Greg: The complete automation engineer has the ability to understand chemical engineering principles and dynamics, improve communication with process engineering and operations, solve measurement and control valve problems, and have the practical knowledge to implement the best distributed control system (DCS) configuration. Companies have enabled this capability by courses and plant assignments. An example of what can be done is seen in the career of Owen Campney, a longtime friend. Owen joined the Monsanto Corporate Engineering electrical and instrument (E&I) design department shortly after I returned from plant assignments in E&I construction. Owen presently contracts as a DCS configuration specialist. Here we ask Owen how he progressed through his career and what type of problems he faced.
Owen got some very focused practical training via a Monsanto E&I school with pH and distillation control labs I put together to provide a hands-on wet lab experience. Hopefully, the attendees didn’t get too wet because I was wet behind the ears when I designed the bench-top scale neutralizer and column. Besides the internal E&I school, what did Monsanto do to help bring you to speed?
Owen: Monsanto partnered with Washington University in Saint Louis (WUSTL) to provide a three-semester intensive course. The course was originally designed for plant production engineers as a refresher in chemical engineering to enable them to become process and mechanical design engineers at the corporate engineering headquarters in Saint Louis. The last couple of years the program was opened up to include E&I engineers. Some of the more notable courses were reaction engineering taught by a professor who would go on to receive numerous awards for collaborative research, and an analyzer course taught by an adjunct professor with extensive plant experience. The course helped me to ask process engineers intelligent questions. I put the knowledge to good use by doing a component balance that showed throwing another recycle stream into a unit operation was not practical because operation was adversely affected.
Greg: We have, in a way, come full circle in that there is an excellent WUSTL hands-on wet lab process control course taught by Monsanto retiree Bob Heider that prepares students for a career in the process industry as highlighted in Jim Cahill's post "Accelerating the Process Automation Learning Curve".
Stan: What were some of your early assignments?
Owen: I was the lead instrument engineer for boiler projects at Monsanto’s Saint Louis and Texas City plants to keep up with the changing steam demand from plant expansions.
Greg: Texas City was preparing for the world’s largest acrylonitrile plant that would end up being my next relocation as the lead E&I design engineer and subsequent construction and start-up engineer. It was the heyday of building chemical plants with a corporate engineering department of 1600 or more engineers. Then the beginning of the end occurred with a decline to a present day staff about 1/100th of the peak. When we stopped building chemical plants, where did you go and what were some of things you did?
Owen: I transferred to an E&I group in the plant engineering department to a plant on the Gulf Coast. I worked on small projects and helped the technicians in troubleshooting loops.
I supported the start-up of a brand new, world-class hydrocarbon production unit and subsequently helped in debottlenecking projects. I improved the split range in the DCS to provide a smoother transition from one stream to another by an innovative strategy to prepare the stream conditions of the next stream to be as close as possible to normal, and the conditions of the existing stream being throttled just before the crossing of the split range point.
There was a whole series of valve problems I worked on. In a new hydrocarbon plant I found a year-old current-to-pneumatic transducer (I/P) for surge valves that was too slow. I adjusted pneumatic switches on the isolation valves on reaction air to slow down the stroke when the valves reached the steep part of the installed characteristic to reduce the upset and potential surge and shutdown of the compressor upon a reactor shutdown. I replaced new positioners that were failing on another compressor due to internal welds in levers breaking from vibration. The supplier provided a different model positioner from overseas. I found two boosters in parallel on surge valves were fighting with each other, an oscillation possibly triggered and aggravated by vibration. I went to one booster. I upsized actuators on the block-and-vent valves on reactor exit to help them break free from the sticking due to the residue from minor chemical components in the reaction gas. In a pressure swing absorption (PSA) unit I found which valve positioners had lost their calibration, a common problem at the time for pneumatic positioners.
Stan: Since valves are mechanical components exposed to the process, they are more vulnerable to problems threatening their precision and reliability. New smart positioners are more sensitive and reliable, holding their calibration and providing diagnostics to alert the user of potential problems. When did you move on from field problems to focus on DCS configuration?
Owen: For the next 12 years I was a specialist in a process control group. In a batch polymer unit I replaced tenor drums that used pneumatic switches for valves and electrical switches from motors and timers with a DCS. I worked mostly on the DCS configuration, but bought special smart transmitters with enhanced stability at 0 psig to improve batch end point control.
I did a lot of debottlenecking projects. In the middle there was a project for 83 megawatts of cogeneration that included replacing burners and furnaces besides the control of three different drum levels for different pressures of the heat recovery steam generator (HRSG). The drum level switched from three-element control (steam flow, feed-forward and cascade control of level to feedwater flow) to single-element control (direct throttling of feedwater valve by level controller) at low steam rates due to limited flow meter rangeability. The gas turbine offered an optimization of the temperature of the low-pressure exit gas that went to the HRSG. We were fortunate our cogeneration unit was big enough to use the smallest gas turbine that runs at 3600 rpm, eliminating a gear box and loss in efficiency.
The most challenging process control applications were in a large intermediates production unit with solids, parallel semi-continuous operations, a high ionic strength, non-ideal equilibrium component, insufficient surge tank volumes and recycle streams. The solids would coat surfaces and plug lines. Periodically unit operations would be shut down and defrosted, a process where the crystal coatings on surfaces were melted and washed away by the injection of a hot fluid, leading to semi-continuous operation. The defrosting by operations was ad hoc due to some unpredictability, although an approach temperature diagnostic was proposed. The surge tank volumes could not absorb all of the variability from upstream units being defrosted because of insufficient size and the need to provide a minimum residence time to react to undesirable components. Consequently, the feed rates to downstream units needed to be changed. A custom FORTRAN program was developed to change these rates to help the coordinate the unit operations. The logic and expertise to adjust and maintain the program resided in a single person. Further complications were caused by recycle streams that created a delayed integrating response in the concentration. A purge stream flow needed to be optimized. If the purge stream flow was too small, the concentration of impurities would build up. Too large of a purge flow would reduce process efficiency by the loss of product and excess reactants in the purge stream. Too large of a recycle flow would cause an increase in diluting component concentration reducing reaction rate.
Greg: In a first-principle simulation I built of a two-stage reaction commonly used in labs for chemical engineering courses, I found out that if I increased the flow of a reactant feed (to match an increase in the other reactant flow or to increase production rate), the reaction rate decreased. Needless to say my director was not impressed when he tried out my simulation by increasing the product concentration setpoint and seeing the product concentration go down and controller output saturate, maximizing the manipulated reactant flow. To make things worse, I had an on-line calculation of yield that showed a loss in reaction efficiency. The whole first principle simulation effort by me suffered a setback. I don’t think I ever fully explained that the decrease in reaction rate was due to the dilution effect of the high solvent concentration in the feed stream. The sign of the process action reversed for the product concentration controller on the outlet of the first stage reactor. I suspect the problem is created in university labs by the use of more dilute streams for student safety. Achieving the desired stoichiometric ratio runs into a limit where the reaction rate and production rate decreases because the molar concentration of reactants decreases from the dilution effect of an increase in recycle stream flow unless the reactant flows are increased. However, an increase in recycle flow increases the amount of reactants and products staying in the system, and the increase in total feed flow decreases the residence time and therefore the available reaction time. Needless to say, recycle streams can have you running around in circles. Studies by Bjorn Tyreus at DuPont and William Luyben at Lehigh University documented in the 1993 Industrial Chemical Research 1993, Volume 32 paper, "Dynamics and Control of Recycle Streams. 4. Ternary Systems with One or Two Recycle Streams," show that a main recycle stream must be on fixed flow control to prevent a snowballing effect. Often the impurities that accumulate are not measured or their effects are unknown. In this plant the build-up of an extremely small impurity concentration identified over 50 years ago in a research report to cause a decrease in reaction efficiency had been forgotten, causing decades of debottlenecking projects to fall short of goals.
Stan: Could you do much to increase production rate by tuning controllers?
Owen: Many of the controller outputs were maxed out because the production unit was running at several times its nameplate capacity. For example, the vacuum system was pushed beyond its limit. Debottlenecking projects moved the bottleneck around.
Greg: Putting in override control would actually decrease production rate if the maxed-out valve position or pump speed does not cause quality control problems. I found this to be the case in an opportunity assessment to increase the production rate of a large continuous polymer plant. When I found out after two hours of discussion that the product pump speed was maxed-out in each polymer line, and the large storage tanks blended out any product variability so quality was not an issue, I decided continuing the assessment was a waste of the dozen or more specialists in the room. I got considerable flack for this attempted short circuit, so the meeting continued. After two days the conclusion was the same. A process control system could not be designed to increase production rate. You needed to increase equipment and piping sizes. What have you been doing most recently?
Owen: I have been doing mostly DCS migration projects. I have gotten to solve the actual problem rather than the rule-of-thumb placeholder in the DCS configuration. People have forgotten what they have. There are details not given that are more important than they know. Some customers have figured this out, and you have a much easier time getting to a successful project. Some production units have good operating descriptions or good piping and instrument diagrams (P&IDs), but rarely both. Some plants want you to create the operating descriptions rather than do a literal translation of the old DCS configuration because of dead code. Some hold back the existing configuration to see what you can do that is better. Real-time simulation is always used for checking out the configuration, except in some small literal translations. Thorough testing is done by engineers before the system is ready for operator training.
Stan: What are the additional requirements for pharmaceutical plants?
Owen: Pharmaceutical plant configurations must be validated before operators see the system. The validation process requires a lot of complex paperwork and a level of detail that is equal to or greater than the configuration.
Greg: This column on the complete engineer is complete. We recommending trying out in the online version the Number 1 App of "Top Ten Apps as Holiday Gifts for the Automation Engineer Who Has Everything"
Top Ten Apps as Holiday Gifts for the Automation Engineer Who Has Everything
(10) Translate common speech to engineer talk
(9) Translate engineer talk to common speech
(8) Explain automation engineering without causing glazed eyes
(7) Guide for grooming
(6) Guide for clothing
(5) Learn to be as funny as "Big Bang" characters
(4) Chill out
(3) Enjoy mindless fun
(2) Forget about logic
(1) Anticipate and understand needs of spouse.