A knowledge-based system approach to helping engineers understand codes of practice

Civil engineers are increasingly called upon to design according to codes of practice which are foreign, or otherwise unfamiliar to the engineer. Some form of 'tool' to aid such engineers in the safe and effective use of unfamiliar codes, is therefore highly desirable. The argument presented in this thesis, is that a specialised 'knowledge-based system' (KBS) can be successfully constructed in order to provide various sons of insights into the design methods used in certain codes of practice. Three key ideas were used in the implementation produced during the research: the development of a novel set of complementary 'facilities' for examining the design methods used in codes; the support of comparison between the examination of two different codes; and ensuring the system's representation of codes and their examination, could be made readily comprehensible to engineers by using familiar human language phrases. Seven different facilities were developed in the research, including: the ability to view the codes installed in the system in a form close to a human language (such as English or French); the ability to perform parts of a code-based design procedure to various levels of detail; and the ability to inspect the interdependences of design parameters within codes. Use of these symbolic and numerical methods could provide the engineer with the information required to understand how and why an unfamiliar code would specify surprising, or otherwise unusual design parameters in some particular situation. They could also be used in a more exploratory fashion, with t1ie same broad aim of greater understanding of an unfamiliar code. A KBS is a sophisticated computer program that uses the idea of processiilg knowledge information. A characteristic feature of KBSs is that one of their primary components is a 'knowledge base' - a store of human expertise. The KBS built in this research, 'COPES' used an existing abbreviated form of the reinforced concrete (RC) beam design codes as its knowledge base. In particular, it contained 'procedural knowledge'. COPES was implemented using conventional computer systems and progranuning languages (pascal and FORTRAN on a Sun workstation). This is in contrast to most contemporary KBSs, which are often built using a 'shell', or an unconventional declarative programming language such as Prolog. One reason for this choice was that COPES used parts of previous computing work done with RC beam design codes, that had also used conventional computing techniques. However, our research did cover an investigation into the prospects for an alternative approach using a modem expert system shell. (It was confirmed that such an approach was generally less suitable in this particular application.) The COPES system proved to be a useful prototype 'toolbox' of various procedural knowledge extraction operations, which could help an engineer's understanding of an unfamiliar code of practice. To provide a practical system, the various explanatory methods developed could potentially be incorporated into an overall CAD (Computer-Aided Design) environment, or alternatively, wrapped up in a more sophisticated interactive program.