Keywords
Wildfire; Water; Contamination; Source; Detection; Inverse Engineering; Theory of Hypernumbers
Abstract
The paper introduces a machine learning method of detecting multiple sources of water contamination caused by wildfire. The method includes changing the water flow regime, monitoring the time series of the contaminant concentration caused by regime changes, and associating the signature of the contaminant changes over time with sources locations. The contaminant signature from multiple sources starting at the moment of changing water velocity are defined by extending the approach for one contamination source. The intensity, location of each source, and diffusion coefficient are defined to satisfy the minimum square between monitoring and theoretical concentrations. The equations derived from the criteria of the best fit between experimental and modeling data are solved using the theory of hypernumbers. The initial values for hypernumber solutions are computed using the transient process of contaminant transport curve analysis. The defined in this paper algorithm can by used for detecting location of the arbitrary impurity in water network system.