ﺗﻌﯿﯿﻦ ﻇﺮﻓﯿﺖ ﺑﺎرﺑﺮي ﻣﺤﻮري ﺳﺘﻮنﻫﺎي ﺑﺘﻨﯽ ﻣﺴﻠﺢ ﺷﺪه ﺑﺎ آرﻣﺎﺗﻮرﻫﺎي ﻃﻮﻟﯽ ﻣﺎرﭘﯿﭻ ﺑﺎ اﺳﺘﻔﺎده از روش اﺟﺰاي ﻣﺤﺪود و ﺷﺒﮑﻪ ﻋﺼﺒﯽ ﻣﺼﻨﻮﻋﯽ

Determining of Load Bearing Capacity of Rectangular Concrete Columns Reinforced with Longitudinal Spirals Using FEM and ANN Methods

ﺳﺘﻮنﻫﺎي ﺑﺘﻨﯽ ﺑﺎ آرﻣﺎﺗﻮرﻫﺎي ﻃﻮﻟﯽ ﻣﺎرﭘﯿﭻ اﯾﺪهي ﺟﺪﯾﺪي در ﻃﺮاﺣﯽ ﻫﺴﺘﻨﺪ ﮐﻪ در ﺳﺎلﻫﺎي اﺧﯿﺮ ﻣﻮرد ﺗﻮﺟﻪ ﻣﺤﻘﻘﯿﻦ ﻗﺮار ﮔﺮﻓﺘﻪاﻧﺪ. اﯾﻦ ﺳﺘﻮنﻫﺎ ﻗﺎﺑﻠﯿﺖ ﺑﺎرﺑﺮي و ﺷﮑﻞﭘﺬﯾﺮي زﯾﺎدي در ﻣﻘﺎﯾﺴﻪ ﺑﺎ ﺳﺘﻮنﻫﺎي ﻣﺴﻠﺢ ﺷﺪه ﺑﺎ آرﻣﺎﺗﻮرﻫﺎي ﻃﻮﻟﯽ ﻣﺮﺳﻮم دارﻧﺪ. ﮐﻠﯿﻪ ﻣﻄﺎﻟﻌﺎت در اﯾﻦ زﻣﯿﻨﻪ ﺑﻪ ﭼﻨﺪ ﮐﺎر ازﻣﺎﯾﺸﮕﺎﻫﯽ ﺑﺮ روي ﭘﯿﮑﺮﺑﻨﺪيﻫﺎي ﻣﺨﺘﻠﻒ آرﻣﺎﺗﻮر ﻣﺎرﭘﯿﭻ ﻣﺤﺪود ﺷﺪهاﻧﺪ. در اﯾﻦ ﻣﻘﺎﻟﻪ ﺑﺮاي ﻧﺨﺴﺘﯿﻦ ﺑﺎر، ﺳﺘﻮن-ﻫﺎي ﻣﺴﻠﺢ ﺷﺪه ﺑﻪ آرﻣﺎﺗﻮر ﻣﺎرﭘﯿﭻ ﺑﻮﺳﯿﻠﻪي روش اﺟﺰاي ﻣﺤﺪود ﺷﺒﯿﻪﺳﺎزي ﺷﺪه اﺳﺖ و ﺳﭙﺲ ﻣﺪل اﺟﺰاي ﻣﺤﺪود ﺑﺎ اﺳﺘﻔﺎده از ﻧﺘﺎﯾﺞ ازﻣﺎﯾﺸﮕﺎﻫﯽ، ﺻﺤﺖ ﺳﻨﺠﯽ ﺷﺪه اﺳﺖ. ﺑﻪ ﮐﻤﮏ ﻧﻤﻮﻧﻪ ﺻﺤﺖ ﺳﻨﺠﯽ ﺷﺪه ﻣﯽﺗﻮان اﻣﮑﺎن ﻣﻄﺎﻟﻌﻪ ﭘﺎراﻣﺘﺮي ﮔﺴﺘﺮده روي رﻓﺘﺎر ﺑﺎرﺑﺮي اﯾﻦ ﮔﻮﻧﻪ ﺳﺘﻮنﻫﺎ ﮐﻪ اﻧﺠﺎم آن در ﺷﺮاﯾﻂ ازﻣﺎﯾﺸﮕﺎﻫﯽ ﺑﺴﯿﺎر ﻫﺰﯾﻨﻪ ﺑﺮ و وﻗﺖ ﮔﯿﺮ ﻣﯽﺑﺎﺷﺪ را ﻓﺮاﻫﻢ ﺳﺎﺧﺖ. ﺗﺎﮐﻨﻮن ﺑﺮاي اﯾﻦ ﺗﯿﭗ از ﺳﺘﻮنﻫﺎ روش ﻋﺪدي ﺑﺮاي ﺗﺨﻤﯿﻦ ﻇﺮﻓﯿﺖ ﺑﺎرﺑﺮي اراﺋﻪ ﻧﺸﺪه اﺳﺖ. در اﯾﻦ ﻣﻄﺎﻟﻌﻪ، ﻫﻤﭽﻨﯿﻦ ﺑﺮاي ﺑﺎر ﻧﺨﺴﺖ ﺗﻼش ﺷﺪه اﺳﺖ ﺗﺎ ﻇﺮﻓﯿﺖ ﺑﺎرﺑﺮي ﺳﺘﻮنﻫﺎي ﯾﺎد ﺷﺪه ﺑﺎ اﺳﺘﻔﺎده از ﺷﺒﮑﻪ ﻋﺼﺒﯽ ﻣﺼﻨﻮﻋﯽ ﺑﺪﺳﺖ آﯾﺪ. ﺑﻪ دﻟﯿﻞ ﮐﻤﺒﻮد اﻣﮑﺎﻧﺎت ازﻣﺎﯾﺸﮕﺎﻫﯽ، از ﻧﻤﻮﻧﻪي ﺻﺤﺖ ﺳﻨﺠﯽ ﺷﺪه اﺟﺰاي ﻣﺤﺪود ﺑﺮاي ﺗﻌﯿﯿﻦ ﭘﺎراﻣﺘﺮﻫﺎي ﺗﺎﺛﯿﺮ ﮔﺬار ﺑﺮ ﻇﺮﻓﯿﺖ ﺑﺎرﺑﺮي و ﻫﻤﭽﻨﯿﻦ ﺗﻮﻟﯿﺪ دادهﻫﺎي ﻣﻮرد ﻧﯿﺎز ﺑﻪ ﻋﻨﻮان ورودي ﺑﺮاي ﺷﺒﮑﻪ ﻋﺼﺒﯽ اﺳﺘﻔﺎده ﺷﺪ. ﭘﺲ از ﻃﺮاﺣﯽ ﺷﺒﮑﻪ، ﻧﺘﺎﯾﺞ ﺣﺎﺻﻞ از ﺷﺒﮑﻪ ﻋﺼﺒﯽ ﺑﺎ ﭼﻨﺪ ﻧﻤﻮﻧﻪ ازﻣﺎﯾﺸﮕﺎﻫﯽ ﮐﻪ در ﻓﺮاﯾﻨﺪ اﻣﻮزش از اﻧﻬﺎ اﺳﺘﻔﺎده ﻧﺸﺪه ﺑﻮد، راﺳﺘﯽ ازﻣﺎﯾﯽ ﮔﺮدﯾﺪ. در ﭘﺎﯾﺎن ﻋﻼوه ﺑﺮ روش ﺷﺒﮑﻪ ﻋﺼﺒﯽ، از روش ﺗﺤﻠﯿﻠﯽ رﮔﺮﺳﯿﻮن ﭼﻨﺪﮔﺎﻧﻪ ﮐﻪ ﺑﺎ ﺑﺮازش از ﻣﯿﺎن ﻧﺘﺎﯾﺞ ﺣﺎﺻﻞ از ﻣﺪلﺳﺎزي ﻋﺪدي راﺑﻄﻪاي ﺑﯿﻦ ﻣﺘﻐﯿﺮﻫﺎي ورودي و ﻇﺮﻓﯿﺖ ﺑﺎرﺑﺮي ﻧﻬﺎﯾﯽ اراﺋﻪ ﻣﯽدﻫﺪ، اﺳﺘﻔﺎده ﺷﺪ. ﻧﺘﺎﯾﺞ ﺣﺎﺻﻞ از دو روش ﺑﺎ دﻗﺖ ﻗﺎﺑﻞ ﻗﺒﻮﻟﯽ ﺑﺮ اﯾﻦ ﮔﻮاﻫﯽ ﻣﯽدﻫﻨﺪ ﮐﻪ اﯾﻦ روشﻫﺎ ﻣﯽﺗﻮاﻧﻨﺪ ﻇﺮﻓﯿﺖ ﺑﺎرﺑﺮي اﯾﻦ ﺗﯿﭗ از ﺳﺘﻮنﻫﺎ را ﺑﺎ دﻗﺖ ﻗﺎﺑﻞ ﻗﺒﻮﻟﯽ ﺗﺨﻤﯿﻦ ﺑﺰﻧﻨﺪ ﮐﻪ ﻣﯽﺗﻮاﻧﺪ در آﯾﻨﺪه ﺑﻪ اراﺋﻪ رواﺑﻂ ﮐﺎرﺑﺮدي ﺟﻬﺖ ﻃﺮاﺣﯽ و اﺳﺘﻔﺎده در ﺻﻨﻌﺖ ﺳﺎﺧﺘﻤﺎن ﻣﻨﺠﺮ ﺷﻮد.

Concrete rectangular columns reinforced with longitudinal spirals are new types of RC columns which their behavior has been investigated by researchers in recent years. These researches are limited to some experimental studies which investigate the effect of different configurations and layouts of the spirals within the cross-section of these columns on the bearing capacity and ductility. In this study, for the first time, the behavior of these columns has been simulated using numerical approaches. Finite Element (FE) models of these columns were developed using ABAQUS/CAE/Explicit v.2016 and then verified against available valid experiments in literature. Subsequently, by performing several sensitivity analyses using verified EF model, the effective parameters on bearing capacity of this kind of columns were detected. By changing the value of these parameters in rational ranges, a comprehensive parametric analysis was done after that using FE models in order to produce necessary input data for training an Artificial Neural Network (ANN). This parametric study was performed because of the lake of sufficient available experimental data. The developed ANN was verified against some experimental data. Finally, in addition to ANN, a regression analysis was performed to obtain a polynomial function can predict the bearing capacity of these type of columns. Obtained results demonstrate that the combination of FE and ANN is an effective method to predict the load bearing capacity of RC columns with longitudinal spiral reinforcements and have a good agreement with the results of regression analysis.