ﻣﺪﻟﺴﺎزي ﺟﺮﯾﺎن اﺳﻼﻣﭗ ﺑﺘﻦ ﺧﻮدﺗﺮاﮐﻢ ﺣﺎوي ﻣﺘﺎﮐﺎﺋﻮﻟﻦ ﺑﺎ اﺳﺘﻔﺎده از روشﻫﺎي ﻣﺤﺎﺳﺒﺎت ﻧﺮم

Modeling the Slump Flow of Self-Compacting Concrete Incorporating Metakaolin Using Soft Computing Techniques

ﻣﯿﺰان ﺣﺴﺎﺳﯿﺖ ﺟﺮﯾﺎن اﺳﻼﻣﭗ ﺑﺘﻦ ﺧﻮد ﺗﺮاﮐﻢ ﺣﺎوي ﻣﺘﺎﮐﺎﺋﻮﻟﻦ ﺑﻪ ﻣﻮاد ﺗﺸﮑﯿﻞ دﻫﻨﺪه آن و ﻧﺴﺒﺖﻫﺎي ﻃﺮح اﺧﺘﻼط، ﻟﺰوم اﺳﺘﻔﺎده از ﻣﺪلﻫﺎﯾﯽ ﺑﺎ دﻗﺖ ﺑﺎﻻ ﺑﺮاي ﺗﻀﻤﯿﻦ وﯾﮋﮔﯽﻫﺎي ﺗﺨﻤﯿﻦ درﺳﺖ و ﺗﻌﻤﯿﻢ آن را ﺿﺮورت ﺑﺨﺸﯿﺪه اﺳﺖ. ازاﯾﻦ رو، در اﯾﻦ ﻣﻘﺎﻟﻪ ﺑﻪ ﺑﺮرﺳﯽ ﻗﺎﺑﻠﯿﺖ روشﻫﺎي اﺳﭙﻼﯾﻦ رﮔﺮﺳﯿﻮﻧﯽ ﭼﻨﺪ ﻣﺘﻐﯿﺮه ﺗﻄﺒﯿﻘﯽ و ﻣﺪل درﺧﺘﯽ در ﻣﺪﻟﺴﺎزي و ﭘﯿﺶﺑﯿﻨﯽ ﺟﺮﯾﺎن اﺳﻼﻣﭗ ﺑﺘﻦ ﺧﻮدﺗﺮاﮐﻢ ﻣﯽ-ﭘﺮدازد. ﺑﺮاي اﯾﻦ ﻣﻨﻈﻮر، ﺗﻌﺪاد 117 داده ﻣﺨﺘﻠﻒ از ﻣﻘﺎﻻت ﻣﻌﺘﺒﺮ ﺑﻪ ﭼﺎپ رﺳﯿﺪه ﺟﻤﻊ آوري و در اﯾﻦ ﭘﮋوﻫﺶ ﻣﻮرد اﺳﺘﻔﺎده ﻗﺮار ﮔﺮﻓﺖ. اﻃﻼﻋﺎت ﻣﻮرد اﺳﺘﻔﺎده در ﻣﺪل ﻫﺎي ﭘﯿﺶﻧﻬﺎدي در ﻗﺎﻟﺐ ﻫﺸﺖ ورودي ﺷﺎﻣﻞ ﺳﯿﻤﺎن، درﺷﺖ داﻧﻪ، رﯾﺰداﻧﻪ، آب، ﻣﺘﺎﮐﺎﺋﻮﻟﻦ، ﻓﻮق روان ﮐﻨﻨﺪه، ﭼﺴﺒﺎﻧﻨﺪه و اﻧﺪازه ﺑﺰرﮔﺘﺮﯾﻦ داﻧﻪ ﺷﺮﮐﺖ ﮐﻨﻨﺪه در ﻃﺮح اﺧﺘﻼط Dmax)( و ﯾﮏ ﺧﺮوﺟﯽ ﺟﺮﯾﺎن اﺳﻼﻣﭗ دﺳﺘﻪ ﺑﻨﺪي ﺷﺪ. ﺑﺮاي ارزﯾﺎﺑﯽ دﻗﺖ ﻣﺪلﻫﺎي ﭘﯿﺸﻨﻬﺎدي، ﻣﻄﺎﻟﻌﻪ اي ﻣﻘﺎﯾﺴﻪ اي در ﻗﺎﻟﺐ ﺷﺎﺧﺺ ﻫﺎي آﻣﺎريMAE ، R ، RMSE اﻧﺠﺎم ﺷﺪ. ﻧﺘﺎﯾﺞ ﺑﺪﺳﺖ آﻣﺪه از ﻣﺠﻤﻮﻋﻪ داده ﻫﺎ در ﻣﺮاﺣﻞ آﻣﻮزش و آزﻣﻮن ﻣﺪل ﻫﺎي ﭘﯿﺶﻧﻬﺎدي و ﻣﻘﺎﯾﺴﻪ آﻧﻬﺎ ﺑﺎ ﻧﺘﺎﯾﺞ آزﻣﺎﯾﺸﮕﺎﻫﯽ ﭘﺘﺎﻧﺴﯿﻞ ﺑﺎﻻي دو روش ﻣﺎرس و ﻣﺪل درﺧﺖ را در ﭘﯿﺶﺑﯿﻨﯽ ﺧﻮاص ﺑﺘﻦ ﺑﺎ دﻗﺖ ﻧﺸﺎن ﻣﯽ دﻫﺪ. آﻧﺎﻟﯿﺰ ﺣﺴﺎﺳﯿﺖ اﻧﺠﺎم ﺷﺪه ﺑﺮاي ﻣﺸﺨﺺ ﮐﺮدن ﺗﺎﺛﯿﺮﮔﺰار ﺗﺮﯾﻦ ﭘﺎراﻣﺘﺮ در ﺟﺮﯾﺎن اﺳﻼﻣﭗ ﺑﺘﻦ ﺧﻮدﺗﺮاﮐﻢ ﻧﺸﺎن داد رﯾﺰداﻧﻪ و ﻣﺘﺎﮐﺎﺋﻮﻟﻦ ﻣﻮﺛﺮﺗﺮﯾﻦ ﻣﺘﻐﯿﺮﻫﺎ درﻣﺪﻟﺴﺎزي و ﭘﯿﺶﺑﯿﻨﯽ ﺟﺮﯾﺎن اﺳﻼﻣﭗ اﯾﻦ ﻧﻮع ﺑﺘﻦ ﺧﻮدﺗﺮاﮐﻢ در ﻗﺎﻟﺐ روش ﭘﯿﺸﻨﻬﺎدي ﻣﺪل درﺧﺖ در اﯾﻦ ﭘﮋوﻫﺶ ﺑﻮده اﺳﺖ.

The sensitivity of slump flow of self-compacting concrete containing metakaolin to its ingredient materials and mixture proportions, necessitate the use of high accuracy models to guarantee both estimation and generalization features. Therefore this paper investigates the potential of multivariate adaptive regression splines (MARS) and model tree (MT) approaches in prediction of slump flow of self-compacting concrete. Total of 117 data collected from the several published literature were used in present work. The data used in proposed models are arranged in a format of eight input parameters including cement, coarse aggregate, fine aggregate, water, metakaolin, super plasticizer, binder and maximum size of aggregates (Dmax) and one output as slump flow. To evaluate the precision of the models, a comparative study has been performed in terms of RMSE, R and MAE indices. The results of training and testing datasets of the techniques are compared with experimental results and their comparisons demonstrate that the MARS and MT models have potential to predict concrete properties with great precision. Performed sensitivity analysis to assign effective parameters on slump flow was indicating fine aggregate and metakaolin is most effective variable for modeling and prediction in this type of the self-compacting concrete using MT technique in this study.