عنوان	مدل جدید برای پیش بینی جمع‌شدگی و خزش بتن با بکارگیری الگوی یادگیری ژرف
نوع پژوهش	پایان نامه
کلیدواژه‌ها	.
چکیده	The aim of this study is to develop a nature-inspired new model for predicting shrinkage and creep of concrete by using deep learning model. The long-term deflection of steel-reinforced concrete members under sustained loads is due to the time-dependent effects of creep and shrinkage of concrete. Both creep and shrinkage are intrinsic material properties that are significantly influenced by the concrete mix design and ambient environmental conditions. However, unlike shrinkage, creep is load-dependent and is therefore influenced by factors pertaining to the loading conditions. There are four widely recognized mathematical models for computing creep and shrinkage strains in plain concrete, that account for different factors and are at least partly empirical. An overview of the factors that influence the creep and shrinkage of concrete was done, as well as the widely recognized mathematical models for computing creep and shrinkage strains. It presents a review of the most recent studies that evaluate the accuracy of the prediction models and provides an overview comparison of the accuracy of the models based on the different studies. The primary objectives of this research are to investigate factors that influence creep and shrinkage of concrete and evaluate existing mathematical models for computing creep and shrinkage strains using the most recent studies, and to evaluate existing empirical methods for computing incremental deflections due to creep and shrinkage, and to propose new models for creep and shrinkage for concrete depending on the deep learning of concrete behavior. For this reason, A corresponding numerical analysis method was implemented into the ABAQUS user-supplied subroutine; this method was verified by comparing the analytical results to the classical experimental results. A plentiful data was obtained from ABAQUS, by changing the parameters that affect creep and shrinkage in concrete. This data was used to optimize new models depending on regression principles that aim to minimize the differencesb etween the proposed model results and the data obtained from ABAQUS. For the models: ACI 209R-92 Model, B3 Model, CEB 90 Model and GL 2000 after a comparison with six statistical methods were. By comparing the calculated results for each model it was found that: The B3 Model performed best in estimating creep compliance, followed by GL 2000 Model and CEB 90 Model. The ACI 209R-92 Model performed worst.
پژوهشگران	محمود عدالتی (استاد راهنما)، شهد حازم خالد (دانشجو)، مجتبی کرمی (استاد مشاور)