A Neural Network Based Framework to Model Particle Rebound and Fracture

Turbomachinery components such as aero engine compressors are subject to performance and lifetime degradation through erosion and fouling, caused by the impact of airborne particles on the blades. The complex nature of the instationary particle-laden flow in partially rotating geometries, which spans over several scales of magnitude in space and time, aggravates the time-accurate prediction of such phenomena. The current state of the art in the simulation of particle-laden flow is a time-resolved particle tracking on a LES field, in which the interaction with walls is handled via rebound and erosion models. Rebound models often require a priori parameter tuning to match experimental measurements. Moreover, the actual stochastic nature of the rebound is neglected and the particle is assumed not to fracture upon impact. However, this affects the resulting particle trajectories and is particularly critical at high (normal) impact velocities, where particles in typical aero engine flow exhibit a high probability of fracture, as illustrated in our previous work. In this work, we propose a method to develop a generalized rebound model which is parameter-free for the user and considers the stochasticity of the rebound. To this end, state of the art methods from function approximation, more precisely, deep dense neural networks are employed. The neural network maps the impacting particles' characteristics to its new particle trajectory after rebound. This rebound model is trained through a supervised learning approach on experimental measurements of the statistical rebound of a subsonic particle-laden jet impinging on a flat, inclined plate. Therefor, we present an efficient method to predict probability distributions in a supervised learning context without a priori parameter tuning of known PDFs. In a second step, we extend the network to account for particle fracture, where the particle breakage is based on a probabilistic approach to determine whether a particle breaks. Experimental measurements of particle fracture under realistic flow conditions in aero engine compressors remain challenging, therefore we limit the particle breakup to a maximum of two parts. The results are compared to existing experimental results and to the rebound trajectories predicted by the model without breakage in order to illustrate the necessity to consider particle breakage. The applicability of the proposed framework is illustrated by a LES of the experimental test rig, and the results are compared with known rebound models