Research On Vibration Measurement And Modal Identification Of Wind Turbine Blade Based On Network Photogrammetry

The blade is one of the most important components of wind turbine generator(WTG).It covers 15%-20% of WTG's cost.The costs of manufacturing,transportation,operation and maintenance increase significantly with the upsizing of the blade.The blade will vibrate under impact of strongly varied wind,which may result in failure or breakage of the blade,even lead to big losses to the wind power plant possibly.The researches on blade's structure feature and vibration characteristics as well as identification of damping ratio,frequency and mode shape etc.will help manufacturers to design better dynamic structure,and help power plants to have more stable WTG.As the blade with features of irregular surface and large size,it is difficult to measure and analyze the vibration response with contact sensors.Therefore,it is quite important to develop non-contact vibration measurement and modal identification means for wind turbine blades.Network photogrammetry is a kind of these non-contact measurement methods which will be studied in this dissertation.In order to perform network photogrammetry for large-scale structures like wind turbine blade,a number of cameras are synchronized and mapped together;binocular vision,point tracking,global calibration are deeply explored;software and extendable measuring platform with multiple cameras for testing experiments are developed.The camera locations,distance and hardware parameters are analyzed to improve the performance of the measuring system.Modal parameters of blade are identified.The main research results are as follows.A method of global vibration measurement for multi-camera parallel distributed network photogrammetry is proposed.Based on binocular rendezvous imaging and registration of the homogenous points from two cameras,a method of multi-camera step by step decoupling beam adjustment is proposed to obtain the optimization parameters Coded marks are used in the photogrammetry.The center point is tracked and matched in the successes images.An improved tracking and matching algorithm for homogenous points is presented based on the traditional KLT with minimized distance criterion.By utilizing key technologies such as global calibration,homonym matching based on coded marks and dynamic feature tracking,high precision dynamic measurement of large field of view can be realized,which meets the engineering application requirements of vibration response measurement of large-scale structures.Two kinds of coded marks are designed and the corresponding location and recognition methods are proposed.One kind mark composed of cross and circles is proposed.Intersection point of cross is extracted based on minimization of curvature changes with sub-pixel cross skeleton.The code information of circle arrangement is decoded with affine inverse transformation and look-up table method.Another mark is composed of characters and circle.The code value is determined by characters.The character region and circular region are obtained by multi-level threshold segmentation.After gray smoothing,the gray centroid method is used to locate sub-pixel center,and the BP neural network training is used to accurately classify,recognize and decode.The example shows that the designs of coded marks are simple and the positioning and recognition accuracies can meet the requirements of large-scale structure dynamic photogrammetry.A target-less approach for vibration measurement with significant linear edge characteristic based on binocular photogrammetry is proposed.In the normal direction of the coarse positioning skeleton,the sub-pixel skeleton is obtained by interpolation search according to the principle of smoothing edge and minimizing the change of adjacent curvatures.Virtual target points are constructed by line segment cross-ratio,and homonym points are tracked and matched based on cross-ratio invariance.Then,vibration of blade edge can be measured.This method does not need to lay cooperative marks in advance,also can meet the dynamic measurement accuracy and has good application prospects.An extensible multi-camera parallel distributed network system is constructed.Combining with the technical indexes of the system and the selection of equipment parameters,the hardware platform is built,and the corresponding software function modules are developed.The relationship between measurement accuracy and position parameters,also the intersection measurement errors are analyzed.The feasibility of six cameras parallel network schemes is further discussed.The analysis results provide reasonable reference for the spatial layout and hardware selection of large-scale structure test system.A 3 kW wind turbine blade was measured by the test system.The relative errors of the first three frequencies identified from circled char marks are less than 0.9% with the results of hammering test using B&K system.Four cameras are combined to form a network to measure the spatial trajectories of all marks under impulse excitation in a global coordinate system.Observation shows that the measured results are consistent with the spatial arrangement and real vibrations of the blade surface marks,and the relative frequency errors are less than 0.9% as well.The above results show that the proposed method is reliable and feasible,and the measurement accuracy is high.An improved method based on Data-SSI is proposed.The Hankel matrix is reconstructed with different dimensions.The Data-SSI process is done many times and produces groups of possible poles.A new index of modal similarities is designed to discriminate spurious poles.The inherent frequencies are acquired by spectral clustering analysis automatically.The results show that most of the spurious poles can be eliminated by means of multiple reconstruction and modal similarity criterion,which can significantly improve the recognition accuracy.The method of vibration measurement and modal identification based on multi-camera network photogrammetry has important academic significance in developing and enriching of the theory and method of vibration modal measurement of large and complex structures.This method can be effectively applied to structures with large-scale or irregular surfaces such as wind turbine blades.The researches in this dissertation can provide basis and technical support for vibration characteristics analysis,quality detection,fault diagnosis and prediction.And it also has good engineering application values.