Association Analysis-based Fault Prevention For The Components Of Metro Overhead Contact System

With the extensive deployment of Metro,its safe operation has become the crucial guarantee for the daily transportation of people and public property.Overhead Contact System(OCS),an indispensable part of Metro,directly influences the power supply of Metro.Therefore,the fault prevention of components in Metro OCS is the key to maintain the safety of Metro.Currently,the daily maintenance of Metro OCS remains the strategy of condition-maintenance,which means the preventive-maintenance is barely achieved.Undoubtedly,it is crucial to reveal the fault mechanism of the Metro OCS components and conduct preventive-maintenance for Metro OCS.Given that the Metro OCS itself contains massive fault data,it provides the valid analysis basis for revealing the fault association relationship in Metro OCS from the data-driven perspective.Through the association analysis between faults of components,the maintenance personnel can be provided with directional maintenance suggestions in the process of maintenance,so as to achieve fault prevention through preventive-maintenance before the fault occurs.Therefore,fault association analysis has become an effective means to prevent the faults of Metro OCS.Based on the fault data of Metro catenary collected on site,this paper reveals the fault mechanism of Metro OCS and realizes the fault prevention through fault correlation analysis.The main work is as follows:Firstly,a data-driven fault prevention model framework of Metro OCS is proposed;then,based on the principle of minimum description length,a multi-layer fault database of Metro OCS is constructed,and a fault transaction database division method based on maintenance period is proposed.Secondly,in order to avoid the need to set a minimum utility value for the algorithm,the Top-k structure is adopted to ensure that the algorithm can automatically generate the K high utility patterns with the highest utility value.In addition,a hierarchical integration strategy based on prefix support degree is proposed to solve the adaptability problem of efficient mining algorithm for the multi-layer structure of Metro OCS network fault database.On the basis of the aforementioned two aspects,an efficient Top-k algorithm,MTHU(Multi-level Top-k High Utility),which can realize the multi-level data mining functionality,is proposed to realize the association analysis of Metro OCS.Then,based on the above MTHU algorithm,a Top-k optimization model based on linear programming is adopted to eliminate the non-objective influence caused by the artificial setting of K value in the algorithm.Then,in order to reduce the subjective influence deviation caused by the artificial setting of weight in the efficient mining algorithm,a utility correction model based on the closed-loop update strategy is proposed in this paper.At last this paper presents an efficient mining algorithm DU-Topk(Dynamic Utility Top-k)based on dynamic utility self-updating strategy;Finally,this paper defines the fault prevention strategy of Metro OCS based on DU-Topk algorithm,defines the difference between narrow fault prevention and broad fault prevention,gives the overall technical framework of Metro OCS fault prevention,and gives the detailed fault prevention priority and maintenance advice based on the actual fault database of Metro OCS.Besides,the proposed MTHU and DU-topk are compared with existing algorithm to verify the effectiveness.