An expert system prototype for construction planning and productivity analysis

The identification and specification of construction operations and the selection of construction methods, including equipment and labour resources, are central activities in the production planning of construction projects. The decisions taken are complex and the choice of construction method and estimate of production output have serious consequences for the effectiveness of programming in achieving the most economic and practicable solution production planning problems. Improving construction productivity by detailed and accurate planning is critically important in the increasingly competitive environment of the construction industry. Mathematical or statistical models of factors affecting production output together with knowledge based systems, available to the planner through the medium of the micro-computer, could be valuable tools by which to improve planning and productivity and hence minimise the costs of construction. In this thesis aspects of construction planning are discussed and opportunities identified for the application of a knowledge based system to interface with conventional network programming computer software. The development of a prototype system to help the planner in generating construction activities and their logical relationships, and in estimating activity durations, is described. The system uses a knowledge base of construction activities and factors affecting productivity to assist the user in the preparation of input files for his CPM program. The knowledge structure of the prototype system is composed of frames, production rules and procedures. The frames consist of facts used to represent the structure hierarchy of activities, the formalised task precedence relationships, description of construction tasks and the information about the productivity factors. The production rules are heuristics used to determine the breakdown of activities into scheduling tasks and precedence of activities, and to quantify productivity factors. The procedures are used for activity durations estimation, generating reports, interfacing with the network program and to carry out a Monte Carlo simulation to determine cumulative probability distributions of activity duration. The results of this pilot system have shown that the system can be an effective productivity estimation tool for construction planners. The ability of the system to improve the quality of construction schedules further suggests that the model developed is rigorous enough to warrant its continued development into a commercial system.