A 3d Stochastic Model for Geometrical Characterization of Particles in Two-Phase Flow Applications

Abstract

In this paper a new approach to geometrically model and characterize 2D silhouette images of two-phase flows is proposed. The method consists of a 3D modeling of the particles population based on some morphological and interaction assumptions. It includes the following steps. First, the main analytical properties of the proposed model – which is an adaptation of the Matérn type II model – are assessed, namely the effect of the thinning procedures on the population’s fundamental properties. Then, orthogonal projections of the model realizations are made to obtain 2D modeled images. The inference technique we propose and implement to determine the model parameters is a two-step numerical procedure: after a first guess of the parameters is defined, an optimization procedure is achieved to find the local minimum closest to the constructed initial solution. The method was validated on synthetic images, which has highlighted the efficiency of the proposed calibration procedure. Finally, the model was used to analyze real, i.e., experimentally acquired, silhouette images of calibrated polymethyl methacrylate (PMMA) particles. The population properties are correctly evaluated, even when suspensions of concentrated monodispersed and bidispersed particles are considered, hence highlighting the method’s relevance to describe the typical configurations encountered in bubbly flows and emulsions.

Using deep learning to retrieve 3D geometrical characteristics of a particle field from 2D projected images: Application to multiphase flows

“…Therefore, modeled images (also called synthetic images in this paper) of multiphase flows could resolve this problem, where the ground truth information is easily accessible and can be generated in large quantities. In order to generate this type of data we refer to stochastic geometry models, where they showed their efficiency to characterize a particle field in a multiphase flow [Langlard et al, 2018b].…”

Estimation of 3D geometrical properties of spheroid-like particle systems using projection images

“…Image analysis has become a powerful tool for monitoring particle systems as it enables to segment the particles projections on 2D experimental images and statistical analysis can be applied to retrieve information on the dispersed phase. Many segmentation algorithms [2]- [4] and modeling methods [5], [6] are reported to geometrically characterize particle systems in the case of sphere-like particles. However, for non-spherical ones, additional problems arise due to the projection mapping which results in a loss of information [7], [8].…”

Bayesian Inference of a Parametric Random Spheroid from its Orthogonal Projections

“…In many situations, ellipsoids arise as a simple, but realistic, model for given objects. For example, ellipsoidal models are frequently encountered in medical areas (Fessler and Macovski 1991;Jaggi et al 1995;Noumeir 1999), and they are also used as generalized model for droplets and/or bubbles in the field of chemical engineering (de Langlard et al 2018b;Kracht et al 2013). More generally, they are fundamental in 3D imaging (Merola et al 2013;Liu et al 2006;Ozturk-Isik et al 2009) for which morphological characteristics (volume, surface, eccentricity, etc.…”