Vibration-based structural health monitoring of cantilever-like structures under varying wind excitation

Neu, E 2016, Vibration-based structural health monitoring of cantilever-like structures under varying wind excitation, Doctor of Philosophy (PhD), Engineering, RMIT University.

Document type: Thesis
Collection: Theses

Attached Files
Name Description MIMEType Size
Neu.pdf Thesis Click to show the corresponding preview/stream application/pdf;... 17.42MB
Title Vibration-based structural health monitoring of cantilever-like structures under varying wind excitation
Author(s) Neu, E
Year 2016
Abstract The ever-increasing price pressure in the commercial aircraft market forces carriers as well as manufacturers to explore further cost-cutting potential. One promising approach for substantial cost reduction is the replacement of interval-based maintenance with continuous Structural Health Monitoring (SHM) systems. Over the last 20 years considerable effort has gone into the development of vibration-based techniques that can derive the current system health state from the structural response to ambient excitation. Many of these approaches rely on Operational Modal Analysis (OMA) techniques, which replace direct load measurements for modal parameter extraction with assumptions about the stochastic properties of the excitation source. In-flight loads could be a suitable source of excitation for vibration-based SHM of aircraft wings, but their eligibility has not been studied yet. Varying flight and operation conditions will introduce considerable variance to the modal properties of a wing, which could hide potentially critical damage. The separation of damage-induced modal parameter changes and flight-related changes introduced by velocity, angle of attack and mass variability, was not thoroughly studied yet. Continuous modal parameter-based SHM presupposes the availability of a robust Automated Operational Modal Analysis (AOMA) methodology. Most of the available AOMA techniques have been developed with regard to applications in civil engineering, where, in contrast to a wing in flight, damping is not dominated by aerodynamic forces and mode shapes do not show significant complexity. The existing AOMA procedures either require modes to have negligible complexity and to be lightly damped or they have to be manually parametrized, which makes them not well-suited for the application of aircraft wing AOMA-based SHM. Finally, no methodology was hitherto proposed to automatize the training set (baseline) preparation procedure of an AOMA-based SHM system. Instead, current practice is driven by iterative data processing and subjective assessment by expert users. This approach is nontransparent, labor intensive and may lead to less than optimal damage detection capability of the SHM system.

Two wind tunnel experiments are evaluated in this work. Data from the High Reynolds Number Aerostructural Dynamics (HIRENASD) experiment are used to investigate transonic wind excitation using measurements from accelerometers, strain gauges and from nearly 200 surface pressure sensors. Furthermore, an experiment with a composite cantilever was conducted to investigate damage detection under wind-, angle of attack- and mass-induced operational variability. Twenty-seven different operational and environmental conditions and one impact damage scenario are investigated. Measurements from Fiber Bragg Grating Sensors (FBGS) and piezoelectric sensors are used to compare multiple data normalization techniques and to validate the automatization techniques developed in this work.

Inflow velocity spectra and surface pressure measurements show that the wind-induced excitation is well-distributed over a wide frequency range, which is one major OMA load requirement. Furthermore, both experiments confirm that wind-loads, even on small-scale structures like the ones investigated in this work, can be considered to be composed of multiple independent sources, which is the second major OMA load requirement. Further discussion reveals that surface pressure variation caused by atmospheric and boundary layer turbulence on the surface of an aircraft wing can be regarded as an appropriate type of excitation for OMA. However, it is also found that narrow-banded transonic disturbances may be present at the wing surface and facility-related narrow-banded disturbances may be present in the incoming flow during wind tunnel testing. These phenomena were falsely identified as structural modes by OMA and further discussion revealed that load measurements must be available to distinguish between these two types of modes.

The damage detection investigation shows that it is possible to detect damage scenarios with modal parameter changes that are nearly an order of magnitude smaller than the Operational and Environmental Variability (OEV)-induced variability. A comparison of data normalization techniques that rely on direct measurement of the OEV shows that a step function approach applied to data that has innate breakpoints performs significantly better than the two other investigated techniques, namely a feature vector extended with information about the encountered operational and environmental conditions and data normalization using linear regression. The application of a Principal Component Analysis (PCA)-based unsupervised dimensionality reduction technique, which currently is one of the most popular approaches to account for unmeasured OEV in SHM, is critically discussed and the limitations of this approach are revealed. A comparison between the wind tunnel results and a numerical model of the investigated specimen shows that the natural frequency shifts introduced by the impact damage not only depend on the damage location, type and severity but also on the currently encountered operational and environmental conditions. The implications of this result for the feasibility of SHM levels that go beyond damage detection are discussed.

A multi-stage clustering approach for automated parametric OMA is introduced. In contrast to existing approaches, the procedure works without any user-provided thresholds, is applicable within large system order ranges, can be used with very small sensor numbers and does not place any limitations on the damping ratios or mode shape complexities of the system under investigation. Furthermore, a novel baseline preparation procedure is described that reduces the amount of user interaction to the provision of a single consistency threshold. The procedure starts with an indeterminate number of operational modal analysis identifications from a large number of datasets and returns a complete baseline matrix of natural frequencies and damping ratios that is suitable for subsequent anomaly detection. The two automatization procedures are integrated into a AOMA-based SHM system and used to detect an impact damage on a composite cantilever under OEV.

This work investigates the stochastic properties of wind-induced loads created by wind tunnels, including transonic flows, and shows that these are a suitable source of excitation for OMA-based modal parameter extraction of wing-like structures. Furthermore, it is examined how OEV as a result of mass, velocity and angle of attack changes influences the damage detection capability of an AOMA-based SHM system. Multiple approaches for OEV-normalization are studied, taking into account scenarios where direct OEV measurements are available as well as scenarios where the OEV influence has to be identified blindly. Finally, the current practice of expert user parametrization is critically discussed and automatization techniques for automated OMA and baseline set preparation are proposed, which overcome the limitations of the previously available approaches with regards to applications in aerospace engineering.
Degree Doctor of Philosophy (PhD)
Institution RMIT University
School, Department or Centre Engineering
Subjects Aerospace Structures
Keyword(s) structural health monitoring
operational modal analysis
operational and environmental variability
Version Filter Type
Access Statistics: 238 Abstract Views, 711 File Downloads  -  Detailed Statistics
Created: Thu, 02 Feb 2017, 13:45:25 EST by Denise Paciocco
© 2014 RMIT Research Repository • Powered by Fez SoftwareContact us