Wheel Condition Monitor (WCM)
WHEEL IMPACTS
Wheel impacts are presented as kilo Newton (kN). This is a force measurement
and it is important to understand how these are calculated and presented.
The very simplified outline of how the system works is that the system measures
acceleration at the rail and translates acceleration in certain frequency
domains into a power reading.
Current WCM models can relate this power rating directly to traditional shear
strain measures but the most common method of translating this to kN is to
scale the result according to the structure type, rail size and then wheel
inspections. Once set, the readings are so stable that you can monitor very
slight shifts in readings over time on a given wheel.
Traditional systems measure kN directly and although this may seem to be an
advantage, the problem is that impact readings vary according to the loading of
the vehicle (the sprung mass). The method used in the WCM is independent of
sprung mass.
Using acceleration data instead of strain gauge data for wheel condition
analysis differs from strain gauge data:
-
Rich frequency domain yields more information where discrete frequency domains
yield information on track/structure, unsprung mass, and sprung mass.
-
For the purposes of wheel condition analysis a particular frequency domain is
used such that a wheel in good condition passing over the array reads as zero.
Impacts that may occur are measured in terms of peak, power, and number of
transitions.
-
Continual linear coverage enables multiple defect identification and defect
characterization instead of fragmented sampling.
-
Rate of change of loading yields rail motion, velocity and displacement data as
well as a count of zero-crossing events. (ie the number of stress reversals)
-
Axle loading (mass) is separate data.
-
Strain data is available in parallel with the acceleration data but is not used
in wheel condition analysis due to patent restrictions, the fact that there is
no additional useful information provided and because the data is highly load
dependent which precludes the trending of defect progression over time.
Strain gage based systems have their uses but they are limited in their ability
to isolate wheel defects. The graphs in the figures below plots the output of
twelve strain gage bridges and ten accelerometers for a 48 axle train
travelling at 140 mph. This data is as recorded and before any conditioning or
calibration and is shown like this to illustrate the raw material used in
subsequent processing.
The top graph is the strain gage data. The axle mass and wheel defects are
combined. It is not clear whether the vehicle of the first four axles is heavy
or has defects. The vehicle of axles 9 to 12 are ambiguous. Clearly there is
are impacting wheels on axles 31 and 32.
The graph below is the acceleration data and shows only the dynamic component
of the wheel passage. It is clear that the first four axles are clean, that the
vehicles of axles 9 to 12 and 31 and 32 are impacting the rail. The improved
signal to noise in this view also shows small defects on axles 19 and 21. All
other wheels record near zero results.
The figure below is a display from the database illustrating multiple passes
over a site of one vehicle. Axle 4 develops a defect on each wheel and the
other wheels on the vehicle are nominal. Alternating passes in this example
alternate between 23 tonne and 136 tonne. Clearly the sprung mass is not a
factor and the wheel defect progression over time is clear and predictable.
The plot shows the isolation and resolution of the system. There is no bleed
through to adjacent axles and for the defective wheelset the defects are
isolated to one section of each wheel. The small black markers at the top of
the graph identify when the vehicle’s wheels change direction of rotation.
The good axles are reported at about 170 kN in this example. This is because
the system can be set to normalize to the fully loaded condition (34 tonne axle
loads). " See the Normalization
section.
