The power of the PID largely remains underutilized. Most of the options and parameters other than scale ranges and tuning settings remain at their defaults. Here we look how to tap into the incredible capability of the PID to minimize batch cycle time and start-up time and to maximize production capacity, flexibility and efficiency while reducing disruption to utility, raw material, and reagent systems. Also detailed is a simple method to compensate for dead time and to eliminate limit cycles.
Overview
Feedforward control can be used to improve regulatory control and optimization by preemptive correction or coordination of process control and valve position control. The feedforward signal gain and dynamic compensation should be designed to provide a correction in the process at the same point and same time as the disturbance. A dead time block and lead-lag block is generally all that is needed for this dynamic compensation of feedforward signals. Given an error in timing is inevitable, preferably the feedforward signal arrives late rather than early in order to prevent inverse response.
The most common feedforward signal is flow because of production rate changes and because nearly all fast and large disturbances originate as flow rather than composition or temperature disturbances. To attain the process flows on the process flow diagram (PFD), the manipulated flow is ratioed to a wild or leader flow. In most cases, the feedforward is the primary feed flow. A ratio station is used to provide the operator with the ability to set the desired ratio and see the actual ratio. The process PID provides a feedback bias correction to the manipulated flow. The actual ratio displayed is the corrected manipulated flow divided by the feedforward flow. For more details on the implementation of ratio control see ISA Interchange 4/18/2014 post “Tip#93 Use a Viewable and Adjustable Flow Ratio with a Feedforward Summer.”
Intelligent integral action is the key to preventing oscillations. The suspension of integral action when final control elements are not moving (e.g. deadband and resolution limits) and measurements are not updating (e.g. communication failure, wireless default update rate, and analyzer cycle time), will stop oscillations and reset windup. The slowing down of integral action when secondary loops or final control elements are slow to respond will prevent the burst of oscillations from large and fast disturbances and setpoint changes.
A signal selection block chooses the lowest or highest of outputs from override PID controllers to honor constraints. Override process controllers prevent violation of process variable constraints. Override valve position controllers prevent violation of control valve position constraints by keeping the control valve position at the most desirable throttle range. Output limits serve as a backup. Intelligent integral action and feedforward action are keys to a fast smooth override response for process protection and optimization. For more details on process optimization see the ISA Interchange 3/7/2014 post “Tip #97: Use Valve Position Control to Optimize Process Efficiency and Capacity.”
Dead time compensation can be simply implemented by the addition of a dead time block in the external reset feedback signal (PID+qo). This PID achieves the about the same performance as a Smith Predictor without having to set a predictor model gain and time constant. The dead time setting must always be within +5% and -25% of the actual dead time. The PID reset time must be significantly decreased toward the limit of the error or module execution time to see an improvement in control. If the reset time is not decreased, the performance is worse than if there was no dead time compensation. Overestimates of the dead time will lead to fast oscillations whereas low estimates just cause a loss of performance advantage. This is the opposite effect for the tuning of the PID where an overestimate of dead time causes sluggish response and an underestimate causes an oscillatory response. Contrary to popular opinion the improvement by dead time compensation is greatest for lag dominant rather than dead time dominant loops. Ironically, lag dominant loops (low dead time to time constant ratio) can have a high controller gain and have less of a need for dead time compensation.
The enhanced PID developed for wireless that suspends integral action when there are no measurement updates can simplify the tuning for at-line and off-line analyzers. If the cycle time is much larger than 63% process response time, the reset time can be set as small as the process dead time and the PID gain can be set as large as the inverse of the open loop steady state gain for self-regulating processes. The tuning is not as simple for integrating processes and the improvement is not as dramatic but the enhanced PID adds a significant degree of robustness. The advantages and rules also pertain to large wireless update times on fast processes. While the control is more robust and setpoint response much faster by the enhanced PID, the peak error and integrated error for unmeasured load disturbances is increased due to the increase in total loop dead time by amount that is 1½ times the analyzer cycle time and ½ of the wireless update time.
Recommendations
- Use feedforward control to give preemptive action for large and fast disturbances.
- Use ratio and bias and gain stations to provide flow ratio control with feedback bias correction and the actual flow ratio viewable and desired flow ratio accessible by operations. Enable startup on ratio control without feedback correction until the process (e.g. distillation column) reaches operating condition.
- Provide dynamic compensation of feedforward signals by dead time and lead-lag blocks to give the right feedforward timing for disturbances.
- If open loop tests cannot be done or the process response is complex (e.g. high order or recycle effects), integrating, or runaway, the dead time and lead-lag times can be adjusted based on oscillation patterns observed.
- Ensure the feedforward correction does not arrive too soon causing an inverse response that can result in the feedforward doing more harm than good.
- If the feedback correction will unavoidably arrive late, decrease the feedforward gain to prevent a secondary oscillation and undershoot that result from feedback control having already done some of the correction.
- Use output tracking or remote output to schedule output changes for open loop backup to protect equipment or bang-bang control to minimize batch cycle times.
- Use a rate limit on valve, VSD, or flow loop setpoints (move suppression) with external reset feedback to the primary PID (e.g. composition, pH, or temperature) manipulating these setpoints to prevent disruption to utility, raw material, and reagent systems from high PID gain or rate time settings or bang-bang control.
- Use intelligent integral action via integral deadband, external reset feedback, and/or an enhanced PID developed for wireless to prevent oscillations from split range discontinuities, deadband, resolution or threshold sensitivity limits, coated electrodes, slow secondary loops or final control elements, communication failures, and large wireless update times and analyzer cycle times.
- If the wireless update time or analyzer cycle time is much greater than the 63% process response time, use an enhanced PID to suspend integral action between measurement updates and increase the PID gain toward a maximum gain that is the inverse of the open loop steady state gain for self-regulating processes.
- Use an innovative use of split range and feedforward control to help a valve position controller deal with large and fast disturbances with parallel large and small valves to achieve best resolution and threshold sensitivity and rangeability.
- Use override control to honor process variable and valve position constraints.
- Use up and down setpoint rate limits with intelligent integral action to provide a directional move suppression to enable a slow approach to an optimum or a split range point and a fast getaway for abnormal operation and to prevent safety instrumentation system (SIS) activation.
- If the dead time can always be accurately set within +5% and -25% of actual dead time, add a dead time block in the external reset feedback signal path and decrease the reset time by a factor of 4 or more to improve performance.
- Use setpoint feedforward to reduce the time to reach setpoint (rise time) when the initial step change in output from a setpoint change is less than 50% of the final change due to PID tuning and structure.
- Move up to model predictive control when the feedforward and feedback dynamics are complex and multivariable and the optimization involves more than one objective.
