Belief networks for engineering applications

A relatively new form of artificial intelligence, namely belief networks, p rovides flexible modelling structures for capturing and evaluating uncertai nty. The belief network consists of nodes to model the variables of the dom ain, and arcs to represent conditional dependence between variables. The de velopment of a belief network requires four major steps: variable definitio n, identification of conditional relationships, definition of the states of the variables, and determination of the probabilistic values of the condit ional relationships. The evaluation of a singly connected belief network is provided. Two applications for belief networks are discussed. One applicat ion involves the integration of a belief network with computer simulation r esulting in an automated system for performance improvement. The second app lication is focused on assessing productivity of construction operations.