Study of the porosity in plasma-sprayed alumina through an innovative three-dimensional simulation of the coating buildup
Résumé
Porosity is a key feature of a thermally sprayed coating microstructure. Within ceramic coatings, porosity is made of pores and cracks of various shapes, dimensions, and orientations. Cracks can be intralamellar or interlamellar due to the buildup of the coating, which leads to piled-up lamellae from impinging and the additional rapid solidification of liquid droplets. Pores are interconnected with cracks, which results in a three-dimensional (3-D) porosity network. Direct observation of this network is an intricate task and current attempts remain somewhat limited. A 3-D simulation of this network was, therefore, developed in this work, based on a stochastic approach to the building up of simulated lamellae in the sprayed microstructure. A library of mathematical objects was achieved from morphological measurements, using confocal microscopy of actual isolated flattened lamellae, i.e., “splats” and scanning electron microscopy (SEM). This stochastic approach to the simulation of hundreds of lamellae also involves the random distribution of cracks and pores. Simulation fit parameters were selected according to the overall characteristics of porosity (i.e., content, orientation, size, etc.) that were determined from the thorough quantitative image analysis (QIA) of cross-sectioned plasma-sprayed alumina coatings. Two plasma modes that varied the atmosphere in a controlled-atmosphere plasma spraying (CAPS) chamber were applied, to produce the microstructures of two different alumina coatings. The 3-D random modeling tool allowed the processing of a volume of digital material through a 3-D simulated binary image of a two-phased composite material. Using one 3-D image result of the simulation, finite element (FE) calculations were performed, in order to study the overall dielectric properties of a plasma-sprayed alumina as a function of porosity. The influence of anisotropy is discussed, in particular, and both analytical and numerical predicted values were compared to experimental ones. The presence of the defects related to both digital coatings is also discussed.