The paper divides railway condition monitoring into fixed
monitoring for immobile infrastructures such as bridges, tunnels,
tracks and associated equipment, and movable monitoring
for vehicles and their mechanics. Fixed monitoring uses sensors
to monitor vibrations, stresses and sound waves passed
through structures (acoustics) caused by passing trains (shortterm
monitoring) and also changes in stresses, pressures and
sound waves passed through structures over the longer term
(long-term monitoring).
One of the key issues for fixed monitoring is network topology.
The topology is constrained by the requirements of the
monitoring and by the physical environment. Sensor nodes can
be arranged in either an ad hoc or a preplanned configuration
[134]. Determining the optimum node placement is a complex
task and often requires a tradeoff. The network configuration
can be optimized against a number of different constraints. A
network may minimize relay nodes, may need to ensure a minimum
level of service (include a certain level of redundancy),
minimize energy usage to preserve battery life, or may need
to ensure accessibility of the nodes [116]. The communication
mechanisms, for example, Bluetooth, Wi-Fi, GSM, or satellite
also need to be evaluated to ensure coverage and reliability and
the routing protocol to ensure data is successfully transmitted
from sensor node to base station to control center. Another
issue is powering the sensors as fixed monitoring often requires
placing sensor nodes in inaccessible locations, for example, in
tunnels, on bridge trusses, or in rail track beds. Many sensors
use batteries but replacing batteries in inaccessible locations
may not be possible. Authors have considered ambient energy
harvesting such as converting vibrations caused by passing
trains into energy or using solar power. These energy harvesters
may be accompanied with energy storage such as capacitors.
Authors have also reduced sensor energy usage using event
detection where a single cheap sensor detects approaching
trains and wakes a larger sensor network to commence taking
measurements of a bridge structure.