Updating Bayesian network for diagnostic failure analysis of construction equipment

Construction equipment is an important type of resources of heavy construction contractors. Since equipment breakdowns can cause project delays and significant financial losses, the contractors are eager to know those factors causing equipment failures directly or indirectly, related to equipment design, maintenance, and operations. Although Bayesian network can be used for diagnostic analysis of failure events or making predictive analysis, building a Bayesian network for such purpose can be difficult as the cause-effect relations can be subjective and their conditional probabilities change with a wide variety of causal factors. A hybrid approach is proposed in this paper to update the Bayesian diagnostic network structures and parameters using real life data, the conditional probabilities and cause-effect relationships can be dynamically updated with observed failure records to reflect the real life situations of a complex equipment system. A case study is conducted to show the benefits of the hybrid approach in construction equipment diagnostic analysis.