Gage R&R on valve pressure loss


Hello all,

I have a test rig designed to measure the pressure loss across valves at various flow rates. An upstream flow control valve slowly closes over the course of several minutes and a flow meter records the flow several times a second. At the same intervals, two transducers, one upstream and one downstream of the valve being tested, record pressure and the difference in pressures is calculated to determine the psid across the valve at a given flow rate.

We are determined to complete a Gage R&R study for this equipment , but we are running into some issues/questions. Due to some hysteresis in the valves being measured, there will be some variation in the Differentia Pressure at a given flow rate, even with a completely perfect measuring system. If you run the test on the same valve hundreds of times, you get different results. The problem is that the difference between different valves of the same batch in generally small enough that the resulting variation in DP is masked by the random variation of each unit. We thought about mitigating this by running the test on fixed size orifices, but we are struggling with orifice size selection. If we pick orifices that are too similar in size, the system will not be able to tell the difference. If we pick greatly varying orifice sizes, the system easily can tell the difference and virtually 100% of the variance is from part-to-part. It seems like with the orifice method, we can essentially choose if it will pass or not by our size selection.

Another question we were having is which DPs to compare since it is collection about 100 datapoints per valve. We initially thought about doing one high and one low flow rate and doing separate studies, but we also thought about considering the DP at a different flow rate to be the "part". With that method, we are running into the same issue of the flow rates we choose determining how different the DPs are. If we choose a flow rate that is very close, the system wouldn't be able to tell a difference in DP. If we choose greatly different flow rates, it can easily tell the difference in DPs.

This is frustrating because common sense is telling us that the system is accurate. When the data points of several valves are overlayed, it shows a pretty clear system pressure vs. flow curve.

Any help is appreciated!

John Predmore

Trusted Information Resource
Welcome to the Cove.

A GR&R is part of a measurement study; as such, it is a study of the measurement system, not the parts themselves. To demonstrate measurements are repeatable and predictable, you generally need conditions which are repeatable and predictable for the study.

If you don't want hysteresis to be a variable in your measurement study, you experimentally constrain your measurement study to only go from large flow to small for your study. If a flow at a fixed flow rate takes several seconds after each transition to reach equilibrium, you can define your measurement routine to always wait the same fixed time before taking readings. Once accuracy is demonstrated under unchanging conditions, your engineers will infer limits of accuracy of instantaneous conditions inside a dynamic situation.

Turbulent fluid flow is inherently a chaotic phenomenon, which may never settle out to a state where instantaneous readings are stable. So demonstrate repeatable measures with laminar flow instead. I once worked at an auto parts factory that did a lot of airflow testing. We fabricated fixed orifice master parts by sealing all leak paths, then drilling one hole of known diameter in the throttle plate. In this way, laminar flow was stable for the purpose of flowtest, and flowrate was calculated as a function of orifice size. Create a family of similar master parts with orifices that span the range of flow rates your gauge typically encounters.

I hope you find something helpful here. Good luck.


Forum Moderator
I feel for you. I have run into a similar situation several times when the product to be tested is mechanically complex and friction causes a lot of within product variation. So far, I have been able to substitute a mechanically simple equivalent for the product in order to focus on the measurement device. This substitute is different depending on what is being measured. For example, if you are measuring force, substitute a spring, etc. Your application appears to be pretty complicated, so I wish you well.
Top Bottom