The impact of component variation in the manufacturing process on variable area (VA) flowmeter performance

Abstract

This paper records a series of flow tests of predicted variable area (VA) meters taken from the production line, to ascertain the effect of manufacturing variation. In order to test the effect of variation of one parameter with others fixed, a set of weights was made which could be inserted in the cut out hole in the top of the float. With this additional weight of the float, an under reading was measured as predicted from the theory. It was also possible to simulate the effect of a misplaced scale. The resultant error was as expected. From the initial tests, it was noted that the amount of variation in the VA meters was small and the effect of this on flow rate might also be small. It was, therefore, decided to measure the effect of different floats with one tube. This procedure was followed for two tubes and for five floats. Tests of random combinations of float and tube are then shown to be well within predicted uncertainty of 2.5 percent FSD. Due to the unsteadiness of the float at high flow rates, it was very difficult to set the flow with confidence. This resulted in an uncertainty of order ±1 mm or more, which is equivalent to about ±0.4 percent. Allowing, in addition, for an uncertainty of between ±0.3 percent and ±0.5 percent depending on the position on the VA tube, the total flow uncertainty was about ±0.6 percent. This could have accounted for the scatter of most of the experimental data. In conclusion it was noted that: • the position of the scale was clearly a dominant cause of manufacturing variation, and while this was unlikely to put the meter out of manufacturer’s specification, it could cause an error of order 0.4 percent for a scale displacement of one millimetre; • the weight was, probably, the next most critical dimension of the manufacture of the VA meter; • the float diameter was likely to result in the lowest manufacturing variation; • the difference in the internal diameters of the glass tubes might have caused some bias in the results. The control of dimensional variation should allow a tight specification for predicted flowmeters.