Research On Analysis Method Of Shear Turbulence Data In Time,Frequency And Wavenumber Domain

When turbulence data is processed in computer, de-noising and wavenumber spectrum ofautomatic matching are improved to be important problems, among them, the wavenumberspectrum is an important physical parameter to evaluate the characteristics of turbulence. Whenshear probe measures turbulence signal, the instrument vibration and environmental noiseinterference will pollute the turbulence signal. Traditional de-nosing algorithm, which used theFourier transform method, only deal with the signal of the statistical average results, it is verysuitable for processing the deterministic stationary signal, but in reality, the signal is usuallynon-stationary, which is difficult to solve the variable of anomaly. Some scholars proposed waveletde-noising method to eliminate the noise signal, and has obtained certain achievements, but due tothe turbulence is a kind of complicated three-dimensional fluid motion, it is difficult to structurewavelet base. In addition, the noise energy will affect the matching precision of wavenumberspectrum, and the domestic spectral matching are using artificial recognition method to judge thespectral quality, which will affect the objectivity and the intelligentialize of data processing andreduce the efficiency and accuracy of data processing. Therefore, turbulence de-noising andwavenumber matching algorithms play an important role in engineering application and theresearch of turbulent theory.In this thesis, the shear turbulence data which is obtained by the microstructure of oceanturbulence observation instrument is taken as the processing object, and the key theoretical issues inthe process of data processing are studied in turn in theory and technology. Based on the researchesat home and abroad, the de-noising in time domain, data statistical analysis in frequency domainand wavenumber spectrum matching algorithm in wavenumber space have carried on as the keyresearch. First, when turbulence data is processed in time domain, considering the vibration signalwill interference turbulence signal, a de-noising algorithm based on wiener filtering is put forward;Then the data in time domain is converted into frequency domain, based on the original turbulencedata conforms to the characteristics of normal distribution, the distribution characteristics infrequency domain are explored; Finally frequency domain is converted into wavenumber space, according to the characteristic that turbulence data conforms to normal distribution, a wavenumberspectrum matching algorithm based on maximum likelihood is proposed. Main innovative aspectsare discussed as follows:(1) Research on de-noising algorithmConsidering the problem that the instrument vibration signal will pollute the turbulence signaldetection and on the basis of the acceleration signal and the shear turbulence signal have a certainintrinsic relationship, the algorithm uses three axis acceleration signal as reference signal, accordingto the principle of minimum mean square error, a de-noising algorithm based on wiener filtering isput forward. Through solving the time domain of wiener filtering to get the optimal weightcoefficient, the vibration noise is maximum eliminated from the turbulence signal. Application theturbulence data which is obtained by fixed-point turbulence instrument observation to validate theeffectiveness of the algorithm, and the results show that the algorithm is effective to improve theprecision of the data.(2) Research on turbulence analytical method in frequency and wavenumberdomainBased on the turbulent statistical average method and probability distribution, the data ofturbulence has been carried on the detailed statistical analysis in the frequency and the wavenumberdomain. Although turbulence is complicated three-dimensional and unsteady fluid motion, itsrandomness still decides that its accord with a certain statistical distribution. This paper applies theideas of the10fold cross-validation to learn the statistical distribution features of power spectrumdensity algorithm in frequency and wavenumber domain. The turbulence data is divided intotraining set and testing set, then the testing set are estimated based on the statistical characteristicsof the training set, and finally the statistical characteristic of turbulence data is evaluated. Throughexperimental data which is got by vertical profiler to validate the feasibility and validity of thealgorithm, and the results show that the algorithm has a certain effectiveness, which can obtain thestatistical distribution features from a large number of seemingly random turbulence data.(3) Research on turbulence wavenumber spectrum matching algorithmOn the basis of the turbulence data complies with statistical characteristics, the automatic matching algorithm of wavenumber spectrum is studied. Considering that large amount of theobserved spectrum are fluctuate around the Nasmyth empirical spectrum, and then supposed that thedistance differentials between observed spectrum and the Nasmyth empirical spectrum accord withthe statistics law of normal distribution. The algorithm takes the turbulence data which wasobtained by the designed turbulence observation instrument (TOI) as the original data and takes thedifferences between the observed spectrum and Nasmyth empirical spectrum in wavenumber spaceas processing object. First, the cross validation method is applied to data-preprocessing to prove thevalidity of the hypothesis. Then the discriminate criterion of characteristic data is given to identifythe invalid data in the observed spectrum through maximum likelihood method and the invalid iswavenumber spectrum eliminated finally. The flume comparison experiment between TOI andacoustic doppler velocimeter (ADV) is used to verify the validity of the algorithm. The results showthat the algorithm is feasible and effective, which improves the accuracy of turbulence data.