Level-set Modeling Simulations of Chemical Vapor Infiltration for Ceramic Matrix Composites Manufacturing

Abstract

Silicon-carbide (SiC) reinforced ceramic matrix composites (CMCs) are a key enabling technology to reduce fuel consumption and emissions of gas turbine engines. In one manufacturing approach, chemical vapor infiltration (CVI) is limited to only coating SiC fibers. The preform is then fabricated using a lay-up of basic plys or 2-D woven sheets composed of the precoated fibers. At the other extreme, CVI is used to completely densify a 3-D woven preform shaped almost like the gas turbine component itself. The latter approach is more suitable for highly engineered components which sit directly in the gas path of the engine, for example, a high pressure turbine blade. In this case, the geometry is necessarily complex for aerodynamic, stress, and lifing (multi-physics) requirements. Presently, optimizing the CVI-dominated manufacturing approach is largely by trial-and-error. In this work, a first-principles modeling of CVI is performed to realize optimization of SiC/SiC CMC manufacturing. The modeling is based on a level-set framework to describe the interface between the vapor and solid phases. A finite-difference numerical scheme using an immersed boundary method is developed for fixed, structured meshes. Massively parallel direct numerical simulations (DNS) of CVI through fiber-woven geometries are performed using one-step chemistry, and over a range of Thiele moduli. Illustrative applications of the resulting large DNS data sets are given, including the development of fiber-weave specific infiltration models and structure functions for mean-field (porous media) Computational Fluid Dynamics (CFD) simulations of CVI.

Chemical vapor infiltration of additively manufactured preforms: Pore‐resolved simulations and experimental validation

“…This effect can be even more significant in the case of complex preform shapes such as fiber weaves. 2 Overall, the model seems to capture the surface area to porosity correlation reasonably in packed porous preforms.…”

“…Numerically, a robust model for representing the deposition front needs to capture the sharp multimaterial interface, account for complex interface motion (deposition kinetics), enable capturing topological changes, and be scalable for large-scale simulations. Recent works 2,3,12 have demonstrated the use of interface tracking methods (particularly the level set method) for capturing dynamic front evolution specifically for CVI. Pore-resolved simulations provide a practical means for development of bet-ter informed closure models to be used in coarse/reactor scale analysis.…”

“…Quilt has been used in a prior study to simulate CVI in Ref. 2, where large-scale simulations were performed to understand the effects of coupled transport and kinetics on the microstructural characteristics of multilayered fibrous preform by resolving the smallest features, that is, individual fibers.…”

“…While these have been widely used in CVI modeling, modern optimization needs have motivated detailed analysis and improvements through highfidelity simulations. 2,3 Pore-resolved direct numerical simulations (DNS) do not depend on porous media models but rather resolve the pore scale geometry explicitly and are based on fundamental governing principles at the smallest resolved scales. As a result, uncertainties arising due to the approximate or empirical representation of the pore scale phenomena are avoided.…”

Microstructural and infiltration properties of woven preforms during chemical vapor infiltration

“…In this paper, we report on recent DNSs of CVI over the complete densification cycle, from dry fabric to the final time, when the ultimate porosity is reached 10 . The DNS approach fully resolves the continuum scales, and thus, a porous media model is not required.…”