Development of data processing techniques for laser doppler velocimeter based flow measurements

Abstract

Several schemes are available for conditional sampling of data on the velocity of a flowing fluid obtained in analogue form. These schemes are usually applied directly to the analogue data or to the digital data obtained by uniformly sampling the analogue data. For non?uniformly sampled digital data, such as those from a Laser Doppler Velocimeter (LDV), hardly any schemes are available except those devised for very specific applications. Hence, there is a need to develop conditional sampling techniques for LDV data. This thesis describes the development and application of certain data?processing techniques for flow measurements using a Laser Doppler Velocimeter. These techniques are developed for non?uniformly sampled data, typical of LDV, in the presence of random noise. The data?processing schemes are designed for: (a) the generation of a detector function (or intermittency function) for conditionally sampling the velocity data obtained by LDV, and (b) the determination of zone averages corresponding to the turbulent and non?turbulent regions of the flow. The generation of the intermittency function is done as follows: The instantaneous flow velocities and the time intervals between consecutive data realisations are first measured by LDV. A criterion function is then obtained by computing the square of the velocity derivatives. This criterion function is then smoothed by low?pass filtering using a scheme called variable window averaging (VWA). This filter is similar to a moving?window function. It computes short?time averages (depending on the averaging?window widths) of the noisy signal and effectively retrieves signals of low frequency. The intermittency function is generated by comparing the criterion function with a fractional multiple (threshold) of the mean square of the velocity fluctuations. The occasional zeros or dropouts in the intermittency function, which occur even during the fully turbulent part of the flow, are well known. An approach to digitally smooth such short?time aberrations in the intermittency function is also developed. This refined intermittency function is then used to compute the zone averages corresponding to the turbulent and non?turbulent regions of the flow. The properties of some of the above techniques, including their sensitivity to parameters, have been studied using both simulated and experimental data. The thesis also examines the nature of the distribution of sample intervals (the time intervals between valid LDV data). It has been reported in the literature that the arrival rates of scattering particles in the measurement volume follow a Poisson process. It is known that the inter?arrival times for such a process follow an exponential distribution. In the present work, the experimentally observed distribution of the inter?arrival times is found to be exponential only under certain conditions. These data?processing schemes are evaluated by applying them to an axisymmetric jet. Intermittency functions, intermittency factors, and zone averages are computed and compared with results obtained by other methods. The present results show fairly good agreement. Thus, this work reports the development of a set of techniques designed for non?uniformly sampled LDV data and demonstrates their application to a standard test case of turbulent?flow measurement.