|
Current work at Schaeffler Technologies AG & Co.KG
Simulations of Flow Systems
- Simulations of control valves, water pumps, injectors, and actuators, for example in the area of supersonic flows in H2 combustion engines are performed. Here, flow simulations (CFD) or thermal simulations (CHT) can provide useful characterizations, which are in turn necessary for 0D/1D system simulations to describe complex vehicle systems.
- Future-oriented H2 developments in example electrolyzers, (convert hydrogen into electricity via electrochemical reactions): the processes within a single cell can often be characterized by system of equations. However, capturing the complexity of an entire stack (scaling up to MW performance through many cells) requires extensive flow simulations.
- For enthalpy-based heat exchangers, complex thermal flow simulations are necessary. Simplified models pose a great challenge in this case, as they depend on large simulation matrices or extensive empirical measurements.
Simulations of Dynamic Multi-Body Systems
-
In the simulation of mechanical, hydraulic, and mechatronic components of internal combustion and hybrid engines, the focus is on the mathematical modeling and analysis of a system’s inherent vibrations. This includes operating vibration analyses, characterization of resonances and consideration of nonlinear effects.
- I am particularly interested in the mathematical description of complex physical processes using systems of partial differential equations, for example, for modeling natural frequencies.
- Modal analysis can be used to study complex systems by measurement and simulation, for instance to characterize noise issues or operational strength under dynamic loads (rainflow counting of stress amplitudes).
- The resonances occurring in the vehicle depend in particular on the modeling of contact and friction between bodies, the influence of elastically modeled components, as well as the hydraulic and mechatronic properties of the systems. In particular, the nonlinear behavior leads to interesting vibration phenomena.
- Validation of the models requires stability analysis of nonlinear systems, and after numerical implementation, the convergence of the integrator method plays an important role. Finally, key parameter sensitivities or optimizations are conducted.
Simulations of Complex Systems
-
In-house developed algorithms and so-called user routines written in C++ or Python are integrated into commercial simulation environments. To ensure performance and traceability, object-oriented programming and version control (e.g., Git) are used.
- Furthermore, simplified models can be used to test a wide range of modeling approaches for robustness or to apply machine learning to them.
- By using modern AI frameworks such as TensorFlow, Keras, NumPy, or scikit-learn, simulation data can be specifically leveraged for training networks. In this way, new relationships can be uncovered and classical optimization methods as well as design-of-experiments (DoE) analyses can be extended. In this field I would like to study in future more actively.
A development process of a simulation model subdivide into several tasks:
a) mathemaical-physcial description of the system, integration into an existing system model.
b) The validation / functionality test of the component model with respect to expected physical effects i.e. by analytical estimations. Numerical stability is checked.
c) Verification is performed by using experimental measurements on test benches and predictability. For this purpose, measurements from various test benches are used, for example to determine hydraulic flow rates of valves or engine test benches to test the individual components and overall behavior of hydraulic systems. It is essential to critically evaluate the measurement data and take test bench influences into account. When transferring results to real systems (engines), interactions and questions of transferability must be clarified — this also applies to AI models.
Hobbies:
Interests:
Is there a connection of quantum gravity, Schrödinger's cat and quantum concepts of machine learning approaches and AI-models?
In the quantum theory of gravity, spacetime becomes a dynamic object subject to the uncertainty relations of quantum mechanics. Comparable to the canonical variables of position and momentum in quantum mechanics, there are corresponding quantities that describe spacetime. These special variables include the parallel transport along closed loops and the fluxes, which are derived objects of the curvature of the spacetime manifold. In the spin network approach, an extension of parallel transport from curves and fluxes to surfaces is preferred. Furthermore, there is an analog to energy quantization: the discretization of area and volume within the quantum gravity framework.
The Schrödingers cat paradox describes the state of a quantum system to be non-classical in the sense that it is a probability of the cat being dead or alive. Only due to the opening of the box (the measurement) the cat (the quantum object) obtains its status (an eigen state of the object). The probability in this experiment can be interpreted as reasonable expectation of the cat representing a state of knowledge (the cat is either alive or dead). Such probabilities can be interpreted in a Bayesian sense, because the observer’s information state changes with the measurement. Looking ahead, I believe this concept is highly suitable for machine learning in the field of simulations and measurements of complex systems. The idea is to make a broadened use of physical knowledge of the system. The advantage of such an approach might be that a smaller amount of data is necessary.
Another aspect of quantization is the non-uniform distribution of the electromagnetic energy in blackbody radiation, depending on the frequency of the radiation. Characteristically, only integer multiples of the energy quantum are allowed. This idea can also be transferred to neural networks. In the case of an optimization of a large number of weigths in different discrete layers unexpected numerical errors might occur. The weigths can be mapped to integer values by using scaling factors. Then the neural network is supposed to be robust under small changes of the weigths ("noise") such that only the essential pattern is recovered by the trained network. This approach might be very interesting whenever many layers are needed.
Ph.D. studies in natural science with focus
on general relativity (GR), quantum mechanics, algebraic quantum field theory and quantum gravity:
One of the most interesting research areas in physics is to find an unifying theory of general relativity and quantum physics. One approach is the so-called Loop Quantum Gravity (LQG),which is based on the quantization of geometric objects. Inspired by the duality of Schrödinger's wave function and the Heisenberg matrix representation formulation, the intenion of the thesis is to derive the algebraic operator formulation of loop quantum gravity approach. It is investigated whether further states (such as thermal KMS states) exist in addition to the well-known Ashtekar-Lewandowski state. Such thermal KMS-states appear naturally in algebraic quantum field theory approach.
Ph.D. studies
Diploma thesis 2006:
Teaching experiences:
International conferences:
Interesting lectures:
Important scientific communities:
|
