How to compensate for asymmetric flow in a square duct?

[hollerg]

I am trying to establish a defensible basis that a sensor has a +/- 2% installed accuracy in a square/rectangular duct, with either my existing meter or a new multipoint meter .

So far I haven’t found data validated L/D requirements for flow meters mounted in rectangular ducts for my complex geometry. (Even ASHRAE data is often only "loss coefficients" not profile data)

I've also had no luck finding validated rectangular duct data on converting a highly asymmetric flow profile to a flatter profile, suitable for measurement with a multipoint array, such as the Ebtron meter. (No source data listing for the L/D recommendations.)

I will admit my experience is with instruments in circular ducts/pipes but given the data and practices of AGA and ISA, I wonder if the L/D requirements quoted for multipoint arrays in rectangular ducting are too optimistic.

IF someone has links to such data, or a referral to a rectangular duct flow expert (wind tunnels?), I would be grateful.

Lastly, it has occurred to me that if I used elbows with turning vanes, would I consider the total straight run as being nearly the same as one straight run? Any idea how poor an approximation that is?

Thanks in advance for your help
Gary

Background:

I performed PITOT traverses @ the meter location for the current configuration using the T-rule:

Centrifugal blower (side outlet @ bottom elevation,
-- a divergent section (~8 deg half angle),
-- a straight section 6D (equiv) of 18x18 duct (meter 2.4D upstream 3.5 D downstream)
-- a double elbow section. The elbows (R=1) form a U w/ 1.5D between them, directing the duct back under the blower
-- a final straight section of 6D before I enter the process equipment

Five sets of 5x5 PITOT traverses established that the profile at the current meter location is highly inconsistent. Sometimes flow hugged the duct so tightly the traverse (per the equal area method) misses the highest flow against the wall. Other times I see the velocity is relatively flat along one axis and parabolic along the other axis.

The vendor for the thermal "mass" meter (FCI) and a competitor (Kurtz) both agree the current location is a tough place for a meter. They think moving it to the last straight section and calibration from traverse data is more than adequate to make a single point meter accurate. However my Pitot traverse makes me believe they trivialize the impact of square/rectangular duct geometry, especially in light of the fact both would recommend installing flow conditioners to reach rated accuracy, if it were circular duct.

Even with the sensor in final straight run, I fear that the skewed flow, the short straight runs and the double elbow will generate a profile too inconsistent to establish a +/- 2% accuracy, without some other mechanical change

 

[Illmatic1]

, such as use of flow conditioners. I am trying to establish a defensible basis that the sensor has a +/- 2% accuracy in this ducting, with either my existing meter or a new multipoint meter.