Research On Data Mining Algorithm For Dynamic Inspection Data Of High-speed Railway Bridge

The Track Dynamic Inspection Data is the data obtained by periodically testing the entire track since the line was opened.With the continuous development and improvement of railway digitalization and informatization,the amount of dynamic inspection data is increasingly rich,containing a lot of potential information and knowledge.In recent years,with the rapid development of artificial intelligence,data mining theory and technology,research on mining and analysis methods based on Track Dynamic Inspection Data,timely grasp the track operation status,and monitor abnormal conditions,not only for timely and effective evaluation of infrastructure status,but also for safe operation.Providing technical support has important application value,and it has important reference significance for discovering the internal laws of infrastructure evolution and tapping potential knowledge.At the same time,the characteristics of high data size,large amount of data and complex noise are also a great challenge for data mining technology research,and it has important theoretical research value.In the construction of high-speed railways in China,the bridges account for more than half of the total mileage.It has a very important position in the railway infrastructure and is also a key area for the operation and maintenance of public works.In this paper,in-depth data mining research on high-speed rail bridge motion detection data,firstly from the time-frequency analysis of orbit dynamic detection data,extract the high-speed rail bridge data information,and then research on the difficult problem that affects data mining research-mileage drift,on this basis Data clustering technology is used to study the evolution law of bridges and its internal relations from two angles,and the data mining results are reasonably and valuablely explained from the engineering perspective.The main work and research results of this paper include:(1)Aiming at the complex problem of Track Dynamic Inspection Data including noise and high-speed railway bridge characteristics,a high-speed bridge motion detection data extraction algorithm based on time-frequency local features is proposed.Firstly,the wavelet analysis theory is used for denoising preprocessing.Secondly,the high-speed railway bridge features are analyzed by using time-domain representation based on domain transformation.Then,the high-speed railway bridge is successfully identified by the two-step segmentation method of local extremum method and local boundary correction.Regional and non-bridge areas.Finally,the experimental results of 24m simply supported bridges and 32m simply supported bridges in the range of 0-300km of Beijing-Shanghai high-speed railway are analyzed.The accuracy of bridge identification is high.The bridge inspection data is successfully extracted from the whole line inspection data of railways and railways.Lay the foundation for data mining of high-speed rail bridges.(2)Aiming at the actual engineering problem of drift of the Track Dynamic Inspection Data,this paper proposes a high-speed railway bridge motion detection data mileage correction algorithm based on local waveform matching.Firstly,the local waveform is firstly matched based on the correlation coefficient,and then the local correction is performed through the curve feature points.Finally,the test data of the 100-200km calendar of the Beijing-Shanghai high-speed railway is tested,and the accuracy of the mileage correction result is proved by professional software.The study of the evolution law of high-speed railway bridges provides data security,and lays a foundation for all research based on Track Dynamic Inspection Data,and has strong engineering practicability.(3)The research on spatio-temporal data clustering algorithm of high-speed railway bridge,in order to discover the evolution law of bridges and its internal relations,from the perspective of time and space,based on different distance metric clustering algorithms.15 data and geographical curves of detection items were found.The relationship between the bridges can be divided into curved regional bridges and non-curved regional bridges.Based on the similarity measure of DTW distance,the evolution law of high-speed railway bridges with different years is found,which indicates that the clustering results have strong engineering interpretability,which is of great practical significance for the prediction of irregularity of high-speed railway bridges.