PhD - Physics informed AI for simulation credibility in Renningen

Simulation plays a crucial role in the development and operational stage of many BOSCH products. However, limited computational resources, uncertain parametrization, and structural model form uncertainty usually restrict the predictive accuracy and reliability of simulation models of physical systems. In order to support decision-making & simulation credibility, such uncertainty sources must be characterized and quantified appropriately. Measurement data from test benches or field data may serve as an additional source of information (e.g. ANN/Deep NN) to infer more knowledge and therefore reduce overall uncertainty/increase overall credibitity.

However, in many application domains it is yet unclear how to combine both the simulation model and available data properly under sometimes very restrictive computational resources. Therefore hybrid approaches from physics-informed AI and AI-informed physics are under consideration.

• Help shape the future: The following research directions will play a fundamental role during  your PhD: Research in methods from applied mathematics in the area of uncertainty quantification (probabilistic programming, multi-fidelity, surrogate modelling, design of experiments, Bayesian inference, p-Boxes, ...), research in artificial intelligence (AI) with main emphasis on physical informed training. In addition, strategies for validation & verification and applications to different use cases: from electrical drives to autonomous driving.

• Education: Excellent university degree (Master) in mathematics, engineering, physics or similar
• Personality and Working Practice: Team-oriented and willing to work in an interdisciplinary research area
• Experience and Knowledge: Proven programming skills, in particular Python
• Languages: Fluent in English (written and spoken)