used standard strain gauges to sense the approaching train and
woke the system by detecting changes in stress on the bridge.
Sala et al. [107] used activation sensors type not specified) to
wake their system for monitoring steel truss bridges. In contrast,
Townsend and Arms [123] used a similar technique but
with only one set of track-mounted strain gauges, which varied
their sampling rate depending whether a train is present. They
operated in low sampling mode of approximately 6 Hz to detect
the increase in strain when a train approached. Once detected,
the sensors increased their sampling rate and generated a strain
waveform for data analysis. The authors stated that this reduced
the power from 30 mA for continuous mode to less than 1 mA
for event detection. Chen et al. [35] introduced a prototype
system that developed this further by detecting approaching
trains, waking the measuring sensors on the bridge trusses and
then downloading the measured data onto the passing train,
which acts as a data mule. Another data muling approach is the
BriMon system [32], where trains were fitted with a beacon. A
bridge-mounted high-gain antenna detected the train’s beacon
30+ seconds before the train approached and woke the batteryoperated
accelerometers on the bridge. The accelerometers
measured the vibrations of the passing train on the bridge and
uploaded their data to the passing train. This data muling only
requires local data transmission, which reduces the power usage
further.