FHI: A Fault Intensity-based Hierarchical Association Analysis Model for Mining Fault Database of Railway OCS

The overhead contact system (OCS) is the power source of electrified railway, it is very important to maintain the OCS safely, stably and efficiently. At present, reliability analysis is the main method to study OCS fault data. Most of fault trees include only part of components of OCS. However, there are few studies from system-wise perspective. This paper applies data mining to OCS for the first time and proposes a hierarchical association analysis model based on fault intensity (FHI) according to the characteristics of OCS fault data (sparsity and hierarchy). Firstly, we proposed method of multi-dimensional partition (MDP) to transform fault records to horizontal data format, which can reduce the sparsity of data. Secondly, we defined the fault intensity to replace the threshold of support and introduced adjustment factor to realize it, which can eliminate the influence of unequal partition and make the mining result more reasonable. Thirdly, we proposed a new pruning strategy to mine cross-level association rules. By mining the historical OCS fault data collected from high speed railway lines of entire province in north-west China, we verify the validity of this analysis model and reveal the inner connections between faults. Based on the analysis results, the detailed suggestions are given to guide the operation and maintenance of ocs.