Two-Dimensional Elastic Space Learning for Fault Diagnosis of Rigid Cage Guides

The safety and operational efficiency of coal mining operations are fundamentally dependent on the reliable performance of mine hoisting systems, where rigid cage guides serve as critical components for vertical transportation. Complex underground environments are characterized by harsh conditions, multi-source noise interference, and variable operational loads, which present significant challenges for accurate fault diagnosis. To address the limitations of traditional fault diagnosis methods in handling complex time-frequency fault patterns of rigid cage guides, a fault diagnosis framework based on Two-Dimensional Elastic Space Learning (2DESL) is proposed in this article. First, one-dimensional vibration signals are transformed into time-frequency images using Continuous Wavelet Transform to capture comprehensive fault characteristics. Next, a unified structure preservation model is constructed that integrates both global statistical patterns and local geometric relationships for multi-scale feature representation. Then, a dual-directional dimensionality reduction strategy is implemented to extract features along both time and frequency dimensions simultaneously, preserving complex two-dimensional fault patterns. Finally, an adaptive Feature Guidance Mechanism (FGM) is introduced to perform intelligent feature selection and weighting, thereby enhancing noise immunity and reducing data redundancy. By incorporating multi-scale structural information and joint time-frequency distributions, the framework enables accurate identification of various fault types including misalignment and gap defects while maintaining robustness under high-noise conditions. The effectiveness of the proposed method is validated using experimental data from a hoisting system testbed, achieving an average fault recognition accuracy of 95.58%.