MICAFI

MICA Final Review Meeting

An abridged version of

Final overview of the Outcomes of the MICA Project

By Brian Spalding, (CHAM Ltd)

June, 1998

Summary

The aim of the MICA project, namely to create, test and validate a Model for Industrial CFD Applications, has been largely achieved. What has been demonstrated is that the use of Virtual-Reality-based user-interface software packages, by persons who may not be "CFD-literate", is practicable, and may often be the most convenient and economical means for solving their design and analysis problems.

Many lessons have been learned, one of the most important being the necessity to create and disseminate a flexible CAD-to-CFD translator program; these lessons will influence the manner in which the commercial exploitation will be carried out, by means of the Simuserve enterprise.

This review concentrates on spelling out the lessons and describing current plans for exploitation via Simuserve.

Contents

Changes in the perceived needs of customers for CFD

The original concept

The spectrum concept

The importance of the CAD-to-CFD connection

The reduced importance of "body-fitted" grids

Differences between the needs of the various application sectors

Text

Changes in the perceived needs of customers for CFD.

The original concept

CHAM's experience of commercial CFD has, in the past, involved the provision of services of two kinds, namely:-

the performance of consultancy work, under contracts requiring CHAM's engineers to use CHAM's software so as to create flow simulations of interest to the customer, who would receive, as the deliverable, summary reports and recommendations; and

sale of the software to the customer, who would then perform the computer simulations himself.

The idea underlying MICA was that many customers would benefit from a service lying between these two extremes. In such a service, CHAM, or some other possessor of powerful flow-simulation software, would perform the calculations, and possibly supply some advice, while the customer would be responsible only for setting up the problem and for reviewing and interpreting the results.

The customer’s tasks were to be made especially easy by the use of virtual-reality techniques, and by the attachment to the flow-simulation software of intelligent self-adaptive data-input modules. Moreover his financial commitment would be small, because he would not need to purchase either expensive software or still more expensive hardware; and he would need no highly trained CFD specialists on his staff.

What has become apparent, during the course of the MICA experience, is that:-

the current VR interface is not yet quite easy enough to use, even for users having the greater-than-average expertise of several of the MICA partners;

the introduction of intelligence into the software has proved to be more difficult than was supposed, because of the great variety of flow situations for which customers require simulations;

even if the interface were as good as it could (and will) be, and if the self-adaptive intelligence were much greater than it is currently, use of the VR-plus-remote-computing facility will still be perceived as excessively daunting by many persons and organisations which could benefit from the use of CFD.

1.2 The spectrum concept

Two conclusions have been drawn from this experience, namely that, if the CFD-using community is indeed to be greatly enlarged by the use of the Internet:-

more work must be done to improve the ease-of-use of the user interface and the intelligence of the software; and

a richer spectrum of services should be provided, in which spectrum the original MICA concept lies somewhere in the middle.

The spectrum would include:

consultancy as described above, whereby the customer would supply the data about the shapes and sizes of the objects and circumstances in question (preferably by way of CAD-package output files of standard format, e.g. STL or DXF);

consultancy with enhanced deliverables, in which the service provider (CHAM, say) would deliver the results of the computations not only by way of a report, but also by way of output files which the customer could import into a viewer-only display package of VR type;

consultancy with assisted problem specification, in which the customer would purchase and use a geometry-only version of the VR data-input package, and would transmit his so-specified problem to the service provider which would conduct the computations, returning the results for the customer to display and interpret;

consultancy with further-assisted problem specification, in which the customer would additionally himself use the data-input interface (i.e. the VR editor) so as to define boundary conditions in physical terms, e.g. inlet-flow rates and temperatures, heat sources, while still leaving the physical modelling and numerical inputs (turbulence-model choice; grid dimensions, iteration numbers, etc) for the service provider to select;

MICA-type computing-plus-advice provision, in which the customer would take over the full responsibility for setting up the problem and interpreting the result, the service provider's role being reduced to that of providing the flow-simulating software, running on a powerful parallel-architecture machine, and advice if it was asked for;

conventional rental or sale of software and hardware, in which CHAM's role would be to provide the software/hardware combination which would enable the customer to perform and interpret the simulations wholly on his own; and

provision by CHAM of the software only, for use on hardware which the customer would select for himself.

In this spectrum-of-services concept, the expertise level of the customer would have to rise, the farther down the list appears the service for which he asks.

His choice would be made, of course, in accordance with his current capabilities, needs, financial circumstances and personnel resources; and the ability to choose differently at different times will, it is to be expected, be a benefit which he will value.

1.3 The importance of the CAD-to-CFD connection

At the time at which the MICA project was conceived and planned, it was considered that the Virtual-Reality interface would be so powerful and attractive that it would meet all the needs of the prospective users.

The fact that these persons would already have spent much time in creating computer representations of their designs by means of CAD packages, and would not wish to start all over again in VR, was not recognised with sufficient clarity or give sufficient weight in the planning.

This fact was however quickly discovered as the interactions with partners began; and it was reacted to (albeit not so quickly) by CHAM.

Specifically, interfaces were created between certain CAD packages and VR, in the form of "file-translators". Two exist at the present time. The first translates STL files into VR-data files; the second does the same for DXF files.

It is true that more work is necessary in respect of the latter; for it has been experienced that the files which suffice for architectural-design purposes may not define building structures (for example) adequately for CFD. There may be gaps between walls and roofs; and inconsistent information may be provided about which is the inside and which the outside of the surfaces which bound a solid object.

The experience afforded by the MICA project has brought these matters to light; and solutions have been arrived at in all individual cases.

Ideally (from the CFD-software point of view), the CAD-package user who intends that his product should be used for CFD would learn how to avoid inconsistencies and omissions. Then the translator's task would be straightforward.

From the CAD-package user's viewpoint, however, it would be ideal if the translator were provided with sufficient intelligence to enable it to guess correctly what the user would have done if he had been "thinking CFD".

CHAM's current DXF-to-VR translator, for which "interpreter would be a better name, does possess some intelligence; and it will be equipped with more as experience is gained of what the most common deficiencies and inconsistencies actually are.

1.4 The reduced importance of "body-fitted" grids.

It was decided at the start of the MICA project that attention would be confined to flow-simulation problems which could be handled by cartesian or cylindrical polar grids. This was thought of as an unwelcome necessity, enforced by the limitations of time and resources.

As things have turned out, the limitation has proved to be advantageous; for it has led to such far-reaching enhancements to CHAM's pre-existing ASAP procedure (for fitting bodies with curved surfaces into regular grids) that, for many problems the solutions are of higher accuracy than body-fitted co-ordinates could provide.

The key elements have proved to be:-

the use of the facet-wise description of objects, used for "clip-art" displays in VR, as the means of determining which edges of the regular-grid cells are intersected by object surfaces, and at what precise locations;

the recently-developed PARSOL" (i.e. partial solid) technique which appropriately modifies the balance equations for the intersected cells; and

the extension of the fine-grid-embedding feature of PHOENICS to enable it to work adequately with the default "staggered-grid" formulation of the balance equations.

1.5 The differences between the needs of the various application sectors

Ten application sectors were selected for attention in the MICA project. broadly classified into two groups, namely: "furnaces" and environment.

Even within these groups, so great a diversity of physical requirements was soon perceived to exist that the decision was taken to provide ten distinct special-purpose programs.

While considering the commercialisation of the products, it is now relevant to remark that differences exist in respect of the extent to which users in the different sectors are likely to benefit from MICA-type services.

For example, the customers for the flow-around-buildings product are entirely different in number, character and expectations from those concerned with explosions in oil platforms.; and the reliability of the predictions is also likely to be entirely different in the two sectors.

In the former, the customers are more numerous, the reliability of the predictions is greater, but the willingness to spend money on CFD predictions is lower. In the latter, the customers are far fewer, the reliability of the predictions is immensely lower, but the willingness of the customers to pay money is satisfactorily great.

In CHAM's view, both sectors of the industry can benefit from MICA-type remote-computing services; but they must be approached differently. The reasons are:

The architects who populate the first sector need quickly-provided and low-cost simulations of scenarios characterised by a great deal of detail (houses of varied shapes, trees, hills, wind-barriers and the like).

Because of the detail, high-performance computers are needed; but, though the architectural profession may find the computing cost a serious deterrent, it is unlikely to, and truly need not, question the validity of the simulations which are provided.

The safety engineers and licensing authorities concerned with explosions in oil platforms, by contrast, are, or certainly should be, greatly concerned about validity.

The truth is, however that there do not exist any certainly correct models for the rate of combustion in turbulent gases in spaces between and around solid obstructions to flow.

Therefore the architectural community can benefit from a MICA-like service because it can potentially offer a flow-simulation service which could not otherwise be afforded at all.

The oil-platform-safety community can benefit from such a service provided that the best of the (scanty) available scientific knowledge is incorporated into the software, and that the associated expert-advice givers are the world's best

The more such distinctions are thought about and expressed, the clearer it becomes that each of the sectors requires special treatment. This does nothing to invalidate the MICA concept; indeed, the opposite is true. What it does do is show that the commercial-exploitation task is not easy.