of the focus of this paper.
The convenience of OMA continues to attract significant interest from both researchers and practitioners since bridge OMA tests will normally present none or only minimal disruption to the normal traffic. However, the natural excitation in ambient vibration testing (AVT) is always coloured noise, and thus the success of OMA techniques obviously depends on how close the excitation is to the principal white noise assumption. Notwithstanding, OMA has been invoked in testing of a wide range of short to medium-span bridges, including suspension structures [51], trusses [52] and arches [19]. However, AVT has shown some inherent drawbacks due to the variable nature of the actual excitation amplitude, direction, duration and frequency content which are, as per definition, not measured. For instance, the level of ambient excitation may be weak, or the frequency content may be narrow-band, and as a result only a limited number of modes can be extracted from the ambient data.
Missing modes have commonly been reported in AVT when compared to numerical structural models or alternative tests. Brownjohn et al. [53] reported difficulty in identifying the lateral modes of the Fatih Sultan Mehmet suspension bridge by utilising ambient traffic and wind excitation due to low vibrational levels. Alwash et al. [54] found that resonant response at the fundamental natural frequency could not be distinguished from background vibration, while higher modal responses could be identified in the case of environmental (wind and flowing water) excitation for testing of a three-span, 100.5 m long continuous reinforced concrete bridge. Chen et al. [36] pointed out that the fundamental lateral mode as well as several high order vertical modes failed to be extracted from the environmental excitation responses of a concrete motorway off-ramp bridge with 11 short to medium continuous spans. The uncertainty introduced by missing modes continues to challenge OMA, because it can have profound effects on the ultimate outcomes such as correct model updating or damage detection [55].
These challenges have seen an emergence and increased interest in hybrid vibration testing (HVT), where an artificial force is applied to the structure in addition to the ambient excitation, and a system model is identified using the response to such combined forcing [55], [56], [57], [58]. The main advantage of HVT is the ability to use excitation devices that are small and practical when compared to the exciters needed for the traditional experimental modal analysis (e.g. eccentric mass shakers), which are heavy and difficult to transport, since the amplitude of the artificial forces needed for HVT can be equal to, or even lower than, the amplitude of the ambient forces. HVT has been explored by Reynders et al. for the modal system identification of the Z24 bridge [57], a steel arch footbridge and a concrete stressed-ribbon footbridge [58]. In-depth comparisons with numerical models and other tests showed that the modal parameters obtained from the HVT are of good quality [58]. HVT-based FEM updating and bridge structural damage identification reported in [57] delivered improved well-posedness of the updating problem and allowed detecting inaccuracies in the FEM.
However, previous research efforts [57], [58] utilized either a drop weight system or a pneumatic artificial muscle to provide excitation for the implementation of HVT. Only in [56] a preliminary effort to use small electromagnetic shakers on a 13-storey reinforced concrete building was reported. Light-weight electrodynamic shakers [56], [59] can easily provide a specific, controllable excitation signal, such as a single or multiple concurrent sines, periodic, random, swept sine etc. We therefore propose herein, as our first contribution, that light electrodynamic shakers be utilized in HVT to provide the controlled artificial dynamic excitation to a multiple-span post-tensioned segmental concrete bridge to enable more robust and reliable dynamic analysis. To the best knowledge of the authors, small shakers with sweep or chirp excitation have not yet been used in HVT.
Furthermore, the previous work on HVT is restricted to only selected system identification methods, such as the reference-based combined deterministic-stochastic subspace identification [55]. The input artificial force needs be measured, which may not be favoured or convenient for the civil engineering practitioner working in challenging field conditions. Thus, there still exists a gap in exploring the HVT combined with output-only identification techniques popular for modal identification in the civil engineering community. The present study, on the other hand, explores output-only identification algorithms for extracting structural modal parameters from HVT, taking the advantages of both AVT (no requirement for input measurement) and increased excitation levels provided by HVT but using only small and practical actuators. Two different output-only time-domain system identification algorithms, i.e. AR and ERA-OKID, were employed to recover modal frequencies, damping ratios and mode shapes from the corresponding vibration response data. The SSI-data system identification algorithm was also applied to experimental data in order to compare the former algorithms’ performance in terms of mode identifiability, result quality and computational effort. Modal identification results are also obtained from pure AVT to highlight the improvements brought about by HVT. Furthermore, a detailed FEM was developed and used to calculate bridge vibration modes theoretically. The correlation analysis between experimentally and numerically derived modal parameters were conducted to obtain a further insight into the modal identifiability and identification reliability of the proposed HVT methodology. The application of OMA system identification techniques to HVT and thorough evaluation of their performance is our second contribution.
The following text is structured as follows. First, the bridge structural system and FE modelling are presented. Then, the choice of actuators and the proposed experimental program of the AVT and HVT are discussed. Next, the theory behind the modal system identification techniques adopted is briefly overviewed. The results of modal identification are then thoroughly discussed. The article ends with the conclusions drawn from the tests that summarise the key findings of the paper.
Section snippets
Description of the bridge and its finite element model
Our structure is the Nelson St. off-ramp bridge located on the southern fringe of the Central Business District of Auckland, New Zealand, as a part of a ‘spaghetti’ junction at the meeting point of three major motorways. Two views of the bridge are shown in Fig. 1, while Fig. 2 is a sketch explaining the overall structural form and arrangement and showing major dimensions and abutment and pier numbers. The bridge has both horizontal and vertical curvature (see Fig. 1a). Its total length is
Hybrid vibration tests
In the HVT of the Nelson St. off-ramp bridge, two light-weight APS Dynamics Model 400 ELECTRO-SEIS® long stroke shakers with APS 0412 Reaction Mass Assembly (Fig. 5) worked in tandem with environmental sources to excite the bridge in the vertical and lateral direction, respectively. The selection of the shaker position is generally a compromise between the theoretically optimal location and in-field practicalities. The theoretically optimal position is the one that excites most strongly all
Auto-regressive method
AR models are stochastic difference equations that are one of the most efficient and popular methods of describing time series data through parametric representation [3], [69], [70]. A multivariate AR model of order q, denoted as AR(), for acceleration response at discrete time k is:
In the above equation, is the matrix of coefficients for lag and is the vector of Gaussian white noise error terms. The parameters in an AR model can be estimated by
Natural frequencies and damping ratios
Table 1 provides a comparative overview of the modal frequencies and damping ratios obtained with the AVT and HVT data and the numerical FEM. In the table, the modes are indicated by symbols V, V/T and L for the vertical, vertical-torsional and lateral modes, respectively, and the mode number (Column 1). All modes below 10 Hz identified from the FEM are listed in the table. Taking the numerical modal analysis results (Column 2) as the reference, the vertical-torsional mode (V7/T1) could not be
Conclusions
In this paper, we present an HVT method for a post-tensioned segmental concrete bridge with eleven short to medium spans. In the HVT, a broad-band linear chirp excitation enhanced the environmental excitations acting on the bridge. This additional excitation was generated by relatively small and light exciters making the method easy to implement in the field. For modal system identification, three OMA techniques were used, namely the classical SSI-data as well as the AR and ERA-OKID methods,
CRediT authorship contribution statement
Ge-Wei Chen: Conceptualization, Methodology, Formal analysis, Investigation, Writing - original draft, Visualization. Xinghua Chen: Validation, Investigation, Resources. Piotr Omenzetter: Conceptualization, Investigation, Writing - review & editing, Visualization, Supervision, Project administration.
Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Acknowledgements
The authors acknowledge the financial supports of the Opening Fund of the Hubei Key Laboratory of Disaster Prevention and Mitigation (China Three Gorges University) (Grant No. 2017KJZ08). Piotr Omenzetter’s work on this paper within the LRF Centre for Safety and Reliability Engineering at the University of Aberdeen was partially supported by Lloyd’s Register Foundation. The Foundation helps to protect life and property by supporting engineering-related education, public engagement and the
References (88)
- et al.
Identification of unusual events in multi-channel bridge monitoring data
Mech Syst Sig Process
(2004) - et al.
Finite element modal analysis and vibration-waveforms in health inspection of old bridges
Finite Elem Anal Des
(2014) - et al.
Development and validation of an automated operational modal analysis algorithm for vibration-based monitoring and tensile load estimation
Mech Syst Sig Process
(2015) - et al.
Finite element modelling and updating of a lively footbridge: The complete process
J Sound Vib
(2007) - et al.
Seismic response trends evaluation and finite element model calibration of an instrumented RC building considering soil-structure interaction and non-structural components
Eng Struct
(2014) - et al.
Damage detection by finite element model updating using modal flexibility residual
J Sound Vib
(2006) - et al.
Damage classification and estimation in experimental structures using time series analysis and pattern recognition
Mech Syst Sig Process
(2010) - et al.
Statistic structural damage detection based on the closed-form of element modal strain energy sensitivity
Mech Syst Sig Process
(2012) - et al.
Structural damage detection using a parked vehicle induced frequency variation
Eng Struct
(2018) - et al.
Technology developments in structural health monitoring of large-scale bridges
Eng Struct
(2005)
Vibration based structural health monitoring of an arch bridge: From automated OMA to damage detection
Mech Syst Sig Process
Long-term monitoring and data analysis of the Tamar Bridge
Mech Syst Sig Process
N4SID: Subspace algorithms for the identification of combined deterministic-stochastic systems
Automatica
Ambient and free vibration tests of the Millau Viaduct: Evaluation of alternative processing strategies
Eng Struct
Operational modal analysis of an eleven-span concrete bridge subjected to weak ambient excitations
Eng Struct
Operational modal analysis by updating autoregressive model
Mech Syst Sig Process
Towards an automatic spectral and modal identification from operational modal analysis
J Sound Vib
System identification of smart buildings under ambient excitations
Measurement
Placement of sensors in operational modal analysis for truss bridges
Mech Syst Sig Process
Reference-based combined deterministic–stochastic subspace identification for experimental and operational modal analysis
Mech Syst Sig Process
Finite element model updating and structural damage identification using OMAX data
Mech Syst Sig Process
Dynamic modelling of post-tensioned concrete floors using finite element analysis
Finite Elem Anal Des
Assessment of optimal ARMA model orders for modal analysis
Mech Syst Sig Process
Dynamic testing and parameter identification of a base-isolated bridge
Eng Struct
Fully automated (operational) modal analysis
Mech Syst Sig Process
Automation of the stabilization diagrams for subspace based system identification
Expert Syst Appl
Experimental investigation into amplitude-dependent modal properties of an eleven-span motorway bridge
Eng Struct
Measured dynamic properties for FRP footbridges and their critical comparison against structures made of conventional construction materials
Compos Struct
Vibration serviceability of footbridges under human-induced excitation: A literature review
J Sound Vib
Model updating using genetic algorithms with sequential niche technique
Eng Struct
Bridge structural condition assessment using systematically validated finite-element model
J Bridge Eng
Application of time series analysis for bridge monitoring
Smart Mater Struct
Nonparametric analysis of structural health monitoring data for identification and localization of changes: Concept, lab, and real-life studies
Struct Health Monit
Recent perspectives in dynamic testing and monitoring of bridges
Struct Control Health Monit
Automated modal tracking in a football stadium suspension roof for detection of structural changes
Struct Control Health Monit
Assessment of highway bridge upgrading by dynamic testing and finite-element model updating
J Bridge Eng
Response surface–based finite-element-model updating using structural static responses
J Eng Mech
Structural health monitoring of civil infrastructure
Philos Trans R Soc Lond A: Math Phys Eng Sci
Structural health monitoring based on vehicle-bridge interaction: Accomplishments and challenges
Adv Struct Eng
Investigation of dynamic properties of long-span cable-stayed bridges based on one-year monitoring data under normal operating condition
Struct Control Health Monit
Development of a hybrid bridge evaluation system. Doctoral dissertation
Modal identification of output-only systems using frequency domain decomposition
Smart Mater Struct
A method for the direct identification of vibration parameters from the free response
Shock Vib Bull
Cited by (17)
Detecting deck damage in concrete box girder bridges using mode shapes constructed from a moving vehicle
2024, Engineering StructuresPrediction of displacement in Reinforced concrete based on artificial neural networks using sensors
2023, Measurement: SensorsCompletely non-contact modal testing of full-scale bridge in challenging conditions using vision sensing systems
2022, Engineering StructuresCitation Excerpt :Experimental modal analysis (EMA) i.e. modal testing with artificial forcing can be applied to estimate the dynamic properties of the structure (e.g., natural frequencies, mode shapes, and damping ratios) through subsequent analysis. Bridge operational modal analysis (OMA) [1–4] uses accelerometers to conduct ambient vibration testing (AVT) i.e. with natural loading. The standard approach in each case involves attaching accelerometers to the structure and wired cables from test points to the location of the data acquisition system.
Reference-free damage identification method for highway continuous girder bridges based on long-gauge fibre Bragg grating strain sensors
2022, Measurement: Journal of the International Measurement ConfederationCitation Excerpt :Information regarding damage to bridges is of great significance for bridge structural diagnoses, condition assessments and maintenance. Thus, to understand the status of a bridge and ensure its structural safety, the timely identification of bridge damage conditions by an effective method is necessary [3-6]. The process of implementing damage detection for bridges is part of structural health monitoring (SHM) [7].
Coupled vortex-induced modeling for spatially large-curved beam with elastic support
2022, International Journal of Mechanical SciencesCitation Excerpt :Simsek et al. [3] analyzed the vibration characteristics of functionally gradient simply supported beams under the action of concentrated moving harmonic loads. However, in actual engineering, elastic structures often have a certain initial curvature, such as bridges [4–7], cables [8–12], belts [13] and helical steam generators [14,15]. In order to simplify the research work, these curved structures were usually modeled as linear structures [16,17].
Output-only modal analysis for non-synchronous data using stochastic sub-space identification
2021, Engineering StructuresCitation Excerpt :In contrast, to the best of the authors’ knowledge, the issue of non-synchronicity of measurement data has not been investigated for the time domain system identification methods. Especially, for the SSI method, which is considered as the most dominant and popular technique in the time domain [29],[36–45], the solution to cope with the non-synchronicity of dynamic response measurements has not yet been proposed. To widen the applicability of SSI algorithm, this study proposes a strategy to extend it to system identification of non-synchronous dynamic response measurements.