Statistical transport theory for turbulent particle dispersion
A statistical transport theory for turbulent particle dispersion is formulated having general applicability to inhomogeneous turbulent flows. The new theory avoids the practical difficulties associated with defining a Lagrangian particle velocity autocorrelation function in the classical approach and circumvents the principle computational disadvantages associated with conventional Lagrangian stochastic discrete-particle techniques. Exploiting the concepts of statistical mechanics, groups of physical particles are represented by phase space distribution functions. In particular, a computational parcel representing a group of discrete physical particles is characterized by probability density functions (pdf) in configuration space and velocity space. Evolution of the first and second moments of these phase space distribution functions is described in a Lagrangian frame. The mean of each pdf is determined by tracking a parcel through a sequence of stochastic eddy interactions. A variance propagation theorem is fashioned from a direct attack on the linearized Lagrangian particle equation of motion (Langevin's equation) for sequential stochastic eddy interactions. Convolution of the parcel pdfs produces the probable instantaneous distribution in physical particle position and velocity. Using Gaussian pdfs, the validity of the new theory was established for dispersed particles in nonreacting and reacting turbulent flows by demonstrating favorable comparison with experiment and with predictions using the conventional stochastic direct modeling method. To improve efficiency, a memory function truncation criteria was developed for the variance propagation theorem demonstrating significant computational savings. In addition, pdf shape sensitivity was examined for uniform and isosceles triangle pdfs revealing that for the same sample size, uniform pdfs yielded highly irregular distributions whereas triangular pdfs yielded distributions virtually indistinguishable from those predicted using Gaussian pdfs. The resulting theoretical construction contributed by this dissertation offers unparalleled economy in pursuing large scale computation of turbulent particle dispersion.
Thesis92b.L583.pdf
7.62 MB
Adobe PDF
ab9d877049cfdad83d085737add085b4