What can be done besides the typical analysis for control benefits that show a setpoint can be moved closer to a constraint by reducing variability? Here we take a quick look at five things that can be done to extend this analysis to unknown effects and recognize and seek benefits from a more extensive view. In many cases, the improvements require simple improvements in the existing DCS configuration and modes of operation.
Figures referenced are slides in the presentation Process-Control-Benefits-Analysis.
The typical analysis of the benefit of process control as seen in Figure 1 shows how tightening the distribution of a process variable enables the setpoint to be moved closer to the constraint associated with more profitable operation.
A neglected opportunity is the margin imposed by the operator because of an incident and/or a lack of process knowledge. The operator is going to take the most conservative approach particularly if there are no metrics or knowledge of the loss in profitability. The benefits from a more intelligent setpoint shown in Figure 2 may be larger than the benefits from reduced variability. First principle models can find the better setpoint and associated economic benefits. A virtual plant using this model can show the control system can achieve this new setpoint and get operations on board. While some of the margin can be eliminated by a simple change in setpoint by the operator, a more accurate solution particularly for changes in production rates and disturbances is gained by advanced process control (APC) to optimize the setpoint. An at-line analyzer or inferential concentration measurement of a key component can be used by the APC to compensate for unknowns in the process and the model.
Another opportunity involves the effect of measurement precision and accuracy shown in Figure 3. The margin imposed by the operator may be due to the variability or offset in the measurement. Thermocouples and pH electrodes are notorious for have offsets and drift. Operators can home in on setpoints to compensate for measurement error but a replacement starts the process over again. The use of resistance temperature detectors (RTDs) and middle signal selection of high temperature high performance glass pH electrodes can go a long way toward being able to operate with better setpoints. APC or cascade control can automatically compensate the setpoint of lower loops.
Often the process variable of interest related to product quality is downstream near the end of the process. How do we know the effect of measured variability of upstream variables on downstream process variables? This is an excellent question because the lack of understanding of the filtering effect from the primary time constant of intervening volumes is The Primary Source of Disagreement in Process Control.
The primary process time constant is small and the filtering effect for plug flow liquid inline systems and sheet lines seen in pulp and paper and gas volumes in the hydrocarbon processes. Process variability goes downstream mostly un-attenuated. In contrast the large liquid volumes of equipment in intermediates and specialty chemical processes offer a huge primary process time constant that filters fast oscillations to the extent where eliminating the upstream variability may have little benefit downstream. A simple equation shown in Figure 4 can be used to estimate the unknown amplitude downstream in a key economic variable from a measured variability upstream. This is useful because other sources of variability may be obscuring the quantification of the transfer of variability despite the use of Power Spectrum analysis. The equation can also be used to estimate the effect of unknown process oscillations in fast unit operations (e.g. phosphorous furnace pressure or compressor surge) or plug flow unit operations (e.g. extruder temperature) filtered by a measurement time constant.
The effect of one variable on another variable in terms of a process gain, dead time, and time constant can be found by the identification software for adaptive tuning and model predictive control. The variability from utility systems and upstream unit operations is multiplied by the process gain and filtered by the process time constant via the Figure 4 equation to predict the contribution to the variability in a key economic variable.
For near and true integrating processes variability is best seen in the PID output rather than in the process variable (PV). Controllers with a high PID gain is indicative of a large primary process time constant of a near integrating process or a slow ramp rate in a true integrating process that enables a maximum transfer of variability from the PV to the PID output. The variability in the output is also the key to process efficiency especially in split range loops that tend to oscillate across the split range point. Alternating steam and coolant flow, nitrogen and vent flow, and acid and base reagent flow drives up operating costs besides introducing variability into utility and raw material systems.
Finally, cost sheets, design of experiments in the actual or virtual plant can be used to find gaps that are often associated with disturbances or changes in operating points. These disturbances can be tracked down to discontinuous operations, manual output changes by operators, and changes in setpoints and hence margins by operators. Often changes can be correlated to a shift change and a sweet spot imposed. Eliminating the need to go to manual, replacing discrete actions and on-off valves with control loops and throttling valves, flow feedforward for production rate changes, and advanced control to find and maintain the best setpoints can eliminate most of the gap identified in Figure 5.
