Building a KBS for Health and Safety Assessment

EASE Estimation and Assessment of Substance Exposure is a knowledge based system for assessing workplace exposure to potentially hazardous new substances It was built for the Health and Safety Executive of the United Kingdom HSE EU regulations require a manufacturer of a new substance to notify the appropriate authority who will carry out a risk assessment of it To aid this process guidance is provided by the regulator for use by both authorities and manufacturers This is usually provided as a paper document but HSE decided that a computer based guidance system would be of bene t Safety related considerations and a desire for quality led to the system being developed in accordance with ISO standards The system guides the user by o ering a menu of appropriate choices whenever it needs information The use of CommonKADS ensured that the problem require ments and the expert knowledge involved were captured within a standard framework which promoted unambiguous communication between the members of the project team and provided a solid base for system design implementation maintenance and enhancements The system was implemented using the NASA CLIPS development tool for the inference engine and knowledge base Initially required to run under MS DOS on a PC AT equivalent with K of RAM a second release to run under Windows reused the inference engine and knowledge base requiring only a revised user interface The system has been widely distributed for use by authorities throughout Europe Also appears in Applications and Innovations in Expert Systems IV Proceedings of BCS Expert Systems Cambridge December SGES Publications Problem Description EASE Estimation and Assessment of Substance Exposure is a knowledge based system for assessing workplace exposure to potentially hazardous substances which was built for the Health and Safety Executive of the United Kingdom EU directives and regulations require manufacturers or suppliers of new or existing substances to notify the appropriate authority who will carry out a risk assessment of it To aid this process guidance is provided by the regulator for use by both authorities and manufacturers This is provided as a paper document but HSE decided that a computer based guidance system would also be of bene t The objective of the application is to enable assessors to produce a reliable estimate of the degree to which workers will be exposed to the new or existing substance Exposure assessment is an expert task and is best performed by an experienced occupational hygienist For the system to be deployed it must incorporate best practice knowledge engineering techniques and as there are safety related considerations and both HSE and the developers of the system are committed to quality the work had to be carried out in accordance with ISO standards A knowledge based system was built which ensures that assessors remember all the nec essary questions to ask when estimating exposure to a substance and then calculates the exposure level The system also incorporates backtracking facilities in order to allow er rors to be corrected and What If analyses to be performed Two versions of the system were built The rst was targeted to run on PCs or higher under MS DOS which was the only hardware available to some manufacturers when the system was built a subsequent version was designed to run under Windows with a revised user interface The system had been designed in a modular fashion so that the second version of the system required only a few changes to the inference engine and knowledge base and most of these changes were in response to users comments Knowledge Acquisition and Analysis A prototype of this KBS had previously been produced using a di erent KBS program ming tool The textual knowledge base from the prototype of the KBS was used as a transcript which was analysed in the same way as any interview transcript fragments of text were identi ed as being relevant to the task of problem solving and were stored in dictionaries of concepts properties tasks or other ontological types In addition a number of decision trees were built e g Figure which represented some of the decision processes which take place during problem solving These decision trees along with the knowledge analysis models which are described later were produced using TOPKAT which is a tool for supporting knowledge acquisition and knowledge modelling Kin The knowledge analysis models which were developed used the CommonKADS method ology Bv CommonKADS is the name of the methodology developed by the KADS II project which was funded under the CEC ESPRIT programme SWd Wie CommonKADS views KBS development as a modelling process Knowledge analysis is performed by creating up to six models which represent the knowledge from di erent viewpoints CommonKADS recommends a number of di erent models which start with the representation of various aspects of an organisation and support the whole knowledge engineering process up to the point of producing a detailed design speci cation Four of the six CommonKADS models are primarily concerned with identi cation of a suitable application for an AI system and so were not considered relevant for this application This was because the application for the KBS had already been chosen and because an earlier attempt at automation had resolved many of the issues connected with the proposed role of the knowledge based system and its required inputs and outputs However the remaining two models the Expertise Model and the Design Model were built A member of HSE s sta cooperated in the development of the models allowing him to understand the models su ciently to criticise them constructively and to modify them in the future A few questions were referred back to a senior expert at HSE who also provided sta to perform user testing of the system Knowledge Analysis CommonKADS Expertise Model The CommonKADS Expertise model is divided into three levels representing di erent viewpoints on the expert knowledge Inhalation Direct handling