基于机器学习的地铁轨道几何劣化规律个性化建模研究

较高的轨道平顺性是保障地铁列车安全舒适运行的基础,准确掌握地铁轨道的劣化规律对保障轨道质量具有重要意义。根据地铁线路特点,选择影响地铁轨道质量劣化的7类异质性因素,给出赋值模型,并基于机器学习方法建立轨道质量指数(track quality index,TQI)短时预测前馈神经网络模型。为了验证模型,采集了北京地铁1号线的线路设备数据及2016年8月15日至2019年2月18日间的17次TQI检测数据,形成训练数据集和测试数据集,并采取深度学习技术,利用训练数据集对该模型进行训练。基于测试数据集的模型预测值的可决系数为0.938,平均绝对百分比误差为4.80%,结果表明该模型是有效的且具有较高的预测精度。

Individualized Modeling of Subway Track Geometry Degradation Law Based on Machine Learning

High quality of track regularity is the foundation for maintaining the safety and comfortable operation of subway trains. An accurate understanding of the degradation law of a subway track is extremely significant in ensuring the quality of the track. According to the characteristics of subway lines, 7 heterogeneous factors that affect the quality degradation of subway tracks were selected and the corresponding assignment model was proposed. Based on machine learning methods, the short-time prediction feedforward neural network model for track quality index (TQI) was established. In order to validate the model, the data of line equipment and 17 TQI detections, from Aug. 15, 2016 through to Feb. 18, 2019, of Beijing subway Line 1 were collected to generate a training and test data set; the former was used to train the model via deep learning technology. The coefficient of determination and the mean absolute percentage error of the prediction values based on the test data sets were 0.938 and 4.8%, respectively. The results show that the model is effective, with a high prediction accuracy.