Jump Over Left Menu
Engineering Computing Newsletter: Issue 52
- Engineering Decision Support: Case-Based Reasoning
- IT Awareness in Engineering Seminars Comes Together
- CSCW Awareness Initiative
- The Intelligent Systems Integration Programme
- Complex Coloured Image Inspection
- Integrated Process Support Environments
- Community Clubs
- Conferences and Meetings Notices
Engineering Decision Support: Case-Based Reasoning
Expert or Knowledge-Based Systems (KBS) are one of the success stories of Artificial Intelligence (AI) research. In a recent survey the UK Department of Trade & Industry found over 2000 KBS in commercial operation (the survey excluded KBS in University research. laboratories). It has been around twenty years since the first documented KBS (the trinity of classic systems: DENDRAL, MYCIN and PROSPECTOR) were reported, yet in that time the basic architecture of KBS has changed little. The early KBS, and today's systems, are based upon an explicit model of the knowledge required to solve a problem - so called second generation systems using a deep causal model that enables a system to reason using first principles. But whether the knowledge is shallow or deep an explicit model of the domain must still be elicited and implemented. However, despite the undoubted success of model-based KBS in many sectors, most developers of these systems have met several problems:
- knowledge elicitation is a difficult process, often being referred to as the knowledge elicitation bottleneck
- implementing KBS requires special skills and often takes many staff years
- once implemented model-based KBS are often slow and have difficulty in accessing and managing large volumes of information
- once implemented these systems are difficult to maintain.
Solutions to these problems have been sought through better elicitation techniques and tools, better KBS shells and environments, improved development methodologies, knowledge modelling languages and ontologies, facilitating the co-operation between KBS and databases and techniques and tools for maintaining KBS.
However, over the last few years an alternative reasoning paradigm and computational problem solving method has increasingly attracted more and more attention. Case-based reasoning (CBR) solves new problems by adapting previously successful solutions to similar problems. CBR is attracting attention because it seems to address the problems outlined above directly, namely:
- it does not require an explicit domain model and so knowledge elicitation becomes a task of gathering case histories
- its implementation is reduced to identifying significant features that describe a case
- it is an easier task than creating an explicit model, by applying database techniques large volumes of information can be managed
- CBR systems can learn by acquiring new knowledge as cases thus making maintenance easier.
The processes involved in CBR can be represented by a schematic cycle (see Figure 1), typically comprising four processes - the four REs:
- RETRIEVE the most similar case(s)
- REUSE the case(s) to attempt to solve the problem
- REVISE the proposed solution if necessary
- RETAIN the new solution as a part of a new case.
Figure 1: The CBR Cycle
A new problem is matched against cases in the case base and one or more similar cases are retrieved. A solution suggested by the matching cases is then reused and tested for success. Unless the retrieved case is a close match the solution will probably have to be revised producing a new case that can be retained.
Although CBR is a relatively new problem solving paradigm (the first academic demonstrators were developed at Yale University in the mid-eighties) there have already been many commercially successful applications. One of the first commercially available CBR applications was produced by Lockheed at Palo Alto.
Modern aircraft contain many components that are made up from composite materials. These materials require curing in large autoclaves. Lockheed, the US aerospace company, produce many such parts. Each part has its own heating and cooling characteristics and must be cured correctly. If curing is not correct the part will have to be discarded. Unfortunately, the autoclave's thermodynamic properties are not fully understood (i.e. there is no model that operators can draw upon). This is complicated by the fact that many parts are fired together in a single large autoclave and the parts interact to alter the heating and cooling characteristics of the autoclave.
Operators of Lockheed's autoclaves relied upon drawings of previous successful parts layouts to inform them how to layout the autoclave. However, this was also complicated by the fact that layouts were never identical because parts were required at different times and because the design of the composite materials was constantly changing. Consequently operators had to select a successful layout they thought closely matched and adapt it to the current situation.
This closely resembled the CBR paradigm and when Lockheed decided to implement a KBS to assist the autoclave operators they decided upon CBR. Their objectives were to:
- reuse previously successful loadings
- reduce the pressure of work on one or two experts
- secure the expertise of the experts as a corporate asset
- help to train new personnel.
The development of CLAVIER started in 1987 and it has been in regular use since the Autumn of 1990. CLAVIER searches a library of previously successful autoclave layouts. Each layout is described in terms of: parts and their relative positions on a table and their relative positions in the autoclave, and production statistics such as start and finish times, pressure and temperature.
CLAVIER finds substitutes for parts in a layout that do not match, and it recommends new layouts to operators. In adapting new layouts from previous ones CLAVIER:
- creates new layouts by adapting pieces of previous layouts
- minimises the number of required parts not included in the layout
- maximises the number of high priority parts included in the layout
- maximises the total number of parts in a layout.
CLAVIER acts as a collective memory for Lockheed and as such provides a uniquely useful way of transferring expertise between autoclave operatives. In particular the use of CBR made the initial knowledge acquisition for the system easier. Indeed, it is doubtful if it would have been possible to develop a model-based reasoning system since operatives could not say why a particular autoclave layout was successful. CLAVIER also demonstrates the ability of CBR systems to learn. The system has grown from 20 to over 150 successful layouts and its performance has improved such that it now retrieves or adapts a successful autoclave layout 90% of the time.
Lockheed's CLAVIER system has become the classic CBR system demonstrating a number of benefits.
That CBR could be applied to solve a problem where no explicit model existed, hence it would have been impossible to have built a model-based KBS.
CBR systems can be built without passing through the knowledge elicitation bottleneck, since elicitation becomes a relatively simple task of acquiring past cases. Implementation becomes a relatively simple task of identifying relevant case features, and moreover a system can be rolled out with only a partial case-base as happened with CLAVIER. Indeed, using CBR a system need never be complete since it will be continually growing. This removes one of the bugbears of KBS - how to tell when a knowledge-base is complete.
CBR systems can propose a solution quickly by avoiding the need to infer an answer from first principles each time, important in many real-time decision support systems.
Individual or generalised cases can be used to provide explanations that are more satisfactory than explanations generated by chains of rules, important in many domains with legal implications.
CBR systems can learn by acquiring new cases, making maintenance easier, as demonstrated by CLAVIER.
Finally, by acquiring new episodic cases CBR systems can grow to reflect an organisation's experience. If a rule-based KBS was delivered to six companies and used for six months, after that time each system would be identical, assuming no maintenance had taken place. If six identical CBR systems were used in a similar way after six months there could be six different systems, as each could have acquired different episodic cases.
The next ITE Awareness seminar, Informing Technologies to Support Engineering Decision Making, to be held at the Institution of Civil Engineers (ICE) on the 21-22 November, will introduce delegates to the principles of CBR. The Seminar will concentrate on applications in use and under development by major companies. It will be supported by exhibitions and demonstrations of CBR software from leading vendors.
Ian Watson, University of Salford
IT Awareness in Engineering Seminars Comes Together
The next lTE Awareness seminar, Informing Technologies to Support Engineering Decision Making, to be held at the Institution of Civil Engineers (ICE) on the 21-22 November, has come together extremely well and you will see the result in the enclosure to this newsletter. You can get a feel for the event by reading the article on the front cover.
We are more than pleased that Professor Fritz Prince, who gave last year's Royal Academy of Engineers lecture, has agreed to join our seminar to give us a flavour of what is happening in the topic on the other side of the Atlantic. His presentations are always exciting and thought provoking, and I am sure he will also have views of where our new IT research strategy on this area should go. We have also arranged with the EPSRC's Head of the Innovative Manufacturing Initiative (IMI) for this event to be the first major occasion when his newly appointed Programme Sector Managers will make a public appearance together. As you will see from the programme they will lead discussion on the way forward for IT research within their IMI themes:
- Construction as a Manufacturing Process
- Aerospace Manufacture
- Responsive Processing.
The Programme Managers will also be there to chair their parallel sessions and listen to you all. We have invited stimulus contributions to get discussions going. These stimuli will come from those with a strong understanding of the industrial need for Engineering Decision Support. This sort of steer has been chosen because IMI will unashamedly be driven by industrial need and demand, and a recognition of potential market opportunity. However, there will be ample opportunity for you to contribute to those strategy developments informally throughout the seminar or in the formally allocated discussion period. If you wish to work on the topic of Engineering Decision Support Systems (EDSS) then you should try and come to this day or at least let us know your views.
We are also well under way in the planning of the next seminar entitled Virtual Reality and Rapid Prototyping, to be held at the University of Salford on 26-27 January 1995. Nigel Birch, Bob Stone and Roy Kalawsky are the driving forces behind this event and we have already lined up some most interesting speakers and demonstrations. However, we may have missed someone or some important ideas on the topic, from either industry or academe. So please contact me at the address below if you have any interest in presenting at the event, which will be in a similar format to the one mentioned above, where the IMI Programme Sector Managers will again play an important role and where those attending can help shape future EPSRC policy in this other important IT in Engineering area.
Finally, for those who have already shown interest in putting in a proposal to develop an open education and management approach in the area of software quality design for engineering research, we have not forgotten you. With the changing structures of EPSRC the development of our tender document seems to have run foul of new internal and European contract law. As a result we have had to go through several bureaucratic hoops to get the document out. It is nearly finished, so those known to me will get their documents soon. If anyone else would be interested in putting in a proposal do let me know.
You can make your views and interests known to the programme by contacting me.
Professor James Powell, ITE Awareness Initiative Office
CSCW Awareness Initiative
The initiative is an integral part of the JFIT Computer Supported Co-operative Work (CSCW) programme. It involves both the CSCW and Teleworking Special Interest Groups and over 40 commercial and academic organisations working together in specialist consortia. Promotional support may be given to organisations active in the field of CSCW or hosting events which relate directly to the aims of the initiative. The awareness initiative reflects the DTI/EPSRC's refocussed science and technology policy. Under this new policy, the initiative aims to transfer CSCW technology and knowledge from the research community to the commercial arena.
Focus on the CSCW Consortia
VIRTUOSI (Support for virtual organisations)
This project will use virtual reality (VR) techniques to represent the organisation in which collaboration takes place. This aspect of the project will be strongly supported by Division Ltd, which is a world-leading UK company specialising in the development of VR systems. The project will develop two application pilots: one for improving communications between BICC Cable factories worldwide; the other enabling Nottingham County Council to assist local SMEs in the design and manufacture of fashion garments.
Both BT who is the lead partner for the project and GPT will be involved in developing the telecomms aspects of the project. Lancaster, Nottingham and Manchester Universities will bring expertise from Computer Science, Sociology and Psychology. Contact: Dr Alistair Rogers.
BFS (The Business Flight Simulator)
Modelling organisations to reflect business options is possible using existing software tools which are adapted and integrated into an environment approximating to the organisation. What if options can then be examined and discussed before committing to a particular organisational structure.
Led by ACT Business Systems, project partners include Bull Information Systems, City University Business School, University of Westminster, Esmerk Informat and Touche Ross. Contact: Mr Paul Ballard.
SYCOMT (Systems Development and CSCW - Methods and Techniques)
Recent research by Digital Equipment Corporation has shown that companies waste millions of pounds developing computer systems that hinder rather than assist work procedures. Groupware implementations run the risk of repeating the same mistakes again. Digital is teaming up with National Westminster Bank, University of Lancaster, and cybernetics specialists Syncho Ltd to look at designing systems to support group working in retail banking. Contact: Mr Geoffrey Darnton.
CD (Collaborative Documents)
Businesses are becoming increasingly mobile. Senior professionals travel at short notice, often being in different time zones from their colleagues. Despite this, they must operate as team players dependent on each other. Three multinational corporations from the foods (Grand Metropolitan), pharmaceutical (Pfizer) and oil industries (BP) are participating in the project as end-users, ensuring that technology serves the real world.
ICW (Integrated Co-operative Workspace)
Project objectives are to counter the current lack of scalability, adaptability and openness which precludes CSCW from being consistently assimilated across all core activities of an enterprise.
The industrial partners' specialities cover the spectrum of standards-based computing architecture and activities: client/server database (Oracle Corporation); directory services (Nexor); process enactment (ICL Openframework); multiple data feeds (Fretwell-Downing). They will use these skills in conjunction with Brainstorm Ltd and the Universities of Sheffield and City, to create a modular, integrated environment to support enterprise-wide CSCW and demonstrate this in health-care and financial applications. Contact: Mr Matthew Caunt.
DUCK (Designers Use of Cooperative Knowledge)
Competitive pressures within the industry place a premium on early completion of design work. Concurrent design and the sharing of information is vital to reduce risk and gain consensus.
A computer tool set which will allow simultaneous multiple access to a design and is a project log is being developed by a consortium comprising the University of Paisley, BAe, SEMA, and MARI Computer Systems Ltd (lead partner).
The tool-set will comprise a shared electronic design Journal running over a distributed network of computers, allowing designers to work as if sharing a single office. Contact: Mr Graeme Oswald.
Special Interest Groups
TSIG - The Teleworking SIG has been established under the auspices of DTI to provide a forum for the exchange of information on all aspects of teleworking, and to promote best practice in this fast growing area. Contact Mr Love.
The CSCW SIG was formed by the UK Human Interface and Communications and Distributed Systems clubs and is supported by DTI. The SIG has taken an active role in support of the pre-programme phase of the CSCW Initiative. Contact Patrick Holligan.
The CSCW Awareness Initiative is managed by Garth Shephard of TEK.nowledge. To register with this initiative contact Jacqueline Whalley at DTI.
Produced by DTI in association with Noiseworks.
Allan Smith, Programme Co-ordinator, Department of Trade and Industry
The Intelligent Systems Integration Programme
This is a brief report on the Intelligent Systems Integration Programme (ISIP). ISIP is a joint Department of Trade and Industry/EPSRC (previously SERC) industry-academic collaborative research programme with an emphasis upon the problems of commerce and industry which second generation Artificial Intelligence/Knowledge Based Systems tools and techniques can help to solve. In particular support for decision making is seen as a major area where these techniques can produce a significant business benefit.
With the refocusing last year of DTI's policy towards technology transfer and best practice, Advanced Technology Programmes (which included ISIP) will not be issuing further calls for research proposals. There are eight Research Projects delivering demonstration applications, with support funding from the DTI and EPSRC: with the possibility of some associated EPSRC uncled research activities. In order to stimulate technology transfer, ISIP project consortia have undertaken to include dissemination and awareness in their work programmes.
In addition to these research projects there is also support for four Special Interest Groups which are investigating specific areas where AI / KBS may provide business solutions. (These areas are Planning and Scheduling, Deep Knowledge, Intelligent Interfaces and Integrated Systems). There is also support for twelve regional Technology Transfer Clubs whose role is to provide expertise to commerce and industry at large on the use of AI/KBS to solve business problems.
A Selection of ISIP Research Demonstration Projects
(Bartlett Research, Richard Rogers Partnership, PowerGen plc, Design London Ltd, Broadgate Properties plc, Criterion Software Ltd, Qualum Ltd, Smallwood Systems Ltd, Avanti Architects Ltd, Bovis Construction Ltd, Department of Computer Science, University College London).
The Intelligent Architecture project proposes the creation of an intelligent, artificial environment for those dealing with the Built Environment, such as architects, planners, engineers and facilities managers, allowing them to visualize structures, manipulate spatial representations, simulate space utilizations and model functional outcomes of physical and spatial form.
CAKE (Computer Aided Knowledge Engineering)
(Cogsys Ltd, Salford University Business Services Ltd, British Steel, University College Swansea, The Cogsys Club Ltd.)
The CAKE proposal concerns research into new methods and techniques for creating real-time knowledge based applications to use as an integrated part of mainstream automation software. The concept of CAKE is to provide software tools and methods which will enable the construction of complex real-time knowledge based applications by engineers and/or technologists with only limited software and knowledge engineering experience.
CHRONOS (Temporal database management system)
(International Computers Ltd, Lloyd's Register of Shipping, Imperial College, UMIST.)
The aim of the CHRONOS project is to develop a temporal database management system which will support the efficient manipulation of temporal information. This is seen as an essential requirement of many, if not most, management support systems. The application demonstrator work will focus upon business activity planning and resource management.
STAR (Assessing risk with multiple sources of knowledge, the STandardised Argument Report)
(Imperial Cancer Research Fund, LHASA UK, Logic Programming Associates, City University).
The STAR project will develop a technology for evaluating and reporting on risk. New materials, industrial processes and other innovations are frequently associated with hazards, but hazards which are ill-defined and difficult to quantify. Amongst the techniques to be used are qualitative and quantitative reasoning under uncertainty. The demonstration application will be in the integration of different information sources to support decision making in the area of risk assessment.
(Artificial Intelligence Applications Institute, Logica Cambridge Ltd, Lloyd's Register of Shipping, IBM UK Ltd, Unilever plc.)
The Enterprise project is designed to support organisations' ability to monitor and improve on their performance against strategic objectives. The aim is to provide an integrated set of computer tools to model, analyse and improve various aspects of how an enterprise works and how it is organised. The emphasis will be upon process modelling, re-engineering and decision support.
The DTI holds a mailing list for people interested in the intelligent systems area and further information on individual projects (including contact points for each project), the Special Interest Groups and the Technology Transfer Clubs can be obtained from them. The Programme Manager for ISIP is Allan Smith and the EPSRC Swindon Office contact for Systems Engineering is Nigel Birch.
Tony Conway Informatics Department
Complex Coloured Image Inspection
Many recognition tasks can be performed satisfactorily by conventional computer vision systems utilising either feature extraction and related measuring algorithms, or by expert systems based upon lists of rules and inferences. However, with reference to coloured and complex images, there are no obvious features or rule sets which are easily applicable. The WISE (Windowed Image Shape Evaluation) system, developed at Brunel University, overcomes many of these problems by windowing images and then classifying them by using weightless logical node nets which compares them with those previously seen during a training phase.
Originally, WISE was developed as part of a low-cost horizontal belt feeder which was designed to interface with a robotic assembly cell. WISE has since been applied to the inspection of piece parts, to the classification and inspection of complex coloured images, and to the recognition of light-scatter patterns from machined surfaces.
In order to classify complex coloured objects, a suitable transformation of colour space must be made before the n-tuple weightless logical node technique is used for the recognition process. Initially, Hue Saturation Intensity (HSI) colour space is used because, compared with Red Green Blue (RGB) colour space, each parameter can be manipulated independently of the others. Three-dimensional graphs are then plotted of H, S and I with the third axis in each case being the frequency of occurrence. These three graphs are summed to produce a three-dimensional representative model as exemplified in figure 1. This model is then sliced at 16 levels to produce 16 sections (called colour planes). Each of the 16 sections gives a shape for the n-tuple system to inspect. An example of the 16 sections and the three-dimensional map from which they are taken is shown in figure 2. These 16 sections are each inspected using 16 discriminators, one corresponding to each section and the results of the 16 discriminators are displayed in the form of a histogram. When various types of images are inspected on a plain background, any individual one will produce the same HSI colour space and hence the system is position invariant.
Figure 1: A HSI three-dimensional colour set
Figure 2: Sixteen two-dimensional cross sections through the HSI colour set
Figure 3 shows the results of the inspection of two credit cards which are identical apart from the fact that they were issued in different years. The various discriminator scores represent differing positions and orientations of the trained card and the inspected cards. Clearly, in each case, irrespective of their position, the two credit cards can be classified and separated.
Figure 3: Discriminator responses for the inspection of two credit cards at ten different positions
Although recognition in colour space is invariant to object position and rotation, in many inspection tasks it is often necessary to ascertain the locations of errors in terms of H, S or 1. In order to provide such facilities, the image is divided into 16 rows, each of which has three separate H, S and I discriminators. (It should be noted that, for higher resolution, more rows could be used and the image also could be divided into columns; however, for illustrative purposes, only 16 rows have been used). Such an example is shown in figure 4 which depicts the results obtained from a flower picture. The diagram represents 16 bars in four groups: the top 16 (i.e. the top quarter) are 16 colour planes which give the position invariant colour content of the image: the next 16 bars (i.e. the upper middle quarter) give the classification of intensity for the 16 rows of the image; the next 16 bars (i.e. the lower middle quarter) give the classification of saturation for the 16 rows, and the next 16 (i.e. the lower quarter) give the classification of hue. Each bar is the discriminator output from the n-tuple classifier appropriate to the colour, row and HSI of the image. The flower picture was altered in hue by manipulating the image in software. Hue pixels in the range of 40-42 were changed globally to 45, representing a 4% change in the colour yellow which is the dominant colour in the upper part of the picture. The image was changed and then inspected and the resulting responses bar chart is shown in figure 4. The upper quarter set of bars is less than 100% illustrating that there has been a significant change in the colour content and the lower quarter, representing the hue variations of the 16 rows, shows significant changes. In particular, the upper bars, corresponding to the major yellow flower, show large errors with bars going down to as low as 7%. To complete the inspection understanding, the actual errors are highlighted by flashing the incorrect pixels on and off.
Figure 4: Flower image & responses bar chart when hue in the range 40-42 is changed to 45
Additionally, in colour inspection tasks, every effort must be made to minimise colour variations due to lighting changes and, consequently, further algorithms were developed which compensate for such adverse effects. In principle, standard colour patches are placed in the scene alongside the object to be inspected and, during the inspection cycle, the colours of these patches are assessed and a transform matrix constructed. This matrix is then applied to the captured image to compensate for the illumination changes prior to inspection. In the case of the two credit cards mentioned previously, after training, the response of card A was 100% whereas that of card B was 20%. When the light level was changed by increasing the supply voltage by 2.5%, card A only gave a response of 48%. However, when the colour invariant transform was applied and card A was again inspected, the overall response increased to 80% whilst the card B response remained at 20%. Therefore, the colour invariant transform restores a higher integrity to the acquired image and demonstrates that the WISE system can be configured to be tolerant to small changes of ambient lighting.
In summary, the WISE system can recognise objects which are both complex and coloured, performance is position-invariant if transformed colour space is used, training can be tailored to provide tolerance to small changes in colour and lighting and, additionally, operational speed is independent of image complexity.
Bruce Wilkie, Brunei University
Integrated Process Support Environments
The concept of Integrated Process Support Environments (IPSE) is not new: it has been used in software development for many years. However, it has generally been restricted to supporting technical management and formal methods. One of the many challenges that face IPSEs is to extend applicability to more general management. Work at Bristol University is focusing on how IPSE technology can be appropriated by senior management in the civil engineering industry. In particular the feasibility of extending the IPSE2.5 technology, (which is now part of the ICL Processwise portfolio) is being studied.
The Processwise environment has several distinguishing features that are essential for effective process support. Probably the most important is the ability to change a user's environment whilst the system is running. The Process Control Engine is able to accept new descriptions of the model that changes the way a role behaves. In most conventional environments to make such changes requires the program to be recompiled. Using Processwise an individual user's environment may be changed, without interfering with other users. Processwise also provides tool integration, allowing users to bring into their environment tools developed elsewhere. Control of activities is provided by an opportunist rule based scheduler. Finally, Processwise provides a persistent data base that logs the current status of the process.
IPSEs are concerned with providing individuals with an appropriate working context, such a context will include integrated tools, access to information and the means of communication. The assertion is that the process will be supported in an 'intelligent' way. Faced with a decision as to how the process should proceed towards its desired state, it uses rules, both algorithmic and heuristic, that fire opportunistically in search of the process goal. These rules represent the collective knowledge of how a process is managed. For example, the conditions that must exist before an activity can commence or the interactions that need coordinating as part of an activity. Wider knowledge about the organisation, its structure, the roles' people play, authority, power and the products it transforms, is represented by various constructs in the Process Management Language (PML). A comprehensive support system is thus a rich source of organisational knowledge. The IPSE provides an 'intelligent' environment that pervades the total process; the result is a shift in paradigm. In the old paradigm people use tools and access information on an individual basis, in the new, the context in which tools and information are used is provided by the IPSE; the users become inhabitants within a process that is being enacted by the computer.
Generally, it has been assumed that processes can be decomposed to a set of pre-defined actions and that the complex behaviour of a process can be reconstructed by their aggregation. A process is largely viewed as objective; independent of the subjective context from which the users may perceive it. However, the more closely one looks at the role of management, the more difficult it becomes to sustain a coherent model focused on activities. Much of the manager's job is context sensitive. In fact, one could argue that in certain respects it is chaotic, in that, seemingly small changes to the context of a problem may result in considerably different actions being taken. Certainly it becomes very difficult to develop an activity model, where the process followed is sensitive to the environmental context in which the decision is made.
An environment suitable for management must be flexible and sufficiently rich in variety to encourage professional judgement, stimulate creativity and not simply perpetuate competent rote decision making. At Bristol we are exploring the possibility of achieving this by focusing attention on the responsibilities associated with a role. The model aims to support culturally sensitive patterns of commitment, accountability and communication and provides a mechanism for individuals to specialise the responsibilities associated with their roles.
Focusing on responsibility transfers the emphasis from determining what has to be done, to asking what state is desired. Reasoning as to how such a state is to be achieved is performed by the manager. What the system does, is to support the commitments and measures of accountability that are agreed. Managers make the decisions and solve the problems, the system provides support by tracking the consequences.
Whilst creative and innovative management clearly requires access to certain information, the problem facing most managers is the wealth of information with which they are presented. Faced with the prospect of information overload, the process of information gathering becomes one of information rejection. It is not only the style in which different cultures reject information that is important, as Powell and Newland have shown, experiential learning, interpersonal interaction and people's world view all play an important part in the transfer of information. Information must be tailored to meet an individual's learning strategies. A goal in the development of process support is to include a system, such as that developed by Powell and Newland, to provide 'interactive sensitivity' and 'semiotic credibility' to the information required for creative management.
At the core of the process support model is the concept of complexity from interacting simplicity. Complex system behaviour is achieved by many relatively simple role objects interacting with each other in a meaningful manner. The complex behaviour, characteristic of the design process, emerges as a result of relationships made possible via the interconnected mesh of interactions. Such a model of a world of roles is intended to exploit the notion of concurrently executing agents all coordinating to achieve a common goal. The basic role is self replicating, enabling it to create sub roles, identical in structure but void of content until occupied by man or machine and to which it can then delegate some of its responsibilities. The model of the organisation is developed by 'growing' the role loops through all of the roles within the organisation.
David Platt, University of Bristol
Parallel Processing in Engineering Community Club (PPECC)
New Parallel Processing Techniques Course
25-26th October 1994, Parallel Processing Techniques, Location: Rutherford Appleton Laboratory.
This is a newly developed course to transfer technology skills in Parallel Processing. The course has been specifically developed from a need expressed by club members.
The course aims to provide engineers with a detailed update on the techniques of Parallel Processing and how to exploit these in practice. A series of case studies and examples will be presented to illustrate the application of the taught techniques to real problems. Training exercises will reinforce and illuminate the techniques discussed.
A limited number of places are available and will be allocated on a first-come first-served basis, so early registration is advised. A registration form is included as an insert to this issue. You may also contact the PPECC secretary (Brian Henderson).
The intention is to take a higher level view of the rapidly changing field of Parallel Processing, and to arm you with the general methodologies and principles that will better stand the test of time. Although no single particular development system or implementation technique will be addressed (for example this is NOT an occam programming course), a survey of those that are available will be provided, with worked examples being presented for a representative sample.
There is an inertial barrier which prevents the uptake of parallel technology - largely based on fear and the past lack of software standardisation. The course can help you overcome the fear, and direct you towards the emerging standards. Parallel computing is already here at the top end - the course should help you exploit the cost effective computing power of parallelism, even if your application is more modest.
If this course proves to be successful then the event will be re-run as part of the 1995/96 event programme.
The programme of future events is currently being organised. This schedule includes seminars (1-day programme of talks on a given topic), courses (1-2 day specific training) and workshops (2-day talks and discussion sessions leading to recommendations on the chosen theme).
The programme of seminars for the next 18 months is now being planned and is expected to include seminars on:-
- Embedded Systems
- Technology Update
- Parallel Finite Elements
- Performance Modelling
Further details and registration forms for the first of these, on Embedded Systems, will be included in the next issue of the ECN (ECN 53, Nov 1994). The other four are being scheduled for the 1995/96 programme and details on their dates will follow in due course.
Three courses are developed for this year's programme:
- Parallel Fortran
- Parallel Processing Techniques
- Parallel C
The first of these, Parallel Fortran, was held in July 1994 and will be repeated next year. The Parallel Processing Techniques course will be run on 25-26 October 1994 (see article below). The Parallel C course will be run in January 1995.
A 2-day workshop on the theme Distributed vs. Parallel: Convergence or divergence is being planned for March 1995. A call for position papers will be included in the next ECN.
Latest details of all these events may also be found via the WWW server for PPECC (announced in ECN51, July 1994) or by contacting the PPECC Secretary. Suggestions on topics/programme for future events are always welcome from members and other interested parties. The PPECC would also be particularly interested in holding joint events with other initiatives where appropriate.
Brian Henderson, Secretary, PPECC, Informatics Department
News from the CFD Community Club (CFDCC)
CFD in Ship and Yacht Design
The simulation of flow past ships and yachts using CFD is increasingly complementing traditional experimental towing tank methods. With the specification of vessels being extended to include not only maximum speeds but also seakeeping and manoeuvring characteristics, the designer must make use of all available techniques in order to make accurate predictions of performance. A one day meeting on the above topic will be held, under the auspices of the EPSRC CFD Community Club, on Wednesday 16 November 1994. The venue will be the Department of Marine Technology in the University of Newcastle upon Tyne. Speakers will address the topic in general and also cover specific topics, such as roll damping, flow past propellers and yacht and sail design. Results will be presented from the on going EPSRC/ Marine Technology Directorate programme of research on manoeuvring, which has been influenced by the new IMO regulations. This is seen as an increasingly important topic following environmental accidents with large vessels. The full programme for this event will be available from Virginia Jones by mid-September. To register please complete the registration form included as an insert in this newsletter.
Dates for your Diary
CFD in Ship and Yacht Design, 16 November 1994 1 day Seminar University of Newcastle
Aerodynamics Research Forum, Joint with MOD AGCFM and R.Aero.Soc, 5-6 January 1995, Royal Aeronautical Society, London
Fourth Introductory School in Computational Fluid Dynamics 9-13 January 1995, One week hands-on School, The Cosener's House.
Improving the Quality of Fortran Programs, 6-7 July 1994 - A Review
Fortran programmers generally have a very different outlook on their programming from other programmers. this is because they are usually Engineers or Scientists who are generating code to solve a particular physically based problem, often based on a numerical approach.
The above course was designed with these people in mind, and was attended by a good mixture of industrialists and academics. On the first day of the course four main topics were addressed. Firstly, the "quality" of a piece of Fortran is rather difficult to quantify, for this reason several metrics that are commonly used by Vendors of software tools were discussed. Next, adhering to the Fortran77 standard in order to give a high degree of portability is particularly useful, since most pieces of code inevitably are required to run on more than one platform. Another important issue in any code development is testing the code. It was stressed that ideally this should be performed by another party, who should be executing the program with the intention of finding errors, and that any test which failed to find an error was an unsuccessful test! Finally, the future of Fortran in the form of Fortran90 was discussed. It is obvious that this is a much awaited development, which deals with most of the short comings of Fortran77. In particular, the introduction of dynamic memory allocation and array handling facilities will greatly enhance the expressive capabilities of Fortran.
The second day of the course allowed some hands-on experience with some of the software tools available from different vendors. It became apparent that each of the tools had some obvious strengths and weaknesses, and that a combination of all the tools would be needed to cover the needs of most software developments. In terms of value for money several of the public domain tools appeared to be very useful. From this experience, it is clear, with the rigorous application of a set of tools of this nature, that the quality of software can be ensured (important for those requiring BS 5750) and that the portability can be maximised.
Phil Mawby, Dept. of Electrical and Elec. Lng., University College of Swansea
An Introduction to Computational Fluid Dynamics
The fourth Introductory School in Computational Fluid Dynamics will be held at the Cosener's House in Abingdon, 9 to 13 January 1995. The course is particularly suitable for those starting in CFD research. The topics to be covered include the N-S equations, explicit and implicit time marching methods, turbulence models, pressure correction methods and grid generation. This course was oversubscribed last year so, if you interested in attending, contact Debbie Thomas to reserve a place.
1995 Aerodynamic Research Forum - Call for Papers
The next Aerodynamic Research Forum to be held on 5-6 January 1995 will be organised by the Royal Aeronautical Society in conjunction with Advisory Group on Computational Fluid Mechanics (AGCFM) and the CFD Community Club.
Abstracts are requested which may be relevant to the directed research programme of EPSRC. They should be around 100-150 words and should be submitted by 1 October. These will include, but are not restricted to, the computation, measurement, prediction and validation of laminar flow, transition, complex configurations, high lift aerodynamics and unsteady aerodynamics with application to both internal and external flows. Abstracts representing current activities and/or status reports from within Industry would be particularly welcome.
Abstracts considering other aspects of aerodynamics which will be of interest and relevance to the forum will also be accepted. Abstracts may also be submitted for poster display only.
As for last year's event, all presenters will be asked to provide a one page summary of their presentation for distribution at the meeting. Presentations will be restricted to 20 minutes with a general discussion following each group of presentations. You will be informed in late October of the final selection of the forum.
Dr N J Wood, School of Mechanical Engineering, University of Bath
Conference on Numerical Methods for Fluid Dynamics
This international conference on CFD, to be held at the University of Oxford on 3-6 April 1995, is the fourth organised by the ICFD (Institute for Computational Fluid Dynamics). The aim of the conference, as in previous years, is to bring together mathematicians and engineers and other scientists in the field of computational aerodynamics and computational fluid dynamics to review recent advances in mathematical and computational techniques for modelling fluid flows.
The conference will cover all areas of CFD, but it is hoped to emphasise three main areas:
- Algorithms and algorithmic needs arising from applications
- Navier-Stokes on Flexible Grids
- Environmental CFD.
Two page abstracts for contributed papers should be submitted by 9 December 1994 stating preference for oral or poster presentation. Notification of acceptance will be given by 31 January 1995. Papers accepted for oral presentation will be required at the meeting for publication in the Proceedings.
Abstracts and enquiries regarding the conference should be addressed to me.
Mrs. B Byrne, Oxford University Computing Laboratory
News from the Visualization Community Club (VCC)
Visualization on the World Wide Web
The Steering Group of the Visualization Community Club has discussed ways of using the opportunity provided by the World Wide Web (WWW), one of whose strengths is the ability to gather references to geographically dispersed contributions into one place. We envisage a page where we can refer to examples of using visualization in the UK engineering research community and naturally these should be visual.
Initially the page might be short while engineering departments are still trying to devise how best to present their work on the Web, but we hope that this will grow into a more comprehensive and useful resource. I should emphasise that the idea is not to gather the information, but the references to the information.
If you have set up information on the WWW about your work and would like this to be more widely known, please let me know sending me an appropriate URL.
The WWW address for the Visualization Community Club is: http://web.inf.rl.ac.uk/vis/vcc/.
Reminder of imminent Community Club events
Thursday 22 September 1994 Visualization in the Built Environment at De Montfort University, Leicester
Wednesday 30 November 1994 Visualization of Experimental Data at Leeds University
Conferences and Meetings Notices
- Neuro-Nimes '94, (Neural Networks and their Applications), 10-14 October 1994, Paris, France
- Virtual Reality and Rapid Prototyping, 26-27 January 1995, University of Salford
- ETCE '95 Conference, (Support of Engineering Processes) 29 January -1 February 1995, Adam's Mark Hotel, Houston
- BCS Conference, (Display Systems) 22-23 February 1995, National Gallery, London
- Second International Workshop on Learning in Intelligent Manufacturing Systems, 20-21 April 1995, Hotel AGRO, H-1l21 Budapest
- 1995 International Conference on Computers in Education, (Call For Papers) 5-8 December 1995, Raffles City Convention Centre, Singapore
- AISA '95, 3 May 1995, Cranfield University
- Expert Systems '94, 12-14 December, 1994, St Johns College, Cambridge
- Fractal '95, 7-10 February 1995, Marseilles