Filtern
Erscheinungsjahr
Dokumenttyp
- Beitrag zu einem Tagungsband (18)
- Zeitschriftenartikel (5)
- Buchkapitel (1)
- Vortrag (1)
- Posterpräsentation (1)
Schlagworte
- Damage detection (6)
- Damage localization (6)
- Subspace methods (5)
- Fault detection (4)
- Model interpolation (4)
- Statistical tests (4)
- Load vector (3)
- Residual evaluation (3)
- Statistical evaluation (3)
- Temperature rejection (3)
Organisationseinheit der BAM
- 7 Bauwerkssicherheit (13)
- 7.2 Ingenieurbau (13)
Eingeladener Vortrag
- nein (1)
Structural health monitoring with statistical methods during progressive damage test of S101 Bridge
(2014)
For the last decades vibration based damage detection of engineering structures has become an important issue for maintenance operations on transport infrastructure. Research in vibration based structural damage detection has been rapidly expanding from classic modal parameter estimation to modern operational monitoring. Since structures are subject to unknown ambient excitation in operation conditions, all estimates from the finite data measurements are of statistical nature. The intrinsic uncertainty due to finite data length, colored noise, non-stationary excitations, model order reduction or other operational influences needs to be considered for robust and automated structural health monitoring methods. In this paper, two subspace-based methods are considered that take these statistical uncertainties into account, first modal parameter and their confidence interval estimation for a direct comparison of the structural states, and second a statistical null space based damage detection test that completely avoids the identification step. The performance of both methods is evaluated on a large scale progressive damage test of a prestressed concrete road bridge, the S101 Bridge in Austria. In an on-site test, ambient vibration data of the S101 Bridge was recorded while different damage scenarios were introduced on the bridge as a benchmark for damage identification. It is shown that the proposed damage detection methodology is able to clearly indicate the presence of structural damage, if the damage leads to a change of the structural system.
Fault detection and isolation can be handled by many different approaches. This paper builds upon a hypothesis test that checks whether the mean of a Gaussian random vector has become non-zero in the faulty state, based on a chi2 test. For fault isolation, it has to be decided which components in the parameter set of the Gaussian vector have changed, which is done by variants of the chi2 hypothesis test using the so-called sensitivity and minmax approaches. While only the sensitivity of the tested parameter component is taken into account in the sensitivity approach, the sensitivities of all parameters are used in the minmax approach, leading to better statistical properties at the expense of an increased computational burden. The computation of the respective test variable in the minmax test is cumbersome and may be ill-conditioned especially for large parameter sets, asking hence for a careful numerical evaluation. Furthermore, the fault isolation procedure requires the repetitive calculation of the test variable for each of the parameter components that are tested for a change, which may be a significant computational burden. In this paper, dealing with the minmax problem, we propose a new efficient computation for the test variables, which is based on a simultaneous QR decomposition for all parameters. Based on this scheme, we propose an efficient test computation for a large parameter set, leading to a decrease in the numerical complexity by one order of magnitude in the total number of parameters. Finally, we show how the minmax test is useful for structural damage localization, where an asymptotically Gaussian residual vector is computed from output-only vibration data of a mechanical or a civil structure.
Structural health monitoring (SHM) of civil structures often is limited due to changing environmental conditions, as those changes affect the structural dynamical properties in a similar way like damages can do. In this article, an approach for damage detection under changing temperatures is presentedand applied to a beam structure. The used stochastic subspace-based algorithm relies on a reference null space estimate, which is confronted to data from the testing state in a residual function. For damage detection the residual is evaluated by means of statistical hypothesis tests. Changes of the system due to temperature effects are handled with a model interpolation approach from linear parameter varying system theory. From vibration data measured in the undamaged state at some few reference temperatures, a model of the dynamic system valid for the current testing temperature is interpolated. The reference null space and the covariance matrix for the hypothesis test is computed from this interpolated model. This approach has been developed recently and was validated in an academic test case on simulations of a mass-spring-damper. In this paper, the approach is validated experimentally on a beam structure under varying temperature conditions in a climate chamber. Compared to other approaches, the interpolation approach leads to significantly less false positive alarms in the reference state when the structure is exposed to different temperatures, while faults can still be detected reliably.
The local asymptotic approach is promising for vibration-based fault diagnosis when associated to a subspace-based residual function and efficient hypothesis testing tools. It has the ability of detecting small changes in some chosen system parameters. In the residual function, the left null space of the observability matrix associated to a reference model is confronted to the Hankel matrix of output covariances estimated from test data. When this left null space is not perfectly known from a model, it should be replaced by an estimate from data to avoid model errors in the residual computation. In this paper, the asymptotic distribution of the resulting data-driven residual is analyzed and its covariance is estimated, which includes also the covariance related to the reference null space estimate. The importance of including the covariance of the reference null space estimate is shown in a numerical study.
Fault detection for linear parameter varying systems under changes in the process noise covariance
(2020)
Detecting changes in the eigenstructure of linear systems is a comprehensively investigated subject. In particular, change detection methods based on hypothesis testing using Gaussian residuals have been developed previously. In such residuals, a reference model is confronted to data from the current system. In this paper, linear output-only systems depending on a varying external physical parameter are considered. These systems are driven by process noise, whose covariance may also vary between measurements. To deal with the varying parameter, an interpolation approach is pursued, where a limited number of reference models -- each estimated from data measured in a reference state -- are interpolated to approximate an adequate reference model for the current parameter. The problem becomes more complex when the different points of interpolation correspond to different noise conditions. Then conflicts may arise between the detection of changes in the eigenstructure due to a fault and the detection of changes due to different noise conditions. For this case, a new change detection approach is developed based on the interpolation of the eigenstructure at the reference points. The resulting approach is capable of change detection when both the external physical parameter and the process noise conditions are varying. This approach is validated on a numerical simulation of a mechanical system.
The local asymptotic approach is promising for vibration-based fault diagnosis when associated to a subspace-based residual function and efficient hypothesis testing tools. It has the ability of detecting small changes in some chosen system parameters. In the residual function,the left null space of the observability matrix associated to a reference model is confronted to the Hankel matrix of output covariances estimated from test data. When this left null space is not perfectly known from a model, it should be replaced by an estimate from data to avoid model errors in the residual computation. In this paper, the asymptotic distribution of the resulting data-driven residual is analyzed and its covariance is estimated, which includes also the covariance related to the reference null space estimate. The advantages of the data-driven residual are demonstrated in a numerical study, and the importance of including the covariance of the reference null space estimate is shown, which increases the detection Performance.