Integration of machine learning methods into numerical simulation
Integration of machine learning methods into numerical simulation
The combination of common physics-based methods with data-driven models to so-called hybrid approaches is a current
topic in the field of numerical simulation. Since hybrid models show great potential to increase prediction accuracy
compared to standard data-driven methods and/or accelerate physics-based numerical simulation, the idea of this approach
is to combine the benefits of both worlds.
Target of this master’s thesis is to elaborate a novel method which directly integrates machine learning methods into the numerical scheme used for solving partial differential equations. Therefore, an existing baseline concept shall be further developed and its application to more complex physical problems shall be investigated.
topic in the field of numerical simulation. Since hybrid models show great potential to increase prediction accuracy
compared to standard data-driven methods and/or accelerate physics-based numerical simulation, the idea of this approach
is to combine the benefits of both worlds.
Target of this master’s thesis is to elaborate a novel method which directly integrates machine learning methods into the numerical scheme used for solving partial differential equations. Therefore, an existing baseline concept shall be further developed and its application to more complex physical problems shall be investigated.