Analyzing localization features of a weighted least-squares technique in a point source reconstruction
Résumé
Identification of unknown contaminant releases is crucial in the events of gas leakage, industrial accidents, safety and security concerns. For a point based release, the nature of a source is determined by fixed number of parameters, mainly its location and source strength. The identification of these parameters is addressed by using limited concentration measurements sampled by a network of receptors. This is formulated as over-determined linear inverse problem. To address this, an optimally weighted least-squares inversion technique is presented here. The inversion technique provides a minimum weighted norm solution and does not require initial guesses for the release parameters. This study proposes the optimality conditions under which a point source is identifiable. The inversion technique and optimality criteria are evaluated using field measurements of a single release trial# 7 conducted during Fusion Field Trials in September 2007. The point source is reconstructed within 3 m from the true release. The source strength is retrieved within a factor of 1.25 to true release mass. The optimality criteria show that the accuracy of reconstruction depends on the nature of model resolution, predictability of measurements and visibility of the true source from the receptors.