While it’s commonly known that calibration management software (CMS) can assist overall instrument maintenance efforts, many may not realize how the software can be used to bring the maintenance of calibration management itself into a predictive mode.
Here’s an actual example of how that could happen. The case is based on deviation analysis using data from a Midwest refinery.
A technician conducted a five-point test of a transmitter configured to operate 0-60 in water and 4-20 mA out, with a +/- 2% recalibration specification (See Figure 1). A graph of the test shows the error at each test point; deviation around the 0% test point (A) represents possible error due to hysteresis.
Figure 1
A drift plot shows the maximum deviation for multiple tests of this tag over time. The green and orange points show a change in Pass/Fail specification from 1% to 2%. This occurred between 2003 and 2004.
Figure 2
In Figure 2, the test referenced above is represented by the blue arrow (C, see figure 2 above)in the drift plot; a technician re-tested the tag the next day as represented by the purple arrow (B, see figure 2 above).
Using statistical analysis, it’s possible from this data to plot a least-squares fit line in order to extrapolate the next time this tag will drift out-of-spec. Alternatively, Weibull distribution with left-censored data could predict, with quantifiable certainty, the next out-of-spec condition.
“It is possible, using statistical analysis, to plot a least-squares fit line to extrapolate the next time this tag will drift out-of-spec,” says Ed Shuler, senior applications engineer with Honeywell Process Solutions (www.honeywell.com/ps), adding that a Weibull distribution with left-censored data may be still better to predict “with quantifiable certainty the next out-of-spec condition.”
