Surrogate Model Development: Creating and refining surrogate models to speed up CFD simulations, leveraging deep learning architectures such as U-Net, RNNs, and Fourier Neural Networks.Innovative Methodologies: Developing new approaches and models for improved fluid simulation solutions, ensuring robust and accurate results across diverse scenarios.Comparative Performance Analysis: Conducting rigorous evaluations of deep learning models against classical fluid solvers to identify strengths and areas for improvement.Optimization and Enhancement: Fine-tuning existing models and algorithms to enhance their performance in surrogate modeling tasks.Collaborative Research: Working closely with interdisciplinary teams at FAU Erlangen Nuremberg, utilizing high-performance computing resources to advance the application of AI in fluid dynamics.

Federated Learning Development: Designed federated learning frameworks for secure, decentralized model training in CFD simulations.Distributed Training: Implemented distributed training techniques to optimize surrogate models and reduce computational overhead.Privacy-Preserving Techniques: Utilized privacy-preserving machine learning methods to safeguard data while developing advanced deep learning models.

Specialized Fluid Dynamics Dataset: Created a comprehensive 3D dataset with diverse geometries, including cubes, cuboids, and cones, tailored for fluid dynamics applications.Extensive Simulation Framework: Conducted 10,000 simulations using WaLBerla, covering Reynolds numbers up to 15,000, to generate training data for surrogate models.Enhanced U-Net Architecture: Improved U-Net models to accurately predict fluid properties in complex flow conditions, integrating advanced features such as additional encoder layers and attention mechanisms.Efficient Data Generation: Achieved a 35-40x increase in data generation efficiency using Python scripts and Fritz computing clusters, optimizing training data preparation.Advanced Training Techniques: Employed multi-GPU training strategies (PyTorch's distributed data parallel, DeepSpeed, FSDP) to manage the computational demands of large models with over 511 million parameters.Improved Predictive Accuracy: Enhanced the predictive performance of the advanced U-Net model, reducing mean absolute error from 0.40 to 0.12, a 70% improvement over the basic model.Optimization of Data Normalization: Demonstrated that standard scaling outperforms min-max normalization, improving the predictive precision of fluid dynamics models.

Lehrstuhl für Technische Thermodynamik [LTT]Innovative Fluid Analysis: Conducted comprehensive studies on novel fluid types for thermal management systems, improving the cooling efficiency of advanced technological applications.Simulation of Heat Transfer Dynamics: Analyzed various materials and channel geometries to optimize heat transfer performance using advanced simulation tools like MATLAB, Particle Image Velocimetry (PIV), and OpenFOAM.Experimental Flow Pattern Investigation: Performed detailed experimental investigations to understand different flow patterns and their impact on heat transfer, employing MATLAB, Python, and PIV for precise measurements and data analysis.Analytical and Numerical Modeling: Developed both analytical and numerical solutions for heat transfer challenges, particularly in complex channel geometries, utilizing cutting-edge computational fluid dynamics software.

Thermal and Structural Analysis: Conducted detailed thermal and structural assessments to enhance the performance and safety of complex engineering systems, focusing on critical components and pressure parts.Fluid Dynamics Assessment: Evaluated fluid behavior and structural integrity under various operating conditions, utilizing advanced simulation tools to ensure reliability and efficiency.Automated Design Solutions: Developed automated design solutions and technical documentation using platforms like .NET (VB.NET, ASP.NET) to streamline engineering workflows and improve process efficiency.Engineering Certification Support: Assisted in engineering certification processes, ensuring compliance with industry standards such as ASME and IBR codes through meticulous analysis and documentation.Software Development for Process Improvement: Created and maintained process improvement software using languages like C++, C#, and VB.NET, driving enhancements in engineering design and analysis workflows.

Boiler Design Support: Assisted in designing boiler pressure parts, ensuring compliance with ASME and IBR codes to meet industry standards and enhance system performance.P&ID Preparation: Prepared detailed P&IDs for various projects, facilitating accurate documentation and design processes.FEA Software Support: Provided coding support for the FEA team, developing new solvers for boilers and heaters using C# and C++, enhancing simulation capabilities.Automated Drawing Solutions: Automated the creation of 2D and 3D boiler pressure part drawings using VB.NET, ASP.NET, and AutoCAD, streamlining the design workflow.Web Application Development: Developed a web application for automatic documentation with ASP.NET and C#, improving efficiency in project documentation and reporting.Internal Software Development: Created internal software for design and thermal calculations using ASP.NET and C++, optimizing engineering processes and analysis.

Industry Standard Methodologies: Worked closely with design and analysis teams to adopt and apply industry-standard problem-solving techniques in engineering projects.Software Development: Developed a software portal for calculating economizer and superheater coil parameters using VB.NET and Excel, streamlining engineering calculations and data management.CFD Techniques for Boiler Analysis: Acquired and applied various industrial CFD, CSD, and CMD techniques for the detailed pressure part analysis of boilers, enhancing accuracy and efficiency in performance evaluations.