结合离线计算与迁移学习的机务大数据闭环整合算法

A Closed-loop Integration Algorithm for Aircraft Maintenance Big Data Combining Offline Computing and Transfer Learning

为了实现铁路机务信息化和自动化管理,保证机务的全面、高效以及协同管控,提出结合离线计算与迁移学习的机务大数据闭环整合算法。聚类机务大数据,形成机务大数据的闭环数据链;机务大数据管理部分基于迁移学习抽取和检索机务大数据序列,识别并获取执行机务管理时所需的机务数据。测试结果显示:算法的扩展性较好,加速度的结果均在0.9以上,复制系数结果均在2%以下,具备良好聚类效果;智能系数和数据访问载荷最高结果分别达到155.7和74.6,整合性能良好。

In order to realize the informatization and automatic management of railway locomotive and ensure the comprehensive,efficient and collaborative management and control of locomotive,a closed-loop integration algorithm of locomotive big data combined with off-line calculation and migration learning is proposed to cluster locomotive big data and form a closed-loop data chain of locomotive big data.The locomotive big data management part extracts and retrieves the locomotive big data sequence based on migration learning,and identifies and obtains the locomotive data required for locomotive management.The test results show that the algorithm has good scalability,the acceleration results are more than 0.9,and the replication coefficient results are less than 2%,which has a good clustering effect.The highest results of intelligence coefficient and data access load are 155.7 and 74.6,respectively,i.e.,a good integration performance.