Modelling translucent plastic layers in WINDOW

[Joe Roberts]

Hi,

I would like to model a series of layers of translucent plastics with translucent insulation, in order to calculate the overall g-value of the build up. If I have the typical properties of each layer (Tsol, Rsol1, Rsol2 etc), can I just create user made items in the "Glass library" then build up my roof from these layers to get the overall properties of the system?

Thanks!

Joe

[Collin Robinson]

Joe, I'm curious what comes out of this discussion. In the past I have relied on the manufacturers for these measurements because I didn't know how to do it in Window. It's a bit unlike the other elements in the IGDB. I only see a couple of translucent glazing layers materials and they are frits. I'm not sure there is translucent plastic facing boards similar to what some manufacturers use, for example. Curious what you find here!

Collin

[Jacob C. Jonsson (LBNL)]

Hi Joe,

What you are describing is tricky, and combined with that the software will give you an answer if you do what you describe it easy to assume that the answer would be correct, but I'd not recommend it. The short answer is that having the whole package measured by a manufacturer is the safest approach. I'll give a paragraph on each challenge with the calculation and data (but can go on and on if encouraged) and why the result will probably be wrong and why I cannot really estimate how wrong. Depending on how much time you have to spend on this there is a lot of testing you can do to experiment and see what kind of g-values you get when you turn different knobs in the data. 

Problem 1: Do you have gaps between the translucent plastics and the insulation? Insulation comes in many forms so contact between the layers could be anything from almost nothing to perfect. Optically layers interact differently if there is air between the surfaces or not (even for a non-scattering  example of glass+pvb+glass laminate behaves quite differently compared to glass+air+pvb+air+glass). If there is a gap you are more likely to get reasonable results from WINDOW than if they are touching. This also ties into the thermal conductivity of the problem, it may or may not be more correct to calculate a single effective conductivity for the whole package if there are no gaps than to model gaps where you would have radiative heat transfer between the layers.

Problem 2: A lot of plastics have substantial absorption in the near infrared range. If the Tsol and Tvis for a product are different you need spectral data rather than just integrated values to do the multilayer calculation correct. The larger the difference and the more layers the more wrong doing multi-layer calculation with Tsol will be. 

Problem 3: How scattering are the samples? The most recent versions of WINDOW 7.8 allow for user created diffuse material and layers to be imported. But spectral data is required for that. Using that approach allow you to "bound" the problem making one version with no haze and another with 100% haze. Impact of this depends on number of layers, how reflective the layers are, and how scattering they are.

Maybe something else I am forgetting, and even though I did try to limit myself it turned into a bit of an essay.

Best,

Jacob

[Collin Robinson]

Jacob, thank you for this amazing response to Joe's question. I just wanted to share transmission spectra for some plastics and glass, from Optics. I didn't know plastics have NIR absorption. - Collin

[Jacob C. Jonsson (LBNL)]

Collin,

You managed to highlight how my brief generalization is not that air-tight. Regular glass has broadband NIR absorption as well and when you go to 10mm thickness the tables have turned. I added low-iron for comparison showing how little that absorbs in NIR.
Numbers from Optics:

Record, Tvis, Tsol, Absolute difference

3mm (larger difference for acrylic)

Clear_3.dat, 0.899, 0.834, -0.065

Acrylic_3.dat, 0.923, 0.849, -0.074

Starphire_3.vta,  0.914, 0.902, -0.012

10mm  (larger difference for clear glass)

Clear_10.dat, 0.860, 0.685, -0.175

Acrylic_10.dat, 0.915, 0.803, -0.112

Starphire_10.vta, 0.907, 0.872, -0.035

[Collin Robinson]

Jacob, 

Interesting about the difference in clear vs low-iron glass, and thick vs thin. I updated the graph sheets.

[Collin Robinson]

Joe and Jacob,

One of my projects started as transparent insulation sloped glazing and changed to glass. The glass had a translucent interlayer and a frit pattern, a custom makeup. It was an unusual beast! Installed over a large area it is delivering very satisfactory results, however. Much to my relief. During design, I was able to get the Tvis, U-factor and SHGC for my makeup in Window to be pretty close to what the manufacturer's techs stated for this. I trusted their numbers, a major manufacturer. It was helpful because I could make a BSDF file and use it in daylight sims and renderings to approximate the beam spread of the solar disk. Based on what Jacob said, however, I now question especially my thermal results. See attached for that makeup in Window.

Collin

[Jacob C. Jonsson (LBNL)]

Collin,

Thanks for sharing that example as it shows how the specifics matters a lot. For that project you have a very strong solar control as the first layer (Tvis 0.2, Tsol 0.08). When so little energy gets through the first layer, the impact of poor approximations/errors behind it are scaled down (the Tsol of 0.08 will severely reduce the SHGC) .  If that final layer is 80% coverage with Viracon V175 white frit I think the numbers are a bit too high, but it cannot have too much impact based on the more exterior layers so it doesn't matter much. And that is one of those "bounding" experiments I like, if you replace layer 4 with first a low-iron 3mm piece of glass, and then a 10mm generic bronze, you could expect that the answer is somewhere in-between and depending on if there is a large range or small range between the edges you have to worry more or less about your data.

For U-value I have less experience so that I do not see anything wrong in that model doesn't mean there aren't issues.