Aiming at the problems of various fault types, such as the great difference in fault knowledge expression and the weak fault causality reasoning ability in hydraulic systems of bridge erecting machines, which lead to a low accuracy of fault component location in hydraulic systems, a hydraulic system fault diagnosis method based on ontology Bayesian networks was proposed. Firstly, by analyzing the fault knowledge of the hydraulic system for a bridge erecting machine in detail, the fault ontology was formally defined, and the fault ontology model of the hydraulic system was constructed with probabilistic extension. Subsequently, the conversion rules for the ontology Bayesian network were established, based on which the automatic transformation from the ontology model to the Bayesian network model was realized by using the Jena API. This conversion process was facilitated by the maximum likelihood estimation algorithm, resulting in an optimal Bayesian network model for fault diagnosis. Finally, a certain model of the hydraulic system for a bridge erecting machine was investigated using this methodology, and the Netica simulation platform was employed to conduct diagnostic reasoning from observed fault phenomena to fault components. The experimental results demonstrate that this approach enhances the accuracy of fault diagnosis and can provide a reference for the fault diagnosis of construction machinery hydraulic systems.