Figure 1 summarizes the nature and applicability of these measurement technologies. Figure 2 gives more details on their use for continuous measurements. Impulse line applications have not been considered for main process applications but can still find use on general services and less critical installations.
Of course, besides technical suitability, it's important to consider economics. Typical comparative costs, from lowest to highest, are: conductivity → capacitance → tuning fork → hydrostatic→ displacer → ultrasonic → load cell → radar → nucleonic.
Selection also must consider both the process and its control.
Process. It's essential to understand the physical property variations of the process fluids and the phase changes that may occur within the process during normal and abnormal conditions.
Boilers, flash vessels and distillation column bottoms involve boiling liquids, resulting in noisy levels. Displacers in external cages frequently are used on steam generators and flash vessels, provided the process fluids are of low viscosity and relatively clean. The non-contact nucleonic method will prove most reliable for distillation column bottoms, where reproducibility is more important than absolute accuracy. While expensive, it can be more than justified given its value in providing stable column operation and in preventing reboiler fouling due to loss of level.
Avoid the use of impulse lines in level systems if the process pressure varies and there's a tendency for solids' formation due to freezing, precipitation or polymerization. Purging the lines with inert gas or process compatible fluids will have limited success and is high maintenance.
Nucleonic level detection provides a powerful tool to perform on-line process diagnostics. Typical applications include monitoring level profiles in tray towers, distribution in packed beds, locating level build-up and blockages in vessels, and general flow studies.
Control. Let's consider a general equation describing the output, m, from a three-mode (proportional-integral-derivative) controller:
m = (100/P)[e + (1/Ti)∫edt + Td (de/dt)] + mo
where P is proportional band, %; Ti is integral action time, min.; Td is derivative action time, min.; mo is steady-state controller output; and e is ±(Xset – Xmeas), the error between set point and process measurement.
Based on its form, we can predict the following behavior:
1. If there's no error the controller output will equal steady-state output, mo.
2. Controller gain is 100/P. So, increasing P decreases the controller gain with % change of output for same % error change reducing and vice versa.
3. The integral term, 1/Ti, indicates that as Ti rises its effect falls. An increase in error results in an increase in rate of change of controller output. Slow processes can use higher Ti, provided the process isn't too slow to absorb the energy change — if it is, cycling will result.
4. Decreasing the derivative term, Td, reduces its effect. Increasing error rate change increases % controller output change. In typical continuous process applications liquid level measurements are noisy; they present rapid changes in error with time, i.e., large de/dt. So, derivative mode never should be used — otherwise equipment damage may occur.
Continuous process applications often rely on surge vessels to minimize flow upsets to downstream units. The level is allowed to float between minimum and maximum values. Use proportional control mode alone with flow cutback override control.
Controlling level at a fixed point, such as for distillation column bottoms, requires proportional and integral control modes.
High integrity protection. For a level measurement deemed critical for plant safety it's common practice to install two or more redundant level systems. Redundancy implies elimination of the likelihood of a common mode failure, which can result when using identical methods, instrumentation and manufacturer.
Inherent in high integrity protection is the principle of fail-safe design. However, the total system needs in-depth study to determine the potential of fail-to-danger scenarios and to ensure testing facilities and procedures are acceptable.
Frequency of testing for satisfactory operation can dramatically impact system reliability. Unfortunately, conducting real on-line testing of level instrumentation generally is rarely possible because creating the process condition required, e.g., high level in a vessel, isn't feasible.
JOHN E. EDWARDS is a senior consultant with P & I Design Ltd., Stockton-on-Tees, U.K. E-mail him at firstname.lastname@example.org.