Research data supporting chapter 'SNN with Time-Varying Weights for Rail Squat Detection' of Dissertation 'AI Solutions for Maintenance Decision Support in Railway Infrastructure'

The data and codes were prepared and uploaded to 4TU.ResearchData by Wassamon Phusakulkajorn to support the results in Chapter 3 (SNN with Time-Varying Weights for Rail Squat Detection) of her dissertation. This chapter has been submitted for publication as Phusakulkajorn, W., Hendriks, J.M., Li, Z., Núñez, A., Spiking Neural Network with Time-Varying Weights for Rail Squat Detection. In this research, we develop a spiking neural network (SNN) with time-varying weights to detect rail surface defects, e.g., squats, of varying severity levels, using ABA measurements. This method aims to improve the detection accuracy of light squats, which present challenges due to their subtle, short-duration responses and typically a low percentage of appearance in ABA signals compared to healthy rails. Instead of using large network architecture, this work uses simple network architecture with no hidden layers to solve a complex spatiotemporal problem presented in early squat detection. The data used in this research contain four UCI benchmarks (Liver disorders, Breast cancer, Ionosphere, and Iris) and real-field ABA measurements from Dutch and Swedish railways. All implementations are done in MATLAB, where (.mat) files are analytical solutions and (.eps) and (.jpg) are figures used in the main manuscript.

本数据集的相关数据与代码由Wassamon Phusakulkajorn整理并上传至4TU.ResearchData，用于支撑其博士论文第3章《基于时变权重脉冲神经网络的轨道扁疤检测》的研究成果。该章节已以Phusakulkajorn, W., Hendriks, J.M., Li, Z., Núñez, A. 为作者，标题为"Spiking Neural Network with Time-Varying Weights for Rail Squat Detection"的论文完成投稿待刊。本研究构建了一款具备时变权重的脉冲神经网络（Spiking Neural Network，SNN），旨在通过ABA（Acceleration-Brake-Acceleration）测量数据检测不同严重程度的轨道表面缺陷，如扁疤；针对因信号响应微弱、持续时间短，且相较于健康轨道，ABA信号中扁疤占比极低而导致检测难度较高的轻型扁疤，本方法旨在提升其检测精度。不同于依赖大规模网络架构的解决方案，本研究采用无隐藏层的简洁网络架构以解决早期扁疤检测中存在的复杂时空问题。本研究使用的数据集包含四类UCI基准数据集，分别为肝脏疾病数据集、乳腺癌数据集、电离层数据集与鸢尾花数据集，同时包含来自荷兰与瑞典铁路的实测ABA测量数据。所有实验实现均基于MATLAB平台完成，其中(.mat)格式文件为分析结果文件，(.eps)与(.jpg)格式文件为主文稿中使用的配图。