LD-RPMNet: Near-Sensor Diagnosis for Railway Point Machines

2025-06-01

AI Summary

To address the dual requirement of model lightweighting and robustness for edge-deployed fault diagnosis of railway point machines using proximal sensors, this paper proposes LD-RPMNet—a lightweight hybrid network. Methodologically, we introduce a Multi-scale Depthwise Separable Convolution (MDSC) module to efficiently extract multi-resolution time-frequency features, and incorporate a Broadcast-style Self-Attention (BSA) mechanism to enhance global dependency modeling, enabling compact integration of Transformer and CNN architectures. Experimental results demonstrate that LD-RPMNet reduces parameter count and computational complexity by 50% compared to baseline models, while achieving a diagnosis accuracy of 98.86%—a 2.9 percentage-point improvement. The model thus delivers high diagnostic precision alongside significantly enhanced edge deployability and interference robustness.

Technology Category

• Machine Learning: Learning on the Edge & Model Compression
• Knowledge Representation and Reasoning: Diagnosis and Abductive Reasoning
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs

Abstract

Near-sensor diagnosis has become increasingly prevalent in industry. This study proposes a lightweight model named LD-RPMNet that integrates Transformers and Convolutional Neural Networks, leveraging both local and global feature extraction to optimize computational efficiency for a practical railway application. The LD-RPMNet introduces a Multi-scale Depthwise Separable Convolution (MDSC) module, which decomposes cross-channel convolutions into pointwise and depthwise convolutions while employing multi-scale kernels to enhance feature extraction. Meanwhile, a Broadcast Self-Attention (BSA) mechanism is incorporated to simplify complex matrix multiplications and improve computational efficiency. Experimental results based on collected sound signals during the operation of railway point machines demonstrate that the optimized model reduces parameter count and computational complexity by 50% while improving diagnostic accuracy by nearly 3%, ultimately achieving an accuracy of 98.86%. This demonstrates the possibility of near-sensor fault diagnosis applications in railway point machines.

Problem

• Lightweight model for near-sensor railway fault diagnosis
• Optimizing computational efficiency with Transformers and CNNs
• Improving diagnostic accuracy while reducing parameter count

Innovation

• Lightweight model combining Transformers and CNNs
• Multi-scale Depthwise Separable Convolution module
• Broadcast Self-Attention for efficient computation