Author(s): Jose Luis Herrera Alanis; Maritza Arganis; Margarita Preciado; Ramon Dominguez Mora
Linked Author(s): Alejandro Mendoza
Keywords: Gumbel gravitatorial optimization annual maximum rainfall AX program
Abstract: Precipitation runoff frequency analysis data is widely utilized for design and evaluation of hydraulic structures. Traditional parameter estimation techniques include moments method application, as well as successive approximations and least squares regression methods. Conversely, likelihood function maximization employs various optimization techniques, with genetic algorithms and gradient searches being particularly notable. In this paper, an innovative SGA gravitational optimization method was employed to derive Gumbel distribution function parameters through likelihood function maximization. A Python program, developed with artificial intelligence support, facilitated this process. Comparisons were conducted between measured and calculated values for analyzed variable, as well as with results obtained using Moments and Maximum Likelihood methods implemented in the AX program, which is widely used in Mexico. Results from gravitational algorithm revealed a significant dispersion between measured and calculated data when compared to an identity function. Furthermore, relative percentage errors between parameters obtained from the AX program and those reported by EMS were 343.5% for location parameter and 18.9% for scale parameter. Standard adjustment error was 34.29 with the SGA method, in contrast to 2.87 presented by the AX program using moments method.
Year: 2025