Adjoint Method Research Articles

Robust Design Optimization of Electric Machines with Isogeometric Analysis

In electric machine design, efficient methods for the optimization of the geometry and associated parameters are essential. Nowadays, it is necessary to address the uncertainty caused by manufacturing or material tolerances. This work presents a robust optimization strategy to address uncertainty in the design of a three-phase, six-pole permanent magnet synchronous motor (PMSM). The geometry is constructed in a two-dimensional framework within MATLAB®, employing isogeometric analysis (IGA) to enable flexible shape optimization. The main contributions of this research are twofold. First, we integrate shape optimization with parameter optimization to enhance the performance of PMSM designs. Second, we use robust optimization, which creates a min–max problem, to ensure that the motor maintains its performance when facing uncertainties. To solve this bilevel problem, we work with the maximal value functions of the lower-level maximization problems and apply a version of Danskin’s theorem for the computation of generalized derivatives. Additionally, the adjoint method is employed to efficiently solve the lower-level problems with gradient-based optimization. The paper concludes by presenting numerical results showcasing the efficacy of the proposed robust optimization framework. The results indicate that the optimized PMSM designs not only perform competitively compared to their non-robust counterparts but also show resilience to operational and manufacturing uncertainties, making them attractive for industrial applications.

Interactive Airfoil Optimization Using Parsec Parametrization and Adjoint Method

In the development of interactive aerodynamic optimization tools, the need to reduce the computational complexity of flow calculations has arisen. Computational complexity can be reduced by estimating the flow variables using machine learning, but that approach has a number of hindrances. Avoiding these hindrances through lowering the computational complexity by stating the assumptions of inviscid incompressible potential flow is the focus of this article. The assumptions used restrict the applicability of this approach to only specific cases, but in engineering practice, these cases are quite widespread. The assumptions allowed the coupling of the adjoint method with parsec parametrization and the panel method, yielding a highly computationally efficient and robust tool for optimizing an airfoil’s lift coefficient (Cy). The optimization of the NREL S809 airfoil was carried out, and the results were verified using the Xfoil 6.99 software. The Xfoil verification showed that by making minimal changes to the airfoil’s shape, the Cy and lift-to-drag ratios were significantly improved. The improvement magnitude was over 94% for a 0 deg angle of attack (AoA) and over 16% for 6.2 deg AoA. This indicates an improvement in performance that is similar to that of some genetic algorithms, but with computational costs that are many orders of magnitude lower.

Automatic design of interpretable control laws through parametrized Genetic Programming with adjoint state method gradient evaluation

This work investigates the application of a Local Search (LS) enhanced Genetic Programming (GP) algorithm to the control scheme’s design task. The combination of LS and GP aims to produce an interpretable control law as similar as possible to the optimal control scheme reference. Inclusive Genetic Programming (IGP), a GP heuristic capable of promoting and maintaining the population diversity, is chosen as the GP algorithm since it proved successful on the considered task. IGP is enhanced with the Operators Gradient Descent (OPGD) approach, which consists of embedding learnable parameters into the GP individuals. These parameters are optimized during and after the evolutionary process. Moreover, the OPGD approach is combined with the adjoint state method to evaluate the gradient of the objective function. The original OPGD was formulated by relying on the backpropagation technique for the gradient’s evaluation, which is impractical in an optimization problem involving a dynamical system because of scalability and numerical errors. On the other hand, the adjoint method allows for overcoming this issue. Two experiments are formulated to test the proposed approach, named Operator Gradient Descent - Inclusive Genetic Programming (OPGD-IGP): the design of a Proportional–Derivative (PD) control law for a harmonic oscillator and the design of a Linear Quadratic Regulator (LQR) control law for an inverted pendulum on a cart. OPGD-IGP proved successful in both experiments, being capable of autonomously designing an interpretable control law similar to the optimal ones, both in terms of shape and control gains.

Adjoint sensitivity kernels for free oscillation spectra

SUMMARY We apply the adjoint method to efficiently calculate sensitivity kernels for long-period seismic spectra with respect to structural and source parameters. Our approach is built around the solution of the frequency-domain equations of motion using the direct solution method (DSM). The DSM is currently applied within large-scale mode coupling calculations and is also likely to be useful within finite-element type methods for modelling seismic spectra that are being actively developed. Using mode coupling theory as a framework for solving both the forward and adjoint equations, we present numerical examples that focus on the spectrum close to four eigenfrequencies (the low-frequency mode, 0S2, and higher frequency modes, namely 2S2, 0S7 and 0S10 for comparison). For each chosen observable, we plot sensitivity kernels with respect to 3-D perturbations in density and seismic wave speeds. We also use the adjoint method to calculate derivatives of observables with respect to the matrices occurring within mode coupling calculations. This latter approach points towards a generalization of the two-stage splitting function method for structural inversions that does not rely on inaccurate self-coupling or group-coupling approximations. Finally, we verify through direct calculation that our sensitivity kernels correctly predict the linear dependence of the chosen observables on model perturbations. In doing this, we highlight the importance of non-linearity within inversions of long-period spectra.

The Cut-Cell Method for the Conjugate Heat Transfer Topology Optimization of Turbulent Flows Using the “Think Discrete–Do Continuous” Adjoint

This paper presents a topology optimization (TopO) method for conjugate heat transfer (CHT), with turbulent flows. Topological changes are controlled by an artificial material distribution field (design variables), defined at the cells of a background grid and used to distinguish a fluid from a solid material. To effectively solve the CHT problem, it is crucial to impose exact boundary conditions at the computed fluid–solid interface (FSI); this is the purpose of introducing the cut-cell method. On the grid, including also cut cells, the incompressible Navier–Stokes equations, coupled with the Spalart–Allmaras turbulence model with wall functions, and the temperature equation are solved. The continuous adjoint method computes the derivatives of the objective function(s) and constraints with respect to the material distribution field, starting from the computation of derivatives with respect to the positions of nodes on the FSI and then applying the chain rule of differentiation. In this work, the continuous adjoint PDEs are discretized using schemes that are consistent with the primal discretization, and this will be referred to as the “Think Discrete–Do Continuous” (TDDC) adjoint. The accuracy of the gradient computed by the TDDC adjoint is verified and the proposed method is assessed in the optimization of two 2D cases, both in turbulent flow conditions. The performance of the TopO designs is investigated in terms of the number of required refinement steps per optimization cycle, the Reynolds number of the flow, and the maximum allowed power dissipation. To illustrate the benefits of the proposed method, the first case is also optimized using a density-based TopO that imposes Brinkman penalization terms in solid areas, and comparisons are made.

Discrete and Continuous Adjoint-Based Aerostructural Wing Shape Optimization of a Business Jet

This article presents single- and multi-disciplinary shape optimizations of a generic business jet wing at two transonic cruise flow conditions. The studies performed are based on two high-fidelity gradient-based optimization tools, assisted by the adjoint method (following both discrete and continuous approaches). Single discipline and coupled multi-disciplinary sensitivity derivatives computed from the two tools are compared and verified against finite differences. The importance of not making the frozen turbulence assumption in adjoint-based optimization is demonstrated. Then, a number of optimization runs, ranging from a pure aerodynamic with a rigid structure to an aerostructural one exploring the trade-offs between the involved disciplines, are presented and discussed. The middle-ground scenario of optimizing the wing with aerodynamic criteria and, then, performing an aerostructural trimming is also investigated.

A zero-net-mass-flux wake stabilization method for blunt bodies via global linear instability

A rectangular cylinder, with an aspect ratio of 5, is a widely used bluff body in engineering practice. It undergoes intricate dynamical behavior in response to minute alterations in the flow angle of attack (α). These modifications invariably precipitate the failure of wake control for classical flow control methods with various α values. In this study, global linear instability, adjoint method, and sensitivity analysis are employed to identify the optimal position for flow control. It is found that the sensitive region gradually transitions from the leeward side to the downwind side of the model as α and Reynolds number (Re) increase. So, we set up airflow orifices for flow control in both positions. Jet flow control on the leeward side effectively inhibits vortex shedding (α ≤ 2°). High-order dynamic mode decomposition is employed to reveal the inherent mechanism of control. Suction control on the downside effectively mitigates the shear layer separation phenomenon induced by the altered spatial structure associated with higher α. A novel zero-net-mass-flux wake control, bionics-based breathe-valve control (BVC), is proposed to optimize the control effect. BVC is applicable for various α and Re, with optimal effectiveness achievable through jet velocity adjustments. The prediction-control approach in this investigation provides a targeted method to mitigate flow-induced vibration.

Aerodynamic/stealth design of S-duct inlet based on discrete adjoint method

Aerodynamic/stealth design of S-duct inlet based on discrete adjoint method

On the utilization of the adjoint method in microwave tomography.

In microwave imaging, the adjoint method is widely used for the efficient calculation of the update direction, which is then used to update the unknown model parameter. However, the utilization and the formulation of the adjoint method differ significantly depending on the imaging scenario and the applied optimization algorithm. Because of the problem-specific nature of the adjoint formulations, the dissimilarities between the adjoint calculations may be overlooked. Here, we have classified the adjoint method formulations into two groups: the direct and indirect methods. The direct method involves calculating the derivative of the cost function, whereas, in the indirect method, the derivative of the predicted data is calculated. In this review, the direct and indirect adjoint methods are presented, compared, and discussed. The formulations are explicitly derived using the two-dimensional wave equation in frequency and time domains. Finite-difference time-domain simulations are conducted to show the different uses of the adjoint methods for both single source-multiple receiver, and multiple transceiver scenarios. This study demonstrated that an appropriate adjoint method selection is significant to achieve improved computational efficiency for the applied optimization algorithm.

Design of metamaterial-based heat manipulators using isogeometric level-set topology optimization

We exploit level-set topology optimization to find the optimal material distribution for metamaterial-based heat manipulators. The level-set function, geometry, and solution field are parameterized using the Non-Uniform Rational B-Spline (NURBS) basis functions to take advantage of easy control of smoothness and continuity. In addition, NURBS approximations can produce conic geometries exactly and provide higher efficiency for higher-order elements. The values of the level-set function at the control points (called expansion coefficients) are utilized as design variables. For optimization, we use an advanced mathematical programming technique, Sequential Quadratic Programming. Taking into account a large number of design variables and the small number of constraints associated with our optimization problem, the adjoint method is utilized to calculate the required sensitivities with respect to the design variables. The efficiency and robustness of the proposed method are demonstrated by solving three numerical examples. We have also shown that the current method can handle different geometries and types of objective functions. In addition, regularization techniques such as Tikhonov regularization and volume regularization have been explored to reduce unnecessary complexity and increase the manufacturability of optimized topologies.

Efficient Jacobian Computations for Complex ECT/EIT Imaging

The reconstruction of the spatial complex conductivity σ+jωε0εr from complex valued impedance measurements forms the inverse problem of complex electrical impedance tomography or complex electrical capacitance tomography. Regularized Gauß-Newton schemes have been proposed for their solution. However, the necessary computation of the Jacobian is known to be computationally expensive, as standard techniques such as adjoint field methods require additional simulations. In this work, we show a more efficient way to computationally access the Jacobian matrix. In particular, the presented techniques do not require additional simulations, making the use of the Jacobian, free of additional computational costs.

The sampling method for optimal precursors of El Niño–Southern Oscillation events

Abstract. The El Niño–Southern Oscillation (ENSO) is a significant climate phenomenon that appears periodically in the tropical Pacific. The intermediate coupled ocean–atmosphere Zebiak–Cane (ZC) model is the first and classical one designed to numerically forecast the ENSO events. Traditionally, the conditional nonlinear optimal perturbation (CNOP) approach has been used to capture optimal precursors in practice. In this paper, based on state-of-the-art statistical machine learning techniques1, we investigate the sampling algorithm proposed in Shi and Sun (2023) to obtain optimal precursors via the CNOP approach in the ZC model. For the ZC model, or more generally, the numerical models with a large number O(104−105) of degrees of freedom, the numerical performance, regardless of the statically spatial patterns and the dynamical nonlinear time evolution behaviors as well as the corresponding quantities and indices, shows the high efficiency of the sampling method compared to the traditional adjoint method. The sampling algorithm does not only reduce the gradient (first-order information) to the objective function value (zeroth-order information) but also avoids the use of the adjoint model, which is hard to develop in the coupled ocean–atmosphere models and the parameterization models. In addition, based on the key characteristic that the samples are independently and identically distributed, we can implement the sampling algorithm by parallel computation to shorten the computation time. Meanwhile, we also show in the numerical experiments that the important features of optimal precursors can still be captured even when the number of samples is reduced sharply.

Manifold-guided multi-objective gradient algorithm combined with adjoint method for supersonic aircraft shape design

The aerodynamic design of supersonic aircraft necessitates the comprehensive consideration of multiple conflicting design objectives and multi-disciplinary characteristics. However, the efficiency of heuristic algorithms in solving multi-objective aerodynamic problems significantly decreases with the increase in the number of design variables and objectives. Although gradient-based adjoint optimization methods can eliminate the constraints of design variables on aerodynamic optimization and have high optimization efficiency, they are prone to getting stuck in local optima and are difficult to apply in multi-objective optimization. To overcome these shortcomings, a manifold-guided multi-objective gradient-based algorithm (MG-MOGBA) is proposed. This algorithm enhances the optimization efficiency of the algorithm by coupling the multi-objective gradient operator with the predicted Pareto front (PF) manifold structure to guide the population search direction and reduce the search space. A multi-objective aerodynamic optimization framework based on the discrete adjoint method is developed to automate the aerodynamic optimization process. Test functions and airfoil optimization results confirm that the proposed algorithm achieves superior results with fewer iterations and is more efficient than heuristic algorithms. The four-objective supersonic civil aircraft example demonstrates that the algorithm can effectively handle high-dimensional multi-objective aerodynamic optimization problems. A potential equilibrium solution increased the cruise factor by 17.5% and 12.8% under supersonic and transonic conditions respectively, reduced the maximum overpressure of the sonic boom by 25.1%, and decreased the root bending moment by 20.3%. Overall, MG-MOGBA can efficiently solve high-dimensional multi-objective aerodynamic optimization problems and has good scalability for the dimensions of design objectives and variables.

A novel directly energy-preserving method for charged particle dynamics

In this paper, we apply the coordinate increment discrete gradient (CIDG) method to solve the Lorentz force system which can be written as a non-canonical Hamiltonian system. Then we can obtain a new energy-preserving CIDG-I method for the system. The CIDG-I method can combine with its adjoint method CIDG-II which is also a energy-preserving method to form a new method, namely CIDG-C method. The CIDG-C method is symmetrical and can conserve the Hamiltonian energy directly and exactly. With comparison to the well-used Boris method, numerical experiments indicate that the CIDG-C method holds advantage over the Boris method in terms of energy-conserving.

Sensitivities of ozone to its precursors during heavy ozone pollution events in the Yangtze River Delta using the adjoint method

Although the concentrations of five basic ambient air pollutants in the Yangtze River Delta (YRD) have been reduced since the implementation of the “Air Pollution Prevention and Control Action Plan” in 2013, the ozone concentrations still increase. In order to explore the causes of ozone pollution in YRD, we use the GEOS–Chem and its adjoint model to study the sensitivities of ozone to its precursor emissions from different source regions and emission sectors during heavy ozone pollution events under typical circulation patterns. The Multi-resolution Emission Inventory for China (MEIC) of Tsinghua University and 0.25° × 0.3125° nested grids are adopted in the model. By using the T–mode principal component analysis (T–PCA), the circulation patterns of heavy ozone pollution days (observed MDA8 O3 concentrations ≥160 μg m−3) in Nanjing located in the center area of YRD from 2013 to 2019 are divided into four types, with the main features of Siberian Low, Lake Balkhash High, Northeast China Low, Yellow Sea High, and southeast wind at the surface. The adjoint results show that the contributions of emissions emitted from Jiangsu and Zhejiang are the largest to heavy ozone pollution in Nanjing. The 10 % reduction of anthropogenic NOx and NMVOCs emissions in Jiangsu, Zhejiang and Shanghai could reduce the ozone concentrations in Nanjing by up to 3.40 μg m−3 and 0.96 μg m−3, respectively. However, the reduction of local NMVOCs emissions has little effect on ozone concentrations in Nanjing, and the reduction of local NOx emissions would even increase ozone pollution. For different emissions sectors, industry emissions account for 31 %–74 % of ozone pollution in Nanjing, followed by transportation emissions (18 %–49 %). This study could provide the scientific basis for forecasting ozone pollution events and formulating accurate strategies of emission reduction.

Shape Optimization of Heat Exchanger Fin Structures Using the Adjoint Method and Their Experimental Validation

The freedom of additive manufacturing allows for the production of heat-transferring structures that are optimized in terms of heat transfer and pressure loss using various optimization methods. One question is whether the structural optimizations made can be reproduced by additive manufacturing and whether the adaptations can also be verified experimentally. In this article, adjoint optimization is used to optimize a reference structure and then examine the optimization results experimentally. For this purpose, optimizations are carried out on a 2D model as well as a 3D model. The material chosen for the 3D optimization is stainless steel. Depending on the weighting pairing of heat transfer and pressure loss, the optimizations in 2D result in an increase in heat transfer of 15% compared to the initial reference structure with an almost constant pressure loss or a reduction in pressure loss of 13% with an almost constant heat transfer. The optimizations in 3D result in improvements in the heat transfer of a maximum of 3.5% at constant pressure loss or 9% lower pressure losses at constant heat transfer compared to the initial reference structure. The subsequent experimental investigation shows that the theoretical improvements in heat transfer can only be demonstrated to a limited extent, as the fine contour changes cannot yet be reproduced by additive manufacturing. However, the improvements in pressure loss can be demonstrated experimentally following a cross-section correction. It can therefore be stated that with increasing accuracy of the manufacturing process, the improvements in heat transfer can also be utilized.

Optimization method of water-quality observation positions in offshore CCS sites using the adjoint marginal sensitivity method

Carbon dioxide (CO2) Capture and Storge (CCS) is recognized as a promising method to mitigate the global warming. In Japan, sub-seabed CO2 storage is considered to be commercialized in a decade. However, offshore CO2 storage has a risk of leakage. Since CO2 seepage into seawater may affect the marine environment, CO2 concentration must be monitored in CO2 storage sea sites under a Japanese law. However, there is no scientifically logical base to determine observation positions. In this study, we therefore developed a method of determining optimal observation positions in CO2 storage sea sites. First, we defined optimal observation positions as those that can estimate seepage positions with high accuracy with the minimal number using the adjoint marginal sensitivity method, even if the seepage positions locate anywhere in a target sea site. Second, non-dimensional parameters were proposed to determine whether a datum at a candidate observation position can be successfully used to estimate a possible seepage position using the adjoint method. Then, we defined a coverage index that indicates how much of a target sea site is “covered” in total by observation points and proposed an algorithm to optimally place observation positions. Finally, we validated the developed method numerically and indicated that, to perform highly accurate estimation, it is desirable that a seepage position be covered by at least three observation points.

A Memory-Efficient Neural Ordinary Differential Equation Framework Based on High-Level Adjoint Differentiation

Neural ordinary differential equations (neural ODEs) have emerged as a novel network architecture that bridges dynamical systems and deep learning. However, the gradient obtained with the continuous adjoint method in the vanilla neural ODE is not reverse-accurate. Other approaches suffer either from an excessive memory requirement due to deep computational graphs or from limited choices for the time integration scheme, hampering their application to large-scale complex dynamical systems. To achieve accurate gradients without compromising memory efficiency and flexibility, we present a new neural ODE framework, PNODE, based on high-level discrete adjoint algorithmic differentiation. By leveraging discrete adjoint time integrators and advanced checkpointing strategies tailored for these integrators, PNODE can provide a balance between memory and computational costs, while computing the gradients consistently and accurately. We provide an open-source implementation based on PyTorch and PETSc, one of the most commonly used portable, scalable scientific computing libraries. We demonstrate the performance through extensive numerical experiments on image classification and continuous normalizing flow problems. We show that PNODE achieves the highest memory efficiency when compared with other reverse-accurate methods. On the image classification problems, PNODE is up to two times faster than the vanilla neural ODE and up to 2.3 times faster than the best existing reverse-accurate method. We also show that PNODE enables the use of the implicit time integration methods that are needed for stiff dynamical systems.

Adjoint Design Optimization Under the Uncertainty Quantification of Reynolds-Averaged Navier-Stokes Turbulence Model

Considering the structural uncertainties of Reynolds-averaged Navier–Stokes (RANS) models, a design optimization method under eigenspace perturbations of the RANS model has been proposed for aerodynamic applications. Optimized geometries with confidence intervals with improved performance have been obtained for U-pipe, Busemann airfoil, and supersonic separator. The perturbations are injected into the eigenvalues and eigenvectors of turbulence anisotropic tensors nonuniformly and adaptively; thus, the uncertainty interval of the RANS model is obtained by six simulations. The adjoint method is employed to perform single-objective optimization on three physical models based on uncertainty quantification. The geometric profiles before and after optimization are presented, and the area surrounded by different profiles reflects the differences in geometric optimization caused by the uncertainty of the model form. Shape optimization within the confidence interval achieves enhanced performance with robust improvements, reducing the sensitivity to manufacturing tolerances. Design optimization under the framework of uncertainty quantification may have great practicability as an engineering application tool.

Physics-agnostic inverse design using transfer matrices

Inverse design is an application of machine learning to device design, giving the computer maximal latitude in generating novel structures, learning from their performance, and optimizing them to suit the designer’s needs. Gradient-based optimizers, augmented by the adjoint method to efficiently compute the gradient, are particularly attractive for this approach and have proven highly successful with finite-element and finite-difference physics simulators. Here, we extend adjoint optimization to the transfer matrix method, an accurate and efficient simulator for a wide variety of quasi-1D physical phenomena. We leverage this versatility to develop a physics-agnostic inverse design framework and apply it to three distinct problems, each presenting a substantial challenge for conventional design methods: optics, designing a multivariate optical element for compressive sensing; acoustics, designing a high-performance anti-sonar submarine coating; and quantum mechanics, designing a tunable double-bandpass electron energy filter.