Development of a knowledge-based system approach for decision-making in construction projects

Developing a realistic cost estimate is vital to the construction project at every stage of planning and construction. Construction costs have traditionally been estimated using two methods: (1) comparative estimating and (2) quantity-take-off estimating. There are inherent limitations in each process that restrict its usefulness in supporting engineering decisions in the early critical stages of project feasibility and design. Neither method is very suitable in forecasting the costs of one-of-a-kind, never-before-built facilities. A knowledge-based system is proposed and prototyped to enhance the construction planner's ability to clarify decisions and isolate the preferred course of design based on both engineering and cost decisions. The basic thesis of the model design is that techniques can be developed to relate construction project requirements to current cost data through accepted engineering practices in order to develop better cost estimates for initial budget and feasibility decisions--without detailed design documents. The modeling concept uses a knowledge-based system to integrate expert judgments, engineering criteria, safety codes, historical data, descriptive factors, and specific site conditions. The knowledge-based system is prototyped and applied, as a case study, to estimating the costs of the Walton County Convalescent Center. A nursing home provides an excellent case study because it demonstrates complex modeling interrelationships and the applicability of the knowledge bases to nonmilitary construction projects. The results of the model were rated against multiple measures of merit: accuracy, specificity, timeliness, and user friendliness. Under each measure, results strongly supported the knowledge-based parametric cost engineering models over traditional methods. The modeling method, with only general information of the facility requirements was within 2. 8% of direct costs and 6% of construction bid costs computed using traditional methods. This accuracy was achieved with less than one-thirtieth the man-hours required using traditional methods. Twenty-nine other accuracy data points comparing final to estimates resulted in an average of 7. 3% using the knowledge-based system compared to 23. 9% using traditional methods. The case study demonstrates the specificity and user-friendliness in exercising the expert system. In addition, six areas for potential future research are discussed.