Greg: The May Puzzler that asked why Stan didn’t need to scurry back to the plant on Monday to finish the start up of a column has a startling explanation.
Stan: To the amazement of everyone, the column was lined out. Someone said to me, "Of course, what did you expect?" There was an explanation. However, I don't remember it.
Greg: This senior moment means we are free to discuss the possibilities. We had interesting replies from Ben Walker and Hunter Vegas.
Ben: There are two possible answers to the May puzzler. If the question refers to what will you find on Monday after lunch, the answer is a pink slip. If it is what you will find when you get home on Friday, the answer is a call to get back to the plant.
Hunter: After watching the control scheme swing the column for a half hour and continuously acknowledge process alarms, the operators would have put the control system in manual and let the column run straight lines for the rest of the weekend!
Stan: Often the most progress on a startup is made when the control room is cleared of spectators and managers and their respective distractions. The weekend is a great opportunity for operations to focus on the shortest route to getting the unit running.
Greg: Smart operators put the temperature controller in manual and set the reflux and steam flows to provide the desired internal traffic in the column to line it out. Automated startups duplicate these actions by sequencing and holding flows in the right ratio. Only when the column is up to operating temperature and pressure should the reflux or steam be manipulated directly or indirectly by changing distillate or bottoms flow by the temperature control. Even when put in automatic, ratio of reflux and steam to feed is retained as feedforward control to deal with the inevitable rate changes, and the temperature controller output is added in as a corrective bias.
Stan: If the temperature control is put in automatic before column temperatures reach the boiling point, the temperature measurement is bogus.
Greg: The temperature is only indicative of composition when the mixture is boiling. The temperature sensor may be seeing vapor because the liquid is below the weir in tray columns or because the down flow is separated from the sensor from channeling in packed columns.
Stan: Hopefully, the levels in the sump and overhead receiver have enough operating room so the level controllers keep the distillate and bottoms flow zero until the column reaches normal conditions to avoid upsetting downstream equipment. It is particularly important not to prematurely send forward a product stream by putting the level controller in automatic too soon.
Greg: If a composition or level controller manipulates reflux or steam, the considerations for ratio and feedforward control are similar to what was mentioned for the temperature loop.
Stan: In general, initializing and holding the key flows in the right ratio is called a “head start” and brings a unit up to set point smoothly at a speed set by the process’ open loop time constant.
Greg: A startup turned over immediately to temperature and compositions controllers without a “head start” and tuned with a Lambda factor (ratio of closed loop to open loop time constant) of 4 will take four times longer even if the measurements were representative and not erratic during the startup. However, some unit operations, such as reactors with large process time constants, can legitimately use a Lambda factor less than 1 in the tuning of the temperature controller. In these applications, as soon as the temperature measurement is indicative of reactor conditions, the temperature controller is put in supervisory control and its set point moved from the current value to the operating point to overdrive the output beyond its resting value and make the startup faster than what can be achieved by a head start. or unstable reactors, the overdrive coolant is important to prevent overshoot and runaway conditions.
Stan: If you consider it takes four time constants plus the dead time to reach 98% of the final value, for a reactor with a 5 min. dead time and a 1 hr time constant without a head start, a startup would take about 16 hr and 2 hr, for a Lambda factor of 4 and 0.5, respectively.
Greg: You can roughly see the maximum potential to reduce batch or continuous unit operation startup time from overdrive by comparing, on a trend chart, the time required for the manipulated variable to settle out near its resting value to the base of a stacked rectangle and triangle for a manipulated variable with overdrive. The rectangle starts with the set point change and ends one loop dead time after four times the process time constant multiplied by the Lambda, and has a height equal to the resting value of the manipulated variable. The right triangle sits evenly on top of the rectangle and has an altitude equal to the output limit minus the resting value. The area of the triangle and rectangle should roughly be equal to the present area of the manipulated variable during startup on existing trend charts.
Stan: If the process equipment depends upon upstream unit operations or utilities running smooth, the vulnerability to upset is in proportion to the equipment’s startup time.
Greg: This leads to the top ten signs a startup has gone wrong.
Top Ten Signs a Startup has Gone Wrong
- Product tanks are empty
- Waste tanks are full
- The startup team and DCS have negative free time
- The entire design team is in the control room
- The whole research team is the control room
- The managers have left the control room
- The consoles are lit up like pinball machines
- Operators keel over from doughnut overdose
- A collection is taken for the startup budget
- The order for souvenir baseball caps is cancelled
This Month's Puzzler:
Why do the dissolved oxygen probes in the lower side nozzles of an aerobic waste treatment tank and bioreactor have a higher reading than the ones in the upper side nozzle?