Uncertainty analysis in probabilistic design of detention rockfill dams using Monte-Carlo simulation model and probabilistic frequency analysis of stability factors

The detention rockfill dams are of promising importance in flood control projects, due to their minimal technical requirement, low cost, minimal environmental side effects, and self-automotive operation process. However, due to the complexity of Non-Darcian flow interactions with stability and uncertainties of dam, the reliable design is a challenging topic. This study aimed to examine the effects of uncertainties in probabilistic design of these dams. We proposed a reliable design framework for detention rockfill dams with a focus on the importance of stability analysis. The effects of design uncertainty sources on the stability of dam, safety factors of overturning, sliding and bearing, along with the hydraulic performance of the dam were examined. The results of the model revealed that the uncertainties in input parameters can effectively regenerate uncertainties in the hydraulic performance ranges from - 53.54 to + 110.11%. The safety factor against the sliding (SFS) has maximum dependencies with the uncertainties ranging - 32.63 to + 87.81%. The Monte-Carlo Simulation (MCS) and fitting probability distribution functions to the safety factor histograms, and uncertainty quantifications results in 88.3%in increasing the safety factors as a reliable methodology for stability design of detention rockfill dams. Thus, the study calls for reliable, certain, and safe design of flood protection rockfill ponds. The ecological evaluation and applying more advanced uncertainty assessment methods remains a future research direction of the current study. The developed framework can be used to acquire future detention rockfill dam design/modeling requirements for reliability-based design optimization as a simulation-optimization model coupled whit MCS.

A novel equation for longitudinal dispersion coefficient prediction based on the hybrid of SSMD and whale optimization algorithm

The purpose of the present paper is improving the accuracy of existing formulas for the longitudinal dispersion coefficient (LDC) prediction based on a novel and simple meta-heuristic optimization method called Whale Optimization Algorithm (WOA). Although several existing formulas calculate LDC in the rivers based on the hydraulic and hydrodynamic specifications, most of them have significant errors in confronting extensive field data. In this study, comprehensive field data, including the geometrical and hydraulic properties of different rivers in the world, were adopted to build a reliable model. Statistical error measures were used to evaluate and compare the results with other studies. Furthermore, the Subset Selection of Maximum Dissimilarity (SSMD) method was utilized for a reputable selection of data for training and testing the WOA model. Subset selection is a critical factor in artificial intelligence (AI) computations. Finally, an integrated model based on the SSMD method and WOA technique has been proposed to develop the high accuracy formulas for the prediction of LDC. According to the results, the developed formulas are competitive or superior to the previous formulas for LDC estimation. Results also indicated that the WOA algorithm could be applied to improve the performance of the predictive equations in other fields of studies by finding the optimum values of coefficients.

Improving one-dimensional pollution dispersion modeling in rivers using ANFIS and ANN-based GA optimized models

Simulation and prediction of the pollution transport is one of the major problems in environmental and rivers engineering studies. The numerical tools have been used in simulation of the concentration profile transmission for description of river water quality. The one-dimensional advection-dispersion equation (ADE) is used in applied water quality modeling and requires the accurate estimation of longitudinal dispersion coefficient (D_x). This paper develops a hybrid numerical-intelligence model for dispersion modeling in open-channel flows. The main contribution of this paper is to improve the results of 1D numerical simulation of pollutant transport in steady flows by estimation of dispersion coefficient (D_x) based on artificial intelligence models and subset selection of maximum dissimilarity (SSMD). The developed hybrid model uses an intelligence module based on optimized adaptive neuro fuzzy inference system (ANFIS) and artificial neural networks (ANNs) for longitudinal dispersion estimation, in which their structures are optimized by genetic algorithm (GA). Intelligence estimates of D_x by ANN, ANFIS, ANFIS-GA, ANN-GA, multiple linear regression (MLR), and empirical equation are compared with observed values of D_x available in 505 river section, and the ANFIS-GA, as the most accurate, is incorporated and integrated with developed 1D-ADE numerical module. The numerical solution of 1D-ADE is done using physically influenced scheme (PIS) for face flux estimation in finite volume method. The performance of hybrid models PIS-ANFIS-GA, PIS-ANFIS, and PIS-empirical is compared using the R², RMSE, MAE, and NSE values in comparison with analytical solution and measured concentration hydrographs. The results revealed that the hybrid numerical-intelligence model is more accurate than the other classical methods for sediment/pollutant dispersion prediction in open-channel flows. The developed hybrid numerical-intelligence model can accurately simulate the dispersion processes in rivers and is a novel step in applicability of ANFIS-GA and ANN-GA models. Graphical abstract ᅟ.