Process efficiency can be increased by eliminating the excess use of reactants, reagents, and energy, eliminating the production of waste and off-spec material, and taking advantage of low energy and raw material sources. Not well recognized is effect of sensor drift and location, missing measurements, abnormal operation, process understanding, control strategy, deadband, resolution, and threshold sensitivity on optimum operation. Often publicized in the literature is the benefit of a reduction in process variability by better tuning, the use advanced process control (APC), such as the use model predictive control (MPC) to reduce a gap between the present and optimum setpoint, and a linear program or real time optimization (RTO) to find the optimum setpoint. Successful APC applications often address the previously mentioned automation system deficiencies in course of the APC implementation. Since APC is beyond the scope of this blog, we focus what makes a good foundation for APC as detailed in the ISA book Essentials of Modern Measurements and Final Control Elements in the Process Industries. This book like my other books was written for the purpose of organizing and sharing knowledge gained, and providing a sense of closure freeing my mind to move on to new ideas and opportunities. The royalties are good for a night out.
Threshold sensitivity and resolution limits in the measurement and final control element causes limit cycles (sustained oscillations) whenever there is one or more integrators in the process or controller (e.g. integral mode). Deadband causes limit cycles when there are two or more integrators. Control valve resolution and deadband is worse near the closed position leading to back and forth crossing of the split range point. Limit cycling at the split range point wastes energy by alternating between heating and cooling for temperature control, wastes gas reactant by alternating between feed and venting for pressure control, and wastes reagent by cross neutralization of acids and bases for pH control. The enhanced PID developed for wireless can prevent unnecessary crossing of the split range point if noise is screened out.
Threshold sensitivity and resolution in the measurement also amplifies rate action that can prohibit the use of the derivative mode. One unfortunate example is the use of wide range 12 bit thermocouple input cards in the 1980s vintage DCS instead of temperature transmitters. The steps from the resolution limit results in a full scale spike in controller output for controller gain and rate time settings permissible for temperature loops on vessels and columns based on process dynamics. Sadly enough, the elimination of temperature transmitters was used as part of the cost justification for the 1980s DCS.
For chemical reactors, every 10 degree centigrade increase in temperature can correspond to a doubling of the reaction rate for low activation energies. However, temperatures too high can trigger reverse reactions and side reactions that reduce yield. Unintended polymerizations of highly reactive highly hazardous components, such as hydrogen cyanide and acrolein, poses safety hazards from high heat releases and plugging.
For distillation columns, temperature is an inference of composition and an error of a fraction of a degree can correspond to undesirable product concentration fluctuations. Slide 27 in the ISA Automation Week 2011 tutorial ISA-AW-2011-Fundamentals-of-Distillation-Column-Control.pdf by Terry Tolliver provides an example of the change in bottoms and distillate product concentration with temperature. If you put your cursor on the note icon in the upper left corner, you can see Terry’s notes on the slide.
For bioreactors, the cell growth rate can decrease by 3% for an operating temperature that is only 1 degree centigrade below the optimum. However, a temperature above the optimum can result in an escalating net decrease in cell growth due to cell death.The pH sensitivity is even greater and often symmetric where a tenth of a pH error can cause a 5% drop in growth near the optimum increasing to a 10% drop in growth per tenth of a pH error as the true pH gets more than 0.2 pH units away from the optimum. Slides 1 and 2 of Bioreactor-Growth-Rate-Kinetics.pdf show an example of the growth rate sensitivity to pH and temperature, respectively.
For neutralization systems, moving away from the neutral point (e.g. 7 pH at 25 degree centigrade) to a flatter portion of the titration curve will save on reagent use and reduce the amplification of fluctuations from non-ideal mixing, imperfect valves, and disturbances. The reduction in process variability allows the setpoint to be put closer to the constraint (high or low pH limit).
Resistance temperature detectors have a factor of 100 less drift and factor of 50 better threshold sensitivity and repeatability than thermocouples. New pH glass designs and wireless transmitters offer a factor of 5 less drift and a factor of 10 better resolution particularly after repeated sterilizations. Solution pH temperature compensation corrects for true changes in the process pH due to changes in the dissociation constants with temperature. Such improvements in sensor performance can have large effects on process efficiency based on examples cited here for bioreactors, chemical reactors, and columns. Often operations are trying to deal with these measurement errors on an ad hoc basis by adjusting setpoints. The setpoint bias is a guess after the fact. The installation of at-line and online analyzers can automatically correct the setpoint by via remote cascade control (supervisory control). The enhanced PID developed for wireless can eliminate the oscillations from sample system delay and analyzer cycle time and prevent reaction to analyzer failures as noted in the InTech July-Aug 2010 article "Wireless – overcoming challenges of PID control & analyzer applications"
Sensor locations with less transportation delay and with high velocities can provide faster responses. Locations with better mixing and no bubbles can reduce noise. Lower deadtime and a higher signal to noise ratio enables a higher controller gain and tighter control. The best location for column temperature control is the tray that shows the largest and most symmetrical change in temperature for a change in the ratio of the manipulated flow to feed flow. On slide 26 of Terry Tolliver’s ISA Automation Week tutorial, tray 6 is the best location for a change in distillate to feed (D/F) ratio. The use of this tray for control not only provides the best sensitivity and linearity but keeps all of the other trays near their best operating point for a change in feed. For more details on the criteria for better sensor selection and location, checkout my blog posted Dec 1, 2011 on the modeling and control website How to Succeed – Part 4
Often the least costly raw material for fuel is a waste or recycle product with variable concentration and slower response. A small MPC readily configurable in a DCS, such as depicted in Advanced-Application-Note-001-pdf, can simultaneously minimally use a less variable and faster but more costly flow for tight control while maximizing the use of the less costly flow.
In part 3, we look at control strategies that can inherently increase production rate or efficiency for batch, fed-batch, and continuous operations.
