Default inner convection coefficients

[PGorzalka]

What is the problem?

• According to ISO 6946, convective surface coefficients are 5.0 m²K/W for upward heat flow, 2.5 for horizontal and 0.7 for downward. In TEASER, inner_convection values default to
  • 2.7 for walls: reasonable because close to horizontal
  • 1.7 for ground floors: somewhat reasonable if basement exterior walls are considered ground floors (which is afaik not the case in archetype buildings)
  • 1.7 for rooftops: maybe I miss a point, but this seems to be a mistake.

Why do we want to solve it?

• The current implementation works fine with the archetype building parts: The layer specifications there seem to be designed to meet the U-values specified in the typologies considering the current convection coefficients. However, a model built from it will overestimate radiation vs. convection for roofs and underestimate it for ground floors. Furthermore, custom layer specification may lead to unexpected results.

How do we want to solve it?

• Correct all default values
• Adjust layer specifications of rooftop archetype building parts (this is a big amount of work which I don't think I could do)

[mlauster]

We defined the upward and downward coefficients similar, because we can't determine if the heat flow is upwards or downwards. This depends on the temperature of the adjacent environment, e.g. ambient or another thermal zone. So a roof can have a downwards oriented heat flow, if the room above is heated to a higher temperature. Anyway, we never found a good solution to handle this, looking forward to your approach! :-)

[PGorzalka]

That definitely does makes sense. However, if one assumes the archetypes e.g. of TABULA are defined for "winter conditions", the simulation results are not consistent with them regarding "summer conditions" in the current implementation anyway, right? If that is the case, I would say changing the "inner convection" attributes of the archetype coefficients to 0.7/2.5/5.0 and adjust the layers would be "more right" (but would not really have an impact on simulation results).

Nevertheless, your comment makes it an AixLib issue rather than a TEASER issue because a solution would imho require using the convection module with changing coefficients AixLib.Utilities.HeatTransfer.HeatConvInside and include the possibility of user-specific coefficients there (so you can have other coefficients than 0.7/5.0 for the case of lumped elements). In TEASER, inner_convection would have to be converted into a two-element list for both cases of air being warmer or colder than the surface.

All in all, a rather complicated solution for a problem with a probably very low positive impact on the results, or even a negative impact because calculation times will be longer.