Thursday, 7 November 2013

Total Process Plant Design


 Pugh Elements of the PDS.jpg



















I recently read a book on product design which drew conclusions which I think apply equally to process plant design - Pugh's "Total Design".

He calls the "design" taught in universities "partial design", resulting from the necessary breaking down of a complex and holistic discipline into things which can be taught in an academic setting by non-practitioners.

He talks of the need to contextualise this necessary simplification as just a small part of the real-world total design activity, and to spell out that what is being taught is not really design itself.

The greatest problems with this approach occur when we mistake the bricks for the building - many universities unfortunately teach process design as if processes do not happen in process plants.

This may be elaborated in academia into approaches which ignore all of the real difficulties of total design in favour of an attempt to make design a sort of applied maths in which one or two elements of partial design are combined, to make what they mistakenly think is an integrated model.

In such approaches, the three most important measures of a design's quality (its cost, safety and robustness) are glossed over, in favour of optimisation of a small number of parameters in a greatly simplified model of just part of the plant.

Process plant design is an art, whose practitioners use science and maths, models and simulations, drawings and spreadsheets only to support their professional judgement. It cannot be supplanted by these things, as people are smarter than computers (and probably always will be).

Our imagination, mental imagery, intuition, analogies and metaphors, ability to negotiate and communicate with others, knowledge of custom and practice and of past disasters, personalities and experience are what designers bring to the table.

If more people in academia understood the total nature of design they would see the futility of attempts to replace skilled professional designers with technicians who punch numbers into computers. Any problem a computer can solve isn't really a problem at all - the problems of real-world design lie elsewhere.

In a related issue, on the latest TCE letters page, an article in the previous issue by a vendor of  modelling software which made claims for their software's abilities to co-optimise two unit operations comes in for incredulous criticism.

Setting aside the issue of whether it is worth modelling so precisely something which cannot be built precisely like the model, (because no real world artefact ever is) - If it is questionable whether we can optimise a simple system of a reactor and associated separation process together, it is ridiculous to think we can model a whole process well enough to substitute modelling for design.

Creating a dynamic integrated system level model of a whole process is exactly what process plant designers get paid to do, and they do it in their heads.

The research programme "towards zero prototyping" which inspires and funds that software vendor is a pipe dream. Designers are necessarily people, and any model which runs on a computer falls far short of a reliable description of the real world.

The letter to the TCE pointed out that extensive pilot plant work generated the data which was fed into the modelling software. As the pilot plant work was essential and the software was entirely optional, no progress towards zero prototyping seems to have been made.

Engineering deals with problems which it will almost certainly always be far quicker to ask an engineer to solve than to program a computer to, even if we had the data (which we can never have on a plant which hasn't yet been built), a computer smarter than a person (which we will probably never have), and a programme which codes real engineering knowledge, instead of a simplified mathematical model with no input from professional designers.

I wonder how doctors would feel if scientists and mathematicians suggested that they could produce an expert system which would exceed their competence without consulting any doctors?

This is a classic academic purist's mistake: The psychologist claims that sociology is just applied psychology, the biologist says that psychology is just applied biology, the chemist that biology is just chemistry with legs, the physicist that chemistry is just applied physics, and the mathematician that physics is just applied maths. Emergent properties are irrelevant to the theorist, but in practice they are everything.

To paraphrase XKCD: Engineering is to maths as sex is to masturbation.

2 comments:

Yixiu said...

we get another text book now!

Sean Moran said...

It's worth a read. Wait until you see my next blog post....