This conference paper was submitted for presentation at the NAFEMS World Congress 2025, held in Salzburg, Austria from May 19–22, 2025.

Abstract

In today'™s competitive automotive landscape, accelerating time to market and optimizing cost efficiency are critical. BMW Group and Dive CAE have examined how advancements in computational fluid dynamics (CFD) can address these challenges, focusing on GPU acceleration, cloud parallelization and Smoothed Particle Hydrodynamics (SPH). The study examined drivetrain design projects, particularly the development of new differential systems. Modern differential systems are becoming increasingly complex, posing significant challenges for design engineers. Additionally, evolving safety, environmental, and performance standards demand iterative redesigns and extensive testing, lengthening development cycles. Several operating points and designs were compared and assessed with respect to oil churning losses and comprehensive oil coverage of system components. The SPH method is particularly effective for modelling problems of this type, involving free-surface or multi-phase flows. Moreover, unlike grid-based methods, its Lagrangian, particle-based framework naturally handles complex geometries and moving components without requiring re-meshing. Additionally, it reduces manual pre-processing work, paving the way for automation of large parallel simulation studies. Dive CAE employs a Weakly Compressible SPH approach (WCSPH), incorporating a variety of measures relevant to industry-level accuracy and usability. Key methods include a semi-analytical integral boundary condition to improve near-wall flow accuracy. This paper outlines the theoretical foundations of the method and provides selected validation results. GPU acceleration of the SPH code demonstrates a runtime reduction by a factor of 5-18 compared to CPU architectures. Cloud parallelization enabled concurrent testing of 12 operating conditions, shortening project turnaround time (TAT) by a factor of 5 compared to an on-premise setup. Eventually, the paper also includes an analysis of the cost effect of migrating the simulations to GPUs and the cloud. In conclusion, the study examines how GPU acceleration, cloud technologies and SPH contribute to overarching goals of accelerating time to market and reducing costs.