Submitted Abstract
Industrial-scale Chemical Vapor Deposition Processes are tackled with a hybrid Workflow that combines Equation-based modelling with Computational Fluid Dynamics software and machine learning algorithms. Data from the Process, i.e. from the reactor sensors are correlated to ex-situ product characterization measurements and simulation results, giving unique predictive capabilities, currently out-of-reach. A simplified CFD process model, with relatively low computational requirements yet limited predictive quality is combined with machine learning algorithms for data compression and regression modelling. The integrated computational Workflow will be able to correlate and data from various sensors and measurements, in-situ and ex-situ and deliver accurate predictions in terms of product quality or the need for equipment maintenance. New product development will be accelerated as the predictive tool limits the number of expensive trials and experiments.