بررسی مدل های ریاضی رسوبگذاری مخازن و مقایسه آنها با روش شبکه های عصبی
/محمد تقی اعلمی
تبریز: دانشگاه تبریز،دانشکده کشاورزی،گروه آب
۱۹۷ ص ، جدول، نمودار، عکس، لوح فشرده
چاپی
واژه نامه بصورت زیرنویس
کتابنامه ص.: ۱۹۳-۱۹۷
دکترا
مهندسی آب- سازه های آبی
۱۳۸۳/۰۶/۱۸
تبریز: دانشگاه تبریز،دانشکده کشاورزی،گروه آب
به منظور پیش بینی مقدار رسوبگذاری در مخازن سد ها و تغییرات بستر رودخانه ها یکی از ابزارهای مهم مدل های ریاضی می باشد.در این رساله ابتدا تغییرات پروفیل طولی بستر یک مدل هیدرولیکی متعاقب احداث یک سد ذخیره ای در عرض آن از دیدگاه تئوری مورد تجزیه و تحلیل قرار گرفته و معادلات حاکم بر پدیده با اعمال روش تفاضلات محدود حل گردیده است.شبیه سازی مخزن سد کارده با استفاده از روش شبکه های عصبی مصنوعی قسمت دیگری از این تحقیق می باشد.
In order to prediction of reservoir sedimentation and changes in the bed profile, the application of mathematical models are one of the main tools. The utilization of these models in solving most problems of river engineering in connection with developmental projects have been extended considerably.In the present thesis, the longitudinal profile of the changes of beds following the construction of a storage reservoir across a hydraulic model has been analyzed theoretically, and the governing equations of the phenomenon, have been solved using the finite differences method.Then the stream tube model for alluvial river simulation (Gstars-2) has been applied for reservoir sedimentation of storage dams and case study has been performed on Kardeh dam reservoir. The results obtained have been compared with the hydrography survey held in 1995 on the site of this dam reservoir and also compared with results obtained by the Hec-6 and River-Intake model. The comparison has shown that the Gstars-2 model has the efficiency and characteristics of a one dimensional mathematical models and it is more accurate because of considering semi two dimensional condition of the flow. Also in the application of this model for reservoirs the nonequidibrium sediment transport has been used because of the importance of spatial-delay and / or time delay.Simulation of Kardeh dam reservoir sedimentation using the method of artificial neural network is another part of the present research. for this purpose a multi layer perceptron (MLP) network has been designed using empirical input and output and for its training error backpropagation algorithm has been used.The results show that the relationship between empirical input and output can be obtained through a trained network and the necessary predictions can be made without using nonlinear relationships that we have to simplify them in practice.