When the Kingsley Report was accepted by Engineering Board (EB) in July 1989 one of its principal recommendations was to improve the awareness of general engineering researchers in current IT developments; there was also the complementary desire to ensure IT researchers became aware of the IT needs of their engineering colleagues so that they could undertake future research of direct value to engineering, and thence the nation. Kingsley suggested one way of doing this was through joint research programmes where researchers representing both the above categories would work closely together. However, due to an inappropriate joint funding mechanism, only a few such interdisciplinary research programmes were actually supported. As a result the need for cross-disciplinary awareness still exists at a time when the Office of Science and Technology, through its White Paper, is demanding more industrially applicable research to help this country's wealth creation.
In recognition of this continued need I have been appointed by EB's Engineering Research Commission (ERC) to promote, co-ordinate and monitor an IT Awareness Initiative covering the cross-disciplinary interests of all researchers supported by the committees of ERC or Information Technology Advisory Board (ITAB). The eventual goal of the initiative will be to develop a portfolio of high quality interdisciplinary IT research having direct relevance to engineering industry, and especially its R and D. Experience within ERC has shown that such an initiative requires careful coordination. It also indicates that relevant programmes of research, suitable for SERC and from 1/4/94, the Engineering and Physical Sciences Research Council (EPSRC) funding, often results from joint workshops between ERC and ITAB researchers. With this in mind I have been charged to organise a series of events in topic areas specified by ERC and ITAB; these will present the current state of relevant IT techniques and try to cover problems that need to be addressed in the engineering disciplines. Day one of these events will be an open meeting introducing all to the possibilities of neural networking in engineering research; day two will be focused on defining future SERC research strategies for this topic. Attendees for day two will be invited to represent those active in the field or those who have important contributions to make to defining such a strategy. It is in this important respect that I ask you to help in deciding the content, nature and style of these IT Awareness events.
ERC, ITAB and I, would clearly wish to initiate events, and their resulting research, on important, relevant and timely topics. We believe those reading this newsletter will have good understanding of what IT research is necessary and possible in the short, medium and long term. They may also have a better idea than us as to what is industrially relevant and timely. We would therefore like to involve you in the process of deciding future directions of the research and the awareness events in this important area; in this respect we would like to make this Initiative, your Initiative.
You can help me best in this task by sending your ideas for the Initiative to the snail or e-mail address shown at the end of this article. Your response should be no more than the equivalent of one A4 page; give topic titles with a short description of the scope of the research on each topic; a short justification of its relevance and timeliness; and finally indicate why you feel success on this topic can be achieved through joint working by UK ERC/ITAB researchers. My task will be to coordinate your responses, to prioritise them, to confirm this priority with ERC/ITAB, and to organise relevant 2-day events as soon as possible to get the ball rolling on appropriate topics.
Members of ERC committees have already suggested suitable cross-committee topics:
Before my appointment the first of these suggestions was developed into an extremely successful two day workshop as part of the RAL EASE programme; as a result Instrumentation and Control Systems Engineering Committee (ICSEC) believes over 30% of those attending will submit grant applications in the area of intelligent modelling and control. Furthermore, in order to give me an appropriate short term task, ERC has asked me to arrange the first two events on the topics of neural networking for engineering research and engineering decision support. Discussions with many researchers across the engineering disciplines lead me to believe these topics are important and timely. However, I wish to reiterate that, once these have been organised, all future events will be structured on the basis of comments received from the community.
The topic for the first event has been decided - "neural networking for engineering research" and will be held at the Institution of Civil Engineers (ICE) 1, Great George Street, London on the 18-19th April. We already have some speakers, but we are still looking for others to give breadth to the range of presentations. If you are interested in attending this meeting for either or both of the days would you please send us an A4 position paper indicating your interest in the topic and what contribution you feel you can make to the debate.
The first event will be held in Central London, for two reasons. Firstly we felt it necessary to have a readily accessible venue that many attendees could easily reach in a day. Secondly, we wanted to reduce to a minimum the travel cost of participants attending such a meeting. This is because ERC has just decided that academic delegates have to pay a fee to cover part of the cost of all future EASE supported meetings; ERC has also decided that in future industrialists would pay full costs plus a small margin. This decision was taken after much careful discussion, however, those concerned with the implications of such a decision might like to make their views about such a strategic change known to me in a separate communication; I am sure ERC would wish to know the potential impact of all its decisions.
Fortunately we have been able to keep the costs for this first event down to £50 for academics and £100 for non-academics for Day 1 and £25 and £50 respectively for Day 2; this will include the costs of all meals and Proceedings, which we intend to publish at the time of the event. Speakers will not pay any fee. We are also presently negotiating special rates with several hotels in close proximity to ICE during the time of the meeting. We hope that ICE will be the home of all our London meetings, however it is our intention to hold at least one meeting per year in the regions. A total of 4 meetings is planned for 1994/95. Please let me know what you feel about this idea and where you would like our first regional meeting to take place.
I hope, in my time as IT Awareness Co-ordinator, to be able to co-ordinate your views to the benefit of IT research for the UK. I am a great listener, but I can only listen and co-ordinate if your views are made known to me. Therefore please take this opportunity to write to me on the basis of any of the above requests, or if you have any other ideas which will improve the development of better IT awareness. I would like to use the Engineering Computing Newsletter (ECN) to develop an open and interactive dialogue. The IT Awareness Initiative will therefore have a regular slot in ECN to keep you in touch with progress. You can also contact me at the:
IT Awareness Initiative Office
24 Blandy Road
Henley-on-Thames
In the next issue of ECN we will begin an exercise to revalidate the ECN Mailing List. This will be conducted as a POSITIVE RESPONSE exercise using a pre-paid reply card. Anyone not returning their card reconfirming they wish to continue receiving ECN will be removed from the Mailing List no later than 1 June 1994. This will mean that you will receive no further information on EASE and its activities (eg events). The current plan is to include the prepaid reply card in two consecutive issues ONLY (ECN 49 and ECN 50).
PHOENICS (an acronym for Parabolic, Hyperbolic or Elliptic Numerical Integration Code Series) is one of the most widely used general-purpose commercial codes for simulating fluid flow. It can be used via a beginner's menu system, a command input language for intermediate users and it is also intended to be expandable, allowing experienced users to interface their own modelling features to the processing module. This article briefly describes the code and some of the steps involved in implementing a user-defined linear equation solver. This uses a conjugate gradient algorithm to replace the existing solver in the flow-solving module of PHOENICS without access to the main source code.
PHOENICS uses a finite volume formulation of the differential equations with a choice of interpolation schemes (upwind, hybrid, QUICK, etc). It can be used to solve problems of up to three dimensions, including those exhibiting time-dependence, compressibility, single-phase and multi-phase flow, free surfaces and turbulence. Cartesian, polar, general curvilinear (including non-orthogonal or moving) and rotating coordinate grid systems can be used. The three main modules which make up PHOENICS are SATELLITE, EARTH and PHOTON, ie the pre-processor, processor and post-processor, respectively.
The problem used as a basis for testing the solver is that of heat conduction in a cube. For the purposes of this work, the main interest in the problem is that its discretisation leads to the solution of a symmetric system of linear algebraic equations, which can be solved by a conjugate gradient algorithm: Ax=b, where A is a symmetric positive definite matrix and x and b are vectors. The dominant computation in the algorithm is a matrix-vector product: although PHOENICS provides a large number of utility routines, there are none for computing matrix-vector products. In order to implement this method, and indeed most other user-defined features, it is first necessary to know how variables are retrieved and stored in PHOENICS. The most general way of doing this (although this is not the simplest) is by directly accessing the large central storage array of PHOENICS, called the F array. A number of differing types of variables are stored in this array, to which reference will be made below.
Access to the flow variables in the F array is made by the use of integer indices which serve as pointers to segments used to store the variables. Thus we need the starting location of the segment and the order in which the values are stored. All the real variables in EARTH are referred to by use of the F location integer index called LF, defined by LF=LOF (variable name) + indicial expression where LOF is an integer function called the zero location index and is the index of the location just before the segment in which the variables of interest lie. These LOFs are allocated by PHOENICS at the beginning of the computation and their values depend on the dimensions of the problem, the variables stored and solved for, whether the problem is steady or transient, elliptic or parabolic in nature, etc. The indicial expression is a linear function of the cell-location coordinate indices, IX, IY and IZ, and the total number of cells in each coordinate, NX, NY and NZ.
In using the conjugate gradient algorithm to solve for the linear system of equations Ax = b we refer to the F array directly for access to x and b vectors, which reside in the F array as the solution variable, PHI, and the source term, SU, as well as for the off-diagonal elements of the A matrix: AP, AN, AE and AH. They are all stored as whole-field-solver variables, i.e. they occupy segments of the F array containing NX*NY*NZ contiguous locations. Working space for the following is also required: the residual and search vectors, storage for the main diagonal of the matrix (as this is not done automatically in PHOENICS) and storage for the result of the matrix-vector multiplication. We can obtain this by declaring arrays explicitly, although this has obvious limitations when the dimensionality of the problem has to be increased. To obtain more flexibility, we make use of some of the working space in the F array, known as full-field variable storage. Full-field variables are those which have a stored value for each grid cell. Thus there are NX*NY*NZ values stored for each variable of a given kind, but they occupy NZ non-contiguous segments, each containing NX*NY locations. There are four classes of full-field variables, of which we use two: the whole-field solver variables described above and dependent variables. These need not necessarily be the dependent variables of conservation equations, such as velocity, pressure, enthalpy or concentration; they may be variables to which users ascribe a significance of their own. The subroutines for the conjugate gradient algorithm, the calculation of the main diagonal of the matrix and for the matrix-vector multiplication can then be written with reference to the central storage array used by the flow-solving module of PHOENICS. The new routines are called from a subroutine that is based on an exemplary routine supplied in EARTH, called GREX3, which either serves to provide access to some utility routines provided by PHOENICS, or can be used as an example to follow when writing our own features. This subroutine is then incorporated into the groundf program which is read by EARTH. Some default settings need to be changed in the SATELLITE module indicating the attachment of the user-written routines before linking and building the EARTH module.
By using the F array for full-field and whole-field variable storage required by the linear equation solver, the desired flexibility of the implementation was achieved. It also provided valuable experience in understanding how the central storage within EARTH operates. The aim in implementing a conjugate gradient solver was not primarily for efficiency, but it has demonstrated how more complex features could be implemented.
A copy of the full report, from which this article has been extracted, can be obtained.
In 1991 the University Funding Council funded a £3m initiative over a three year period (ITTI) to improve the availability of training materials for the use of IT in UK Higher Education Institutions.
The Initiative has funded 29 projects at universities throughout the United Kingdom. These projects are, in the main, producing quality paper based and computer based products in areas such as:
To date, 31 computer or paper based training products have been produced. In addition, 13 reports or booklets have been prepared, mainly directed towards providing information and advice on the development of multimedia and hypertext applications. Some 40 courses/workshops have also been held.
The Table gives the institutional uptake of each training product. The vast majority of products which have been available for more than 6 months have been ordered by more than 30 institutions. Almost all institutions have ordered at least one product.
| Product | Uptake Institute |
|---|---|
| Guide-lines Materials Production | 70 |
| ISI Data Service Training Pack | 75 |
| GRASS Training & Software Pack (GIS) | 46 |
| Getting Started IDRISI Exercise (GIS) | 65 |
| Introduction to Data Analysis, CBT Using Minitab | - |
| CBT Introduction to SAS | - |
| GIS: CBT Awareness Demonstration | all |
| Understanding IT: CBT on How to use your PC | 54 |
| Online UNIX Help (prototype) | 30 |
| CBT UNIX Document - Guide Converter, CBT UNIX Document - X-Tutor | 25,30 |
| CBT Guide Tutor | 5 |
| Biomedical Videodisc Database | - |
| BRS Search Primer BRS Search Database Design & Mtce Guide |
32 26 |
| CBT for Basic IT Skill Training 4 Products | 8, 4, 4, 11 |
| EXAMINE Computer Based Quiz System (prototype) | 20 |
| Authorware Tutorial Toolkit (prototype) | 20 |
| X-Windows User Training | 37 |
| Motif Training | 37 |
| Motif Training | 18 |
| X-Windows Systems Training Getting Started with ARCIINFO (2 products) Introduction to LATEX C to Fortran Programmer Conversion Moving On with IDRISI Exercises (GIS) Training Materials for Computer Centre Advisers CBT Generating Descriptive Statistics using SAS |
Recently or not yet announced |
It is premature, for most products, to evaluate their level and pattern of usage at institutions. The Coordinator, with help from the Universities and Colleges Information Services Association (IT Directors are normally the institution's representative), is about to commence an assessment of the level and pattern of usage of announced products.
Detailed descriptions of announced products are located in Sections H7D-H7H of the NISS Bulletin Board.
Engineers are increasingly using graphical workstations and the sophisticated yet intuitive point and click interfaces which they provide. In this environment, hypertext is a method for presenting documents interactively on the screen. Instead of having to read scrolling text in batches of 25 lines at a time, the reader of a hypertext document is presented with buttons - words or pictures - on the screen. When the reader selects one of these buttons by pointing at it with the mouse, further information is revealed either in situ or by following a link to the relevant part of a related document. Thus readers can move quickly and directly to the information they seek. The University of Kent is currently involved in the third year of a three year project investigatingHypertext Techniques for On-line Documentation and has produced two products, gman and Xtutor, which may be of interest to readers of this newsletter. Both are available free by anonymous ftp from the HENSA Unix archive at the University of Kent. They are written using the Unix Guide hypertext system.
This product makes it easier to peruse UNIX manual pages. It converts any standard UNIX manual page (written using -man macros) into Guide hypertext on the fly. Section headings are presented as hypertext buttons and their contents revealed by selection with the mouse. A search option locates text strings. Related manual pages can be displayed through hypertext links built into the SEE ALSO section.
Xtutor is an interactive hypertext tutorial for new users of the X Window system. It demonstrates the use of windows, explains how to manipulate them (move, resize, iconise, etc.) and involves the user in activities to practise these skills. Xtutor covers both mwm and twm window managers.
Both products are available by anonymous ftp from unix.hensa.ac.uk in the directory:
misc/unix/unix-uide/guide_docs
If you do not have Unix Guide at your site, you can ftp a free Sun4 version of Guide that contains no authorship facilities. This is called guide_reader, and is in the directory:
misc/unix/unix~uide/guide_reader
The full UNIX Guide (one of the Kent Software Tools from the University of Kent) is available for Sun4, HP, Silicon Graphics and IBM RS6000.
The Hypertext Techniques for Online Documentation project is funded by the Information Technology and Training Initiative.
Around 45 people, including me, gathered in the Senate Room at the University of Liverpool on 7 October 1993 for the Opening Ceremony of this new Institute. The official opening was performed by the Vice Chancellor, Professor Philip Love, who expressed his personal pleasure at being asked to launch this important new venture. He wished Professor Mike Delves, the Director of the Institute, and everybody connected with it, every success in this exciting venture.
Mike Delves then described how the Institute had been formed from the merger of two very successful predecessors, the University's Centre for Mathematical Software Research and the North West Transputer Support Centre, established in 1988 under the SERC/DTI Transputer Initiative. He outlined the work of the new Institute which promises to build on the foundations set by the two previous Centres and, in particular, increase the already significant number of collaborations with industry.
Dr Allan Robinson (Shell Research), who has had a long association with the University, including the chairmanship of the Management Advisory Board of the former North West Transputer Support Centre, was the final speaker at the Ceremony. He addressed the topic of University/Industry collaboration.
He believed that change was the theme of the day and, in discussing the response of industry, placed great emphasis on value and affordability. Industrial Research and Development Laboratories are subject to this change and, being an integral part of the business world, are also subject to the same free market forces. Higher education is part of this changing world with the interactions remaining an integral part of industrial R & D for recruitment, research support and training.
The driving force of the new Institute is a desire to solve practical problems, and the new parallel machines mean, for example, that modelling techniques have the potential for use in the design and improvement of products, with the promise of savings over a purely experimental approach. The range of the Institute's services in research and development, scientific interactions and training should contribute to industries' wider efforts on continuous improvement. The new Institute's expertise offered industry an opportunity to get advanced computing technology right first time.
The European Research Consortium for Informatics and Mathematics (ERCIM) is an organisation dedicated to the advancement of European research and development, in the areas of information technology and applied mathematics. Founded in 1989, this non-profit organisation comprises, at present, 10 leading IT research organisations from different European countries.
Rutherford Appleton Laboratory is the UK's representative in ERCIM. The raison d'etre for ERCIM is the growing need for consistency of research strategies at National, European, and International levels. ERCIM addresses this need in its role as a registered European Economic Interest Grouping (EEIG). Recent activities have included critical appraisals of CEC Framework programmes, contacts with non-European research initiatives (e.g. Japan's Real-World Computing Initiative), and consultations with CEC officials at senior level. The aim is to enhance European research policies in collaboration with industry and legislative bodies. ERCIM issues regular reports on these activities.
For researchers and potential technology users, ERCIM's main functions are:
Individuals or organisations interested in ERCIM can subscribe (for free) to the in-house Newsletter ERCIM News, published quarterly. ERCIM News publishes brief (typically 400 words long) descriptions of ongoing or recently finished IT research / application projects, informs on issues of research policy, and carries announcements of conferences and workshops. Each issue also has a special theme " section reporting on projects in a particular research area. Recent special themes included advanced networking, parallel processing, software quality, and advanced database technology. If you wish to subscribe to ERCIM News then please contact me.
Early in 1989, the Energy & Thermofluid Mechanics Group (ETMG) of the Institution of Mechanical Engineers perceived a need to increase awareness in the engineering industry of CFD and its relevance to engineering design. National and Regional Seminars have since attracted over 650 delegates, bringing together the whole range of interests from academics and analysts to industrial designers, managers, and software vendors.
In addition to this general awareness programme, the Group released three engineering test cases to CFD users which were the subject of a Seminar in November, 1991. The Test Cases were chosen to be representative of real engineering applications and consisted of:
In each case a specialist in the field prepared data for which he had high quality experimental results. This data has since been released to the EASE CFD Community Club and is being made available to interested analysts.
Collaboration with the Community Club is continuing, most recently in the co-sponsored Conference, Engineering Application of Computational Fluid Dynamics, which was held in September 1993, and attracted an audience of 75.
The importance which ETMG attached to developments in CFD has been more than borne out by events. Already several of the Institution's other Groups and Divisions have launched CFD events relating to their individual subject areas. Topics covered have been the prediction of the consequences of fire and explosion, application to building services, and developments in reciprocating engine and turbomachinery design. The next CFD seminar, The Validation of Computational Techniques in Vehicle Design - Stage One: CFD is being organised by the Institution's Automobile Division on Monday, 18 April 1994.
It is hoped that the present links with the CFD Community Club will also generate further events of interest to the CFD Community.
In addition to the fluid dynamics software described in ECN Issue 45: Phoenics, FEAT, Fluent, Star-CD, and Flow3D, we have recently added another package NEKTON to the list of CFD codes available for use on the Atlas Cray Y-MP/8. NEKTON (a trademark of Nektonics Inc) has been obtained from Fluent Europe and is available to users who have their own workstation licences.
NEKTON is a general purpose fluid flow modeling package based on a high-order finite element technique known as the spectral element method. It solves steady or unsteady incompressible fluid flow and heat transfer problems, in 2-D, asymmetric, and 3-D geometries. The Spectral Element Method uses a variational domain decomposition technique. The computational domain is broken up into large macro-elements within which solution variables and geometry are represented as high-order tensor-product polynomial expansions. The grid can be unstructured and the elements can have sides which are non-linear. This technique lends itself to the vectorisation and parallelisation which are possible with the Cray Y-MP/8.
As with the other CFD codes, a grant from SERC or another Research Council is necessary to provide the cpu time allocation needed to use the Cray. If you need advice about applying for a grant, contact myself or Margaret Curtis in the Atlas Centre.
If you are currently using any other commercial CFD codes and would like to harness the power of the Atlas Cray YMP8 to your problems, then contact me to discuss the possibilities. Similarly, if you know of other CFD codes that could benefit the UK research community by tackling problems not solvable by other codes mentioned in this article then again I would like to discuss them.
Another CFD program has been added to the Common Academic Software Library on HENSA. It is a vortex code developed at the University of Manchester by P A Smith and P K Stansby.
The vortex code solves for the 2D impulsively started, viscous, incompressible laminar flow around a circular cylinder. A wide range of Reynolds numbers can be used (at least 10 to 10\ The code uses the discrete vortex method, the diffusion terms being simulated by a random walk and the convective terms by a vortex-in-cell algorithm.
Flow solutions can be obtained for a number of shapes (circle, square, ellipse, triangle) with rounded corners. Some GKS routines are implemented for obtaining streamlines around the body as well as vorticity plots.
Access to the archive can be made via the JANET X.25 network with the address unix.hensa.ac.uk. The vortex code has been placed under the directory /misc/cfd/software, together with the other CFD software (described in ECN issues 42 and 47 this year).
The above report has been produced as part of the work of the Community Club. It attempts to suggest an approach to software development that involves the use of a practical subset of quality assurance (QA) methods for the academic community. The goal is to ensure that the software is easier to develop, maintain, re-use and distribute to colleagues without an excessive overhead in time and effort.
The guide covers a simple approach to software design, implementation and documentation. It also suggests a number of inexpensive software tools that can help the developer produce good software. An example of the application of the proposed methods is included to demonstrate the approach.
The report may be obtained from Debbie Thomas.
The Club is organising a one day meeting on current research in Aeronautical CFD which will be held on Wednesday, 20 April 1994 in the recently refurbished Lecture Theatre at RAL. The chairman will be Steve Fiddes from the University of Bristol and the speakers will include holders of current SERC grants in the area of aeronautical CFD. The full programme and registration form will be distributed with the March edition of the ECN.
On Wednesday 17 November 1993 the Club held a one day seminar on the above topic at University College London. The event proved popular with 52 attendees coming from as far away as Glasgow and Belfast. The topics covered included loading on offshore structures, water quality modelling, ship design and various aspects of wave simulation. The attendees particularly appreciated the opportunity to get together with other researchers in the area during the refreshment breaks to discuss their mutual interests.
Copies of some of the speakers overheads and related papers are available from Debbie Thomas.
