Q: I wish to know about pH measurement and control systems for a biphasic (organic-plus-water-based) reaction. The slurry is a salt in the organic phase. The salt is tiethylamine HCl, and the solvent is EDC.
Is it possible to sample material and measure externally. How do we do it online?
The second issue is this: If we measure the pH and control by an external loop, the measurement may be different from what is actually in the reactor.
Can we have a pH measurement system mounted from the top of the reactor inside?
A: It is possible to detect pH inside a tank, but only in highly agitated vessels, because if the velocity is under 1 ft/s, it will drastically slow the measurement response because of the increase in the boundary layer on the glass. Maintenance of these probes is also difficult, although you can reduce the frequency of probe removal for cleaning and recalibration by using automatic cleaners. (See Figure 1).
I prefer measuring the pH in pumped recirculation lines with the probe inserted in a flow through sight glass and provided with automatic cleaner for keeping it clean (See Figure 2). You will find a detailed treatment of pH control by D.L. Hoyle, G. K. McMillan and F.G. Shinskey in the Instrument Engineer’s Handbook, p. 2044, Vol. 2, 4th ed. The available pH detector probe designs are described in Volume 1.
A: A pH measurement can be made of a two-phase mixture, but it first must be separated by decanting so that the electrodes are immersed in only the aqueous phase. It also can be made in a polar organic solution, such as an alcohol. The reference electrode may have to be replaced by a carbon or platinum electrode if the organic solvent contaminates it. The system will have to be specially calibrated, as the ionization constants for water do not apply.
pH is often measured by inserting electrodes directly into a stirred tank. However, maintenance and calibration of the electrodes is awkward if the extension pipe is long and rigid.
Q: Suppose a controller scans very fast (is not very busy), and the analog measurement of the PV of a PID loop is done with good equipment and good practices.
Under those conditions, is there ever a reason to lower the frequency of execution of the PID execution (in other words, scan the PID instruction less often by increasing the loop update time)? Are there some types of loops that will always perform better if not scanned too often, such as perhaps a jittery PV or a PV that changes rapidly and often?
A: Well, the scan period increases dead time, so your question can be reframed by asking, “Is increasing dead time is ever helpful?” That is like asking, “How frequently should you open your eyes to best observe the surroundings?”
The shorter the scan period, the better the control—but there are practical equipment limitations, such as overloading the computer or finding that the final control element is not fast enough. Increasing sampling speed does no good if the measurement is intermittent.
A: I cannot think of a process-control reason to scan slower. Sampling adds half the sample interval of dead time to the control loop—never a good thing. If things are jittery, then use a filter.
However, I think, scanning slower is helpful if:
- The measurement is sampled as in an analyzer (refer to Greg Shinskey’s process control book). Match the scan time to the analyzer sample time
- The control equipment has numerical problems with a fast scan—not a good situation.
- Scanning fast causes loading problems with the controller.
A: The only time when it is helpful is in digital control systems, where a cost is associated with processor time, and the user has control over those tradeoffs. Honeywell’s Process Manager controllers provide a budget of processing resources. Each controller sharing a time slice within the processor will consume “processor units,” more units for a slot configured to execute more frequently. The total number of control loops accommodated depends upon their execution times, so increasing execution time increases cost per loop. It is still very important to set processing frequency sufficiently fast relative to the process time constant plus dead time.