Drop collisions are critical in determining drop-sizes and mixing processes in multi-phase flows. However, existing spray drop collision models exhibit numerical dependencies and lack comprehensiveness in the outcomes that are considered. The present work attempts both to improve the physical accuracy and to reduce numerical dependencies in spray drop collision calculations.
As a first step, the range of collision outcomes was extended to consider bounce, coalescence and, fragmenting and non-fragmenting separation processes. Fragmentations were modeled assuming that the interacting drops form an elongating ligament that either breaks up by capillary wave instability, or retracts to form a satellite droplet. Comparisons of deterministic model predictions with experiments for colliding mono- and poly-disperse drop streams showed that the outcome of the collision and post-collision characteristics, such as drop-sizes, velocities and spatial distributions of drops, are predicted with acceptable accuracy.
Next, a radius-of-influence of collisional interaction approach was proposed to reduce numerical dependencies. In the method, discrete particle resolution in space was controlled either by manipulating the collision volume, or through selective parcel splitting/merging. To obtain fidelity in the simulations, a mean collision time obtained from the specifics of the collision process was formulated and used. The KIVA-3V CFD code with the new collision model and a gas-entrainment model that corrects the momentum exchange between the drops and the gas was used to simulate a high velocity, non-vaporizing diesel spray. Tests demonstrated adequate insensitivity of the model to numerical parameters.
The improved collision probability and collision outcome predictions were coupled to form a comprehensive collision model, which was used to simulate non-vaporizing diesel sprays produced under conditions relevant to diesel engine common-rail injection systems. Spray penetration results from visualization studies and, drop-size and velocity data from PDPA measurements were used to assess the capability of the model for different injection pressures and various chamber densities. The model gave reasonably good predictions under these conditions. The present improvements to the collision and spray models have increased the fidelity of spray modeling predictions, and are consequently expected to help improve the accuracy of multi-phase flow simulations.