P.C. Coles

Towards increased speed computations in 3D moving eddy current finite element modelling

A finite element scheme used to model eddy current problems with motional velocity is described here. The formulation is a variation on the A-/spl psi/ method. An additional Minkowski-transformation term is required to take into account the velocity. However, computational instability arises when the velocity increases to the point that the first order velocity terms severely dominate the second order diffusion terms. The method presented here uses upwinding to help regain stability. An additional degree of stability is inserted at higher speeds by using a lower speed result as an initial vector. This leads to a reduced permeability in saturated regions which counterbalances to some extent the increase in velocity. The method is validated by experimental measurement.

Finite element modelling of magnetic hysteresis

This paper describes an implementation of static magnetic hysteresis using 3D finite elements and a scalar Preisach hysteresis model. The scheme uses a new method for finding the consistent magnetisation state for a given history of applied sources. Experimental results are compared with predictions for a simple 3D model in which we have only considered one direction of hysteresis. >

Full-scale testing of a high speed linear synchronous motor and calculation of end-effects

A description is given of apparatus constructed to perform full-scale tests on linear synchronous motor (LSM) designs suitable for 100-passenger vehicles, traveling at speeds up to 400 km/hr. Test results are presented that illustrate the terminal characteristics of the machine, which are shown to be predictable using a simple mathematical model. In all forms of linear machine the air-gap flux takes time to develop at the entry end of the stator and a further time to decay at the exit end. This effect, which is due to induced secondary currents, is known as the longitudinal end effect. A simple method is described for predicting the effect, and the results are confirmed by measurements on a full-scale test rig. >

Numerical models of an induction machine

Measurements of the performance of a 5-kVA induction machine are compared with predictions carried out using finite elements and semianalytic techniques. Both transient and steady-state performance results are compared and sufficient dimensions are provided that others may use the experimental results as a benchmark.

Finite element calculation of forces on a DC magnet moving over an iron rail

This paper describes results taken from a test rig consisting of a DC magnet over a 0.35 m radius spinning iron wheel. The magnet is excited by two coils. The iron parts are unlaminated. Eddy currents are induced in the wheel by virtue of the relative motion of wheel and magnetic field. All iron parts have a nonlinear B-H characteristic. Forces on the magnet are compared with 3D finite element predictions. The results are of relevance to the design of MAGLEV vehicles which are supported by DC magnets. >

A finite element scheme for colliding meshes [EM device modelling]

This paper describes a scheme which allows two separate finite element meshes to slide into and thus overlap each other. The Lagrange multipliers method is used to couple the two meshes together at their boundary. Overlapping elements are dealt with by adjusting the contributions of the elements of one of the two meshes to the system matrix according to the extent of overlap in such a way that at total overlap, those elements are effectively decoupled from the model. Applications include the modelling of electromagnetic plunger type devices.

Comparison of short primary linear machines for high speed maglev vehicles

A brief review of analytical techniques for linear induction machines describes both Fourier based techniques and the more recent 3D finite element methods. A discussion of linear induction machine topology then follows in which conventional axial flux machines are contrasted with transverse flux motors. Primary side excited short stator linear synchronous machines are described and the features of a rotating rig to test high speed versions are explained. Some initial test results from this rig are then given together with the results from a Simple analytical method. The paper finishes by comparing two high speed designs contrasting a homopolar synchronous machine with an axial flux linear induction motor.

A 3-D overlapping finite-element scheme for modeling movement

This paper describes a three-dimensional (3-D) overlapping finite-element scheme for modeling movement in electromagnetic devices, which results in a continuously changing air gap. The method uses separate finite-element meshes to model the individual parts of the device. These meshes are coupled together electromagnetically on an interface using the Lagrange sliding interface technique. Treatments of overlapping mesh volumes are also discussed. The method was used to model a jumping eddy-current disc experiment. The results of the simulations compared favorably with measurements.

Finite element modelling of thin skin depth problems using magnetic vector potential

A technique for modelling thin skin depth eddy currents in 3D devices using the magnetic vector potential A is presented. An analytic solution for the magnetic field in the thin skin is used to develop a surface impedance which appears as a surface integral at faces of elements which interface the thin skin eddy current region. The method has been implemented using classical nodal variable finite elements. Results are compared with an analytic solution and measurements from a nonlinear test rig.

Finite element modelling of 3D moving conductor devices with low conductivity

Devices which include moving conductors which are invariant in cross section normal to the direction of motion may be modelled using a finite element scheme which incorporates the Minkowski transformation. Three components of magnetic vector potential and electric scalar potential are required inside the moving medium. When the conductivity of the moving medium is very low, the magnetic field induced by the eddy currents is relatively small and can be neglected. In this case the problem may be formulated more economically in terms of two static scalar solutions. The two methods are compared with each other and with analytic solutions and experimental results.