On this Thanksgiving Day I am particularly thankful for the advances in technology, low friction packing, smart transmitters, and digital valve controllers. These advances have reduced the measurement and valve position installation errors by an order of magnitude. Most of the effects of installation (process and ambient conditions and installation method and location) have been addressed. Many of these effects were previously unknown. Positive thinking may have lead us to think the installed accuracy was the catalog accuracy (performance for very controlled lab conditions). Operations and process technology adjusted set points and just shook their head when the material and energy balances did not close. We often did not find out about the errors until some advance by the manufacturer addressed the problem. Then the benefits were detailed to help sell the new measurement or valve. 99%+ of literature simply described the principle of operation and typical applications. There were some research articles, but the practical significance was often lost in the math, theory, or setup. There were a few articles that gave a hint of the problem. My list is at the end of this blog.
What you really want most often is a mass flow measurement. However this depends upon the density of the mixture. The fluid density variation not only depends upon temperature and pressure but also composition. The effect of composition can be estimated based on the pure component densities and concentrations via Amagat’s law, which works well for liquid mixtures despite being technically based on ideal gas partial volumes. The coriolis meter uses Amagat’s law for two components to provide a relatively accurate inferential measurement of fluid composition. Thus, the mass flow measurement provided by pressure and temperature compensation of a differential head, magmeter, or vortex meter can have an unknown error due to variations in process composition. Many users probably don’t realize that even the volumetric flow measurement by a differential head meter is affected by fluid density, and thus composition.
Everyone is probably cognizant of the effect of velocity profile and hence the upstream piping system and know not to put a control valve upstream of a flowmeter. Swirl is particularly detrimental. Less known is that variations of a percent or more in the discharge or meter coefficient can occur from changes in Reynold’s number and orifice edge wear. A transition to laminar flow is disastrous. The tolerance of inside pipe diameter and surface roughness can introduce several percent uncertainty. Flow conditioners, honed meter runs, flow nozzles, and venturi tubes and smart transmitters help considerably.
Upstream piping and changes in kinematic viscosity affect the vortex meter coefficient. The effects of installation on coriolis meters are typically negligible. The noise from vibration and dissolved gas has been essentially eliminated by new Coriolis designs.
Changes in process pressure and temperature can introduce an error of several percent in 1980s and earlier vintage instrumentation DP transmitters used for flow and level measurement. Changes in ambient temperature can affect capillary and diaphragm seals. Solutions are equal length capillary with the same sun exposure on the high and low side or separate smart transmitters mounted on the equipment or piping with digital computation of the differential pressure. Often not recognized are the transient errors in DP measurements that use bubblers and purged lines from the changes in process pressure that cause changes in the purge gas compression.
For additional practical information on avoiding flow measurement installation problems checkout the November Control Talk interview with Hunter Vegas titled “Retrofit Projects – Getting Flows Right”
For DP level measurements, fluid density and bubbles and hence the composition besides the temperature and pressure of the process affect the measurement. For ultrasonic level measurements, changes in the velocity of sound and scattering cause errors. Thus changes in vapor composition and temperature, entrainment of liquid droplets, and foam can be a problem. For radar, if the dielectric constant is large enough and geometry for the vessel and source installation is defined properly, the installation effects are typically negligible.
New pH electrode designs are much less sensitive to sodium ion error, contamination, plugging and the loss of efficiency and response time from premature aging of the glass from high temperature. New high temperature designs has doubled the life expectancy of the electrode and returned the response time from an hour or more to a matter of seconds. I didn’t know the response time could get so bizarrely slow at temperatures above 50 degrees centigrade until new technology solved the problem. Similarly I did not know a drift of several tenths of a pH could occur after sterilization until an electrode design essentially eliminated the drift. Smart transmitters now have process temperature compensation built in to account for the changes in solution pH from changes in the dissociation constants with temperature. Most users only know about the standard electrode temperature compensation for the changes in the millivolt potential developed by the glass electrode per the Nernst equation. Velocity errors are still largely unknown and the error introduced by changes in the activity of the hydrogen ion with ionic strength is largely ignored but quantified in Chapter 2 of Advanced pH Measurement and Control – 3rd Edition.
For temperature measurements, there are thermal errors from heat conduction from the thermowell tip to the flanged or threaded process connection, dynamic error from thermowell lags, nonlinearity error (solved by sensor matching and smart transmitters), lead wire errors, insulation errors, radiation errors in furnaces, velocity errors in high flow gas streams, and sensor decalibration errors as detailed in Chapter 2 of Advanced Temperature Measurement and Control – 2nd Edition.
For pH and temperature, non ideal mixing introduces significant process measurement errors. The concentrations and temperatures in a vessel or over the cross section of a pipe are not uniform. Very little attention is paid to this. The effect is thought to be more significant for highly viscous flows. The significant effect of composition and viscosity on temperature profile has been studied for extruders.
The effect of coatings is sketchy. We know it can be profound for pH electrodes where an almost imperceptible coating can increase the response time from 12 to 120 seconds. Similar but not as dramatic effects should occur for coatings on thermowells depending upon the conductivity of the coating. Low velocities increase the response time for both pH and temperature besides increasing the likelihood and rate of coating formation.
Finally for control valves, pipe swage and changes in available pressure drop, density, viscosity, and flashing impact the actual flow delivered. The effect of upstream piping geometry is not publicly known but suspected to be a factor since the use of smart control valves as inferential flow measurements was adversely affected. Actuator, connections, linkages, packing, internal seal, and internal clearances (a big problem at high temperatures and for particles) determine the backlash (deadband) and stick-slip resolution and sensitivity) in the response of the internal closure element (plug, disk, or ball). Smart digital positioners can solve most of these problems if the feedback measurement is actually the internal closure element position and a diaphragm actuator is used. Flexible tuning methods with pressure and rate feedback can improve precision. However, rotary valves with an on-off heritage and/or a scotch yoke or rack and pinion actuator have inherent problems not solvable by a positioner.
Manual on Installation of Refinery Instruments and Control Systems, Part I – Process Instrumentation and Control, API RP 150, 2nd Edition, March, 1965
Loeffler, R.F., “Thermocouples, Resistance Temperature Detectors, Thermistors – Which?”, Instruments & Control Systems, May, 1973
Anderson, R.L. et.al., “Limitations of Thermocouples in Temperature Measurements”, ISA, 25th International Instrumentation Symposium, May, 1979
Richmond, D.W. “Selecting Thermowells for Endurance and Accuracy”, InTech, Feb, 1980
Demorest, William J., “There’s more to transmitter accuracy than the spec”, Instruments & Control Systems, May, 1983
Behrmann, W.C., Thermocouple Error due to Sheath Conduction”, InTech, Aug, 1990
Stichwey, L., “Gas Purged DP Transmitters for Liquid Level and Flow”, InTech, Nov, 1992
Yazbak, Gene and Diraddo, R.W., “An Inside Look at Extrusion Melt Temperatures, Plastics Technology, May, 1993
Sanders, Fred, “Watch Out for Instrument Errors”, Chemical Engineering Progress, July, 1995
Ginn, Peter, L. “It’s the Physics!”, InTech, Feb, 1996