Models and Testing

Posted in real world problems on Friday, October 02 2015

About this time last year I wrote up a post on storage tank venting and how to size a gooseneck. Well I was called upon recently to collect my notes and calculations into a guideline to go with my code and it forced me to reevaluate the assumptions I baked into my calcs.

The main thing that I've been thinking about is the choice of model when thinking about the vent. It is tempting to model the tank as it actually is, with the real fluid that will be inside and the temperatures and pressures it will actually experience, and model flow through the vent based on that. But maybe this is wrong headed?

Ok let's re-wind. There is an industry accepted set of models for the required venting from a tank, with industry accepted safety margins baked in, and those models are in API 2000. The code both has the models for tanks (in section 3) and a standard for rating vents (in section 5) and I think that is important.

The first part of the code is trying to take every special snowflake of a tank + process fluid and math it into the flow-rate you would need through an air tank in a testing shop. The second part is about standardising the testing apparatus to rate those flowrates through a vent. That tank on the test bed is the common ground between all the different special snowflake tank systems and all the special snowflake vent systems. At the end of the day the question you are answering is not "How much vapour will vent from my tank?" it is "What is the rated size of a vent that can vent my tank?" and those are very different questions. The confusion arises because the rating is in terms of a flowrate ($\frac{Nm^{3}}{hr}$).

So fast-forwarding to the gooseneck. We know our vent is destined for a specific tank, but I think we should pretend we don't actually know that. We are building a mathematical model of the tank on a test bench and our vent attached, it is our job to math our way into the rating the vent would get were we to plop it on an API 2000 compliant test apparatus. With that logic in mind we pretend our vent is atop a tank full of air at about ambient conditions and determine its rating and flow-capacity curve using appropriate engineering assumptions. (That these vents might end up being field constructed and not actually tested is another ball of wax entirely, and worthy of some serious chin-scratching and gum flapping with engineers and the ultimate owner of the tank).

Two things I've gone back and forth on, assumptions wise, is whether or not one should assume the air outside the tank is "Normal" (i.e. at 0C and 1atm) and whether or not entrance effects should be taken into account for the vent.

For the first case I say yes and mostly as a matter of convenience. For one the ambient pressure exiting the vent being 1atm makes sense, for two assuming normal flow at the exit and back calculating a different pressure and temperature at the entrance is just easier. You could try to more closely model the test set up, where the flow meter is immediately after the blower and before the actual air tank, but I don't think you would gain anything in that analysis. Any slight gains in accuracy would be washed out by the inaccuracy in the pipeflow correlations (I would imagine).

For the second case I'm more on the fence. The code makes great pains to have the vent isolated from entrance effects when being tested, so you test the vent itself and not the interface between the vent and the tank. Presumably the entrance effects are taken into account in the tank modeling equations(?) and including them in testing would "double count' it. On the other hand, taking into account entrance effects as just being part of the vent (esp. when you know it is going to be just plopped right on top of the tank) is more conservative, it errs on the side of safety and really nothing is lost by doing this. It's not like the sky will fall if you make your goose-neck slightly larger than necessary.