نوع مقاله : مقاله پژوهشی
نویسندگان
1 مربی، گروه مهندسی عمران، دانشکده مهندسی ساختمان و معماری، دانشگاه فنی و حرفه ای، تهران، ایران
2 دانشگاه فنی و حرفه ای- دانشجوی دکتری عمران سازه های هیدرولیکی دانشگاه تبریز
3 گروه عمران، دانشکده فنی و مهندسی، واحد اهر، دانشگاه آزاد اسلامی، اهر - ایران
چکیده
کلیدواژه‌ها
عنوان مقاله [English]
نویسندگان [English]
Local energy loss is one of the basic design parameters of underground aqueducts or culverts. So far, numerous studies have been performed in this regard, and several formulas have been developed to predict the local energy loss coefficient. Previous studies have shown that the energy loss coefficient is a parameter dependent on the hydraulic variables of the flow and culvert geometry. However, due to the uncertainty in the phenomenon, the existing relationships have not led to comprehensive and accurate results. In this study, several models were developed by considering different input parameters by using experimental data to predict the local energy loss coefficient in slope-tapered culverts and the efficiency of the intelligent method of Gaussian Process Regression (GPR) as a kernel-based approach was evaluated. The results were compared with Gene Expression Programming (GEP) method. Also, for determining the effect of input variables, sensitivity analysis was performed. The results proved the high efficiency of the method used in the research in estimating the energy loss coefficient. The results of GPR modeling showed that the model with input parameters of Froude number (Fr), ratio of water depth to culvert diameter (Hw/D), and reducer length (Lr) is the superior model. In this case, the evaluation criteria values for the test data series were obtained as CC=0.85, DC=0.799 and RMSE=0.2. Also, by performing sensitivity analysis, it was observed that Froude number has the most impact on the local loss coefficient and could cause a significant increment in model efficiency. The use of the Froude number increased the accuracy of modeling by almost 40%.
کلیدواژه‌ها [English]
)
