Simulation of sedimentation and fluidization of polydisperse suspensions via a Markov model

Abstract The particle-based approach to sedimentation is extended to include velocity fluctuations that result in hydrodynamic diffusion. The vector process describing the joint values of position and velocity is Markov. Thus, no integration of velocity is required. Height–velocity “skeletons” for each particle are generated from a bivariate-normal distribution with means, variances, and covariance that depend on three parameters. For each particle, there is a unique region in which the vector of species concentrations determines that particle's parameters and hence its Markov process, but the concentrations in that region depend on the Markov processes of neighboring particles. Though only discrete values of height and velocity are generated, the model ensures that sample paths and particle velocities are continuous. Furthermore, steady-state velocities are normally distributed and velocity autocorrelations decay exponentially. Published experimental results indicate that both are excellent approximations. For polydisperse suspensions, the Markov model is much simpler than the standard hydrodynamic-diffusion model and represents the actual process much better. We simulate the sedimentation and fluidization of polydisperse suspensions and study the effects of two additional parameters: variance and autocorrelation decay rate of particle velocities.