Cogging Torque Computation and Mesh For Non-Radial Electrical Motor in Flux FluxNews66june2014

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 2

Flux News

Cogging torque computation and mesh for non-radial


electrical motors in Flux®. Sylvain Perez - CEDRAT.

A
ll electrical motor designers know that the computation Mechanical set/Magnetic

of cogging torque is a tricky task, particularly in 3D.


torque [ROTOR] (N.m)
4’ TORQUE_ROT
-4’ TORQUE_STAT

Indeed, the amplitude of this variable is almost the


same as numerical noise. In most cases, conventional mesh
methodology is not sufficient and specific methodology must
be used. At CEDRAT, thanks to its experience, the application
team has developed methodologies to successfully compute
cogging torque in most cases.
This article presents a specific mesh methodology to compute Figure 3: Cogging torque calculation with a classical mesh
cogging torque for 3D non-radial electrical motors. with first elements order.

Specific mesh methodology for non- Fig. 3 presents three different calculations for cogging torque
evaluation. All calculations are based on a virtual works method
extruded 3D electrical motor but the contours under consideration are different. The first
calculation considers the contour of the sliding cylinder (blue
To explain this mesh methodology, the 3D motor presented is an curve), the second the contour of the rotor and magnets (magenta
“Axial Flux Motor” (Fig. 1). This kind of motor is generally used in curve) and the third the contour of the stator (green curve). All
environments involving compact lengths, like elevator systems. results are bad, cogging torque is not centred 0 N.m. However,
these results give us some information. We can see that cogging
torque evaluations considering stator contour are the most
accurate. This suggests that mesh is more appropriate surrounding
the stator than the rotor. For further information, we need to use
dedicated tools to perform mesh analysis.

» Analysis of mesh defect with Flux tools

To identify where the mesh can be improved, Flux offers dedicated


tools based on error criterion calculation. In this example, cogging
torque is dependent on a derivate of co-energy that means
Figure1: Axial Flux motor. selecting the magnetic co-energy gradient as an error criterion
(Fig. 4).
The specific mesh methodology is based on Flux 3D mesh tools We display error criteria on each volume element of magnets
and some physical considerations. We will look at the various steps (Fig. 5) and of the sliding cylinder (Fig. 6). The greater the number
to follow. The first one is the starting point, whether, a conventional of red spheres, the bigger the criterion error. Another important
mesh imposed by “Mesh Point Discretisation”. Then, we will see point to consider is the density of spheres which may also offer
how to use the Flux tool for analysing mesh quality. Afterwards, a good indicator.
on the basis of analysis and some physics angles, we will apply a
new mesh imposed by “Mesh Line Discretisation” with relaxation.
Finally, we will compare results for the different meshes in terms
of “Accuracy/Computation time”.

» Starting with a conventional mesh imposed by “Mesh Point


Discretization”

The conventional mesh method is to use discretisation imposed


by point, i.e.: imposing mesh specifications at each point of the
Figure 4: Error criterion for cogging torque evaluation.
geometry. With Flux, we currently use this mesh method coupled
with aided mesh options. In this case, we obtained the mesh
presented in Fig. 2. From this conventional mesh with first-order
elements, computation results are given in Fig. 3.

Figure 5: Criterion error display on sliding cylinder.

Figure 2: Classical mesh


imposed by “mesh point
discretization”.

Figure 6: Criterion error display on magnets.


(continued on page 11)

11 | Flux News | N° 66 | June 2014


Flux News

» Final comparison
We note that the criterion error display on the sliding cylinder
indicates that bad mesh elements are only in the airgap below the It is important to get a good precision, bearing in mind that a good
magnets. That tells us that shadow should be higher and that face “Precision/Computation time” trade-off is the objective. At this
relaxation is too fast. The criterion error display on magnets is less time, we have not considered computation time, this paragraph
clear to analyse, the red sphere indicates that volume relaxation is compares results and time computation for each mesh described
too high and should be reduced. We also note high sphere density previously with first-order elements and with second-order
(not necessarily red ones) along magnet edges which suggests the elements (Tab.1, Fig. 9 and Fig. 10).
mesh on edges should be refined.
This information allows us defining a new efficient mesh.
Mechanical set/Magnetic
torque [ROTOR] (N.m)
4’ TORQUE_ROT
-4’ TORQUE_STAT
(a)
» Define a new efficient mesh based on “Mesh Line
Discretization”
Figure 9: Cogging torque
In the previous paragraph we highlighted the need to reduce calculation with a
relaxation and increase shadow. Based on CEDRAT application classical mesh with first
team knowledge and confirmed by the previous mesh analysis, elements order (a)
we can conclude that for volumes bordering the air gap (magnets Mechanical set/Magnetic
and with second
and stator tooth), mesh on edges has to be refined. torque [ROTOR] (N.m)
4’ TORQUE_ROT
elements order (b).
Thin mesh on edges seems to be explained by the fact that where
-4’ TORQUE_STAT
(b)
the magnetic flux density changes quickly, mesh has to be refined
in order to precisely compute magnetic field paths. Accordingly,
the mesh method consists of finely mesh edges of volumes next
to the air gap (Fig. 7). Results of this second mesh with first-order
elements are given by Fig. 8.

Cogging torque is centred on 0 N.m whatever the contour Mechanical set/Magnetic


torque [ROTOR] (N.m)
4’ TORQUE_ROT

considers. Even though there are some differences, the 3 cogging -4’ TORQUE_STAT
(a)
torque calculations are now compliant. We can consider that the
second mesh is efficient. Figure 10: Cogging
torque calculation with
the new mesh with first
Figure 7: New efficient elements order (a)
mesh based on“Mesh Mechanical set/Magnetic
torque [ROTOR] (N.m)
and with second
elements order (b).
4’ TORQUE_ROT

Point Discretization”. -4’ TORQUE_STAT


(b)

Method Mesh element Number of Number Computation


order nodes of volumic time
elements
Classical 1st 8685 38270 2’
mesh
2nd 59556 38270 10’
New 1st 36444 174892 8’
mesh
2nd 258699 174982 72’

Table 1: Mesh characteristics and computation time.

According to Tab.1, Fig. 9 and Fig. 10, the new mesh with first-
order elements offers the best compromise between accuracy
and computation time.

Conclusion
Cogging torque computation has always been a complex exercise,
especially in 3D. Nowadays, Flux tools combined with CEDRAT’s
experience in this domain make life easier. In this article we have
reviewed a methodology to obtain efficient mesh for your device
to rapidly produce a reliable estimate of cogging torque.
Figure 8: Cogging torque calculation with the new mesh
In a next issue we are proud to present you a new article "Cogging
with first elements order. torque computation and mesh for radial electrical motors".

12 | Flux News | N° 66 | June 2014

You might also like