摘要：AbstractReliability and dependability of the infrastructure is a must for any railway asset manager to guarantee both safety and capacity of the network. To avoid operational downtime and, even more, accidents timely maintenance of the railway infrastructure becomes a crucial aspect. Current maintenance policies are mostly reactive or periodic, which give surge to a high O&M cost. Reducing maintenance cost while enhancing asset reliability may be achieved through the adoption of predictive maintenance policies. This requires the availability of a condition monitoring system able to assess the infrastructure health state through diagnosis and prognosis of degradation processes occurring on the different railway components. Central to any condition monitoring system is the a-priori knowledge about the process to be supervised in the form of either mathematical models of different complexity or signal features characterizing the health state. This paper proposes a statistical model for the switch panel of railway turnouts that characterizes two key components: railpad and ballast. Exploiting vibration data collected during train passages their natural frequencies are estimated through an estimation scheme based on empirical mode decomposition and subspace identification. By analysing vertical acceleration data corresponding to 400 train passages the estimated resonance frequencies associated with the ballast and the railpad have been well characterized by normally distributed random variables. The proposed estimation architecture and the resulting low-complexity statistical model opens an opportunity for the monitoring of developing degradation processes in the railway’s turnout.
