BUILDING RESEARCH ASSOCIATION OF NEW ZEALAND REPRINT. Development of marketable knowledge-based systems. A.H.R.Fowkes, W.R.

Transcription

1 BUILDING RESEARCH ASSOCIATION OF NEW ZEALAND REPRINT Development of marketable knowledge-based systems A.H.R.Fowkes, W.R.Sharman, and

2 NZES 88 THE DEVELOPMENT OF MARKETABLE KNOWLEDGE-BASED SYSTEMS A H R Fowkes W R Sharman A H Dechapunya Building Research Association of New Zealand (BRANZ) Judgeford, New Zealand. ABSTRACT Co-operation between BRANZ and Auckland University has brought knowledge engineering techniques from a research topic to a useful tool for information dissemination. Selection criteria for the subject domains of marketable knowledge-based systems, which have been developed using this team approach, are given. Resource requirements for production and field trials of two knowledgebased systems are detailed. The criteria for the field trials of BRANZ knowledge-based systems are discussed. 1.0 INTRODUCTION - a The Building Research Association of New Zealand (BRANZ) has a commitment to provide technical information to the building industry and the public (BRANZ, 1987). One of the vehicles for disseminating information is the online computer service BRANZLINE. The reasons for the development of this service were detailed in a paper to a recent international conference on building research and information (Fowkes, et a1 1986). BRANZLINE is split into two sub groups of online services, BRANZINFO and BRANSWERS. BRANZINFO is intended to contain systems which provide information passively, such as the bibliographic database which can be searched or browsed through by the user. Systems which give specific answers, actively leading the user to the information that they require, will be contained in BRANSWERS. The latter is the service where KBS is expected to play a major role as more building technology information is processed into a form more usable than at present by the building industry and the public.

3 It is important that the user is aware of the status of the information presented. Is it information BRANZ holds without being able to vouch for its accuracy or appropriateness (such as library information)? or is it the considered view of BRANZ? This grouping of information systems is designed to clarify the status of the information presented. The basis of BRANZINFO is the library database which is a bibliographic database of approximately 24,000 items on building technology held in the BRANZ library. The system utilises an easy to use menu driven front end program for database searching which incorporates an online ordering option. This part of the online computer service has been in public use since October BRANSWERS is a recent addition to the service, it contains LINBEAM, an application program for the design of lintels and beams supporting light timber frame construction. BRANZLINE has been designed to allow systems to be added to the service with no further restructuring. Knowledge-based systems (KBS) are thus aimed at possible future inclusion in that service. The delivery route for KBS is thus initially intended to be by terminal access to the BRANZ information computer via PACNET. The development of information systems including KBS are seen as part of an ongoing commitment to provide building technology information to the building industry and the public. The decisions as to which systems are to be developed for inclusion on the online service are made on the basis of the value of the information to the user not on the sophistication of the technology used. Systems at present under development for inclusion on BRANZLINE are being developed using conventional application programing, info-base (page display), database or knowledge-base technology as appropriate for the application. Cooperative research in information technology between BRANZ and Auckland University has provided BRANZ with another potential means of processing information so that it is more useful to the building industry and the public. At the centre of the cooperative research is the development of the CLASS language (Hamer et a1 1988) which is proving useful for developing building code and design applications. An environment now exists where KBS can be efficiently produced at BRANZ (Dechapunya 1987). The combination of hardware, software, BRANZ staff and Auckland University staff allows: - KBS and conventional work to coexist - The benefits of teamwork to be realised By late 1986, several KBS had been developed as part of BRANZ research into the uses of Information Technology. As these systems were evaluated it was realised that KBS techniques were emerging as useful tools for the dissemination of building technology information.

4 2.0 ISSUES IN THE PRODUCTION OF KBS FOR FIELD TRIALS A team of BRANZ staff comprising computer scientists, building scientists, liaison staff, editorial staff and management was invited to participate in addressing the issues involved in producing KBS for general use outside BRANZ. A working group was formed which identified the issues in the areas of production, marketing and corporate policy. The strategic aspects that were considered were detailed in a paper to this conference last year (Whitney 1987). The report submitted by this committee was used as a starting point for the planning of a production trial of KBS. Planning for the production trial could not proceed before the issues relating to system development, policy and marketing had been defined sufficiently for a user group and subject domains to be selected. 3.0 CRITERIA FOR PRODUCTION TRIALS Criteria were established by the BRANZ working group as follows: System development criteria The system must be based on knowledge which has: - high degree of technical information reliability - good presentation level - available human experts - a reasonable life expectancy Marketing criteria The system must be perceived as offering a significant benefit in its use. The possible benefits are an enhancement in performance over other methods with respect to: - speed - cost - effort - reliability - problem solving capability - flexibility The users of the system should have: - high volume of repetitive tasks - capability to change their work habits - a compatible working environment

5 Policy requirements - comprehensive documentation must be produced on the system and its technical audit. - the project must be included in the BRANZ management plan so that sufficient resources are applied to it and all staff are aware of the intentions of the trial. - a record of all sessions conducted during the field trials must be kept to enable an evaluation of the nature of the future uses of KBS to be carried out. 4.0 SELECTION OF MARKET SEGMENT AND KBS SUBJECT DOMAINS To find a market segment for trial and two systems which fit the criteria above needed an intricate juggling of the possibilities. In mid-1987 information systems in use, under development, e ither for the online service or for research purposes, were as follows: SYSTEM NAME FUNCTION TYPE OF SYSTEM REFERENCE - Search the library database Database LINBEAM - Design lintels or beams supporting Conventional light timber frame construction program FIREPROP GLASS CLIMDATA FIRECODE DAMP SEALANT - Fire properties of building materials. database of SANZ MP9 - Calculate the glass thickness for a window - Search through climatic data for New Zealand - Check designs comply with part of a draft fire code - Diagnose common moisture problems - Select a building sealant Database Conventional program Database Knowledge-base Knowledge-base Knowledge-base ADHESIVE WALL-BRACE - Select an adhesive for wood based products - Design wall bracing to the light timber frame code Knowledge-base Knowledge-base

6 To choose an appropriate user group required an analysis of the building industry use and likely use of computer systems. The analysis was based on personal experience of the industry built up by BRANZ in promoting its online information service since October BUILDING INDUSTRY BY DISCIPLINE Architects Engineers Quantity surveyors Architectural draughtsmen Design draughtsmen Draughtsmen COMPUTER USE Little direct use Heavy users often using complex programs, but hardly ever online Growing use hardly ever online Starting to use CAD Starting to use CAD Starting to use CAD APPROPRIATE SYSTEMS REFERENCE, LINBEAM, GLASS, FIREPROP, ADHESIVE, SEALANT DAMP, WALL-BRACE FIRECODE All REFERENCE REFERENCE, LINBEAM, GLASS, FIREPROP, ADHESIVE, SEALANT DAMP, WALL-BRACE FIRECODE All REFERENCE, LINBEAM, GLASS, FIREPROP, ADHESIVE, SEALANT DAMP, WALL-BRACE FIRECODE Building Inspectors Libraries in above Rapidly increasing use REFERENCE, LINBEAM, of computers in some local GLASS, FIREPROP, Authorities WALL-BRACE, FIRECODE Very knowledgeable and have REFERENCE access to online services. It would have been useful if all of the KBS that had been produced were suitable for the production trial. However, for a variety of reasons, only WALL-BRACE fulfilled all the criteria. The criteria for the system and policy led BRANZ to select a domain where there was very well defined information. Possibly the best defined and most used information in the building industry is that contained in building control and related documents. BRANZ has a commitment to provide technical information for inclusion in such documents and has a number of experts available in a wide range of building technology subjects.

7 The application of technical standards to building design and construction is often done by an iterative process of design followed by checking compliance with the standard. The groups most involved with this are building designers and design checkers. A checking system is inherently an easier process than design since no creative thought is required. A KBS for checking designs would therefore be easier to create than a design system. From the earlier analysis of the building industry it can be seen that the checking department of Local Authorities, and in particular building inspectors would be a suitable user group. In addition most of the programs under development contain some information of interest to this user group. Even though most of the systems do not fit the criteria for the trial, they could be shown to this group as a means of explaining what Information Technology is trying to achieve. To keep resources within reasonable bounds for the trial it was decided that two KBS should be produced and released for field trials to two Local Authorities. Four Local Authorities were visited and all were keen to participate. Manukau City and Wellington City were selected. During discussions with the Local Authorities, ideas for subject domains were pursued. BRANZ staff and the Local Authorities independently selected the wall bracing and sub-floor bracing parts of NZS 3604:1984 Code of practice for light timber frame buildings not requiring specific design as useful subject domains. The general area of light timber frame building design is one that BRANZ has been heavily involved in for some time. BRANZ has overlap in technical expertise in this area which fulfilled another of the criteria. WALL-BRACE is a system to check the wall bracing provided in a light timber frame house to resist the loads imposed on the house by wind and earthquake. The system will also assist the designer in meeting the requirements of the light timber frame code. WALL-BRACE was being developed as part of the continuing research into KBS by BRANZ and Auckland University, so the resources for its development were already established and allocated. SUBFLOOR-BRACE is a system to design the sub-floor framing (which includes all framing below ground floor level, braces, piles and foundations) needed in a light timber frame house to resist the loads imposed by wind and earthquake. The system will also assist the designer in meeting the requirements of the code. The resources required for the development of SUBFMOR-BRACE had to be established and obtained. The environment for the production of this KBS at BRANZ was considered to be excellent. Hardware, software and subject domain expertise were all available. However, in-house computing staff were fully committed on other projects. Fortunately, the BRANZ contact with the Universities led to Victoria University recommending a diploma student in computer science who turned out to be a suitable knowledge engineer.

8 The first assessment of time required to complete the checking part of SUBFLOOR-BRACE is approximately 750 hrs. Most of the time is allocated to the knowledge engineer, but time is also included for the subject expert, project manager and the information system coordinator. The role of the latter is to manage the online service and in this context represents the interests of the eventual users of the systems. 5.0 FIELD TRIALS The general aim is to determine the suitability of online information systems for providing information to the Building industry and the public. The objective of the trial is to obtain industry user's reactions to knowledge-based systems. This feedback will be used to determine future research and development emphasis in Information Technology. Planning for the field trials is currently under-way. It is envisaged that the trial will commence in late 1988 and proceed for 6 months. The systems are to be delivered via a dial-up PACNET connection in the same way as BRANZLINE, the online computer service to the building industry. Training of Local Authority personnel will be given, although its need is expected to be minimal. Monthly visits to note progress in use are to be made. Criteria for trial For the systems to be judged completely suitable as a means of specific information dissemination, the users should, by the end of the trial: Be able to use the system adequately. Prefer the KBS to a completely manual system in respect of its, speed, effort required in use, reliability, problem solving capability and flexibility. Monitoring Monitoring of the use of each KBS will be used to verify the above criteria. A questionnaire will be issued to the users to isolate online access and other extraneous problems from the trial.

9 6.0 CONCLUSION BRANZ is developing KBS and other information systems in response to users specific information needs. A multi-disciplinary approach was successful in selecting appropriate subject areas to meet the needs of the system developers, users, researchers and BRANZ strategy for information dissemination. Field trials of two KBS are due to start in late ACKNOWLEDGEMENTS The authors would like to acknowledge the contribution made by colleagues at BRANZ and Auckland University. REFERENCES BUILDING RESEARCH ASSOCIATION OF NEW ZEALAND Annual Report. Building Research Association of New Zealand. Judgeford. Dechapunya, A. H Environment for BRANZ Research and Development in KBS, The Second New Zealand Conference on Expert Systems, NZES'87 Auckland. Fowkes, A. H. R., Dechapunya, A. H., and Smith, A. G The Development of the BRANZ Computerised Information Service. Proceedings of the 10th Triennial Congress of the International Council for Building Research, Studies and Documentation, Volume 2, Washington DC. Hamer, J., Hosking, J. G., Mugridge, W. B The evolution of the Class Language. To be presented at the Third New Zealand Expert Systems Conference, NZES188 Wellington. Standards Association of New Zealand Code of practice for light timber frames buildings not requiring specific design. NZS Wellington Whitney, R. S Knowledge-based Systems for Building Technology: Strategic Aspects, The Second New Zealand Conference on Expert Systems, NZES187 Auckland.

10 Corv 2 B The develornlent of market able kr.iowlelge-based svst

11 BUILDING RESEARCH ASSOCIATION OF NEW ZEALAND INC. HEAD OFFICE AND LIBRARY, MOONSHINE ROAD, JUDGEFORD. The Building Research Association of New Zealand is an industry-backed, independent research and testing organisation set up to acquire, apply and distribute knowledge about building which will benefit the industry and through it the community at large. Postal Addrc>ss BRANZ, Fr~vate Bail, PCI~II~I~