Why Academia Keeps On Solving The Wrong Problems #2: “Flowsheeting” and “Process Design”

I’m usually not at all keen on the term “flowsheeting”, a term which tends to be used in professional practice mainly for something done by software engineers. I’ve defined it in my Dictionary of Chemical Engineering Practice as follows: 

Flowsheet: aka flow chart. A term covering a wide range of different diagrams of shapes, connected by lines to represent a sequential series of operations or processes; used by doctors as well as software, electrical and process engineers. A deprecated term for a PFD for this reason (though) the term is popular in academia, often used to describe non-professional textbook illustrations or modeling program outputs.

Notwithstanding my objections to the term in this context, I’m going to use it in this article to discuss the things which academia does under this banner, i.e. the production of the “non-professional textbook illustrations” which they like to call PFDs or P+IDs.

On social media recently, I came across an attempt by academics to produce what they referred to initially as a “Google translate” for process engineers, intended to translate PFDs to P+IDs (and later as “ChatGPT” to try to stay current).

Of course, actual process engineers know that P+IDs are not simply PFDs ‘in another language’. The two diagrams are a summary of two completely different aspects of process plant design. Neither does a process engineer need an analogue of “google translate”, since they speak both ‘languages’. Anyone who needs Chat GPT to produce a PFD needs firing.

The availability of “P+IDs” in academia is a longstanding problem. Unlike PFDs, P+IDs are largely to do with process control, but university degree modules on “process control” are actually applied mathematics courses.

Students come out of them familiar with Laplace transforms and using Matlab, but without any real understanding of what a control loop is. Since so many current academics never practise as engineers, they never discover that knowing what a control loop is, is a great deal more useful to practitioners than the Laplace transform. Or indeed Matlab.

Anyhow, those “Google translate for process engineers” academics offered an illustration of a “PFD” which had supposedly been translated, using their approach, into something they thought was a “P+ID”. In fairness to them, they did not know that P+ID stands for piping and instrumentation diagram, rather than “process and instrumentation diagram”, or indeed (it seems by their product) anything about what a professional PFD or P+ID looks like, contains, or is for. They had turned one non-professional textbook illustration into another, and very simple examples at that.

What was most impressive was just how wrong it was possible to get this simple task, and how unselfconscious they were about sharing with the world just how very bad their “solution” was.

Lots of professional engineers weighed in to explain the problems with their approach, but we were all resolutely ignored: only the positive comments were acknowledged by the academics. I know social media is all about affirmation, but surely academic “chemical engineers” should address professional concerns about a product which ultimately appears to be intended for professional use (however wrongheadedly)? As the Captain says in Cool Hand Luke, “What we've got here is failure to communicate. Some men, you just can't reach.”

Academia exists in its own little bubble. I know that they think the same of practice, but whenever academia develops something which helps us come up with a safer, more robust, more cost-effective design, we adopt it pretty quickly. They just don’t do this very often. We aren’t ignoring academia (though most professional engineers have neither access to scientific literature nor the time and inclination to read it). They simply fail to understanding the very nature of our problems.

Let’s turn now to “process design”, and the idea that process simulation will replace the design process in future. This has been the fond hope of academics since computers were first devised, and forms the basis of much so called “process design” teaching.

Simulations are made of mathematics. Mathematics is perfect, but the real world is made of rather more complex and imperfect stuff, and contains even less perfect, even more complex people.

Materials and feedstocks are never perfect. Equipment is never perfect and tends to become increasingly imperfect over time. Operators are never perfect and also tend to become increasingly imperfect over time unless well managed. Plants are never constructed exactly as per the original design.

I cannot find any research papers in which modeling and simulation programs are used in their straight-out-of-the-box format to design a plant, (as is standard in academia) and the predictions of the model then validated against the real plant. The most impressive support I can see for the validation of this approach is that after such programs have been “calibrated” with large volumes of full- or pilot-scale plant data, they can predict performance of small sections of plant to a reasonable degree of accuracy.

This is, however, simply using the program to contain empirical data, with the program itself only filling in small gaps in the data, and even this approach has only “worked” with a couple of linked unit operations at a time, rather than a complete plant. A professional designer could probably have produced the design of a complete plant in far less time than it took to achieve even this limited success.

So even if it became possible in the future to accurately model the full complexity of the real world, it would take longer to program the model than to simply design the plant, and the part of the process design which a simulation describes is at best the mass and energy balance, and Process Flow Diagram (PFD).

Only someone who thought that process control, safety, plant layout, hydraulic considerations, and cost were trivial side-issues, would think it plausible that simulation could replace process design. Process design is in any case one small part of process plant design.

Even if this hurdle was overcome, plants designed by computer would be understood by no one, and would therefore be incapable of verification as sufficiently safe, robust, and cost-effective. Following such an approach would be based on a misplaced faith rather than reasoned professional judgment.

The following anecdote from a fellow engineer, involved in writing simulation software after significant operational experience, is illustrative:

In 2002, I sat in a seminar at an Aspentech conference in Washington DC, where most of the presentations were from academics, although I recall BP Chemical presenting something.  This would have been the beginning of what has come to be called “generative” design – whereby you give a computer model constraints and goals and let it come up with lots of options and solve towards an “optimum”. They were presenting about using this kind of solver to come up with novel separation arrangements to reduce capital and operating costs, and lots of academic case studies on how this could be used. The chair of the session, a well-known professor from a highly regarded US university, lamented that industry wasn’t adopting this technology and he didn’t know why. I pointed out that the average engineer in industry didn’t understand how the math worked and couldn’t explain the resulting design. The presentation from BP Chemical even laid it out – they used the technology to devise a new separation scheme from some purification, but then still have to tear the result apart and use traditional design methods to validate that it could be done (and correct the places where the mathematical model provided something unworkable).

Even if all of the issues with this approach were overcome, these programs do not produce the engineering deliverables which designers do, so they are therefore not design tools.

This wraps up my mini-series on academia, but not because I have run out of ways in which I believe academics get it wrong. Two articles on this theme is enough to be getting on with!

Read Why Academia Keeps On Solving The Wrong Problems #1: P+IDs, PFDs and Plant Layout