As I finish writing this blog, it is the morning after landfall of Hurricane Zeta, the 27th named storm in the 2020 Atlantic hurricane season and the 11th to make landfall in the US. It has been a record-breaking year for hurricanes with significant production impacts on oil and gas platforms in the Gulf of Mexico and many industrial process facilities across the Gulf Coast. Zeta was just one more thing to make 2020 an unprecedented year in so many ways.

I’ve worked in the process industry for years, many as a DCS programmer. I would often be asked by others something like “How long did it take to program that?” In my response, I would explain that it only took me X hours to program that sequence, but it took me 3X hours to program all the hold conditions, interlocks, and what-if logic.  

Programming how things should work when everything goes according to plan is easy. Programming something to be bullet proof, to account for all the possible things that could go wrong, that’s what takes time. What if that valve doesn’t open, that pump doesn’t start, that vent line plugs, the agitator fails, the reaction doesn’t proceed as normal, the operator doesn’t act when prompted? A good process control engineer thinks through all the possible scenarios they can conceive based on their experience. It takes time, and even in the best of situations it’s an imperfect exercise. After working through all the scenarios, interlocks are the last line of defense. Interlocks don’t care how you got into a possibly hazardous situation, they just recognize it and act. They keep you safe, but they aren’t great for process uptime. And they won’t keep you from getting that control room call at 3 a.m. or ten minutes into dinner out.

Process design engineers encounter similar issues when designing a new process or trying to increase capacity of an existing process. Given desired production capacity, it is relatively easy to determine the flowrate and horsepower requirements for a pump, how many trays need to be in a column, or how big a reactor needs to be. Historically, the next round of design involves experience and intuition. Determining what could go wrong, how likely is it to happen, and what can be done about it. Designing for the what-ifs. Is a single pump here good enough, or should I put in a parallel spare? How big should I make this buffer tank? Should I just oversize everything so that when they tell me I have to cut 20% of the capital budget I’ll have fat built in? At least with an expansion project you have a little history of where the current pain points seem to be.

Mark Twain popularized a saying that I have always liked, "There are three kinds of lies: lies, damned lies, and statistics."  There is a similar quote by statistician and economist Robert Giffen. Perhaps even better for consultants like me, it reads, “There are three degrees of comparison among liars. There are liars, there are outrageous liars, and there are scientific experts.” Technically, I’m an engineer, not a scientist, but I still like that one.

Many people seem to have a general distrust of statistics these days. It isn’t unwarranted. Selective choice of datasets, charts with vague or missing axis labels, and lack of context are routinely used by content creators with hidden or even obvious agendas to push their messages. However, when used for good statistics can have a great impact. Statistics is the ultimate tool to see the forest for the trees. And a particular kind of statistics, the Monte Carlo analysis, can help us design and run our industrial plants more efficiently than ever before.

When you first hear the term Monte Carlo method, you may think about James Bond rolling dice at the craps table in a posh casino. You wouldn’t be far off, at least in terms of rolling the dice. Based on statistics, computer simulations can roll the dice hundreds or thousands of times to determine how often our favorite spy will roll an eleven. You can never predict what the next roll of the dice will be, but if you roll the dice enough you can predict how often you will roll any combination very precisely.

Monte Carlo simulations can also compute the value of Pi by randomly throwing dots into a square with a circular sector drawn in. Pi is a great use of statistics to provide context and meaning to a rather abstract concept, an irrational transcendental number we all memorized in school, but likely never fully understood. This example also illustrates visually how randomness can lead to accurate and precise results through the power of statistics.

Similarly, you can use a Monte Carlo simulation to determine how often you will fill a storage tank, a pump will fail, or the 11th hurricane of the season will cut power to your plant. More powerfully, you can simulate how often two, or three, or more of these disruptive events will happen simultaneously allowing you to predict increasingly complex scenarios, the seemingly “unpredictable.” This sort of simulation would give you a sensitivity analysis indicating what was truly most critical in your plant. It would show where small changes would have big impacts, and where other changes may have no negligible effects at all.

If you were to add in some economic data, like capital cost data for equipment, cost of utilities, selling price of products, labor rates, third party expenses, fines, input and output rates, you could not only determine where changes would have the most impact but do a cost/benefit analysis of each of those changes. It would take a team of math PhDs and programmers to pull off something like that, but that’s the sort of technical challenge AspenTech specializes in, and just what Aspen Fidelis™ is designed to do.

Aspen Fidelis can help you make smart capital investment choices during the initial plant design or for a later expansion.  It can also give recommendations on how many spares you should keep in the warehouse versus when it would be okay to wait for spares to be shipped from the vendor. You can even determine how best to run a plant with partial failures or if you need to turn down production volumes.

 

To learn more, read our white paper, Process Plant Design: The High Cost of Slow Design.