Scenario: A guy walks up to a wind engineer and says….
I just got off the phone with a grad student at the local university who was breathless with good news for me – with the new wind deflector he’s just designed for my roof-mounted solar racking system, the wind actually pushes the panels down into the roof! I won’t need any ballast at all!
Is that so?
Yes, when the student first observed this result in his computer CFD steady-state simulations, he wasn’t sure it was real. So, with his advisor’s help, he put a full scale prototype of my new 9° tilted rack in the university’s high-speed aerospace wind tunnel and measured the lift and drag forces. They were even able to test at 90 mph, the design wind speed in my target market. “The tests confirm what we saw with the CFD,” the student told me. “The panels were pushed down into the floor of the tunnel.”
So now it looks like I’m ready to win jobs and begin installations on commercial roof tops across the country! What more could I need?
Well, a PE stamp, for starters. And no knowledgeable PE should stamp a design based on these results.
First, not all wind tunnels are the same. Aerospace wind tunnels, like the ones used to test cars and aircraft, produce “smooth” wind at a constant speed. Boundary layer wind tunnels create gusty, turbulent wind. Wind flow near the edge of a building roof is nothing like the smooth wind along the floor of an aerospace wind tunnel. In fact, the test lab would have done better to mount panels on the wings of a fighter jet, and then place the whole jet in the tunnel in takeoff position. This is because powerful swirling winds, or vortices, form above the corners of flat roof buildings, the same vortices that provide so much lift for delta-wing aircraft. If you don’t test the panels under these vortices, you have no idea how much lift the panels will see near the edges of the roof. This is especially true of panels with slopes of 10 degrees or less that feature wind protection deflectors.
Later, the same guy walks up to the same wind engineer and says…
“My PE knew that rule you mentioned (apparently, most do), and refused to OK the use of the downforce idea in the design. But I think my project is different…
You see, I’m bidding on a big job – several distribution center roofs, 1000 ft-wide nearly square boxes that are only 35 ft tall. Out in the center of such a large roof, far from the edges, there are no vortices, and the wind flows along the roof surface. Perhaps I can use the test results there, and will only need ballast for panels near the edges of the roof?
Well, there are three real problems with using the aerospace tunnel results for such a “middle of the roof” test. All three problems are the result of ignoring the critical nature of wind turbulence. Aerospace wind tunnels go to great lengths to prevent or remove turbulence (or eddies), since they need smooth flow to test their aircraft. Turbulent eddies are the dominant feature of the atmospheric boundary layer in which your panels will sit, so ignoring them is a bad idea.
To understand the first problem with a smooth flow test, we need to discuss what the “90 mph design wind speed” implies. This speed is specified in your local IBC code, and it is the wind speed averaged over 3 seconds. A 3 second gust is a very big eddy, maybe 300 feet across, much bigger than your panels. During this 3 second span, your panels can expect to see shorter bursts over 110 mph which will fully load the panels.
Does that mean I should retest my model at 110 mph?
No. Running the tests at 90, 100 and 110 mph will not yield any new information because for a square shape like a solar rack, the lift and drag coefficients are independent of wind speed over this range of speeds. The lab could have saved energy and money by running the tests at 40 mph. Having 110 mph gusts does, however, mean that the loads during a real world design wind event will be over 50% higher than what the university reported, because the aerospace tunnel flow did not have any of these shorter, faster gusts.
But that’s just 50% more downforce! Even better, right?
That might be true if eddies did nothing more than just speed up and slow down the wind speed. But they don’t. They buffet objects in their path from all directions, pushing from side to side and up and down. Even if the panels are being pushed down into the roof some (or even most) of the time, they will certainly not be pushed down into the roof all of the time, but they certainly need to stay on the roof all of the time.
The third problem with a smooth flow test is the air flow around the object can differ fundamentally between smooth flow and turbulent flow, and the smallest turbulent eddies influence the nature of the wind wake, and the wake in turn can determine whether the panels stay in place.
The importance of turbulence is the reason why the ASCE Manual of Practice for Wind-Tunnel Studies of Buildings and Structures stipulates that all important eddy lengths be modeled in the approach wind, and provides guidance for how this can be assured during a boundary layer wind tunnel study. This is the only kind of study explicitly permitted by the building code and it’s the kind of study used to determine all of the wind pressure values in the building code.
But what about the CFD? Couldn’t the CFD be modified to include turbulence?
Unfortunately, ASCE 67 only provides advice for wind tunnel testing, not CFD. There is currently no equivalent to ASCE 67 for CFD. If there were to be such a manual, however, it is likely that the first recommendation in this situation would be to run an unsteady simulation (preferably a large eddy simulation) so that peak forces could be measured. The grad student at the beginning of our discussion ran a steady-state CFD simulation, which is by far the most common, quickest, and cheapest computer simulation. A steady-state simulation predicts average wind forces. Unsteady simulations are much more time consuming, require more computing power, and are more expensive. But they come closer to modeling the environment you’d get in a boundary layer wind tunnel, and a boundary-layer tunnel environment is as close to a real-world simulation as there is right now.
The CFD discussion also raises an issue important enough to merit its own rule. The grad student only simulated one wind direction. Just like the roof itself, the wind loads on tilted panels can be worst for cornering winds.
Okay, I’m starting to get the picture. Any other issues I should consider?
There are. For example, how will parapets of various heights affect the wind loads? In our experience, parapets reduce loads for some designs and increase them for others. How will the panels fail? Will they tip over, or slide along the roof as a group? If they move as a group, how big does the group need to be to avoid sliding? In our experience, this group size will vary depending on the array position on the roof, the height of the building, and of course the shape of the panels.
We’ve talked to many imaginative people who talk about the various ideas they came up with to determine the ability of their design to resist winds. “Put the panels on the back of a truck and take it out on the interstate,” for example. We’ve seen video of panels arranged behind a turboprop, with winds over 100 mph blowing across the panels. Such tests demonstrate something (the ability for small planes to take off in front of a solar array, perhaps?) but they don’t demonstrate how the panels will behave on the roof of a building.
The knowledge and methods that wind engineering experts the world over use to determine wind loads have been developed over the past 50 years. It is likely that the building where your panels will be placed has been designed to these standards. If it has been designed correctly, the windows will not fly off the face of the building even if the 500-year wind storm hits it directly. The solar panels on the roof should have the same level of safety and reliability. The success of your business may depend on it.