A Unified PID Control Methodology to Meet Plant Objectives

Plant objectives for basic and advanced regulatory control can be quite diverse. PID control plays an important role in maximizing safety, environmental and equipment protection, process efficiency and process capacity. There can be many different sources of process variability that control loops need to deal with such as raw material and recycle composition, weather, utility temperature and pressure, operators, interactions, optimization, on-off actions, startups, transitions, shutdowns, measurement and process noise, and limit cycles. Process nonlinearities, process non-self regulation, deadtime dominance, slow measurements and valves, analyzer cycles times, deadband, and threshold sensitivity limits can make the task much more difficult. The techniques for applying PID control to achieve plant objectives are largely individualistic than systematic. Each application tends to be treated as unique requiring special expertise. The result has been over 100 tuning rules to meet different perceived and actual needs of the process. Consultants are often required and PID features are not effectively utilized. In some plants PID loops have been pushed into model predictive control (MPC) because PID function application practices are largely heuristic and undocumented.

There are five key related PID control features; external-reset feedback, setpoint filters and rate limits, output tracking, and an enhancement for wireless that synergistically provide capabilities that not have not previously been fully realized. These key features used in conjunction with a simple identification of process variable (PV) ramp rate provide a unified general approach to meeting diverse plant objectives. This paper shows a methodology can be developed that minimizes implementation and maintenance efforts. 

The major types of open loop process responses are lag dominant self-regulating, integrating, and runaway. All of these processes have a common initial response that is a ramp rate. The PID controller accomplishes most or all of the feedback correction during the initial ramp rate. These processes can be treated as near or true integrators. The use of integrator tuning rules with the closed loop arrest time set equal to the total loop deadtime provides maximum disturbance rejection and inherently prevents violation of the both the low and high controller gain limits for the integrating and runaway processes. The use of deadtime block in the identification of the ramp rate to create an old PV one deadtime in the past extends the applicability of the method to deadtime dominant self-regulating processes. The peak and integrated errors for unmeasured disturbances are minimized. 

The use of a setpoint filter equal to the reset time enables the same tuning be used for setpoint changes without overshoot. Previously, the controller was detuned or an alternate PID structure such as proportional on PV rather than error was used to reduce overshoot. The result was a slower feedback action and special tuning methods.

The enabling of external-reset feedback (ERF) and analog output (AO) setpoint rate limits eliminates special tuning to prevent oscillations from slow secondary loops and slow final control elements, minimize interaction, maximize coordination and maximize equipment and environmental protection. ERF prevents the primary PID output from changing faster then the secondary loop, valve, or AO rate limits will allow the use of maximum disturbance rejection tuning. The AO setpoint rate limits provide a move suppression that is different for increasing and decreasing PID outputs. The directional moves suppression minimizes interaction, provides the same rate of change of flows for blending and reaction, offers nearly identical models when manipulated by MPC, and gives a slow approach to an optimum and a fast getaway for a disturbance or abnormal operation. The rate limit is set to the desired maximum rate of change for each direction.  

The old PV from the deadtime block subtracted from the new PV to create a delta PV is simply added to the current PV to provide a PV one deadtime into the future for each execution of the PID. This future PV can be used with the output tracking feature to make the approach to a new setpoint as fast as possible with negligible overshoot. The future PV and output tracking can also be used to proactively protect against safety instrumented system (SIS) activation, relief actuation, compressor surge, and RCRA pH violation. Directional move suppression complements the preemptive action from output tracking when the PID is returned to feedback control. 

The identification of deadtime and the ramp rate computed by dividing the delta PV by the deadtime can be used to compute the dynamic compensation needed for feedforward signals. The dynamic compensation helps the feedforward action arrive at the same time and same place in the process as the disturbance. For a half decoupler, the disturbance can be taken as the offending controller output.

The ramp rate and future PV can be used to adapt the reset time to prevent faltering or overshoot in the approach of the PV to the setpoint. The deadtime and ramp rate combined with capture of the PV and PID output for two different operating points can provide an adapted tieback model for operator training and process control improvement development and prototyping.

If a fast readback of the actual final control element PV (e.g., control valve position or variable frequency drive speed) is available, simply enabling ERF will stop limit cycles because the PID will recognize the position or speed is not changing.

Finally, the addition of a feature to suspend PID action on the PV when there is no update of the PV will enable the enhanced PID to be used for wireless and analyzers without having to detune the controller. A conventional PID would need to have the gain significantly decreased and reset time greatly increased and rate time set to zero if the time between updates approaches the 63% response time. For the enhanced PID, the controller gain can be set to the inverse of the open loop gain when the update time is larger than the response time to provide an almost complete correction for a setpoint change or disturbance in one execution of the PID. The extremely large and variable time of off-line analyzers with lab samples is inherently handled offering this one execution correction. If a PV threshold sensitivity setting is used to ignore noise, limit cycles will be stopped without the need for final control element feedback.

To summarize, the synergist generalized use of five key PID features and a deadtime block to create a ramp rate and future PV can help the process engineer to get the most out of the PID and to think in terms of the process needs rather than specialized tuning.