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

This is the first of two articles looking at the shortcomings of academia in the context of problem solving in process engineering. My title is possibly a bit generous, since it implies that academia has actually solved some real-world problems, which I consider pretty unlikely (unless we are counting self-defined problems such as maximising the number of fee paying students per hour of expended lecturer time, but I digress).

I suppose the best place to start here is to differentiate between problems and tasks, and between commonplace and what are known as “wicked” problems.

Practising engineers understand that many engineering problems are either too complex or too vaguely defined (or both!) to be analysed rigorously within the resources available. In other words, the challenges of most engineering problems can be split into two components: uncertainty and complexity. Engineers mostly cannot take a ‘rules-based’ approach to problem solving because of these challenges; insufficiency of data being a key issue which many engineers face.

On the other hand, if a “problem” isn’t so uncertain or complex that it can’t be solved using a rules-based approach, then it is probably not a genuine problem at all, but merely a task.

In process engineering design, we are faced with combinations of problems and tasks. The tasks might include activities such as checking the sizes of unit operations using heuristics, but these are mathematically trivial, and the product may well be ambiguous. When it comes to the problems, there is no opportunity in professional life to start thinking that engineering is a branch of applied mathematics, or that a computer can solve them — computers can only carry out tasks. Common sense is what is needed, and a feeling for ambiguity, qualitative knowledge, and multidimensional evaluation of options: in short, professional judgment.

Many of our engineering challenges are not just problems, they are “wicked” problems. “Wicked” problems are (amongst other things) unique and ill-defined, and have no true or false answers. Some of our problems are the worst kind – “superwicked problems” – such as global warming, because engineers are both arguably their cause and their only hope for resolution. Both these kinds of problems also include people issues. (Many engineering problems are in my experience actually entirely people problems, but that would be another article entirely …)

Coming back to the point of this article, let’s look at how academia has fared so far in its attempts to solve the everyday problems of process plant design. Its main contributions are in what is most broadly described as “process flowsheeting” and “plant layout”. However, do bear in mind that many academics have never seen a real set of P+IDs and PFDs – some of them don’t even know what P+ID stands for – so, in case any academics have read this far, let’s consider how P+ID is (and isn’t!) defined in my Dictionary of Chemical Engineering Practice:

piping and instrumentation diagram aka P&ID, P+ID: The process engineer's signature drawing. A drawing setting out the physical and logical interrelationships between process plant components in the form of a topologically correct symbolic drawing which shows the unit operations, piping, and instrumentation of a process plant.

process and instrumentation diagram: A strongly deprecated PFD/P&ID hybrid

process flow diagram. A diagram which shows in outline the main unit operations, piped interconnections and mass flows of a process plant. It represents the mass balance, and resembles in many ways a simplified P+ID.

So, a P+ID is NOT a process and instrumentation diagram, and any AI program which supposedly translates a process flow diagram into a process and instrumentation diagram is barking up the wrong tree, because “How do I turn a process flow diagram into a process and instrumentation diagram?” is simply not a question which chemical engineers ever seek to answer. I’ll come back to this in a later article but let’s just say for now that there are many other fundamental misconceptions in the academic approach to “flowsheeting”, although they do at least give it some research time and attention.

When it comes to plant layout, however, the majority of academics consider it to be so trivial an activity that they ignore it altogether, whilst the brave few who do get involved in plant layout research have turned it into a “problem” (actually a task) so far divorced from its real-world counterpart as to be of zero practical utility.

I wrote an article in PII a while back describing how an anonymous academic had accused me of being horribly out of date with the latest literature on academic attempts to address automation of plant layout. It had only been a few years since I last took a hard look at it (at which point it was at best no further forward from a practical point of view than it was in the 1980s). Despite said academic’s brazen tone, I nonetheless felt it important to double check whether academia had indeed finally cracked this recalcitrant problem in the last five years, so I contacted the researchers responsible for the ten most cited papers in the field. I didn’t have high hopes, because their papers suggested that they had simplified the problem to the extent necessary to allow a computer to solve it, rather than trying to make a computer good enough to solve the actual problem, but I also didn’t want to judge them unfairly. I’m not prejudiced against such attempts – I just knew that they had made no real progress for decades.

One thing which was notable in my literature search was that the series of seminal articles on plant layout written by practitioners (most notably Kern) published over the years in Chemical Engineering Magazine – which form the basis of discussion in all chemical engineering textbooks – were entirely uncited in the chain of references leading to current academic papers.

In addition, what I found out from those willing to talk to me (most were not, but the most highly cited authors were) was that no-one has ever used such software in the real world, and no-one plans to. There is commercial software which assists with layout, developed without academic input, which carries out repetitive, mindless rule-based scutwork such as clash detection, and materials takeoffs. It takes a great deal of setting up, but on large complex projects, it is worth the upfront investment of time. Such software is intended to assist the piping engineer with the stupid, patient type of processing which computers clearly excel at, but nobody expects it to replace a “piper”.

It seems as though academics are, as usual, trying to turn engineering into applied mathematics, because that is what they know best. It is telling that there has been no progress in codifying professional heuristics since people like Jim Madden were attempting this back in the 1980’s. Madden even developed software based on his approach, which eventually grew into PDMS, one of the commercial products currently available. Now, I’m not sure the new commercial software does all that much of importance that the original did not, it just does it way faster, and with prettier graphics. Jim Madden’s rules and methodology can be followed by people as well as computers, so it is included in my plant layout book as one of the five alternative approaches.

Thankfully, since plant layout is pretty much off academia’s radar, at least they have not yet misappropriated commercial software in the way they have with simulation and modelling software. I’ll look at how and why this has happened in my next article.