«

»

Dec
23

XY Plots for Operations and Process Analysis

Process visualization and interaction for operations are about to enter the 21st century after being stuck in the 1980s. Emerson and major companies in the process industries and universities through organizations such as the Center for Operator Performance are doing research on how to increase the effectiveness of operators to match the increased capability of a modern DCS.

The operator interface hasn’t changed much in the last 30 years. The standard fare includes graphical depiction of a simplified process flow diagram, color changes for when valve is or pipe inline open, bar charts for vessel levels, faceplates for PID controllers, numerical numbers for process variables and sometimes for valve positions, and trend charts. As a control engineer, I never could understand why there are no XY or 3D plots, no future trend plots like we have for model predictive control, no measurement locations, and no signals between controllers, measurements, and valves. Wouldn’t the operators benefit from such simple additions in terms of understanding relationships? The measurement location has implications as how representative and how slow or delayed the measurement is. The 3D plots would be useful for visualization of temperature profiles in plug flow and fluidized bed reactors and in distillation columns and absorbers (Process Analysis Using 3D Plots), and for thickness profiles in paper sheets and plastic webs as mentioned in the Dec 28, 2007 entry on this website “Biggest Opportunities for Process Control Improvement – The Operator (Online Metrics)”. Presently, some DCS, such as DeltaV provide a convenient export of data and interface to software, such as MATLAB that can be used to generate 3D plots. Version 12 DeltaV will have the 3D plot capability built-in.

XY plots offer even more intriguing possibilities.

The most obvious need for an XY plot is for analysis and early prediction of compressor surge. A succession of operating points on a plot of compressor pressure rise versus suction flow will show the approach to the surge line and the characteristic cycles of surge versus stall as shown in Section 3 of the online e-book “Centrifugal and Axial Compressor Control”. The operator points could be increasing in dot size or intensity with sporadic numbers. Clicking on the dots would show the time.

Limit cycles other than surge also show up as a doughnut on plots of the process variable and controller output for PID loops. The most common source is valve stiction (for 1+ integrators) and backlash (for 2+ integrators) The size of the doughnut hole is indictive of the amount of stick-slip or deadband. The doughnut hole is lopped sided due to nonlinearities and disturbances.

Other uses include worm plots of batch end points. If the worm is coiling, batch variability is decreasing. The tail and the head of the worm show what where the batch has been and where it is going. Traditionally the XY coordinates have been the two most important principal components at the end of the batch. Expanding our thinking you could visualize how product concentration versus process efficiency (raw material or energy use) for a batch or a shift of continuous operations would be useful.

The nonlinearity of process composition, temperature, and pH are best seen on a plot of these process variables versus the manipulated flow to feed flow ratio as discussed in “What Have I Learned? – Ratio Control (Part 2)”. The most obvious application is pH control where the current operating point would be placed on the titration curve that is a plot of pH versus the ratio of reagent to feed flow. With these plots the process variable must be synchronized with the manipulated flow. This can readily be done for plug flow reactors, blenders, static mixers, desuperheaters, and heat exchangers by the simple use of a deadtime block to delay the flow ratio by the total loop deadtime.