Refinement of breakup models for compact powder agglomerates exposed to turbulent flows considering relevant time scales

Abstract

Breakup processes of dry powder agglomerates exposed to turbulent flows are investigated based on Euler–Lagrange predictions relying on the large-eddy simulation technique and the hard-sphere method. For this purpose, recently developed breakup models for turbulent, drag and rotary stresses [1] are refined taking relevant time scales of the physical processes into account. Depending on the effective breakup mechanism a breakup time is determined defining a time lag between two subsequent breakup processes of the same agglomerate. This physically motivated measure ensures that the methodology is independent of the applied time-step size and avoids unphysical collisions of the fragments and their re-agglomeration. To study the modeling approach in a wider range of applications, six different cases of the particle-laden flow in a generic dry powder disperser are investigated: Three different powders with varying strength due to different sizes of the primary particles forming the compact agglomerates and two different Reynolds numbers differing by a factor of eight. Thus, the entire range between a very mild dispersion rate and a complete disintegration of the agglomerates is covered by this matrix of operating conditions. After analyzing the most critical flow regions where breakup processes are expected beforehand and finally found in the simulations, detailed statistics concerning the different breakup mechanisms are generated and evaluated. That includes their percentages of the total number of breakup events and the number of events normalized by the number of released agglomerates. Furthermore, the dispersion rates achieved at the outlet of the disperser are analyzed and compared with experimental measurements [2]. It is shown that the suggested enhancement of the breakup models by taking physically relevant time scales into account significantly improves the predicted results especially in the high-Re case and that overall a good agreement with the measurements is achieved.