铁路病害预测算法的研究
首发时间:2020-04-03
摘要:近年来人工智能研究取得了巨大成果,给各行各业带来新的机遇和挑战。但是在突破传统的人工检测从而实现快速识别铁路病害仍存在相当的挑战。由于常年暴露在复杂恶劣的自然环境中,同时承受列车荷载的反复作用,轨道结构存在多种类型的变形病害,这些病害会引起轨面平顺状态恶化,不仅影响列车运行的平稳舒适性,甚至威胁行车安全。为了实现在保证一定精度的前提下快速定位问题铁路点,本文通过提出利用stacking融合算法级联多种机器算法以及神经网络对已有的铁路数据进行回归预测,再通过阈值审查的方式定位问题铁路点,从而实现在保证一定准确率的前提下,快速定位铁路问题点。本文的模型使用了三个传统的机器学习算法,同时改进了用于文本分类的RCNN网络用于预测。在使用均方根误差的评价标准下,本文最终结果接近0.04。
For information in English, please click here
Research on Algorithm of Railway Disease Prediction
Abstract:However, there are still considerable challenges in breaking through the traditional manual detection to quickly identify railway diseases. Due to the long-term exposure to complex and harsh natural environments and the repeated effects of train short circuits, there are many types of deformation diseases in the track structure. These diseases will cause the smoothness of the track surface to deteriorate, which will affect the smooth and comfortable operation of the train, Threatening driving safety. In order to realize the rapid location of the problem railway point with a certain accuracy, a stack fusion algorithm is used to cascade multiple machine algorithms and neural networks to perform regression prediction on existing railway data, and then locate the problem through threshold review. Railway points, so as to quickly locate railway problem points with a certain accuracy in the guarantee. The model here uses four traditional machine learning algorithms, while improving the RCNN network for text classification for prediction. By comparison, under the evaluation standard of mean square error, the result of a certain model is 0.13, and the result of this paper is close to 0.04, which is an increase of 9 percentage points.
Keywords: Neural Networks Railway disease Model fusion
基金:
引用
No.****
动态公开评议
共计0人参与
勘误表
铁路病害预测算法的研究
评论
全部评论0/1000