[esp-r] Re: Using mrt for view factor calculation

[Jon Hand (clcv10)]

  A recent question about the use of the MRT module to calculate
viewfactors and how complex the zones geometry can be....

> has anybody used mrt for ray tracing based view factor calculation  
> (as was suggested a little while ago) on any problems with many  
> surfaces? Are there known limits? I tried this in a zone with 24  
> surfaces (see attached files, if interested). Mrt does not seem to  
> be able to cope with this based on default settings for compile (I  
> set the two parameters “grid division” and “patch division” to  
> maximum values – in this case 32 and 50). The results are mainly  
> zeroes … for the zone view factors themselves and for the MRT sensor.
>
> Is there any experience on this out there? What / where would one  
> need to change settings in the headers and recompile for higher  
> resolution? Has that been done / checked in the past?

The MRT module has been tested for zones up to about 55 surfaces and  
with gridding
division set within the range of 12-25. Usually the patch division  
value does not get
changed (i.e. I do not bother changing it).

The interface should let you know the maximum values it will allow  
for the number
of sufaces in the zone.

The patch division is intended to put an array of points distributed  
across the surface
and this usually works ok UNLESS some of the dimensions are small -  
for example if
a surface has had a window inserted and the edge of the window and  
the nearest
edge of the bounding surface is less than 40mm then there may be no  
points
falling in that part of the bounding surface and the method will  
loose accuracy.

The other case where the calculation is less accurate is if a surface  
has more than
about 18 edges and in that case the accuracy is improved if such  
surfaces are
split into two surfaces.

And MRT does get confused if someone has defined a zone which is  
uniformally
inside-out (which can happen if someone defines a zone as an extruded  
floor plan but
puts the list of X-Y coordinates in reversed order.

If the sizes of surfaces within a zone include some with only a few  
square mm and
others with say 1000m2 then we would suggest some subdivision of the
really big surfaces.

The model that was attached to the original question does have a  
number of
surfaces that have small dimensions in comparison to the primary  
surfaces
of the zone.  An initial suggestion is to subdivide some of the  
larger surfaces,
especially the floor and ceiling.

-ESRU


-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.strath.ac.uk/archives/esp-r/attachments/20061219/48442034/attachment.html