Meter mayhem, I thought to myself. One magnetic flow meter showed 18 gal/min, another read 138 gal/min. However, the magnetic flow meter monitoring the combined flow to a reactor indicated 380 gal/min. Okay, it’s not a mass balance but the temperatures only were 60°F apart. The total flow should be 156 ±3 gal/min. Other flows did go into and out of the feed tank but all were for recirculation except for the one that went to the reactor. The pump for the recirculation had a design flow of 900 gal/min; some people still think you can agitate a 40,000-gal tank with a pump.
Obviously, most of the flow from the large pump went for recirculation, with only about 156 gal/min sent to the reactor. The errant measurement pointed to an instrument problem: magnetic flow meters read high when they become plugged. As I pointed out in a previous column (“Match the Flow Meter to the Service”), flow meters rely on inference: magmeters measure velocity for an assumed density and then infer flow rate. Plugging restricts the pipe diameter and increases the velocity, at least according to the mass flow equation: Mass = Density × Velocity × Area.
Another indication of a problem with that meter was its trend line. The trends for the 18-gal/min and 138-gal/min meters jumped around. However, the trend line for the large flow meter almost was flat. Often — but not always — that’s a sign of a dead instrument. Sometimes, though, it stems from over-tuning to avoid oscillation of trend lines; that oscillation is a pulse.
Generally, level and temperature measurements don’t have a pulse. If a level measurement is leaping around on a control system faceplate, it usually indicates an electrical problem, not an instrument one. Temperature measurement coupled with flow, such as what you’d see in a furnace or vaporizer, can have a pulse; the trouble is that oscillation from the flow measurement almost certainly will overwhelm the temperature fluctuation.
An excessively high trend in temperature measurement obviously could result from a loose connection (high resistance); this applies to either thermocouples or RTDs. This probably will stand out from the background noise caused by flow measurement oscillation.
Thermocouples fail high, so do RTDs although sometimes the resistance is only somewhat higher than normal. Temperature sensor probe failures often are pretty spectacular (in the sense that high temperature may lead to fire, explosion, reactor breach, pump failure or other catastrophic and unpleasant events).
One good technique for diagnosis is to use the instrument itself. Compare the pump rates into and out of a vessel against the flow calculated from the level trend data. For example, I measured levels of 20.6% and 26.8% only a minute apart for a tank that needed 308.7 gal to alter the level 1%. So, the indicated level change would have required a flow of 1,917 gal/min — stupendously more than the 150-gal/min pump ever could provide!
This wasn’t the only time a level measurement let me down. On one assignment, several hundred pounds of material seemed to appear and disappear in seconds on a faceplate in the control room. The mass measurements were made from level readings. I conducted a level measurement experiment and then looked at the trend data I downloaded: the data showed that nothing could be gleaned from any comparison until the level transmitter gain was dampened.
Diagnosing problems with control valves often isn’t too difficult. Control valves are considered unreliable when they don’t do what they’re supposed to do. Look at the response time between the command and the response: a good valve opens quickly while a bad valve sticks, causing the controller to read the failure of the process variable (PV) to respond to change as a need to open the valve further. When the valve finally does spring open, the controller sees the change in PV and tries to crank the valve closed. Because the valve is open now, it slams closed and the process begins again. Often, an engineer will dive into this kind of problem thinking it’s a tuning problem and not a hardware problem, being misled by focusing only on the controller and the measuring instrument.
As I’ve said before, don’t believe information unless you can compare it to other data. Take the time to take the long view.