650

Variables

27

4.74%

Sources







Data

These data correspond to the power consumption of the needle motor of a rail switch. This power consumption evolves along the movement (10s approx) in function of the mechanical actions. Between two movements, these time series are similar but vary in function of operating conditions and in function of the health of the switching unit. In our study, we had approx. 20000 movements/time series and we knew for each curve the switching unit health status: Healthy or when Unhealthy, if there was too much friction, an obstacle hindering the movement, or screw loose.

Maneuvers

~20000

Types of malfunction

3

Goal

Detection and diagnosis of rail switch malfunctions


Functional Data Example

Raw Data


Clustering step

Sort the clusters according to cluster's


How's diagnosis work ?


Diagnosis results

Confusion Matrix



Variables


Données