FWFabian Waschkowski

Fabian Waschkowski

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Fabian Waschkowski has a total of 17 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    A Surrogate-Augmented Symbolic CFD-Driven Training Framework for Accelerating Multi-objective Physical Model Development

    by Yuan Fang, Fabian Waschkowski, Maximilian Reissmann, Richard Sandberg, Takuo Oda, Koichi Tanimoto

    Computational Fluid Dynamics (CFD)-driven training, which combines machine learning (ML) techniques with fluid dynamics software that provides CFD feedback, has shown great potential for developing implementable and physically consistent closure models with improved accuracy compared to conventional baseline models. In the original CFD-driven training framework, each ML-generated candidate model is embedded in a CFD solver, and its performance is evaluated by comparing computed quantities of interest against reference data. However, exploring optimal models requires hundreds to thousands of CFD evaluations, resulting in prohibitive training costs for complex industrial flows, where a single CFD calculation can already be computationally expensive. To address this limitation, we propose an extended framework that, for the first time, integrates surrogate modeling into symbolic CFD-driven training in real time, aiming to reduce the overall training cost. The surrogate model learns to approximate the errors of the ML-generated models based on previous CFD evaluations, and this surrogate mapping is continuously refined throughout the training process. Newly generated models are first assessed by the surrogate. Only those predicted to yield small errors or high uncertainty are subsequently re-evaluated with full CFD computations. Specifically, the generated discrete expressions from symbolic regression are mapped into a continuous space using the averaged values of the input symbols as inputs to the probabilistic surrogate model. To accommodate multi-objective model training, particularly in situations where assigning fixed weights to competing quantities is challenging, the surrogate model is further generalized to multi-output settings. This is achieved by extending the kernel function to a matrix form, enabling one mean and variance prediction per training objective. Based on theses probabilistic outputs, various selection metrics and thresholds are evaluated to identify the optimal setup. The proposed surrogate-augmented CFD-driven training framework is developed and demonstrated across a range of cases, encompassing statistically one- and two-dimensional flows, as well as single-expression (targeting only the turbulence model) and multi-expression (simultaneous optimization of both turbulence and heat-flux models). Across all cases, the framework achieves a substantial reduction in training cost while the resulting models maintain comparable predictive accuracy to those obtained with the original CFD-driven approach.

  • Open Access

    Gradient Information and Regularization for Gene Expression Programming to Develop Data-Driven Physics Closure Models

    by Fabian Waschkowski, Haochen Li, Abhishek Deshmukh, Temistocle Grenga, Yaomin Zhao, Heinz Pitsch, Joseph Klewicki, Richard D. Sandberg

    Learning accurate numerical constants when developing algebraic models is a known challenge for evolutionary algorithms, such as Gene Expression Programming (GEP). This paper introduces the concept of adaptive symbols to the GEP framework by Weatheritt and Sandberg (J Comput Phys 325:22–37, 2016a) to develop advanced physics closure models. Adaptive symbols utilize gradient information to learn locally optimal numerical constants during model training, for which we investigate two types of nonlinear optimization algorithms. The second contribution of this work is implementing two regularization techniques to incentivize the development of implementable and interpretable closure models. We apply $$L_2$$ L 2 regularization to ensure small magnitude numerical constants and devise a novel complexity metric that supports the development of low complexity models via custom symbol complexities and multi-objective optimization. This extended framework is employed to four use cases, namely rediscovering Sutherland’s viscosity law, developing laminar flame speed combustion models and training two types of fluid dynamics turbulence models. The model prediction accuracy and the convergence speed of training are improved significantly across all of the more and less complex use cases, respectively. The two regularization methods are essential for developing implementable closure models and we demonstrate that the developed turbulence models substantially improve simulations over state-of-the-art models.


  • Evolutionary neural networks for learning turbulence closure models with explicit expressions

    by Haochen Li, Yaomin Zhao, Fabian Waschkowski, Richard D. Sandberg

    Developing physical closure models with explicit expressions based on a given dataset is essential to science and engineering. For such symbolic regression tasks, biology-inspired evolutionary algorithms are most widely used. However, typical evolutionary algorithms do not utilize any structural information inherent in training data, which limits their performance in finding accurate model structures and coefficients. By combining one evolutionary algorithm, gene expression programing (GEP), with an artificial neural network (ANN) for symbolic regression, we propose a novel evolutionary neural network method, in which candidate expressions are specifically designed so that they can be transformed between the GEP and ANN structures during training iterations. By combining the GEP's global searching and the ANN's gradient optimization capabilities, efficient and robust convergence to accurate models can be achieved. In addition, sparsity-enhancing strategies have been introduced to improve the interpretability of the trained models. The present method has been tested for finding different physical laws and then applied to turbulence modeling problems with different configurations, showing advantages compared to the existing GEP and ANN methods.


  • Open Access

    Gradient Information and Regularization for Gene Expression Programming to Develop Data-Driven Physics Closure Models

    by Fabian Waschkowski, Haochen Li, Abhishek Deshmukh, Temistocle Grenga, Yaomin Zhao, Heinz Pitsch, Joseph Klewicki, Richard D. Sandberg

    Learning accurate numerical constants when developing algebraic models is a known challenge for evolutionary algorithms, such as Gene Expression Programming (GEP). This paper introduces the concept of adaptive symbols to the GEP framework by Weatheritt and Sandberg (2016) to develop advanced physics closure models. Adaptive symbols utilize gradient information to learn locally optimal numerical constants during model training, for which we investigate two types of nonlinear optimization algorithms. The second contribution of this work is implementing two regularization techniques to incentivize the development of implementable and interpretable closure models. We apply $L_2$ regularization to ensure small magnitude numerical constants and devise a novel complexity metric that supports the development of low complexity models via custom symbol complexities and multi-objective optimization. This extended framework is employed to four use cases, namely rediscovering Sutherland's viscosity law, developing laminar flame speed combustion models and training two types of fluid dynamics turbulence models. The model prediction accuracy and the convergence speed of training are improved significantly across all of the more and less complex use cases, respectively. The two regularization methods are essential for developing implementable closure models and we demonstrate that the developed turbulence models substantially improve simulations over state-of-the-art models.

  • Open Access

    Multi-Objective CFD-Driven Development of Coupled Turbulence Closure Models

    by Fabian Waschkowski, Yaomin Zhao, Richard Sandberg, Joseph Klewicki

    This paper introduces two novel concepts in data-driven turbulence modeling that enable the simultaneous development of multiple closure models and the training towards multiple objectives. The concepts extend the evolutionary framework by Weatheritt and Sandberg (2016), which derives interpretable and implementation-ready expressions from high-fidelity simulation data. By assigning a shared fitness value to the evolved closure models and utilizing the CFD-driven training approach by Zhao et al. (2020), the multi-expression training concept introduced here is able to account for the coupling between the trained models, i.e. Reynolds stress anisotropy, turbulent heat flux and turbulence production correction models. As a second concept, a multi-objective optimization algorithm is applied to the framework. The extension yields a diverse set of candidate models and allows a trade-off between the training objectives after analyzing the training results. In this study, the novel concepts are applied to a benchmark periodic hills case and a vertical natural convection flow. The predictions of mean flow quantities are improved compared to decoupled training strategies with distinct and robust improvements for strongly coupled momentum and thermal fields. The coupled training of closure models and the balancing of multiple training objectives are considered important capabilities on the path towards generalized data-driven turbulence models.

  • Open Access

    Transition Modeling for Low Pressure Turbines Using Computational Fluid Dynamics Driven Machine Learning

    by Harshal D. Akolekar, Fabian Waschkowski, Yaomin Zhao, Roberto Pacciani, Richard D. Sandberg

    Existing Reynolds Averaged Navier–Stokes-based transition models do not accurately predict separation induced transition for low pressure turbines. Therefore, in this paper, a novel framework based on computational fluids dynamics (CFD) driven machine learning coupled with multi-expression and multi-objective optimization is explored to develop models which can improve the transition prediction for the T106A low pressure turbine at an isentropic exit Reynolds number of Re2is=100,000. Model formulations are proposed for the transfer and laminar eddy viscosity terms of the laminar kinetic energy transition model using seven non-dimensional pi groups. The multi-objective optimization approach makes use of cost functions based on the suction-side wall-shear stress and the pressure coefficient. A family of solutions is thus developed, whose performance is assessed using Pareto analysis and in terms of physical characteristics of separated-flow transition. Two models are found which bring the wall-shear stress profile in the separated region at least two times closer to the reference high-fidelity data than the baseline transition model. As these models are able to accurately predict the flow coming off the blade trailing edge, they are also able to significantly enhance the wake-mixing prediction over the baseline model. This is the first known study which makes use of ‘CFD-driven’ machine learning to enhance the transition prediction for a non-canonical flow.