Varying-parameter modeling and Extended Kalman Filtering for reliable leak diagnosis under temperature variations

International audienceIn this work the problem of robustness to temperaturevariations in pipeline leak detection and isolation isconsidered, extending preliminary attempts in this direction.The approach is restricted to pipelines short enough so thatthe spatial variations of the temperature can be neglected, andonly time variations are addressed. The methodology is herebased on a model taking into account the variations in allthe parameters subject to temperature changes, like frictionfactor, water density, etc. A temperature measurement is thenused to compensate for such variations in an Extended KalmanFilter used as a state observer, and real-time experimentaltests are provided showing how an accurate leak identificationis achieved when this compensation is done, while it failsotherwise