Search results

The synthesis of electro‐mechanic actuators is formulated as a constrained optimization problem where some performance function of the device is to be met, subject to the satisfaction of some constraints about its dimensions and supply conditions. The optimization problem is tackled by means of a genetic algorithm coupled to a multi‐objective definition of the objective function that merge together objectives and constraints in one single scalar objective function. A fast magnetic analysis tool has been developed so that the computational cost of the genetic optimization run is acceptable. Some results about the synthesis of a tubular linear motor in two sizes are presented and discussed.

The purpose of this paper is to investigate the extension of Delaunay refinement algorithms to work directly with a curved geometry in arbitrary dimensional spaces, which is also able to refine geometry pieces of different dimensions altogether.

The extension of Delaunay refinement is based on ideas of the Bowyer‐Watson algorithm and Ruppert algorithm.

The attempt to extend the fundamental ideas of Delaunay refinement to cope with curved geometries led to an algorithm whose performance in practice, regarding speed and mesh quality, is comparable to classical Delaunay refinement for flat geometries. Unfortunately, there are only theoretical guarantees that the refinement itself works under some conditions. No theoretical mesh quality bounds are provided.

A mesh refinement algorithm that deals with curved geometries is a key feature for adaptive mesh generators, so that points are inserted properly in the curved pieces instead of in linear approximations of them. For instance, it is well known that sharp edges are singular points of finite element formulations. This singularity fulfills in practice as mesh is refined around them. Those corners can be rounded up to avoid singularities. Furthermore, with this kind of tool, for instance, a user could start to mesh a disc from a single triangle representing it. Points would be efficiently inserted in the circle as needed during refinement.

This paper introduces the concept of manifold complex and also an extension of Delaunay refinement algorithm to deal with curved geometries.

Grain‐oriented steel has a distinctly anisotropic and nonlinear behaviour. Only in rare cases is the magnetisation curve known for directions other than the principal ones. The paper aims at providing a model to obtain these curves for any direction if those in the easy and hard directions are only given.

The well‐known elliptic model is modified in order to correctly mimic the typical behaviour of grain‐oriented steel which is not described correctly by the original elliptic model. An additional condition is introduced to fix the angle between the flux density and magnetic field intensity.

The model is found to yield good agreement with measurements in case of a special material for which measured curves for intermediate angles are available.

Further research is necessary to establish whether the model is applicable to other materials.

The new model can be used in numerical analyses of devices comprising saturated grain‐oriented steel material if the magnetisation curves are given in the principal directions.

The classical ϕ‐a formulations for numerical dosimetry of currents induced by extremely low frequency magnetic fields requires that the source field is provided through a vector potential. The purpose of this paper is to present a new formulation t‐b which directly takes the flux density as source term.

This formulation is implemented through finite element and validated by comparison with analytical solutions. The results obtained by both formulations are compared in the case of an anatomical computational phantom exposed to a vertical uniform field.

A good agreement between the t‐b formulation and both numerical and analytical computations was found.

This new formulation seems to be more accurate than the ϕ‐a formulation, and is more suited for situations where the magnetic field is known from experimental measurements, as there is no need for a magnetic vector potential.

Supplying remote wireless sensors is not an easy task if the site where the device is located is not easily accessible. In order to obtain direct measurements of the road-vehicle interactions, sensors must be placed inside the tyre environment thus a power supply must be available for their working there without any wire connection with the car main power. The paper aims to discuss these issues.

An electro-mechanical energy harvester has thus been developed for supplying an automotive wireless sensor of pressure, temperature and acceleration to be placed on the inner line of a tyre. The primary energy source is the vibrations or variable accelerations imposed to the device and induced in the tyre by the wheeling.

The harvester has been designed by means of a multi-physics optimisation based on an integrated electromagnetic-mechanical circuit simulator. Thus an automated optimisation of the device with respect to volume constraints, magnets dimensions, induction coils placement and size have been performed to increase the average power extracted from the device at different wheeling speeds.

The use of the multi-physics environment together with automated optimisation technique has been tested for the first time on the electromagnetic harvester structure.

To propose a novel 3D hybrid approach, based on a discrete formulation of Maxwell equations (the cell method – CM), suitable for solving eddy current problems in unbounded domains.

Field equations for magnetodynamics are expressed directly in algebraic form thanks to the CM. The eddy current problem inside bulk conductors is formulated in terms of discrete modified vector potential, whereas magnetic scalar potential is used in order to model the free space. The CM is coupled to the boundary element method by using a surface boundary operator, which maps the surface magnetic fluxes to the surface magnetic scalar potentials. This leads to a unique set of linear equations to be solved in terms of discrete potentials. The eddy currents in bulk conductors are then obtained from discrete potentials.

It is shown that formulation of hybrid approaches can be simplified by expressing field equations directly in algebraic form without need of weighted residual techniques. An original strategy, based on Green's formula for the magnetic scalar potential, is proposed in order to couple conducting parts to the exterior domain.

Conducting bodies with multiply connected parts cannot be modelled by the proposed approach, since it is based on the magnetic scalar potential. The resulting global matrix is partially dense and non‐symmetric; therefore, standard iterative solvers such as GMRES have to be used.

The proposed approach can be suitably used for analyzing eddy current problems involving models with high degree of complexity, large air domains and moving parts. These are typical of induction heating processes.

This paper proposes a new 3D hybrid approach, based on a discrete formulation of Maxwell equations. A novel coupling strategy relying on integral electromagnetic variables, i.e. magnetic fluxes and magnetic scalar potentials, is devised in order to solve uniquely for eddy currents inside conducting bodies.

The purpose of this paper is to simulate passive proton exchange membrane fuel cells (PEMFCs) for portable electronic devices by means of a non‐linear lumped circuit based on electrical, mass transfer and electro‐kinetic equations.

Electrical, mass transfer and electro‐kinetic equations are combined in order to derive a non‐linear lumped circuit. The dynamic circuit model is tested in realistic operating conditions.

An original equivalent circuit model for simulating the transient behavior of passive PEMFCs is proposed. The PEMFC is represented as a non‐linear equivalent circuit with controlled lumped parameters depending on pressure, temperature, hydration, and system capacity.

Lumped parameters are synthesized assuming a one‐dimensional fuel cell model since layer thicknesses are much smaller than other dimensions. Heat generation and transfer are not modeled even though lumped parameters depend on temperature.

The proposed circuit model can be implemented directly in circuit simulators for designing power management units needed to interface small‐passive PEMFCs and portable electronics such as PDAs, laptops, or mobile phones.

The fuel cell is represented as a non‐linear controlled generator whose parameters are derived directly from multiphysics equations rather than empirical relationships. The dynamic behaviour of PEMFCs can be simulated on completely different times scales, i.e. during transients or during the discharge phase.

The purpose of this study is to investigate and compare the ability of a new optimization technique based on the emulation of the immune system to detect the global maximum with multimodal functions and to test the capability of exploring the parameter space with respect to clustering enhanced Genetic Algorithms (GA).

Both algorithms have been tested on analytical test functions and on numerical functions of applicative interest. A set of performance criteria has been defined in order to numerically compare the performances of both optimization strategies.

Results show the great ability of Artificial Immune Systems (AIS) in thoroughly exploring the space of variables. On the other side, GA are faster to converge to the global optimum, but selection pressure can reduce the number of detected local optima.

This work is an attempt to assess the performances of a relatively new optimization algorithm based on AIS and to find its behavior on multimodal test functions, using GAs as reference optimization technique.

The aim of this paper is the experimental validation of an original time‐domain thin‐shell formulation. The numerical results of a three‐dimensional thin‐shell model are compared with the measurements performed on a heating device at different working frequencies.

A time‐domain extension of the classical frequency‐domain thin‐shell approach is used for the finite‐element analysis of a shielded pulse‐current induction heater. The time‐domain interface conditions at the shell surface are expressed in terms of the average flux density vector in the shell, as well as in terms of a limited number of higher‐order components.

A very good agreement between measurements and simulations is observed. A clear advantage of the proposed thin‐shell approach is that the mesh of the computation domain does not depend on the working frequency anymore. It provides a good compromise between computational cost and accuracy. Indeed, adding a sufficient number of induction components, a very high accuracy can be achieved.

The method is based on the coupling of a time‐domain 1D thin‐shell model with a magnetic vector potential formulation via the surface integral term. A limited number of additional unknowns for the magnetic flux density are incorporated on the shell boundary.

The purpose of this paper is to address the formulation, implementation, and adaptation of closely coupled multi‐physics problems with h‐ and p‐adaptive finite element methods. A general formulation is chosen allowing for coupled problems of various types. Adaptation algorithms for h‐ and p‐refinement are given.

A generic system of second‐order differential equations is used, where the field of each individual problem is represented as an entry in the list of field variables. Specific problems are implemented by mapping material coefficients to the coefficients of the generic form. An example with four natures is investigated with close coupling between electric, mechanical and thermal fields. h‐ and p‐refinement using a single mesh is considered, where the element order may differ for individual fields.

In coupled problems, the error in each single field is affected by approximation properties of all other field quantities. In order to allow for optimal convergence order in the number of degrees of freedom, the error contributions of all fields have to be considered. Separate error estimation in each field is needed especially in h‐adaptation on a single mesh. Energy coupling coefficients were introduced to derive an adaptation criterion. Convergence analysis of h‐ and p‐adaptation proves the feasibility of the approach.

Piezopyroelectricity considers thermal effects in high‐frequency piezoelectric materials, which is a coupled problem of four natures. The paper introduces an adaptation criterion for such complicated coupled problems and proves feasibility.