Jan Backhaus | German Aerospace Center (DLR) (original) (raw)
Papers by Jan Backhaus
18th AIAA/ISSMO Multidisciplinary Analysis and Optimization Conference
18th AIAA/ISSMO Multidisciplinary Analysis and Optimization Conference
The adjoint method has already proven its potential to reduce the computational effort for optimi... more The adjoint method has already proven its potential to reduce the computational effort for optimizations of turbomachinery components based on flow simulations. However, the transfer of the adjoint-based optimization methods to industrial design problems turns out to pose specific requirements to both the adjoint solver as well as the optimization algorithms which utilize the gradient information. While the construction of the adjoint solver through algorithmic differentiation is described in a parallel publication, we focus here on the robust application of the gradient information in a high-dimensional multi-objective optimization with several constraints including non-differentiated mechanical constraints. We describe the optimization methods, which comprise the use of gradient-enhanced Kriging meta-models, and subsequently apply these to the design optimization of a contra-rotating fan stage. The results show that through the described combination of methods the adjoint method can be used in practical design optimizations of turbomachinery components.
2018 Multidisciplinary Analysis and Optimization Conference
Journal of Turbomachinery
This paper describes the development and initial application of an adjoint harmonic balance (HB) ... more This paper describes the development and initial application of an adjoint harmonic balance (HB) solver. The HB method is a numerical method formulated in the frequency domain which is particularly suitable for the simulation of periodic unsteady flow phenomena in turbomachinery. Successful applications of this method include unsteady aerodynamics as well as aeroacoustics and aeroelasticity. Here, we focus on forced response due to the interaction of neighboring blade rows. In the simulation-based design and optimization of turbomachinery components, it is often helpful to be able to compute not only the objective values—e.g., performance data of a component—themselves but also their sensitivities with respect to variations of the geometry. An efficient way to compute such sensitivities for a large number of geometric changes is the application of the adjoint method. While this is frequently used in the context of steady computational fluid dynamics (CFD), it becomes prohibitively e...
In this paper we discuss the usage of finite differences for the computation of the flux Jacobian... more In this paper we discuss the usage of finite differences for the computation of the flux Jacobian in the framework of a discrete adjoint or time-linearised flow solver, in particular the associated choice of an appropriate step size. For comparison, we apply algorithmic differentiation to obtain an exact flux Jacobian. It turns out that the results depend strongly on the choice of the slope limiter. A careful choice of this function is crucial for computations with exact flux linearisations as well as for finite difference approximations.
An adjoint preprocess for an adjoint-based turbomachinery design process is described and applied... more An adjoint preprocess for an adjoint-based turbomachinery design process is described and applied. Shape sensitivities are calculated using an adjoint elliptic mesh deformation tool. Calculation of sensitivities for the CAD parameters is performed using these shape sensitivities and deformed surface grids. The method is applied to the shape optimization process for a counter-rotating fan. Integrating the adjoint preprocess into the gradient evaluation results in a significantly reduced dependency of the computation time on the number of design parameters.
Volume 8: Turbomachinery, Parts A, B, and C, 2012
ABSTRACT This paper studies the use of adjoint CFD solvers in combination with surrogate modellin... more ABSTRACT This paper studies the use of adjoint CFD solvers in combination with surrogate modelling in order to reduce the computational cost of the optimization of complex 3D turbomachinery components. The method is applied to a previously optimized counter rotating turbofan, with a shape parameterized by 104 CAD parameters. Through random changes on the reference design, a small number of design variations are created to serve as training samples for the surrogate models. A steady RANS solver and its discrete adjoint are then used to calculate objective function values and their corresponding sensitivities. Kriging and neural networks are used to build surrogate models from the training data. To study the impact of the additional information provided by the adjoint solver, each model is trained with and without the sensitivity information. The accuracy of the different surrogate model predictions is assessed by comparison against CFD calculations. The results show a considerable improvement of the fitness function approximation when the sensitivity information is taken into account. Through a gradient based optimization on one of the surrogate models, a design with higher isentropic efficiency at the aerodynamic design point is created. This application demonstrates that the improved surrogate models can be used for design and optimization.
Based on the results of a prior study about fan blade degradation, which state a noticeable influ... more Based on the results of a prior study about fan blade degradation, which state a noticeable influence of small geometric changes on the fan performance, an adjoint computational fluid dynamics method is applied to systematically analyze the sensi- tivities of fan blade performance to changes of the leading edge geometry. As early as during manufacture, blade geometries vary due to fabrication tolerances. Later, when in service, engine operation results in blade degradation which can be reduced but not perfectly fixed by maintenance, repair and overhaul processes. The geometric irregularities involve that it is difficult to predict the blade’s aerodynamic performance. Therefore, the aim of this study is to present a systematic approach for analyzing geometric sensitivities for a fan blade. To demonstrate the potential, two-dimensional optimizations of three airfoil sections at different heights of a transonic fan blade are presented. Although the optimization procedure is limited to ...
Lecture Notes in Computer Science, 2013
Competence in High Performance Computing 2010, 2011
The objective of the German BMBF research project Highly Efficient Implementation of CFD Codes fo... more The objective of the German BMBF research project Highly Efficient Implementation of CFD Codes for HPC Many-Core Architectures (HICFD) is to develop new methods and tools for the analysis and optimization of the performance of parallel computational fluid dynamics (CFD) codes on high performance computer systems with many-core processors. In the work packages of the project it is investigated how the performance of parallel CFD codes written in C can be increased by the optimal use of all parallelism levels. On the highest level MPI is utilized. Furthermore, on the level of the many-core architecture, highly scaling, hybrid OpenMP/MPI methods are implemented. On the level of the processor cores the parallel SIMD units provided by modern CPUs are exploited.
Volume 7: Turbomachinery, Parts A, B, and C, 2011
ABSTRACT This article describes how to extend the dsicrete adjoint method to functionals that are... more ABSTRACT This article describes how to extend the dsicrete adjoint method to functionals that are evaluated on arbitrary rotational control surfaces that intersect the flow domain at a position specified by the user, e.g. the pressure loss coefficient of a single blade in a multi-stage configuration. The definition and implementation of the mixed-out states on such surfaces is revisited. The calculation of the corresponding right-hand sides in the adjoint system is explained. These techniques can be used to specify functionals that quantify the deviation of the radial distribution of the flow angles, relative mass flow, etc. from a given target distribution. Sensitivity studies using the conventional approach, i.e. by means of finite differences of many steady solutions, are compared to results based on the adjoint method. The applications demonstrate that the agreement between adjoint and conventional sensitivity predictions is excellent, if the exact definition of the surface functionals is taken into account.
18th AIAA/ISSMO Multidisciplinary Analysis and Optimization Conference
18th AIAA/ISSMO Multidisciplinary Analysis and Optimization Conference
The adjoint method has already proven its potential to reduce the computational effort for optimi... more The adjoint method has already proven its potential to reduce the computational effort for optimizations of turbomachinery components based on flow simulations. However, the transfer of the adjoint-based optimization methods to industrial design problems turns out to pose specific requirements to both the adjoint solver as well as the optimization algorithms which utilize the gradient information. While the construction of the adjoint solver through algorithmic differentiation is described in a parallel publication, we focus here on the robust application of the gradient information in a high-dimensional multi-objective optimization with several constraints including non-differentiated mechanical constraints. We describe the optimization methods, which comprise the use of gradient-enhanced Kriging meta-models, and subsequently apply these to the design optimization of a contra-rotating fan stage. The results show that through the described combination of methods the adjoint method can be used in practical design optimizations of turbomachinery components.
2018 Multidisciplinary Analysis and Optimization Conference
Journal of Turbomachinery
This paper describes the development and initial application of an adjoint harmonic balance (HB) ... more This paper describes the development and initial application of an adjoint harmonic balance (HB) solver. The HB method is a numerical method formulated in the frequency domain which is particularly suitable for the simulation of periodic unsteady flow phenomena in turbomachinery. Successful applications of this method include unsteady aerodynamics as well as aeroacoustics and aeroelasticity. Here, we focus on forced response due to the interaction of neighboring blade rows. In the simulation-based design and optimization of turbomachinery components, it is often helpful to be able to compute not only the objective values—e.g., performance data of a component—themselves but also their sensitivities with respect to variations of the geometry. An efficient way to compute such sensitivities for a large number of geometric changes is the application of the adjoint method. While this is frequently used in the context of steady computational fluid dynamics (CFD), it becomes prohibitively e...
In this paper we discuss the usage of finite differences for the computation of the flux Jacobian... more In this paper we discuss the usage of finite differences for the computation of the flux Jacobian in the framework of a discrete adjoint or time-linearised flow solver, in particular the associated choice of an appropriate step size. For comparison, we apply algorithmic differentiation to obtain an exact flux Jacobian. It turns out that the results depend strongly on the choice of the slope limiter. A careful choice of this function is crucial for computations with exact flux linearisations as well as for finite difference approximations.
An adjoint preprocess for an adjoint-based turbomachinery design process is described and applied... more An adjoint preprocess for an adjoint-based turbomachinery design process is described and applied. Shape sensitivities are calculated using an adjoint elliptic mesh deformation tool. Calculation of sensitivities for the CAD parameters is performed using these shape sensitivities and deformed surface grids. The method is applied to the shape optimization process for a counter-rotating fan. Integrating the adjoint preprocess into the gradient evaluation results in a significantly reduced dependency of the computation time on the number of design parameters.
Volume 8: Turbomachinery, Parts A, B, and C, 2012
ABSTRACT This paper studies the use of adjoint CFD solvers in combination with surrogate modellin... more ABSTRACT This paper studies the use of adjoint CFD solvers in combination with surrogate modelling in order to reduce the computational cost of the optimization of complex 3D turbomachinery components. The method is applied to a previously optimized counter rotating turbofan, with a shape parameterized by 104 CAD parameters. Through random changes on the reference design, a small number of design variations are created to serve as training samples for the surrogate models. A steady RANS solver and its discrete adjoint are then used to calculate objective function values and their corresponding sensitivities. Kriging and neural networks are used to build surrogate models from the training data. To study the impact of the additional information provided by the adjoint solver, each model is trained with and without the sensitivity information. The accuracy of the different surrogate model predictions is assessed by comparison against CFD calculations. The results show a considerable improvement of the fitness function approximation when the sensitivity information is taken into account. Through a gradient based optimization on one of the surrogate models, a design with higher isentropic efficiency at the aerodynamic design point is created. This application demonstrates that the improved surrogate models can be used for design and optimization.
Based on the results of a prior study about fan blade degradation, which state a noticeable influ... more Based on the results of a prior study about fan blade degradation, which state a noticeable influence of small geometric changes on the fan performance, an adjoint computational fluid dynamics method is applied to systematically analyze the sensi- tivities of fan blade performance to changes of the leading edge geometry. As early as during manufacture, blade geometries vary due to fabrication tolerances. Later, when in service, engine operation results in blade degradation which can be reduced but not perfectly fixed by maintenance, repair and overhaul processes. The geometric irregularities involve that it is difficult to predict the blade’s aerodynamic performance. Therefore, the aim of this study is to present a systematic approach for analyzing geometric sensitivities for a fan blade. To demonstrate the potential, two-dimensional optimizations of three airfoil sections at different heights of a transonic fan blade are presented. Although the optimization procedure is limited to ...
Lecture Notes in Computer Science, 2013
Competence in High Performance Computing 2010, 2011
The objective of the German BMBF research project Highly Efficient Implementation of CFD Codes fo... more The objective of the German BMBF research project Highly Efficient Implementation of CFD Codes for HPC Many-Core Architectures (HICFD) is to develop new methods and tools for the analysis and optimization of the performance of parallel computational fluid dynamics (CFD) codes on high performance computer systems with many-core processors. In the work packages of the project it is investigated how the performance of parallel CFD codes written in C can be increased by the optimal use of all parallelism levels. On the highest level MPI is utilized. Furthermore, on the level of the many-core architecture, highly scaling, hybrid OpenMP/MPI methods are implemented. On the level of the processor cores the parallel SIMD units provided by modern CPUs are exploited.
Volume 7: Turbomachinery, Parts A, B, and C, 2011
ABSTRACT This article describes how to extend the dsicrete adjoint method to functionals that are... more ABSTRACT This article describes how to extend the dsicrete adjoint method to functionals that are evaluated on arbitrary rotational control surfaces that intersect the flow domain at a position specified by the user, e.g. the pressure loss coefficient of a single blade in a multi-stage configuration. The definition and implementation of the mixed-out states on such surfaces is revisited. The calculation of the corresponding right-hand sides in the adjoint system is explained. These techniques can be used to specify functionals that quantify the deviation of the radial distribution of the flow angles, relative mass flow, etc. from a given target distribution. Sensitivity studies using the conventional approach, i.e. by means of finite differences of many steady solutions, are compared to results based on the adjoint method. The applications demonstrate that the agreement between adjoint and conventional sensitivity predictions is excellent, if the exact definition of the surface functionals is taken into account.