You are on page 1of 4

Notes on FE Analysis of Heat Flow Using Point, Line, Surface,

Volume Moving Heat Source


By Dr.O.P.Gupta
June 2013
It seems that the users of FE software sometimes find it difficult to model and
analyse the heat flow and temperature distribution produced by moving heat source
encountered in welding, laser heating, surface heat treatment etc. This note is
prepared, based on author’s experience with modelling such problems, with the
objective that it may be useful in such modelling efforts.
Although the author has worked with some commercial software, such as
ANSYS, in order to solve such problems in an efficient manner, he has found some
open source software quite efficient and useful for FE analysis in general and current
problem in particular. The strategy explained in this note can be used with any FE
software, commercial or otherwise.
For the purpose of reference, the geometry in following example has been
modelled using ‘gmsh’ open source software, which is capable of modelling and mesh
generation in variety of 2D, 3D objects (or a mix of several bodies with compatible
mesh). It is available at the site http://geuz.org/gmsh/ .
The FE analysis was performed using ‘Elmer’ FE software, which is a very
powerful software for analysis of multi-physics problems of elasticity, heat flow, fluid
flow, MHD, electro-magnetic field etc. It is also an open source software available at
www.csc.fi/elmer1 .
Modelling Strategy for Moving Heat Source
Although it is possible to model the moving heat source in transient analysis
by supplying information on the current position of the heat source (which is
changing position continuously) for each time step, with the input of heat distribution
in the heat source, it is not a convenient method and it will require a very fine mesh
all along the path of the heat source. A better and cleaner approach is to model it
indirectly using the formulation of heat conduction-convection analysis in moving
fluid. The governing equation is
 T    
c    u   T     (kT )   q (1)
 t   

where u is velocity vector, T, temperature, t, time , k,,c are thermal
conductivity, density and specific heat respectively. The right hand side of the
equation contains volumetric heat input term (here q is rate of heat input per unit mass
per unit time – sometimes the whole term is denoted by Q (=q), the rate of
volumetric heat input).
In order to model the moving heat source, in welding etc., if the heat source
moves in a straight lie (say along x-axis), the model has to be provided the input of

only x-component of u , equal to the speed of welding. Other two components will be
1
Elmer is developed and maintained mainly at CSC - IT Center for Science (CSC).
CSC is a non-profit governmental organization owned by the Ministry of Education
and Culture in Finland. Elmer is mainly used as a collaborative tool in academic
research. There are also a number of other developers who have or are contributing to
Elmer development. The author of this note is also developing a module for analysis
of sheet metal forming operation.
zero. The fluid in this analysis, is replaced by the plate object (with input of material
properties for it). The object should be modeled as two bodies, one being the weld
pool (or hot spot) somewhere in the middle (say) and the other being the rest of the
plate (or object). The external heat input is to be supplied as volumetric heat (q or Q-
explained above-the software Elmer accepts q only) in the body representing the weld
pool only. When the transient analysis is performed using small time steps, the
situation is that of plate moving under the heat source (or equivalent to the moving
heat source over the plate.
An Example
Following are some screen shots of the mesh used (developed using gmsh) and
resulting temperature distribution (obtained using Elmer).
The data for the problem are:
Plate size - 0.5x0.25 m (2D analysis)
Weld pool – half ellipse for edge heating (with major and minor axes 0.010 and 0.008
m), modeled as 2nd body (Fig 1)
Heat input – 1.274x 106 W/kg( or 10x109W/m3), equivalent to 20V and 200 Amp
(4000 W) for whole plate , which is equivalent to 2000 W for half plate here.
Ultimately it is equivalent to approximately 10x109W/m3.
= 7850.0 kg/m3
c= 1265.0
k= 44.8
, c, k are inputs from the material data available in the data library of
software.
Speed of heat source= 2.5x10-3 m/s

Fig. 1 Mesh used (heat source is in the middle)


Fig. 2 Zoomed mesh, showing the 2nd body as half ellipse

Fig. 3 Temperature distribution


Fig.4 Temperature isotherms

Observations
The whole analysis involved 100 time steps of 1 s each with about 2 iterations
in each step. The total CPU time taken in execution was 81.5 sec. The data
preparation might have taken another 10 min. and object and mesh generation another
15 min in ‘gmsh’. Of course, this assumes complete familiarity with the GUI of the
software. Those who are not initiated in this software will require learning time,
which is the case with any software. The software provides good users’ and
programmers’ (models) manual.
Modelling of line, point or volumetric heat source is possible as thin rectangle,
very small circle and half ellipsoid respectively.

You might also like