A purity-driven hierarchy fault diagnosis of Multifunction Vehicle Bus

Abstract Existing diagnostic methods for Multifunction Vehicle Bus (MVB) network faults face inherent trade-offs among lightweight, classification accuracy, and computational efficiency. To address this limitation, we propose an Interval Purity-Guided Hierarchical Decision Tree (IPGDT). The framework dynamically constructs a comprehensive indicator system. This streamlines decision paths and constrains splitting complexity. Crucially, an interval purity-driven criterion governs node partitioning, significantly enhancing fault discriminability.Furthermore, we design a hierarchical screening strategy for tree construction. This approach mitigates cross-type interference by isolating overlapping thresholds based on purity magnitude. Consequently, the method satisfies the millisecond-level real-time constraints of onboard systems. Extensive cross-validation results demonstrate that IPGDT achieves a fault diagnosis accuracy of over 99% with a single inference latency below 10 ms. The proposed method also maintains high diagnostic performance under electromagnetic interference conditions (500 V/1000 kHz).