First off you have to define "accuracy". No one in the industry has done that yet, so good luck. What is important to flow-meter performance is uncertainty, repeatability, hysteresis, and bias. A useful test should evaluate each of these parameters.
Hysteresis is the one we mess up the most on tests. For those who aren't familiar with this term it can be thought about as "the amount that the previous state impacts the present state". For example, a turbine meter takes more force to start its wheel spinning from a dead stop than it does to change the speed of a spinning wheel which gives you a variable hysteresis that is very difficult to account for in a test. For each test you have to ask if the meter is being asked to evaluate continuous flow or batch flow. For a continuous process, you are often safe just chopping the first few seconds of data off of your evaluation. For a batch process you need to figure out a way to deal with the hysteresis.
The other parameters also have to be considered in the test. Someone would have to know a lot more about your process than "Mag meter measuring flow out of a Tank".
In addition to paying close attention to the superb advice by zdas04, you should carefully consider the magnitude of accuracy that you truly require and the costs that you are willing or able to endure to accomplish your needs. You may want to consider how effectively you can use other available process or system data to cross-check the flow rate in question without incurring the problems and costs of additional testing or instrumentation.
Valuable advice from a professor many years ago: First, design for graceful failure. Everything we build will eventually fail, so we must strive to avoid injuries or secondary damage when that failure occurs. Only then can practicality and economics be properly considered.
The last item in the article is a test kit that the vendor offers that evaluates and verifies the performance of the meter tube and the electronics. The assumption is that if the performance is the same as when the unit left the factory, then the values on the factory's flow test certification are valid, having been checked with the test tool. It's a paper trail generator.
I know having a verification test done annually is a biggie in some municipal water treatment or waste treatment plants.
The test boxes are expensive and need their own certification so it's not uncommon to hire the vendor's service or an independent to do the checks.
So if you're looking for a paper trail and you've got a major name magmeter, the vendor might have a test box solution.
Siemens calls theirs a 'Verificator', ABB call it "CalMaster" (I think "Scanmaster" is software only) and Emerson incorporates verification into the meter itself, the 8714i "smartmeter verfication" option (they recommend an optional LCR meter, which is what other hardware test kits provide). Other vendors might have verification, you'd have to inquire.
