UC Riverside

The deleterious impact of erosion due to high-velocity particle impingement adversely affects a variety of engineering/industrial systems, resulting in irreversible mechanical wear of materials/components. Brute force computational fluid dynamics (CFD) calculations are commonly used to predict surface erosion by directly solving the Navier Stokes equations for the fluid and particle dynamics; however, these numerical approaches often require significant computational resources, limiting the scope of observations. Furthermore, when different initial conditions are needed to analyze the system, the whole procedure of CFD calculation should be restarted de novo without recourse to previously converged cases.In contrast, recent data-driven approaches using machine learning (ML) have shown immense promise for more efficient and accurate predictions to sidestep the computationally demanding CFD calculations. This thesis proposes an ML approach using CFD data to predict erosion on a complex boiler header of an industrial coal plant. I developed a hybrid ML approach to predict particle trajectory and surface erosion rate based on initial particle parameters and positions in the OP-650 industrial boiler header. Specifically, I integrated the time-series models, such as LSTM and GPT-2, with the CNN model to predict the surface erosion rate based on the five initial parameters only. The hybrid ML architecture uses the predicted trajectories from the time-series models as input data for a CNN model to predict the surface erosion rate.