When Tuning PID Alone is not the Answer

Emerson's Mark Coughran


Proportional-Integral-Derivative (PID) control is the workhorse control algorithm for process manufacturers. When tuned properly, it seeks to move a loop’s process variable (PV) to the setpoint (SP) as quickly as possible without overshoot or instability.

But, in some cases, even proper tuning is not enough. External factors such as equipment failures, mechanical performance, or incorrect control system configuration can affect to loop’s ability to perform as it is designed.

In an ISA Interchange article, How to Fix Process Control Loop Problems That PID Tuning Cannot Correct, Emerson’s Mark Coughran opens by explaining:

…in many instances, no amount of adjustment of the proportional-integral-derivative (PID) parameters will make the loop behave as desired. When tuning fails, it is time to look outside the PID for a solution.

Mark shares where you should begin:

…a proper loop tuning exercise—including measuring the process dynamics—should be the first step when approaching most loop performance problems.

He shares 10 examples where he found and solved problems outside of PID tuning. These include:

  • Control valve response
  • Valve/actuator connection
  • Control valve flow characteristic and deadband
  • Valve positioner tuning
  • Valve positioner assembly
  • Transmitter failure
  • Signal communication
  • Extra filter in logic
  • Low cutoff in logic
  • Master loop parameters for cascade

Fired Heater OptimizationI’ll share one of these examples and invite you to read the article for the rest. The response in a control valve can greatly affect the performance of a PID loop. Mark describes a fired heater where the operators could not control draft pressure.

The main symptom of the PC-2 loop was that the PV could not reach SP for any value of output. A consultant worked with the plant operator to put the controller in manual mode and make steps on the output, but still saw no recognizable pattern of the PV responding to the output.

A large butterfly valve was being used as a damper. Diagnosis was performed by putting the loop in manual mode and stepping the output driving the valve’s actuator.

Large output steps drove the actuator pressure—displayed on the local gauges—from zero to full supply and back again, showing correct response from the positioner. But the valve did not move, because it was jammed in the fully open position.

The solution was to repair the valve and to bring:

…the HART position signal into the control system, so that diagnosis of control valve problems would be faster in the future.

If you’ll be joining us at the Oct 1-5 Emerson Exchange conference in San Antonio, Texas, there is a really good Educational Services course on tuning, Practical Loop Tuning using Lambda Tuning Method. Visit the Consulting Services-Production Performance area on Emerson.com for more on ways to minimize process variability and optimize control performance.

Posted Monday, September 10th, 2018 under Process Optimization.