I have always favored online analyzers because they offer a continuous measurement with minimal deadtime. These analyzers usually have sensors in a pipeline or vessel to directly measure a property of the process, such as capacitance, conductivity, color, density, oxidation reduction potential, particle size, pH, turbidity, and viscosity. Expensive analyzers such as mass spectrometers that serve many vessels do not offer a continuous measurement because a sample or slip stream is cycled through them.
Despite the promise of new technologies, such as Near Infrared (NIR), the workhorse of the process industry is still the gas chromatograph, as discussed in the Control Talk column Process Analyzers. Analyze This! The deadtime introduced by these at-line analyzers is 1½ times the analyzer cycle time (Tip #90) plus the sample transportation delay and multiplex time. The total deadtime from a chromatograph can range anywhere from 30 to 60 minutes.
At-line analyzers may be more accurate but present more challenging maintenance and control than online analyzers. At-line analyzers require sample system maintenance and special technician expertise for troubleshooting. Failure to update causes the controller to ramp off toward an output limit when an analyzer measurement is used for closed loop control.
Most raw material and batch compositions are measured off-line in the laboratory. The time it takes the operator to take the sample and the lab technicians to do the analysis and enter the result is typically long and variable.
As a consequence these results are essentially useless for closed loop control with a traditional PID. We have seen the consequences of deadtime (Tip #70 and #71). A traditional PID requires the controller gain to be decreased and the reset time to be increased as the deadtime is increased. Composition controllers that use at-line analyzers must be tuned much slower than temperature controllers on the same equipment. Composition controllers are usually relegated to trimming temperature setpoints. The temperature controller takes care of most disturbances and the composition controller slowly corrects for temperature sensor drift and changes in the equilibrium relationships between temperature and composition. Most of the applications are on slow unit operations, such as distillation that have a time to steady state of 10 hours or more. An enhanced PID opens the door for the use of at-line analyzers and even off-line analyzers on even fast processes and improves the loop performance for all analyzers.
Concept: An enhanced PID does not compute the integral mode contribution until there is an update of the measurement or there is a setpoint change. The exponential response of the external-reset signal is used. The contributions from the proportional and derivative modes also do not change. Consequently, between analyzer updates or for a failure of an analyzer to update for a constant setpoint, there is no change in the controller output. When the deadtime from the analyzer system is greater than 95% of process response time, the controller gain can be as large as the inverse of the open loop gain, and the reset time can be based on the original process dynamics. For the accurate identification of the open loop gain, the controller can make a single correction that compensates for a setpoint change or an analyzer update. Long and variable update times from off-line analyzers do not affect the tuning.
Details: For analyzer applications, use an enhanced PID developed for wireless that utilizes external-reset feedback. If nearly the full process response is seen within the analyzer update interval (analyzer update time > 95% of process response time), the controller gain can be increased toward a maximum that is simply the inverse of the open loop gain. The open loop gain is the percent change in measurement divided by the percent change in controller output after all transients have died out (Tip #89). For cascade control, the open loop gain takes into account the effect of secondary controller scale, process gain, and the analyzer primary controller scale. The reset time can be based on the process deadtime. Use feedforward control for measured disturbances since there is little to no attenuation by feedback control because the effect of the disturbance is not seen until after the analyzer deadtime. Feedforward signal changes immediately change the output. If the analyzer fails to update, the enhanced PID output will not change.
Watch-outs: A bizarre analyzer value or an upscale or downscale failure must be screened out. While the enhanced PID simplifies the tuning, the response for unmeasured disturbances still depends upon the total loop deadtime. For fast unmeasured disturbances, the peak error is the open loop error, that is, the error that would appear if there was no feedback control. The minimum integrated absolute error is the peak error multiplied by ½ of the total loop deadtime. For setpoint changes, the rise time will be increased by the analyzer deadtime if the controller gain is less than the inverse of the open loop gain.
Exceptions: Closed loop control will not work for erratic analyzer values or poor signal-to-noise ratios.
Insight: An enhanced PID can simplify the tuning and improve the stability for loops with large and/or variable analyzer update times.
Rule-of-Thumb: Use an enhanced PID with feedback plus feedforward control for at-line and off-line analyzers.