Predicting and Validating Spatial Distributions of Particulate Deposition in Gas Turbine Components

Abstract

Abstract Cooling passages and the secondary air system of gas turbine components are prone to blockage from sand, dust and ash. The ability to model deposition accurately using CFD would allow the prediction of in-service performance degradation and the design of deposition-tolerant hardware. Predictions of deposition require accurate modelling of particle transport and wall interactions. Bounce stick models predict whether a particle will bounce, stick, or shatter upon impact and calculate rebound trajectories. Turbulent transport models allow the interaction of particles with near-wall turbulent structures to be predicted in RANS CFD solutions. This paper implements a Continuous Random Walk (CRW) transport model with a Discrete Elements Methods (DEM) based bounce stick model to predict deposition in commercial CFD software. The resulting spatial trends of deposition are compared with experimental data of particle deposition in an S-bend. The case considered is representative of the flow and metal conditions, particle size and loading distribution, and passage geometry seen in high pressure turbine blade cooling passages. Numerical predictions of deposition distribution show good qualitative agreement with experimental data. This validates the combined CRW/DEM approach for modelling the location of particulate deposition in gas turbine components. The approach is therefore suitable to assess and compare the susceptibility of components to particulate damage in industry. To the author’s knowledge, this is the first example in the open literature of experimentally-validated deposition trends in CFD. It was found that a deposit evolution and erosion model is essential to predict deposit height numerically. Particle tracking simulations without turbophoresis modelling showed significant differences in particle impact velocities and locations, justifying the use of the CRW model. The inclusion of the Magnus force produced considerable differences in quantitative and qualitative measures of deposition throughout the domain, demonstrating the importance of modelling collision-induced rotation in particle tracking. The performance of the DEM bounce stick model was favourable compared to energy-based alternatives which showed reduced accuracy in inertia-dominated regimes where multiple bounces occurred per particle. On average, eleven impacts were observed for each injected particle, reinforcing the dual importance of accurate bounce stick and turbulent transport models.