| Turbulent collision of inertial particles, a process relevant to many natural and industrial applications, has been investigated for over 50 years. However, since the collision rate is governed by a wide range of scales of fluid motion and a direct means of quantifying this process had not been available, progress in its understanding had been very slow. This dissertation represents a systematic effort to advance our understanding of turbulent collision of inertial particles using a tool known as direct numerical simulations (DNS).; We have successfully built a DNS database for the geometric collision rate of inertial particles. With the help of DNS data, we resolved several open questions on turbulent collision. The first question concerns the statistical description of average collision kernel. Two formulations known as the spherical formulation and the cylindrical formulation had been used in the past. We showed that the two formulations are not equivalent and the spherical formulation is the only correct formulation.; Next we confirmed the well-known Saffman & Turner (1956) formulation for the collision rate of zero-inertia particles to within a numerical uncertainty as small as 1%. More importantly, the underlying assumptions in their theory were clarified.; For finite-inertia particles, we found that the collision rate is governed by two physical mechanisms, namely, the turbulent transport effect and the accumulation effect. They represent the effect of large-scale and small-scale fluid motion on the collision rate, respectively.; The accumulation effect was found to reach a maximum when particle inertial response time is on the order of flow Kolmogorov time. It can increase the collision kernel by an order of magnitude even at moderate flow Reynolds number. We found that the accumulation effect scales linearly with Reynolds number for both monodisperse and bidisperse systems in our simulations.; The turbulent transport effect is dominated by the large-scale energetic eddies. A simple model based on an eddy-particle interaction concept was developed for this effect and shown to give a good prediction of the net collision kernel at large particle inertia.; Finally, quantitative models for both effects were developed. An integrated model was shown to predict the net collision kernel to within 25% relative error for all parametric ranges.; In summary, we were able to resolve several issues in the microphysical modeling of geometric collision rates. The effect of fluid turbulence on the collision rate of finite-inertia particles, we believe, can now be modeled with a reasonable certainty for all flow Reynolds numbers and particle parameters. |
