Rapid Detection of Rotating Machinery Anomalies Based on Gramian Angular Field With Gray Level Co-Occurrence Matrix

To address the challenges associated with the early detection of rotor faults, this paper proposes a Dynamic Time-Frequency Feature Mapping Analysis (DTFMA) algorithm that integrates time-frequency feature fusion. Specifically, DTFMA first transforms rotor vibration signals into the time-frequency domain and reconstructs the transformed signals into two-dimensional (2D) feature images— a key step that significantly enhances the distinguishability of fault-related features. Building on this, a dedicated feature fusion strategy is employed to extract and integrate critical information from these 2D feature images; this strategy is further coupled with an adaptive threshold update mechanism, which enables rapid fault alarming. Compared with traditional approaches, this method exhibits superior performance: through the dimensional reconstruction of vibration signals, it achieves fast and accurate capture of early-stage rotor faults. Notably, the novel application of time-frequency feature fusion in the DTFMA framework effectively improves the robustness of early fault features, while the embedded adaptive alarm mechanism ensures both high detection sensitivity and reliable performance—two core requirements for practical early fault detection systems.