“Computer simulation software - modern oracle”

1. “Computer simulation software-modern oracle” A lecture at the Congress-Exhibition by Brian Spalding

2. 1. Outline of the argument • foresee, and so avoid, dangerous events; • select, from the options which are open, those which best promote happiness and well-being; and • create opportunities which never before existed. «What will happen if .. » is the most important question which a conscious being can ask. Answering it rightly more times than not is what keeps most of us safe, healthy and reasonably prosperous; for it allows us to:

3. 1. Outline of the argument • Society seeks, through education, • to inculcate in the young the habit of asking thisquestion; • and to convey how the right answers can be arrived at, • which means to teach to all: techniques of prediction. In essence, all such techniques are the same: examine the past; and if elements of the present are seen there, suppose that what transpired before is likely to happen again. Thus: “When last I pulled the tail of a cat, it scratched me; so, if I do it again, another scratch is what I must expect”. It is a sound principle.

4. 1. Outline of the argument In ancient times, oracles were consulted on matters of importance, as being best fitted by age, experience or connections to foresee, what the past implied about the impending future. That too was a sound principle, for those who could afford the oracle’s fees!

5. 1. Outline of the argument How are these principles applied in engineering? If the task in hand involves little novelty, as when one more engine is to be built for an established and satisfactory production line, simple repetition of past actions is what the principle dictates. But when the performance requirements have changed, exceeding what the old engine is capable of, novelty is needed; and what is new has, by definition, no past to be examined. What to do?

6. 1. Outline of the argument • There is however a more general encapsulation of the past, which we call science; • and for engineers it takes the form of: • laws of conservation of mass, momentum and energy; LomonosovNewton Joule

7. 1. Outline of the argument • laws of transport of those same entities by • diffusion, viscous action and heat conduction ; FickNewtonFourierHooke • laws of deformation of solids in response to mechanical and thermal stresses (Hooke);

8. 1. Outline of the argument • laws governing rates of chemical transformation, and electrical and magnetic interactions. Arrhenius Faraday It is such laws,to which the engineer must turn,whenever actions without precedent are contemplated, in order to answer “what will happen if ” questions

9. 1. Outline of the argument • The present lecture explains: • how simulation-by-computer has become the engineers’ favoured prediction technique, and • how specialised software packages have become the oracles which they consult. • Two things the modern oracles share with the ancient ones: • they cost money(or sheep, oxen or other currency); and • their pronouncements are never 100% reliable. The reasons for both will be explained. First, however a single example will be shown.

10. 2. Example: What happens if a fire starts in a building? Though their causes are various (electrical faults, carelessness, arson, ‘spontaneous combustion’), undesired conflagrations are a fact of life, for which building administrators must prepare by providing means for: • extinguishment (e.g. sprinklers, foam canisters) • prevention of spread (e.g. fire doors); and • escape for personnel and livestock.

11. 2. Example: What happens if a fire starts in a building? But how can they determine whether their preparations will be adequate? Certainly they can check them against the requirements laid down in the municipal or state Building Regulations; and their compliance may save them from prosecution even if the preparations failed. But that will be a poor comfort. They can do better: consult a ‘modern oracle’. Why? However they are started, fires spread in obedience to the general physical laws which were listed above, as constrained by the particular circumstances in question.

12. 2. Example: What happens if a fire starts in a building? • The relevant circumstances include: • the locations and thickness of the walls which impede the flow of air; • the locations and sizes of the apertures (doors, windows, air vents) which allow air (and smoke and flames) to pass through; • the amount and location of combustible material, not only that in the source of the fire, but also that in furniture and furnishings (e.g. chairs, tables, curtains); • the positions of water sprinklers, and the flow rates of liquid through them; and • external conditions such as the strength and direction of the wind and, if it arrives in time, the water injected by the firemen’s hoses.

13. 2. Example: What happens if a fire starts in a building? • During the last thirty years computer-software packages have been created which embody both the general physical laws and (templates of) the particular circumstances. • Their users can, in effect: • open a ‘store cupboard’ containing • fires, • walls, • windows, • tables, • sprinklers, etc.; • place them on a ‘stage’,in appropriate relative positions ; • declare which physical laws are to be obeyed; and then • observe, as though in a theatre, how the ‘play’ develops. • The users have supplied the “if”; then the package pronounces upon • the “what will happen ”, just as did the ancient oracles.

14. 2. Example: What happens if a fire starts in a building? Computational Fluid Dynamics,abbreviated toCFD,is the name which has been given to the body of knowledge and skill which forms the basis of such packages. Those who participate in its various aspects include mathematicians, computer programmers, physicists and engineers. Since last 80-ties it has grown into a multi-million dollar industry; its potential being used particularly for aircraft.

15. 2. Example: What happens if a fire starts in a building? Some pictures will now be displayed whicharise from the solution of a very simple fire-simulation problem. It is the one which has been prepared for display in the Internet-café of this Exhibition. • The pictures show the scenario which is to be studied, namely: • an office containing a standing man, some desks and some computers. • The store-cupboard from which chairs, desks and other objects can be extracted is visible near the top of the picture.

16. 2. Example: What happens if a fire starts in a building? Store-cupboard Scene of the ‘virtual theatre’

17. Object‘People’ 2. Example: What happens if a fire starts in a building? Another man can be easily added to the ‘scene’ by clicking on the appropriate image; here it is the object ‘People’.

18. 2. Example: What happens if a fire starts in a building? One can hide or remove objects (here several objects disappeared from the scene).

19. 2. Example: What happens if a fire starts in a building? The scene can be viewed from any point of view.

20. 2. Example: What happens if a fire starts in a building? It is possible tomove and rotate objects, and

21. 2. Example: What happens if a fire starts in a building? It is also possible to bring in a third man with a chair for him to sit upon.

22. 2. Example: What happens if a fire starts in a building? Here are some results of calculation. The next pictures show colour contours of temperature and arrows indicating air motion under normal conditions on a horizontal and vertical planes.

23. 2. Example: What happens if a fire starts in a building? Finally, an animation is shown of the spread of smoke and flame when combustible material under a chair is suddenly set alight.

24. 2. Example: What happens if a fire starts in a building? • This rather trivial simulation was created in such a way that it could be executed in a few minutes on a lap-top computer. • In engineering practice, of course, situations of much greater magnitude are in question: and CFD simulations may take many hours of computation on powerful computer clusters working in parallel. • Such simulations are routinely • conducted for: • multi-storey car parks;

25. 2. Example: What happens if a fire starts in a building? • super-markets; • concert-halls;

26. 2. Example: What happens if a fire starts in a building? • under-ground railway stations; • tunnels; • airplanes; and • ocean liners.

27. 2. Example: What happens if a fire starts in a building? • In many countries such computer simulations are already the necessary intermediate stage for every application for a licence to build. • However, the licensing authorities need to be appropriately educated. • They must know that: • such simulations can be made; • the expense of making them, although not negligible, is tiny in comparison with the cost of allowing an unsafe building to be erected; and • that, although the simulations cannot be relied upon to predict • with certainty ‘what will happen if.. ’, enough tests have been made to persuade even sceptics they are good indicators of what is probable.

28. 3. Why the “Oracles” are not completely reliable • Since: • the quantitative laws of Newton, Joule, Hooke and Fourier have stood the test of time, • computers are becoming increasingly more powerful, and • physicists and chemists have accumulated extensive knowledge of the properties of materials, • it is natural to ask: • why it should be necessary to issue the warning that CFD can predict only what is probable, not what is certain?

29. 3. Why the “Oracles” are not completely reliable The answer is three-fold. First, computers are still not powerful enough; for CFD simulations proceed by representing what is in reality continuous with a discontinuous near-equivalent. In this example the computational grid was regarded asan assembly of one hundred thousand imaginary boxes, in each of which conditions were treated as being uniform.

30. 3. Why the “Oracles” are not completely reliable This is too coarse a grid to represent the solid objects or fluids. If a hundred million boxes had been used, the boxes would still be of the order of one centimetre in length, width and height; and so hardly small enough to represent, for example, the leg of a burning chair. Only the largest computer clusters in the world would have been able to handle so many; and the computation would take far too long for its outcome to remain of interest.

31. 3. Why the “Oracles” are not completely reliable Secondly, although indeed chemists have accumulated immense amounts of information about how combustion occurs, the information is too immense to be useful. Why? Engineers like to think that fuel and oxygen combining form ‘combustion-product’ gases but chemists have discovered thata great many of intermediate productsare formed in combustion, as the scheme shows. Therefore, engineers create simplified combustion models and their accuracy always raises doubts.

32. 3. Why the “Oracles” are not completely reliable Among intermediate products are those carbon-containing particles which we call ‘smoke’. Its presence has a great influence on the intensity of radiative heat transfer, which, in turn, has a great effect on the rate with which fire spreads. As if this were not enough difficulty to contend with, it has also to be admitted that, chemists have mainly confined their researches to the behaviour of pure substances. Consequently, even if one could compute with accuracy the intensity of the radiation reaching the curtains, the chemical literature contains nothing from which one could compute the speed with which the fabric would burst into flames.

33. 3. Why the “Oracles” are not completely reliable Thirdly, chemistry and radiation apart, even the fluid-flow aspects of the simulation are, in most circumstances, subject to doubt; and the reason is: turbulence. The smoke from a chimney, blown by the wind, although it certainly moves mainly in the wind direction, also exhibits seemingly random motions at right angles to it. Such randomness, which is called ‘turbulence’, pervades all flows, whether of gases or of liquids when their velocity is such as to make inertia forces exceed viscous ones. (This is the so-called Reynolds-number criterion.)

34. 3. Why the “Oracles” are not completely reliable Although turbulence has been much studied, and is represented to some extent in the CFD packages, none of those representations are known to correspond with reality in all circumstances. This regrettable fact seems likely to remain until some Newton reduces chaos to order. Until then, all CFD predictions of turbulent flows must be regarded as no more than probable forecasts of “what will happen if.. ”. When chemical reaction and two-phase effects are present (as they are when water from sprinklers interact with burning gases), the margin for error widens.

35. 3. Why the “Oracles” are not completely reliable What is to be done? The optimists and those who make their living by selling CFD packages and services based on them, find it easy to be impressed by the plausible-seemingattractively-coloured images which the packages produce. They look realistic; and often packages from different vendors give results which are qualitatively and even quantitatively similar. The second fact especially easily dismisses doubts in the accuracy of the results obtained.

36. 3. Why the “Oracles” are not completely reliable The pessimistsargue that the agreement between the packages from different vendors means nothing; for all use the same dubious models of turbulence, etc, and all are compelled to use far-too-coarse grids. They are mainly influenced by the above arguments on the inaccuracy of models.

37. 3. Why the “Oracles” are not completely reliable Aristotle’s advice is here appropriate: • The best lies between the extremes. • It entails recognising that • the CFD-based predictions are no more than indicators of probability; but • they are immensely better than the mere guess-work which is mankind’s only alternative.

38. 4. The cost of CFD The new oracles demand their sacrifices. What are they? 1. The software In the first two decades of the CFD industry, licences to use the software packages could be sold for tens or even hundreds of thousands of dollars. Nowadays their price is hundreds times less. There isno obstacle to purchase them by industrial organizations. A minor impediment to academic ones still remains. Students can usually acquire low- or zero-cost versions … and not only by ‘pirating’.

39. 4. The cost of CFD 2. The hardware Nowadays hardware costs have decreased dramatically; so much so that, were ‘parallel-computing’ more widely promoted by the CFD vendors and its merits appreciated by customers, many of the latter would acquire clusters of computers, using thus more adequate computational grids. Another possibility is to use remote clusters via Internet paying for actual services just as we pay for piped gas or water. In this case it is only a laptop that a user needs to solve great multifactor problems.

40. 4. The cost of CFD 3. The personnel Nowadays, therefore, it is the cost of hiring CFD-literate personnel which is the most serious impediment to the extension of computer simulation. However, although suitably competent personnel are indeed in short supply, that is in part because what that competence comprises has not been adequately defined. Consider the situation of a Building-Licensing authority which, having been impressed, by the demonstration of section 2 of this lecture, decides: “Yes, it’s good. We must have it. Let’s hire some staff.”

41. 4. The cost of CFD What kind of staff should it hire? Not, I would suggest, a person who has just completed a PhD study concerned with turbulence or radiative heat transfer modelling. Such specialists are expensive; but their prolonged specialisation is likely to cause them to over-estimate the importance of one small aspect of the subject, and to lack well-balanced common-sense. Intense enthusiasts for a particular software package are also to be employed only with caution.

42. 4. The cost of CFD • Better are those, • who have experience of several packages, • who recognise that most of their claims to superiority are ill-founded, and • who understand the limitations from which they all suffer. • Best are those who are pragmatically sceptical: • who can extract value from order-of magnitude predictions, • who have so strong a grasp of physical reality that, when the occasion arises, they can conclude that • ‘A particular computer simulation simply must be wrong’.

43. 4. The cost of CFD Such a person seeing the animation of the fire in a computer room, would say: • “Why does the hot air not rise vertically from the burning chair?” • “Surely the arrows appearing above the man are too big!” • “I can not believe that simulation.” He would be right and his doubts would impel him to search for, and find the human error which causes the computer to produce unrealistic simulations.

44. 4. The cost of CFD Once the error was corrected, the simulations became more reasonable, thus: But human error must always be suspected and guarded against.

45. 5. Should computer simulation be used in education? • The above questionis the most important one for our Congress. • Arguments in favour: • computer-simulation packages do embody physical principles which all students should study; • they are cheap to obtain and to run; • they convey a sense of immediacy akin to that of hand-on experimentation; • their use allows some burdensome topics to be omitted from the teaching syllabus.

46. 5. Should computer simulation be used in education? Arguments against: There are some persons for whom every novelty is bad per se. It is not necessary to consider their opinions. Others may adduce the expense of introducing computer simulation as a teaching tool; and it is not negligible because first it is the teachers who will have to be taught. However, the most-frequently-propagated negative view is that computer simulation produces such attractive and plausible-seeming output that students will be persuaded to believe that they correspond with reality. That is perhaps possible; but only if the teachers are themselves naive, having not been well-taught.

47. 6. Concluding remarks • The use of CFD has been increasingsteadily in the last three decades for the design of: • vehicles for land, sea and air travel; • their engines, and those of stationary power plants; • chemical and nuclear reactors; • gas and oil pipelines and the pumps and compressors connected with them; • kinds of electrical equipment including computers themselves, • and in many other directions. Its use seems certain to continue to grow. It is therefore obvious that with the framework of the «Global Education» project the educational systems of the world should prepare their students to participate and to contribute to it.