• A novel data-driven method based on Bayesian inference is proposed for accurately localizing intermittent connection (IC) faults in DeviceNet networks with complex topologies.
• The method integrates observation symptoms and network topology information to derive suspected IC faults and compute posterior probabilities without interrupting normal system operation.
• A maximum likelihood-based fast diagnosis algorithm enables rapid identification of fault locations in multi-fault scenarios.
• Experimental case studies on a laboratory testbed under various topologies demonstrate that diagnosed IC fault locations agree well with experimental setups.