work’s initial assumption is that the sensor will flow down the
pipe with the highest velocity, but future models should account
for the uncertainty in the sensor’s route in junction
nodes.
The algorithm requires hydraulic data (i.e., time series of
link velocities) and network structure (i.e., connectivity and
pipe lengths). It utilizes the EPANET toolkit (USEPA, 2002) for
extended period hydraulic simulation to compute the time
series of link velocities u(t). The network is discretized in space
x1,..., xN to get the location of the mobile sensor and contaminant
concentrations along the pipelines. Time steps are
dictated by hydraulic or physical changes, for example, if
there is a change in velocity or a sensor has reached the end of
a pipe and is moving toward another one. Sensor location is
updated at each time step considering the given flow direction
and rate, and pipe length.