Effect of Mechanical Design – Equipment

There are numerous examples where equipment design by not taking into consideration dynamics caused poor control. A good equipment design should reflect an understanding of how dynamics affect control loops and the efficiency and capacity of the production unit. Most advanced control opportunities arise from equipment deficiencies that become particularly evident as plants are pushed beyond nameplate capacity or are required to turndown more than the intent of the original design. Many of the chemical plants and refineries in the USA were built in the 1970’s and 1980’s.

Engineers are needed with both process design and process control knowledge to ensure equipment in debottlenecking and process improvement projects and new facilities can deal with disturbances and market demands. By far the outstanding examples of the practical integration of process and control are the prolific Greg Shinskey and William Luyben. The result is a conceptual knowledge with far reaching consequences as outlined in “Top Ten Limitations – Concepts.”  Based on what I see in the literature despite book titles to the contrary, there are no successors on the horizon.

For example, I used Shinskey’s Controlling Multivariable Processes and Luyben’s Chemical Reactor Design and Control for my tutorial on Reactor Control at ISA Automation Week. My understanding of how the window for allowable controller gains closes for highly exothermic reactions as the UA secondary or sensor time constant approaches the positive feedback time constant is based on a 1970s article by Luyben. Similarly, Shinskey’s “Controlling Unstable Processes” from the same era provided me key insights to runaway conditions. I bought some newer books on reactor control with interesting titles only to find out the control section was the typical academic spiel on internal model control and single loop model based algorithms posed as a replacement for the PID. No insight, no application expertise, and no practical value. The process part of the book seemed useful although deficient in the guiding principles I am accustomed to from Shinskey and Luyben. Just think what we would learn if these guys expanded their scope of processes and applications studied. At ages 77 and 80, we can pray they stay actively involved and we take every opportunity to savor their wisdom. I will be presenting a recorded interview of Shinskey that will part of an ISA Automation Week event titled Shinskey and Best of Process Control

Astute engineers will use first principle relationships and dynamic simulation to study and define the requirements not only for different production rates but also for changes in raw materials, recycle, and utilities. For control considerations, the process equipment with automation system connections should provide a process gain as linear, a loop deadtime as small, a process time constant as large, and measurements as noise free as possible. The ultimate limit to the integrated error for unmeasured disturbances is proportional to the ratio of the deadtime squared to the 63% process response time (process time constant plus total loop deadtime). Loop deadtime delays the recognition and correction. The process time constant slows down the excursion giving the chance for the loop to catch up with the disturbance assuming the process time constant is downstream of the disturbance (normal case). For details on how equipment design affects process dynamics take a look at the application note “First Principle Process Deadtimes, Gains, and Time Constants

In general, the equipment design and measurement location should provide the greatest sensitivity (greatest process gain). The classic case is the distillation column where the best location for temperature measurement and control is the point where the change in temperature is largest for an increase and a decrease in the ratio of the manipulated flow to the feed flow for a range of feed concentrations that includes abnormal operation and upstream equipment limitations. The manipulated flow is most often distillate or reflux flow but can for low reflux ratios be steam or bottoms flow.

For neutralizations, the process gain can be too large for strong acids and bases. Here stages and setpoints are selected to decrease the process gain to reduce the amplification of valve stick-slip, pressure disturbances, and concentration gradients from non-ideal mixing.

Some examples of how equipment design including process connections limit control system performance besides the ones discussed in my entries on heat transfer and mixing on this website are:

(1) Variable cross sectional area causing nonlinear level gain (horizontal reflux drum)
(2) Obstructions in path of radar level “line of sight” causing erratic measurement
(3) Missing connections for wireless sensors preventing online diagnostics & metrics
(4) Surge tank volume too small for changes in upstream and downstream rates
(5) Vessel volumes too small to provide residence time for reactions at high rates
(6) Vessel volumes too small to prevent entrainment and foam in vapor space
(7) Sparger design causing interaction of pH and dissolved oxygen loops
(8) Electrodes too close to sparger causing erratic measurement from bubbles
(9) Thermowell or electrode connection close to feed entry causing short circuiting
(10) Glass lined thermowell for temperature sensor adding excessive lag
(11) Excessive dip tube size and length adding huge reagent delivery delay
(12) Thermowell or electrode connection behind baffle causing stagnation