Aral Sarrafi

Mode extraction on wind turbine blades via phase-based video motion estimation

In recent years, image processing techniques are being applied more often for structural dynamics identification, characterization, and structural health monitoring. Although as a non-contact and full-field measurement method, image processing still has a long way to go to outperform other conventional sensing instruments (i.e. accelerometers, strain gauges, laser vibrometers, etc.,). However, the technologies associated with image processing are developing rapidly and gaining more attention in a variety of engineering applications including structural dynamics identification and modal analysis. Among numerous motion estimation and image-processing methods, phase-based video motion estimation is considered as one of the most efficient methods regarding computation consumption and noise robustness. In this paper, phase-based video motion estimation is adopted for structural dynamics characterization on a 2.3-meter long Skystream wind turbine blade, and the modal parameters (natural frequencies, operating deflection shapes) are extracted. Phase-based video processing adopted in this paper provides reliable full-field 2-D motion information, which is beneficial for manufacturing certification and model updating at the design stage. The phase-based video motion estimation approach is demonstrated through processing data on a full-scale commercial structure (i.e. a wind turbine blade) with complex geometry and properties, and the results obtained have a good correlation with the modal parameters extracted from accelerometer measurements, especially for the first four bending modes, which have significant importance in blade characterization.

A Comparison of Computer-Vision-Based Structural Dynamics Characterizations

As a specific modern non-contact sensing technology, optical/video information is getting more and more attention employed to interpret structural responses and system status awareness. By means of processing the acquired video, a full-field system information is available which may be applied later to Experimental Modal Analysis (EMA), Structural Health Monitoring (SHM), System Identification (SI), etc., while at the same time, there is no influence to the structural testing such as mass loading and stiffness change. There are numerous technologies to extract the dynamic response of structures from acquired videos. In this paper, several point tracking algorithms are particularly compared, including Lucas-Kanade tracker, Hungarian registration algorithm and particle filter. These computer vision algorithms are implemented to extract the natural frequencies of a lab-scale structure, and the efficiency of each method is investigated regarding the consistency in estimating the natural frequencies and computational time. The recorded video contains external noise caused by lighting change during the experiment, as well as the intrinsic uncertainty on the photosensitive devices. Therefore, the natural frequencies estimated via different algorithms will have different values. An overall comparison between several computer vision algorithms are made in this paper in terms of precision, and computational load.

Vibration-based damage detection in wind turbine blades using Phase-based Motion Estimation and motion magnification

Abstract Vibration-based Structural Health Monitoring (SHM) techniques are among the most common approaches for structural damage identification. The presence of damage in structures may be identified by monitoring the changes in dynamic behavior subject to external loading, and is typically performed by using experimental modal analysis (EMA) or operational modal analysis (OMA). These tools for SHM normally require a limited number of physically attached transducers (e.g. accelerometers) in order to record the response of the structure for further analysis. Signal conditioners, wires, wireless receivers and a data acquisition system (DAQ) are also typical components of traditional sensing systems used in vibration-based SHM. However, instrumentation of lightweight structures with contact sensors such as accelerometers may induce mass-loading effects, and for large-scale structures, the instrumentation is labor intensive and time consuming. Achieving high spatial measurement resolution for a large-scale structure is not always feasible while working with traditional contact sensors, and there is also the potential for a lack of reliability associated with fixed contact sensors in outliving the life-span of the host structure. Among the state-of-the-art non-contact measurements, digital video cameras are able to rapidly collect high-density spatial information from structures remotely. In this paper, the subtle motions from recorded video (i.e. a sequence of images) are extracted by means of Phase-based Motion Estimation (PME) and the extracted information is used to conduct damage identification on a 2.3-m long Skystream® wind turbine blade (WTB). The PME and phased-based motion magnification approach estimates the structural motion from the captured sequence of images for both a baseline and damaged test cases on a wind turbine blade. Operational deflection shapes of the test articles are also quantified and compared for the baseline and damaged states. In addition, having proper lighting while working with high-speed cameras can be an issue, therefore image enhancement and contrast manipulation has also been performed to enhance the raw images. Ultimately, the extracted resonant frequencies and operational deflection shapes are used to detect the presence of damage, demonstrating the feasibility of implementing non-contact video measurements to perform realistic structural damage detection.

Uncertainty quantification of phase-based motion estimation on noisy sequence of images

Optical measurement and motion estimation based on the acquired sequence of images is one of the most recent sensing techniques developed in the last decade or so. As a modern non-contact sensing technique, motion estimation and optical measurements provide a full-field awareness without any mass loading or change of stiffness in structures, which is unavoidable using other conventional transducers (e.g. accelerometers, strain gauges, and LVDTs). Among several motion estimation techniques prevalent in computer vision, phase-based motion estimation is one of the most reliable and accurate methods. However, contamination of the sequence of images with numerous sources of noise is inevitable, and the performance of the phase-based motion estimation could be affected due to the lighting changes, image acquisition noise, and the camera’s intrinsic sensor noise. Within this context, the uncertainty quantification (UQ) of the phase-based motion estimation (PME) has been investigated in this paper. Based on a normality assumption, a framework has been provided in order to characterize the propagation of the uncertainty from the acquired images to the estimated motion. The established analytical solution is validated via Monte-Carlo simulations using a set of simulation data. The UQ model in the paper is able to predict the order statistics of the noise influence, in which the uncertainty bounds of the estimated motion are given, after processing the contaminated sequence of images.

Probabilistic uncertainty quantification of wavelet-transform-based structural health monitoring features

In the application of Structural Health Monitoring (SHM), processing the online-acquired data plays a very important role, among which wavelet transform is an outstanding tool and compared to Fourier transform, it handles the nonstationary behaviors in the time series in an adaptive fashion. When dealing with time-variant data, there are uncertainties from numerous resources inherent to the feature estimation, such as measurement noise, operational and environmental variability, hardware limitation, etc. The corruption from uncertainty will make the data interpretation ambiguous and thereby dramatically degrades the decision quality with regard to the occurrence, location, severity, and extent of damages. This paper derives a probabilistic model to quantify analytically the uncertainty of wavelet transform feature as a random variable, and variance is derived analytically in this work. Considering central limit theorem, Gaussian probability density function characterizes the distribution and this has been validated via Monte Carlo testing. By fully characterizing the uncertainty, the damage detection implementations may be facilitated with a quantified false alarm rate and miss catch rate.

Statistical Modeling of Wavelet-Transform-Based Features in Structural Health Monitoring

In this paper, a probabilistic model is established in quantifying the uncertainty of wavelet-transform-based features in structural health monitoring (SHM), thus the decision-making in regard to damage occurrence will be conducted under a quantified confidence. Wavelet transform has been adopted in processing online-acquired data for decades, and the adaptability of wavelet transform in handling time and scale resolutions make it a powerful tool to interpret the time-variant data series. For the complexity of real SHM problems, uncertainty from both the operational/environmental variability and the inaccuracy of data acquisition hardware degrades the SHM performance. This paper aims to derive a probabilistic uncertainty quantification model to describe the distribution of wavelet-transform-based features, to facilitate more reliable SHM decision-makings, and uncertainty-induced false-positive (Type-I error) and true damage detection rate will be traded-off in a confidence-quantified sense. The distribution derived in this paper is validated via Monte Carlo simulation.

Extracting High Frequency Operating Shapes from 3D DIC Measurements and Phased-Based Motion Magnified Images

The three-dimensional digital image correlation (3D DIC) method in conjunction with a stereo-vision system can provide the full-field dynamic displacements of a structure with sub-pixel accuracy. However, stereo-photogrammetry systems are limited by camera resolution and intrinsic noise of the acquired images. Thus, in order to use optical sensing techniques to identify dynamic characteristics of a structure at high frequencies, the signal-to-noise ratio (SNR) in the sequence of images taken with a stereo-vision system needs to be improved. Within this paper phase-based video magnification, in conjunction with 3D DIC are used to visualize the high frequency operating shapes of a cantilever beam. The magnified sequence of images using motion magnification technique are post-processed using 3D DIC to quantify infinitesimal deformation that is not recognizable using only digital image correlation. The results obtained within this paper reveal the great potential of extracting 3D operating shapes of a high frequency structure using the motion magnification and stereo-photogrammetry techniques. Moreover, results of this paper indicate that using the motion magnification technique increases the SNR of the measurements, and could be used as a new approach to extract more information about the structure than previously possible compared to using 3D DIC alone.

Using 2D Phase-Based Motion Estimation and Video Magnification for Binary Damage Identification on a Wind Turbine Blade

Videos (sequence of images) as three-dimensional signals may be considered as a very rich source of information for several applications in structural dynamics identification and structural health monitoring (SHM) systems. Within this paper high-speed cameras are used to record the sequence of images (video) of a baseline and damaged wind turbine blade (WTB) while vibrating due to the external loadings. Among several computer vision algorithms for motion extraction from the videos, phase-based motion estimation technique is used to extract the response of both the baseline and damaged wind turbine blade. Modal parameters (natural frequencies and operating deflection shapes) were used as damage sensitive features in order to detect the occurrence of damage in the wind turbine blade. The first four natural frequencies of the both baseline and damaged wind turbine blade are extracted by analyzing the estimated motion provided by the phase based motion estimation in the frequency domain. The motion magnification algorithm is also utilized to visualize and extract the operating deflection shapes of the wind turbine blade which may be used later as an indicator of the presence of damage. It has been shown that changes in the dynamic behavior of the wind turbine blade will result to deviations in the nominal natural frequencies and operating deflection shapes, and the damaged wind turbine blade can be differentiated from the baseline WTB using this non-contact measurement approach.

Detection of Natural Frequency and Mode Shape Correspondence Using Phase-Based Video Magnification in Large-Scale Structures

Structural dynamics identification is an important part of both the design and certification process for large-scale structures and specifically utility-scale wind turbine blades. Finding the correspondence between the estimated natural frequencies and the mode shapes of interest can be a very challenging due to the sheer size of the structures and the large amount of instrumentation required. The state of the art methods in experimental modal analysis (EMA) and operational modal analysis (OMA) require attachment of numerous accelerometers along the test structure to extract the natural frequencies and the mode shapes. Instrumenting large structures with accelerometers and handling the wiring and the connections can be a very labor-intensive task; therefore, alternative methods should be considered to address this problem. Within this paper, the capabilities of phase-based video magnification and motion estimation are investigated to find the correspondence between the natural frequencies and the mode shapes. The sequence of images (video) is recorded from the vibrating wind turbine blade and then processed using the phase based motion estimation to extract the spectrum of the response of the wind turbine blade to the impact excitation. Afterward based on the obtained spectrum the recorded videos are magnified to visualize the operating deflection shapes. The motion magnified videos represent the visual perception of the operating deflection shapes, which can be used to find the correspondence between the natural frequencies and the mode shapes. The results of this method have also been validated using the benchmark modal data from the accelerometers as well as the point tracking optical measurement method.

Feasibility of Applying Phase-Based Video Processing for Modal Identification of Concrete Gravity Dams

Hydraulic structures have been considered as one of the most essential civil infrastructures, and play a critical role in developing countries throughout the history for water storage and electricity generation. Due to the importance and the catastrophic consequences of unexpected failures in hydraulic infrastructures, monitoring and maintenance of dams should be handled very meticulously and with high precision. Among several measurement techniques as a specific modern non-contact sensing technology, optical/video information is getting more and more attention to interpret structural responses and system status awareness. By means of processing the acquired video, a full-field system information is available which may be applied later to Experimental Modal Analysis, Structural Health Monitoring (SHM), System Identification, etc. Such non-contact full-field sensing technologies avoid the installation of a gigantic number of conventional sensors in the occasions of large dimension. Within this context, the feasibility of applying Phase-Based Motion Estimation (PME) and video magnification has been studied for structural identification purposes on the concrete gravity dam subject to white noise excitations. Firstly, the PME and motion magnification algorithms are validated by the comparison of a lab-scale cantilever beam test and the numerical simulation. Next, the modal dynamic procedure in ABAQUS is carried out and the time history response of the dam is obtained. Then the simulated motion video of the dam is exported and processed using PME and magnification. The video processing results are finally compared with the results from frequency procedure in ABAQUS. The results obtained prove the concept of using PME and video magnification as a successful methodology in the modal identification of large-scale concrete gravity dams.