Research On Multi-source Heterogeneous Data Fusion Technology In Railway Intelligent Operation And Maintenance Systems

In the current era of rapid development of railway intelligence,the construction of railway intelligent operation and maintenance systems can further enhance the core competitiveness of railway technology and reduce railway operation and maintenance costs,which has a profound impact on the promotion of China’s economic development.The railway intelligent operation and maintenance systems converges and fuses the equipment monitoring data in accordance with the "platform + application" model,facilitating the subsequent data application.As the aggregated equipment monitoring data has complex data sources,different semantics and is prone to conflicts.Therefore,it is imperative to effectively solve the problem of fusing multiple sources of heterogeneous data in the railway intelligent operation and maintenance systems.Aiming at the problem of insufficient fusion of the underlying monitoring data under the railway intelligent operation and maintenance system,this thesis takes the centralized signal monitoring data in the railway intelligent operation and maintenance systems as the main research object,analyzes the characteristics of each monitoring equipment,and extracts the characteristic attributes of the data monitoring.The experimental data conversion was completed,and the corresponding relationship between the attribute characteristics of the monitoring data and the data fusion decision was established.On the basis of in-depth research on data fusion algorithms,a data fusion model based on D-S evidence theory and a data fusion model based on rough set theory are proposed to be applied to the railway intelligent operation and maintenance systems to effectively improve the performance of intelligent decision-making capability.The main research contents of this thesis are as follows.(1)Building a railway multi-source heterogeneous data fusion framework.Currently,there is no unified standard for the overall design of railway multi-source heterogeneous data fusion and intelligent maintenance decision-making framework.This paper proposes a new railway multi-source heterogeneous data fusion framework design scheme by studying and analyzing the characteristics of each equipment inspection data in railway intelligent operation and maintenance system,adding the underlying data fusion mechanism in this framework,completing the fusion between the underlying data and providing powerful data support for subsequent intelligent decision-making.(2)Establishment of fusion model.Through the deep learning research of data fusion algorithms,the most effective way to solve the insufficient fusion of the underlying data is to improve the data quality and ensure the formation of data sharing and reuse between multi-source heterogeneous data.To this end,this thesis proposes a D-S evidence theory fusion model and a rough set data fusion model.Among them,the D-S evidence theory fusion model is filtered by the D-S evidence theory to ensure the improvement of the quality of the underlying data,form an effective combination of rules,and reduce conflicts between rules.The rough set data fusion model optimizes the underlying data of each device by improving the rough set attribute reduction method of the particle swarm optimization algorithm.(3)Experimental verification and analysis.Through a certain section of railway signal monitoring data test,the two data fusion models are verified and analyzed by experiments.The experimental results prove that the D-S evidence theory fusion model and the rough set data fusion model have good applicability and scalability,maximize the use of the underlying data,and improve the real-time performance of the algorithm and the accuracy of fault diagnosis.Compared with the analysis of the results,under a certain scale of data set,the data fusion model under the D-S evidence theory adopts the idea of fault classification,which provides more effective decision indicators for subsequent intelligent decision-making,and the average delay time of the algorithm is lower than that of the rough set data fusion model.The rough set data fusion model increases the reduction of the underlying data attributes and effectively reduces the data redundancy,making the decision accuracy higher than the D-S evidence theory data fusion model.Both fusion models have their own advantages.