«

»

Nov
08

Can Measurement Delay be Helpful?

A delay of any sorts in a control loop is thought to be detrimental. This is true for an unmeasured load upset because the integrated error for a controller tuned for maximum disturbance rejection is proportional to the deadtime squared (see slide 12 of my ISA Automation Week 2010 Presentation)

How-to-Get-the-Most-Out-of-Your-PID-Slides.pdf.

I can’t think of a scenario where final control element delay is anything but a drag on control system performance since it delays the controller’s feedback, feedforward, and setpoint response correction actually getting into the process. The predominant sources of delay for a ramping controller output is a variable speed drive 8 bit A/D input card and the stick-slip and backlash of an on-off valve posing as control valves. These sources of delay cause other problems such poor sensitivity and resolution and limit cycling.

This might lead one to think delay is always bad, but “never say never.” There are some cases where measurement update delay can be helpful, which is good news if you want to take advantage of wireless measurements.

An increase in measurement update time interval (wireless “default update rate”) can increase the signal to noise ratio for column and vessel level and temperature control. Since the change in level and temperature is often very slow, the actual true change in these process variables in a fast update time (a few seconds) is often less than the noise in the measurement. Common initiators of noise for level are turbulence, bubbles, foam, and pressure fluctuations and backlash for mechanical devices. Common sources of noise for temperature are imperfect mixing and bubbles for all temperature sensors and thermal noise, electromagnetic interference, and poor repeatability for thermocouples. For all measurements, the true change in the process variable must be greater than the sensitivity limit and resolution of the measurement. In older DCS, wide range thermocouple cards with a 12 bit A/D caused a resolution error of about 0.25 oF, which is larger than the temperature change seen over a period of 5 minutes in many large vessels and columns.

Even more impressive is the improvement in setpoint response and feedforward control that is possible with a large measurement delay when a smart PID is used (see InTech article “Wireless – Overcoming Challenges of PID Control”.. In this PID enhanced for wireless control (DeltaV PIDPlus), the PID takes immediate action for a change in setpoint or feedforward signal but only makes an integral mode (reset) correction when it sees a measurement update.

If the measurement delay is larger than the time to steady state of the process, the smart PID will see the full effect of the process response to the change in controller output. In this case, the feedback controller gain and feedforward signal gain can be set to provide complete correction. In a traditional PID, the output is continuing to ramp and the controller gain and feedforward gain have to be backed off from the optimum to avoid an over correction. For a cascade of static mixer pH to reagent flow control, a 60 and 15 second default update rate for pH and flow creates this case where the process response is complete within the measurement update time interval. In the following charts the orange and black trends are the setpoint responses of the true (not measured) process variable for the smart PID and traditional PID, respectively. Note that the setpoint response for the smart PID is immediate and complete, faster than what is seen by the wireless measurement. In fact, you need to have a wired measurement or some other indication to know how fast the response truly is.

WirelessReagentFlowResponse.jpg

WirelessStaticMixerpHResponse.jpg

Finally, patience can be the better part of valor particularly for split range control. A measurement default update rate of 15 to 30 seconds for jacket temperature can reduce unnecessary excursions across the split range point and overreaction to steam shock besides improving its setpoint response for vessel temperature control.

To summarize, an increase in wireless update rate can not only extend battery life but improve signal to noise ratio of slow loops, the setpoint and feedforward response of fast loops, and reduce the cycling of split range loops. Check out the Oct 28, 2010 entry and the Aug 5, 2010 entry on this website.

For more details on DeltaV PIDPlus, checkout the white paper I wrote in August

DeltaV-v11-PID-Enhancements-for-Wireless.pdf

As always you need to be intelligent about your application. For highly exothermic reactors, update rates slower than 15 seconds for temperature could create a runaway. Even damping settings of a few tenths of a second in flow and pressure loops can cause compressor and furnace shutdowns. For these applications, wireless is not advisable.