AI Accelerated Engineering - Data-Driven Approaches to Complex Engineering Challenges

Matthias Bauer's presentation, titled "AI Accelerated Engineering: Data-Driven Approaches to Complex Engineering Challenges," showcased the advancements in AI and machine learning within the engineering sector. The presentation focused on NAVASTO's journey in integrating AI into engineering workflows, particularly in the development of their flagship software product, NAVPACK. By utilizing AI-models, NAVPACK enables rapid predictions of simulation results, cutting down solution times from days to mere seconds. This acceleration facilitates real-time collaboration and optimizes design processes, turning dormant data into active organizational knowledge. Real-time results were demonstrated through examples in crash simulation, external aerodynamics, and CFD predictions. A Graph Neural Network (GNN) trained on 350 samples of car shapes was showcased, highlighting the software's predictive capability in aerodynamics. Transfer learning was a key feature of the presentation. Bauer explained how NAVPACK's deep learning models could be fine-tuned with a small number of samples, enhancing their predictive scope without starting the learning process from scratch. Examples of transfer learning included different car models like DrivAer with and without spoilers and the VW Jetta. The presentation also addressed model confidence and uncertainty, crucial factors in predictive modeling. Bauer discussed how NAVASTO ensures prediction confidence and manages uncertainty in model outputs.Finally, Bauer presented several success stories illustrating NAVASTO's impact in the industry. Challenges such as finding optimal positions for active spoilers, designing with limited CFD runs, and assessing various configurations under diverse conditions were addressed using NAVASTO's AI models. These cases demonstrated significant reductions in compute time and lead time for results, underscoring the efficiency of AI-accelerated engineering processes.