Abstract: For the massive data of settlement observation of high-speed railway under construction, in order to quickly, batch and automatically identify abnormal data that can not reflect the real settlement state, based on the historical big data of settlement observation and mathematical statistics methods, the reference threshold for identifying abnormal settlement data of subgrade, bridge and culvert, and tunnel under each main working condition is formulated. After removing invalid observation data, a strategy of prioritizing sensitivity and assisting in verifying stability with calculation results is adopted to compare the fluctuation of average deformation rate with the reference threshold, achieving automatic identification of abnormal settlement data. Using these methods, the railway settlement deformation observation information system is designed and developed and popularized in the Xi'an-Ankang and Xi'an-Yan'an high-speed railway projects. Empirical calculation shows that between the single deformation rate and the average deformation rate, the method's misjudgment rate can be suppressed and the stability can be improved by setting sensitivity adjustment coefficient. In the project, it is possible to quickly identify abnormal settlement data and the location of the mileage.

Key words: high-speed railway, identification method, model test, settlement observation, abnormal data, sensitivity