An automated modeling approach for construction performance improvement using simulation and belief networks

Abstract

An automated modeling approach was developed for the improvement of construction operations by integrating computer simulation and belief networks. Computer simulation is used to model the construction operations while the belief network provides diagnostics to evaluate the simulated construction project performance. Belief networks, also called Bayesian networks, are a form of artificial intelligence (AI) that incorporate uncertainty through probability theory and conditional dependence. While the objective of most construction operations is either reduced cost or shortened duration, a surrogate objective, namely improved performance as measured by performance indices, has been identified to focus the recommendations of the belief network. Five domain-generic performance measurement indices were developed to facilitate the analysis of simulated construction operations: the Queue Length Index (QL), the Queue Wait Index (QW), the Server Quantity Index (SQ), the Server Utilization Index (SU), and the Customer Delay Index (CD). Where a performance index falls outside the acceptable limits or bounds, remedial actions are evaluated by the belief network. Remedial actions include modifying the number of servers or customers, and/or modifying the capacity of either the customer or server. The model has many advantages including: (1) the ability to compare various construction methods or operation strategies; (2) the ability to present solutions even if all user-defined constraints are not met; and, (3) the ability to present more than one solution. The contributions of this research are (1) the development of an automated approach for improving simulated operations, (2) the identification of a surrogate objective, performance improvement, that directs the improvement search toward changes in resource capacities, and, (3) the introduction of belief networks to construction research.