A Semi-Supervised Adaptive Matrix Machine Approach for Fault Diagnosis in Railway Switch Machine.

Area of Science:

• Railway Engineering
• Machine Learning
• Signal Processing

Background:

• Switch machines are critical for railway safety, but traditional fault diagnosis requires extensive labeled data.
• Existing semi-supervised learning (SSL) methods struggle with imbalanced pseudo-labels and poor data representation.

Purpose of the Study:

• To introduce the Semi-Supervised Adaptive Matrix Machine (SAMM) model for improved switch machine fault diagnosis.
• To address the challenges of limited labeled data and data imbalance in semi-supervised learning.

Main Methods:

• SAMM integrates SSL with adaptive technologies, using an adaptive low-rank regularizer for matrix data and adaptive penalties for category imbalance.
• The model expands labeled datasets via probabilistic outputs and semi-supervised learning, with automatic parameter adjustment.
• Optimization is achieved using the alternating direction method of multipliers (ADMM).

Main Results:

• SAMM demonstrated superior performance compared to existing baseline models on a switch machine current signal dataset.
• The model shows exceptional status diagnostic capabilities, particularly with scarce labeled samples.
• Experimental results validate SAMM's effectiveness in semi-supervised classification tasks with matrix data.

Conclusions:

• SAMM provides an innovative and effective approach to fault diagnosis for railway switch machines.
• The model successfully overcomes limitations of traditional methods and existing SSL techniques in data-scarce scenarios.
• SAMM offers robust semi-supervised classification for matrix data, enhancing railway operational safety.