Search results

Accurate numerical differentiation of approximate data by methods based on Green's second identity often involves singular or nearly singular integrals over domains or their boundaries. This paper applies the finite part integration concept to evaluate such integrals and to generate suitable quadrature formulae. The weak singularity involved in first derivatives is removable; the strong singularities encountered in computing higher derivatives can be reduced. To find derivatives on or near the edge of the integration region, special treatment of boundary integrals is required. Values of normal derivative at points on the edge are obtainable by the method described. Example results are given for derivatives of analytically known functions, as well as results from finite element analysis.

The gradient recovery method proposed by Zhu and Zienkiewicz for one‐dimensional problems is generalized to two dimensions, using quadrilateral elements. Its performance is compared with that of conventional local smoothing techniques and of direct differentiation of the finite‐element solution, on finite‐element approximations to analytically known polynomial and transcendental functions on a quadrilateral second‐order finite‐element mesh. The new method appears to be reliable and more stable than local smoothing, and to provide better accuracy than direct differentiation, at low computational cost.

This paper aims to investigate the convergence and error properties of a finite volume-based heat conduction code that uses automatic differentiation to evaluate derivatives of solutions outputs with respect to arbitrary solution input(s). A problem involving conduction in a plane wall with convection at its surfaces is used as a test problem, as it has an analytical solution, and the error can be evaluated directly.

The finite volume method is used to discretize the transient heat diffusion equation with constant thermophysical properties. The discretized problem is then linearized, which results in two linear systems; one for the primary solution field and one for the secondary field, representing the derivative of the primary field with respect to the selected input(s). Derivatives required in the formation of the secondary linear system are obtained by automatic differentiation using an operator overloading and templating approach in C++.

The temporal and spatial discretization error for the derivative solution follows the same order of accuracy as the primary solution. Second-order accuracy of the spatial and temporal discretization schemes is confirmed for both primary and secondary problems using both orthogonal and non-orthogonal grids. However, it has been found that for non-orthogonal cases, there is a limit to the error reduction, which is concluded to be a result of errors in the Gauss-based gradient reconstruction method.

The convergence and error properties of derivative solutions obtained by forward mode automatic differentiation of finite volume-based codes have not been previously investigated.

The gradient recovery method proposed by Zhu and Zienkiewicz for one‐dimensional problems and extended to two dimensions by Silvester and Omeragi? is generalized to three‐dimensional solutions based on rectangular prism (brick) elements. The extension is not obvious so its details are presented, and the method compared with conventional local smoothing and direct differentiation. Illustrative examples are given, with an extensive experimental study of error. The method is computationally cheap and provides better accuracy than conventional local smoothing, but its accuracy is position dependent.

Investigates the efficiency of hybrid solution methods when incorporated into large‐scale topology and shape optimization problems and to demonstrate their influence on the overall performance of the optimization algorithms. Implements three innovative solution methods based on the preconditioned conjugate gradient (PCG) and Lanczos algorithms. The first method is a PCG algorithm with a preconditioner resulted from a complete or an incomplete Cholesky factorization, the second is a PCG algorithm in which a truncated Neumann series expansion is used as preconditioner, and the third is a preconditioned Lanczos algorithm properly modified to treat multiple right‐hand sides. The numerical tests presented demonstrate the computational advantages of the proposed methods which become more pronounced in large‐scale and/or computationally intensive optimization problems.

The purpose of this paper is to analyse the accuracy of the thrust force of a linear actuator computed with different finite elements models.

A series of 2D and 3D models corresponding to different levels of approximation of the original problem are considered. A reliable error estimator based on dual magnetostatic formulations is used.

A 3D model does not necessarily ensure more accurate results than a 2D model. Because of limitations on the number of mesh elements, the discretisation error in 3D can be of the same order of magnitude as the error introduced by the 2D approximation.

The results emphasise the need to consider errors arising from different simplifications with respect to one another, in order to avoid improvements of the model increasing the complexity but not improving the accuracy of the results.

The purpose of this paper is to illustrate automatic differentiation (AD) as a new technology for the device sizing in electromagnetism by using gradient constrained optimization. Component architecture for the design of engineering systems (CADES) framework, previously described, is presented here with extended features.

The paper is subject to further usage for optimization of AD (also named algorithmic differentiation) which is a powerful technique that computes derivatives of functions described as computer programs in a programming language like C/C++, FORTRAN.

Indeed, analytical modeling is well suited regarding optimization procedure, but the modeling of complex devices needs sometimes numerical formulations. This paper then reviews the concepts implemented in CADES which aim to manage the interactions of analytical and numerical modeling inside of gradient‐based optimization procedure. Finally, the paper shows that AD has no limit for the input program complexity, or gradients accuracy, in the context of constrained optimization of an electromagnetic actuator.

AD is employed for a large and complex numerical code computing multidimensional integrals of functions. Thus, the paper intends to prove the AD capabilities in the context of electromagnetic device sizing by means of gradient optimization. The code complexity as also as the implications of AD usage may stand as a good reference for the researchers in this field area.

The purpose of this study is to propose a framework for expedited antenna optimization with numerical derivatives involving gradient variation monitoring throughout the optimization run and demonstrate it using a benchmark set of real-world wideband antennas. A comprehensive analysis of the algorithm performance involving multiple starting points is provided. The optimization results are compared with a conventional trust-region (TR) procedure, as well as the state-of-the-art accelerated TR algorithms.

The proposed algorithm is a modification of the TR gradient-based algorithm with numerical derivatives in which a monitoring of changes of the system response gradients is performed throughout the algorithm run. The gradient variations between consecutive iterations are quantified by an appropriately developed metric. Upon detecting stable patterns for particular parameter sensitivities, the costly finite differentiation (FD)-based gradient updates are suppressed; hence, the overall number of full-wave electromagnetic (EM) simulations is significantly reduced. This leads to considerable computational savings without compromising the design quality.

Monitoring of the antenna response sensitivity variations during the optimization process enables to detect the parameters for which updating the gradient information is not necessary at every iteration. When incorporated into the TR gradient-search procedures, the approach permits reduction of the computational cost of the optimization process. The proposed technique is dedicated to expedite direct optimization of antenna structures, but it can also be applied to speed up surrogate-assisted tasks, especially solving sub-problems that involve performing numerous evaluations of coarse-discretization models.

The introduced methodology opens up new possibilities for future developments of accelerated antenna optimization procedures. In particular, the presented routine can be combined with the previously reported techniques that involve replacing FD with the Broyden formula for directions that are satisfactorily well aligned with the most recent design relocation and/or performing FD in a sparse manner based on relative design relocation (with respect to the current search region) in consecutive algorithm iterations.

Benchmarking against a conventional TR procedure, as well as previously reported methods, confirms improved efficiency and reliability of the proposed approach. The applications of the framework include direct EM-driven design closure, along with surrogate-based optimization within variable-fidelity surrogate-assisted procedures. To the best of the authors’ knowledge, no comparable approach to antenna optimization has been reported elsewhere. Particularly, it surmounts established methodology by carrying out constant supervision of the antenna response gradient throughout successive algorithm iterations and using gathered observations to properly guide the optimization routine.

A technique for accelerated design optimization of antenna input characteristics is developed and comprehensively validated using real-world wideband antenna structures. Comparative study using a conventional trust-region algorithm is provided. Investigations of the effects of the algorithm control parameters are also carried out.

An optimization methodology is introduced that replaces finite differentiation (FD) by a combination of FD and selectively used Broyden updating formula for antenna response Jacobian estimations. The updating formula is used for directions that are sufficiently well aligned with the design relocation that occurred in the most recent algorithm iteration. This allows for a significant reduction of the number of full-wave electromagnetic simulations necessary for the algorithm to converge; hence, it leads to the reduction of the overall design cost.

Incorporation of the updating formulas into the Jacobian estimation process in a selective manner considerably reduces the computational cost of the optimization process without compromising the design quality. The algorithm proposed in the study can be used to speed up direct optimization of the antenna structures as well as surrogate-assisted procedures involving variable-fidelity models.

This study sets a direction for further studies on accelerating procedures for the local optimization of antenna structures. Further investigations on the effects of the control parameters on the algorithm performance are necessary along with the development of means to automate the algorithm setup for a particular antenna structure, especially from the point of view of the search space dimensionality.

The proposed algorithm proved useful for a reduced-cost optimization of antennas and has been demonstrated to outperform conventional algorithms. To the authors’ knowledge, this is one of the first attempts to address the problem in this manner. In particular, it goes beyond traditional approaches, especially by combining various sensitivity estimation update measures in an adaptive fashion.

In electric machines, the electromagnetic torques and forces are developed as a result of the interaction of the magnetic fields. These forces can be computed from the results of field analysis, using numerical or analytical methods. Describes a hybrid technique, which is suitable for the calculation and analysis of electromagnetic torques and forces. This method exploits the advantages of numerical and analytical methods for field analysis, and thus provides a more efficient and effective tool for torque/force computation, in comparison with traditional methods. The computational examples presented show that the method is particularly useful for accurate prediction of the electromagnetic torque and unbalanced magnetic pull in electric machines.