【摘要】 为准确描述各种条件下轨道不平顺复杂劣化过程,本文基于最小描述长度准则,建立一套动态检测数据驱动的轨道不平顺劣化自适应分段建模方法(Minimum-DescriptionLength-Based Rail Track Deterioration Adaptive Segmentation Framework, MDL-RTDAS),将维修作业导致轨道状态劣化过程突变的识别问题转化为模型选择问题,并设计求解算法.根据昌福高速铁路下行方向K21+184～K220+308路段近5年的历史动态检测数据,验证MDLRTDAS的有效性;从识别准确度,模型拟合的残差和容忍检测数据异常干扰方面验证了MDL-RTDAS优于同类模型.结果表明:在缺乏完整、准确维修作业信息的情况下,MDLRTDAS能够克服检测数据异常的干扰,感知劣化趋势变化,自动识别出维修作业造成的轨道不平顺劣化趋势突变,将劣化过程准确分段;相比于同类模型,MDL-RTDAS能更精确、有效地实现轨道不平顺劣化过程的自适应分段建模.更多还原

【Abstract】 To precisely describe the complicated track irregularity deterioration under the varying circumstances,this paper proposes a rail track deterioration adaptive segmentation framework based on minimum description length principle, referred as MDL-RTDAS. In MDL-RTDAS, the identification of the maintenance activities that result inmutations in the deterioration process is reformulated as a model selection problem. The algorithm is also proposed to solve the models. The effectiveness of MDL-RTDAS is verified by using the recent five years measurement data from the mileage K21 + 184 to K220 + 30 on Nanchang-Fuzhou railway. The MDL-RTDAS is compared with other similar algorithms in the accuracy, fitness and robustness. As the results indicate, under the conditions that the information of maintenance operations is incomplete and inaccurate, MDL-RTDAS is able to overcome the interference of contaminated measurements, precisely identify the mutations in deterioration rate caused by maintenance activities, and create a piecewise fitting model for track irregularity deterioration.Compared to other algorithms, MDL-RTDAS owns better performances in rail track deterioration adaptive segmentation.更多还原