Simulating Industrial Electrophoretic Deposition on Distributed Memory Architectures

The application of coatings by employing Electrophoretic Deposition (EPD) is one of the key processes in automotive manufacturing. Here, car assemblies or entire car bodies are dipped into a tank of liquid aimed for preventing the object from future corrosion. However, this process is highly non-trivial. In fact, it has to be ensured that no air bubbles emerge during the dipping which may lead to an incomplete coverage of the coating. Moreover, entrapped liquids that remained after dipping out may lead to corrosion in the consecutive manufacturing process. To detect such problems in an early development stage, simulation methods based on Computational Fluid Dynamics (CFD) are utilized. Additionally, employing a dedicated volumetric decomposition method, this has led to a tool chain ALSIM which allows to simulate the process of EPD with significantly reduced complexity as compared to standard CFD tools. However, despite these benefits, the method still suffers from large execution times. In this work, we are proposing a parallel scheme which allows for an execution on distributed parallel memory architectures. To that end, dedicated workload distribution and memory optimization methods are presented, which eventually allow for an efficient simulation of EPD coatings. Experimental evaluations based on industrial use cases confirm the obtained benefits: While a serial simulation required more than 8 days, the parallel method proposed in this work allows to complete the simulation with 32 processes in less than 15 hours.