Application of Gene Expression Programming and Nonlinear Regression in Determining Breach Geometry and Peak Discharge Resulting from Embankment Failure Using Laboratory Data

Accurate prediction of dam breach parameters in embankment dams is an essential step in the risk management plan. Overtopping and piping are the leading causes of embankment failure in the world. The failure of this type of dam is typically proposed by hydrological and hydraulic computational models of the dam (Wahl, 1998). The relationships for assessing the breach and flow characteristics are generally obtained by artificial intelligence and regression analysis from case studies of historical dam failure. These models relate the input parameters such as the dam height (Hw) and the flow volume through the breach (Vw) to the observed breach parameters resulting from the actual failures. Several relationships have been proposed to calculate Qp as a function of Hw and Vw (De Lorenzo & Macchione, 2014; Hagen, 1982; Kirkpatrick, 1977; Singh & Snorrason, 1984; Hakimzadeh et al., 2014). Downstream sediment transport studies show that breach geometry directly affects the output hydrograph. Investigations on historical records for Qp determination show that Hw and Vw could provide more accurate results than El and Ew. Moreover, the combination of these parameters significantly increases computational accuracy (Thornton et al., 2011; Wang et al., 2018). Several laboratory and field studies have been performed to investigate the hydraulic properties of the breach and the output hydrograph in overtopping failure cases (Dhiman & Patra, 2017; Sadeghi et al., 2020; Vaskinn et al., 2004). Determination of the average breach width (Bave) is an essential factor in determining progressive erosion (Von Thun & Gillette, 1990; Froehlich, 1995) as well as the height of breach (Hb). The range of variation in Bave as a function of the dam height (Hd) is an important issue in the breach lateral evolution process (Johnson & Illes, 1976; Singh & Snorrason, 1984).