A loop is only as good as its setpoint. Most loops are not operating at the best setpoint for 9 major reasons. For sustainable manufacturing we need to understand and address these reasons to achieve better setpoints to reduce the energy, raw material, recycle, and waste and provide more flexible production while maintaining product quality.
Reducing variability often does not translate into benefits until the setpoint is moved closer to a constraint as shown in slide 2 of Advanced-Control-Unleashed-Chapter-3-Figures. There is even a bigger issue. I would maintain that the offset from the optimum is greater from the choice of setpoints by the operator as shown in slide 3. This can be a quick hit in that you may just need to move the setpoint. Better process knowledge or a simple trial run at a lower or higher setpoint might be sufficient. I have seen cases where the running of models provided the knowledge and confidence to find a better setpoint. In other cases, big improvements were made by engineers with practical process and control capability. In Belgium plants, the operators who were engineers found the better setpoints by simply working with the process and being proactive.
A common example is temperature control. Distillation columns rely on temperature for an inferential measurement of composition (Temperature Measurement – the Most Important of the Common Measurements). Temperature determines both production rate and quality in reactors. Reaction rate increases with temperature but just past the optimum, reverse reactions, side reactions, and product degradation may occur. Consequently, there is a peak in the production rate as shown in slide 6. For bioreactors, the specific cell growth rate and product formation rate increases with temperature but cells start to shut down and die for temperatures to the right of the peak in the plot of specific rate versus temperature. The peak is different for cell growth and product formation, so temperature shifts are made generally after the exponential growth phase is fully established. However, the best size and timing of the shift is often not known. A similar situation exists in bioreactors for pH with an even sharper peak. Measurement accuracy and control requirements of a few hundredths of a pH of setpoint are stated. The question is do we really know the best setpoint to the same precision?
The major reasons loops are not operating at more optimum setpoints are:
(1) Measurement error
(2) Loop variability
(3) Process variability
(4) Abnormal situations
(5) Equipment deterioration and modifications
(6) Process modifications
(8) Process knowledge
(9) Tactical variability
The solution is the proper design, implementation, commissioning, and maintenance of the technologies shown in the pyramid on slide 1. A detractor for the advanced process control (APC) technologies is tendency not to address the maintenance issues. Many perform great when commissioned but steadily degrade over time due to changes in raw material (process variability), equipment deterioration and modifications, process modifications, and changes in economics (tactical variability). Plant cultures such as seen in refineries and petrochemical industry that understand the value of APC have the resources to tune and continuously improve the APC strategies. In most other plants, the APC is turned off within 3 years after the APC expertise leaves the control room. The reputation of real time optimization (RTO) has particularly suffered. Charlie Cutler, the father of model predictive control and real time optimization, will be presenting how to sustain APC in his keynote talk and tutorial at ISA Automation Week.
How to best use the pyramid technologies to address the reasons for sub-optimum setpoints are beyond the scope of this blog but some interesting ideas come to mind at the pyramid bottom. Measurement error (principally from drift) is a major source of error in the setpoint. Often setpoints are adjusted empirically to compensate for measurement error. Perhaps least understood is the concept that the concern about measurement accuracy moves to the higher level control when a loop gets its setpoint from higher level control (e.g. cascade control or model predictive control). I am not advocating installing measurements with less accuracy because process modeling, metrics, and analysis benefit from accuracy of process inputs (for importance of flow measurements check out Monitoring Process Performance Online, Flexible Manufacturing, and Advances in Flow and Level Measurements Enhance Process Knowledge, Control)
APC will correct through feedback control the setpoints in lower level loops for measurement errors. This is why we don’t require valve positioners to have an incredibly accurate position measurement and zero offset from the position setpoint of valves under closed loop control. An interesting caveat is that the obsession with digital displays has lead to integral action being offered and turned on in positioners (in the old days positioners only had proportional and rate action). Integral action in the positioner creates limit cycles from stiction. I would advise one who has a valve with a new positioner cycling faster than the loop, to check whether turning off the reset action or putting in an integral deadband in the positioner kills the limit cycle.
The capability of APC depends heavily upon the availability of online and at-line analyzers. I am torn between choosing between valve stiction and backlash or analyzers for the last of the top ten limitations. Check in next week for my decision and reasoning.