Abstract: In high speed railway catenary system, hanging strings are impacted by vehicle pantograph and external environment day after day, they are prone to fracture faults. In this paper, we propose an $\mathrm{L}_{1/2}$-LR model of hanging string fault detection based on the dynamic simulation model of pantograph catenary system. Firstly, we generated the acceleration data of contact wire and messenger wire under the multaneous action of pulsating wind and pantograph. Secondly, EEMD (Ensemble Empirical Mode Decomposition) is used to calculate their features, such as energy entropy, singular entropy, average entropy. As a result, an $\mathrm{L}_{1/2}$-LR model is established by the features. The experimental results show that the accuracy of the detection model is 97.25% under the condition of fluctuating wind and pantograph. As $\mathrm{L}_{1/2}$-LR model has variable selection effect, it is found that detecting whether hanging string is broken with high accuracy and calculation efficiency only by collecting vertical acceleration and horizontal acceleration at the middle position of catenary carrier cables.

Key words: hanging string, EEMD, entropy of information, $\mathrm{L}_{1/2}$-LR regularization