بررسی رفتار ژئوشیمیایی عنصر مس به روش کا-میانگین و پیش‏بینی آن با شبکه عصبی مصنوعی در منطقه کیوی، استان اردبیل

اﯾﻦ ﭘﮋوﻫﺶ ﺑﺮ روی ﺑﺮﮔﻪ ﯾﮏﺻﺪﻫﺰارم ژﺋﻮﺷﯿﻤﯿﺎﯾﯽ ﻣﻨﻄﻘﻪ ﮐﯿﻮی اﻧﺠﺎم ﺷﺪ. ﻣﻨﻄﻘﻪ ﮐﯿﻮی در اﺳﺘﺎن اردﺑﯿﻞ ﻗﺮار دارد. اﯾﻦ ﻧﺎﺣﯿﻪ ﺷﺎﻣﻞ ﺳﻪ واﺣﺪ ﺳﻨﮕﯽ رﺳﻮﺑﯽ، آذرﯾﻦ و دﮔﺮﮔﻮﻧﯽ اﺳﺖ. ﻗﺪﯾﻤﯽﺗﺮﯾﻦ واﺣﺪ رﺳﻮﺑﯽ ﻣﻮﺟﻮد، ﺳﻨﮓﻫﺎی ﻗﺒﻞ از ﮐﺮﺗﺎﺳﻪ و ﺟﺪﯾﺪﺗﺮﯾﻦ آن، ﻣﺮﺑﻮط ﺑﻪ ﮐﻮاﺗﺮﻧﺮ و ﻋﻬﺪ ﺣﺎﺿﺮ اﺳﺖ. ﺑﺎ ﺗﻮﺟﻪ ﺑﻪ اﺳﺘﻌﺪاد ﮐﺎﻧﯽﺳﺎزی ﻓﻠﺰی، ﺑﻪ وﯾﮋه ﻋﻨﺼﺮ ﻣﺲ در اﯾﻦ ﻣﻨﻄﻘﻪ، ﺑﺮرﺳﯽ دﻗﯿﻖ آن ﺑﺎ اﻫﻤﯿﺖ اﺳﺖ. ﺑﺮ اﯾﻦ اﺳﺎس، ﯾﺎﻓﺘﻦ اﻃﻼﻋﺎﺗﯽ در ﻣﻮرد ارﺗﺒﺎط و رﻓﺘﺎر ﻋﻨﺎﺻﺮ ﻃﻼ، ﻧﻘﺮه و ﻣﻮﻟﯿﺒﺪن ﻧﺴﺒﺖ ﺑﻪ ﻋﻨﺼﺮ ﻣﺲ در اﯾﻦ ﻣﻨﻄﻘﻪ اﻫﻤﯿﺖ ﻣﯽﯾﺎﺑﺪ؛ ﻫﺪف از اﯾﻦ ﺑﺮرﺳﯽ، رﻓﺘﺎرﺳﻨﺠﯽ ﻫﺎﻟﻪﻫﺎی ژﺋﻮﺷﯿﻤﯿﺎﯾﯽ در ﻣﻨﻄﻘﻪ اﺳﺖ. در ﭘﮋوﻫﺶ ﺣﺎﺿﺮ ﺑﺎ ﻫﺪف رﻓﺘﺎرﺳﻨﺠﯽ ﻋﻨﺎﺻﺮ ﻧﺎم ﺑﺮده، از روش ﻣﺸﻬﻮر و ﻣﻔﯿﺪ ﮐﺎ ﻣﯿﺎﻧﮕﯿﻦ اﺳﺘﻔﺎده ﺷﺪ. اﯾﻦ روش از روشﻫﺎی ﺧﻮﺷﻪﺑﻨﺪی اﺳﺖ ﮐﻪ ﺑﺮ ﮐﻤﯿﻨﻪ ﮐﺮدن ﻣﺠﻤﻮع ﻓﻮاﺻﻞ اﻗﻠﯿﺪﺳﯽ ﻫﺮ ﯾﮏ از ﻧﻤﻮﻧﻪﻫﺎ از ﻣﺮﮐﺰ دﺳﺘﻪﻫﺎﯾﯽ ﮐﻪ ﺑﻪ آن ﺗﺨﺼﯿﺺ ﻣﯽﯾﺎﺑﺪ، اﺳﺘﻮار اﺳﺖ. در اﯾﻦ ﭘﮋوﻫﺶ از ﺗﺎﺑﻊ ﮐﯿﻔﯿﺖ ﺧﻮﺷﻪﺑﻨﺪی و ﻣﯿﺰان ﻣﻄﻠﻮﺑﯿﺖ ﻧﻤﻮﻧﻪ در ﺧﻮﺷﻪ ﻣﻮرد ﻧﻈﺮ ((s(i) ﺑﺮای ﺗﺸﺨﯿﺺ ﺗﻌﺪاد ﺧﻮﺷﻪ ﺑﻬﯿﻨﻪ اﺳﺘﻔﺎده ﺷﺪ، ﺳﭙﺲ ﺑﺎ در ﻧﻈﺮﮔﺮﻓﺘﻦ ﻣﺮاﮐﺰ ﺧﻮﺷﻪﻫﺎ و ﻧﺘﺎﯾﺞ ﺣﺎﺻﻞ، ﻣﻌﺎدﻻﺗﯽ ﺑﺮای ﭘﯿﺶﺑﯿﻨﯽ ﻣﻘﺪار ﻋﻨﺼﺮ ﻣﺲ ﺑﺮ ﺣﺴﺐ ﭼﻬﺎر ﭘﺎراﻣﺘﺮ ﻋﯿﺎر ﻃﻼ، ﻧﻘﺮه، ﻣﻮﻟﯿﺒﺪن، ﻃﻮل و ﻋﺮض ﻧﻘﺎط ﻧﻤﻮﻧﻪﺑﺮداری ﺑﺎ ﻧﮕﺎه وﯾﮋه ﺑﺮ روش ﯾﺎد ﺷﺪه اراﯾﻪ ﺷﺪ. اﯾﻦ ﻧﺘﺎﯾﺞ ﻧﺸﺎن داد ﮐﻪ در روش ﺧﻮﺷﻪﺑﻨﺪی ﮐﺎ- ﻣﯿﺎﻧﮕﯿﻦ ﺗﻌﺪاد ﺧﻮﺷﻪ ﺑﻬﯿﻨﻪ ﺑﺮای رﻓﺘﺎرﺳﻨﺠﯽ ﻣﺲ در ﻣﻘﺎﺑﻞ ﻫﺮ ﯾﮏ از ﻋﻨﺎﺻﺮ ﻃﻼ، ﻧﻘﺮه و ﻣﻮﻟﯿﺒﺪن ﺑﺎ ﺗﻮﺟﻪ ﺑﻪ ﻣﻮﻗﻌﯿﺖ ﻧﻤﻮﻧﻪﺑﺮداری آنﻫﺎ ﺗﻌﺪاد 3 ﺧﻮﺷﻪ اﺳﺖ ﮐﻪ ﻣﺸﺨﺼﺎت آنﻫﺎ اراﯾﻪ ﺷﺪ. ﭘﺲ از ﺑﺮرﺳﯽﻫﺎی رﻓﺘﺎری ﻋﻨﺎﺻﺮ، آزﻣﺎﯾﺶ ﺷﺒﮑﻪ ﻋﺼﺒﯽ ﻣﺼﻨﻮﻋﯽ ﺑﺮای ﺗﺨﻤﯿﻦ ﻣﯿﺰان ﻣﺲ ﺑﺎ اﺳﺘﻔﺎده از روشﻫﺎی رﮔﺮﺳﯿﻮن ﻋﻤﻮﻣﯽ و ﭘﺲ اﻧﺘﺸﺎر ﺧﻄﺎ اﻧﺠﺎم ﺷﺪ. ﻣﻘﺪار ﺻﺤﺖ (r) ﺗﺨﻤﯿﻦ در دادهﻫﺎی آزﻣﺎﯾﺸﯽ در ﺷﺒﮑﻪ ﻋﺼﺒﯽ ﻣﺼﻨﻮﻋﯽ رﮔﺮﺳﯿﻮن ﻋﻤﻮﻣﯽ و ﭘﺲ اﻧﺘﺸﺎر ﺧﻄﺎ ﺑﻪ ﺗﺮﺗﯿﺐ 0/77 و 0/74 ﮔﺰارش ﺷﺪ. در اﻧﺘﻬﺎ ﻣﺸﺨﺺ ﺷﺪ ﮐﻪ روش ﺷﺒﮑﻪ ﻋﺼﺒﯽ ﻣﺼﻨﻮﻋﯽ رﮔﺮﺳﯿﻮن ﻋﻤﻮﻣﯽ در ﺗﺨﻤﯿﻦ ﺑﻬﯿﻨﻪ ﻋﻨﺼﺮ ﻣﺲ در ﻣﻨﻄﻘﻪ ﻣﻮرد ﻣﻄﺎﻟﻌﻪ ارﺟﺤﯿﺖ دارد.

Geochemical Behavior Investigation Based on Kmeans and Artificial Neural Network Prediction for Copper, in Kivi region, Ardabil province, IRAN

Kivi region is located in Ardabil province of Iran. This research is on kivi geochemical sheet (on scale 1:100000) which is investigated by geological survey & mineral explorations of Iran (GSI) using stream sediment analyzes. This region consists of sedimentary, igneous and metamorphic rock units. The oldest existing sedimentary unit, the preCretaceous rocks and the newest, is related to Quaternary and the present. Due to the ability of metal mineralization, especially the copper element in this region, it is important to study it carefully. Accordingly, finding information about the relation and behavior of the elements of gold, silver and molybdenum to the copper element in this region is important. The purpose of this study is to determine the behavior of geochemical halos in region. In this study with the aim of geochemical behavior investigating the mentioned elements Kmeans method was used. This method is based on clustering methods that minimize the total Euclidean intervals of each sample from the center of the groups to which it is assigned. In this research, the clustering quality function ( p(k) ) and the desirability of sample in the desired cluster ( S (i) ) were usedto determine the optimum numberof clusters. Then, taking into account clusters centers and results, equations were provided to predict the amount of copper with a special look at the method. After elemental behavioral studies, an artificial neural network test using general regression and backward propagation of errors was conducted to estimate the amount of copper. The accuracy value (R) of the estimation in the experimental data in the artificial neural network of general regression and backward propagation of errors was 0.77 and 0.74, respectively. Finally, it was determined that the general regression artificial neural network method has an advantage inThe optimal estimation of copper element in the studyarea.