Modeling flow in pipes to detect and locate leaks using a state observer approach
Résumé
This paper presents the aspects concerning the modeling of flow in pipes to detect and locate leaks,
using a state observer based approach. The pipeline model is based on the equations of water hammer,
discretized by finite differences. This discretization requires two boundary conditions: the pressure or
flow at the inlet and outlet of the pipe. To improve the model, the effect of friction and how to introduce
the boundary conditions were modified. Similarly shown, comparisons of simulations with real data, the
importance of considering the variation of the friction coefficient, and also that the correct choice of boundary
conditions, including the equations of the pump and flow restrictions, allow a better representation of
the behavior dynamic flow. Also shown that the model can be used to estimate the leakage flow and position
using state observers, and takes as an example the Extended Kalman filter type. It is further noted that the
model obtained, it can work with only two measured variables: output pressure and input stream, instead
of the four considered in previous work.