Towards condition monitoring of railway points: Instrumentation, measurement and signal processing

This paper presents the instrumentation and signal processing required to measure the strain occurring on the frog of a railway turnout. The aim is to enable the automatic estimation of the cumulative load exhibited on the element; this is relevant with respect to perdurant changes, safety and maintenance planning. Furthermore, the measurement system enables the identification of irregularities at rolling stock. The frog is instrumented with four strain gauges, housed in robust stainless steel casings. The measured strain signal is partitioned into three portions: a) a section prior to the train reaching the frog; b) the time during which the train traverses the turnout and c) a post train period. Statistical analysis is applied to the signal prior to the train arriving to determine characteristic perturbation behavior. This also enables a numerically efficient method of identifying irregularities at rolling stock. The specific measured strain pattern permits the grouping of the measurements according to train type. Dynamic time warping with local polynomial resampling is applied to each measurement so that a consensus curve can be computed for the strain associated with each train type. Long-term changes in the consensus curve reveal perdurant changes on the frog; while single trains with a significant deviation from the consensus curve can also be identified. The pre- and post-train measurements are compared to determine if there is a durative change in residual-strain. All methods used are derived and pseudo-code is presented for the critical algorithms. Measurement and analysis results are presented for six types of train. An exemplary detailed analysis is presented for multiple traversing of a Talent train.