Tuesday, 6 May 2014

Modelling and Simulation in Process Plant Design

I'm writing a bit of the textbook dealing with use of modelling and simulation for process design, and there seems to be a problem.

I don't seem to be able to find the scientific papers where someone showed that process modelling and simulation programmes are capable of even a fair match with real full scale plant without you inputting lots of data from a real plant just like the one being designed.

What is not well known in academia is that such data is very hard to get if you work for a process design company, as it is jealously guarded by operating companies. Operating companies also tend to only record data which is useful to them. This may well not include the data modellers need.

I can however find lots of papers (and IChemE guidance) which are concerned with the paramount importance of undertaking model verification (debugging) and validation (making the model match reality), stages which seem to be left out of academic use of such models.

Including these stages would draw attention to three things:

1. Even commercial models are full of bugs and bear little resemblance to reality before real operating data is added. Models you write yourself with Matlab etc. are far worse.
2. Genuine plant operational data is very hard to come by
3. It is almost always quicker to design a plant than to model it

The supposed advantage of modelling is speeding up the design process - has anyone else noticed that all of the articles in the TCE plugging modelling for design are written by modelling programme vendors?