‫ﻣﻄﺎﻟﻌﻪ )ﭘﻴﺶﺑﻴﻨﯽ( ﻧﺮﺥ ﺷﮑﺴﺖ ﻟﻮﻟﻪﻫﺎﯼ ﺍﺻﻠﯽ ﺁﺑﺮﺳﺎﻧﯽ ﺷﻬﺮﯼ‬

‫ﺑﺎ ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ ﺭﻭﺷﻬﺎﯼ ﻫﻮﺷﻤﻨﺪ ﻭ ﺭﮔﺮﺳﻴﻮﻧﻲ‬

‫‪۳‬‬ ‫‪٢‬‬ ‫‪۱‬‬

‫ﺟﺎﺑﺮ ﺳﻠﻄﺎﻧﻲ‬ ‫ﺁﺭﺵ ﺁﻗﺎﻳﻲ‬ ‫ﻣﺴﻌﻮﺩ ﺗﺎﺑﺶ‬

‫ﭘﺬﻳﺮﺵ ‪(٨٩/٢/١٧‬‬ ‫)ﺩﺭﻳﺎﻓﺖ ‪٨٨/١/٢٨‬‬

‫ﭼﻜﻴﺪﻩ‬
‫ﺑﺎ ﺗﻮﺟﻪ ﺑﻪ ﺳﻴﺎﺳﺖﻫﺎﻱ ﺍﻓﺰﺍﻳﺶ ﻛﺎﺭﺍﻳﻲ ﻭ ﻛﺎﻫﺶ ﻫﺪﺭﺭﻓﺖ‪ ،‬ﻳﻜﻲ ﺍﺯ ﺍﻭﻟﻮﻳﺖﻫﺎﻱ ﺭﺍﻫﺒﺮﺩﻱ ﺗﻮﺳﻌﻪ ﭘﺎﻳﺪﺍﺭ ﻣﻨﺎﺑﻊ ﺁﺏ‪ ،‬ﺑﻬﺮﻩﺑﺮﺩﺍﺭﻱ ﺑﻬﻴﻨﻪ ﺍﺯ ﺷﺒﻜﻪﻫﺎﻱ ﺗﻮﺯﻳﻊ ﺁﺏ‬
‫ﺷﻬﺮﻱ ﺍﺳﺖ‪ .‬ﻟﺬﺍ ﺭﺍﻫﺒﺮﺍﻥ ﺳﻴﺴﺘﻢ ﺑﺎﻳﺪ ﻫﻤﻮﺍﺭﻩ ﺷﻨﺎﺧﺖ ﺩﻗﻴﻖ ﻭ ﺻﺤﻴﺤﻲ ﺍﺯ ﻗﺎﺑﻠﻴﺖ ﮐﺎﺭﺑﺮﻱ ﺷﺒﻜﻪ ﻭ ﺍﻧﺰﻭﺍﻝ ﻭ ﺧﺮﺍﺑﻲ ﺳﺎﺯﻩﺍﻱ )ﺷﮑﺴﺖ( ﺩﺭ ﻟﻮﻟﻪﻫﺎﻱ ﺍﺻﻠﻲ‬
‫ﺁﺑﺮﺳﺎﻧﻲ ﺩﺍﺷﺘﻪ ﺑﺎﺷﻨﺪ‪ .‬ﻳﻜﻲ ﺍﺯ ﺭﺍﻫﻜﺎﺭﻫﺎ ﻭ ﻛﻠﻴﺪﻫﺎﻱ ﻣﻬﻢ ﻣﺪﻳﺮﻳﺖ ﺑﻬﻴﻨﻪ ﺑﻬﺮﻩ ﺑﺮﺩﺍﺭﻱ‪ ،‬ﺗﺪﻭﻳﻦ ﺍﺳﺘﺮﺍﺗﮋﻱﻫﺎﻱ ﻧﻮﺳﺎﺯﻱ ﻭ ﺑﺎﺯﺳﺎﺯﻱ ﺩﺭ ﺷﺒﻜﻪﻫﺎﻱ ﺗﻮﺯﻳﻊ ﺁﺏ‬
‫ﺷﻬﺮﻱ‪ ،‬ﭘﻴﺶﺑﻴﻨﻲ ﻧﺮﺥ ﺷﻜﺴﺖ ﻟﻮﻟﻪﻫﺎ ﻭ ﺍﺭﺯﻳﺎﺑﻲ ﻗﺎﺑﻠﻴﺖ ﮐﺎﺭﺑﺮﻱ ﺁﻧﻬﺎ ﺍﺳﺖ‪ .‬ﺭﺍﻩ ﺣﻠﻬﺎﻱ ﮔﻮﻧﺎﮔﻮﻧﻲ ﺑﺮﺍﻱ ﻣﻄﺎﻟﻌﻪ ﭘﻴﺶﺑﻴﻨﻲ ﺷﻜﺴﺖ ﺩﺭ ﻟﻮﻟﻪﻫﺎ ﺗﻮﺳﻂ ﻣﺤﻘﻘﺎﻥ‬
‫ﻣﺨﺘﻠﻒ ﺧﺼﻮﺻﺎﹰ ﺩﺭ ﺳﺎﻟﻬﺎﻱ ﺍﺧﻴﺮ ﻣﻄﺮﺡ ﺷﺪﻩ ﺍﺳﺖ ﻛﻪ ﻫﺮﻳﻚ ﻧﻴﺎﺯﻣﻨﺪ ﺩﺍﺩﻩﻫﺎﻱ ﻣﺨﺼﻮﺹ ﺧﻮﺩ ﻣﻲﺑﺎﺷﻨﺪ‪ .‬ﻣﺪﻝﻫﺎﻱ ﻗﻄﻌ ﻲ ﺗﻮﺍﻧﻲ ﺑﺮ ﺣﺴﺐ ﺳﻦ‪ ،‬ﻣﺪﻝﻫـﺎﻱ‬
‫ﻗﻄﻌ ﻲ ﺧﻄﻲ ﺑﺮ ﺣﺴﺐ ﺳﻦ‪ ،‬ﻣﺪﻝﻫﺎﻱ ﻗﻄﻌﻲ ﭼﻨﺪﻣﺘﻐﻴﺮﻩ ﻭ ﻣﺪﻝﻫﺎﻱ ﮔﺮﻭﻫﻲ ﺗﮏ ﻣﺘﻐﻴﺮﻩ ﺍﺣﺘﻤﺎﻟ ﻲ‪ ،‬ﺍﺯ ﺟﻤﻠﻪ ﺭﺍﻩ ﺣﻠﻬﺎﻳﻲ ﻫﺴﺘﻨﺪ ﮐﻪ ﺧﺮﺍﺑﻲ ﻭ ﺯﻭﺍﻝ ﺩﺭ ﺳـﺎﺧﺘﺎﺭ‬
‫ﻟﻮﻟﻪﻫﺎ ﺭﺍ ﺩﺍﺋﻤﻲ ﻭ ﺑﻪﺻﻮﺭﺕ ﻳﻚ ﺭﻭﻧﺪ ﻳﻜﻨﻮﺍﺧﺖ ﺑﺎ ﭘﺎﺭﺍﻣﺘﺮﻫﺎﻳﻲ ﻣﺴﺘﻘﻞ ﺍﺯ ﺯﻣﺎﻥ )ﻛﻪ ﺷﮑﺴﺖ ﻟﻮﻟﻪﻫﺎ ﺭﺍ ﺑﺎ ﺳﻦ‪ ،‬ﺟﻨﺲ ﻭ ﻗﻄﺮ ﺁﻧﻬﺎ ﻣﺮﺗﺒﻂ ﻣﻲﺳـﺎﺯﺩ( ﻣﻄـﺮﺡ‬
‫ﻣ ﻲﮐﻨﻨﺪ‪ .‬ﺩﺭ ﺍﻳﻦ ﻣﻘﺎﻟﻪ ﻋﻼﻭﻩ ﺑﺮ ﺑﻬﺮﻩﮔﻴﺮﻱ ﺍﺯ ﭘﺎﺭﺍﻣﺘﺮﻫﺎﻱ ﺑﺎﻻ ﺑﻪﻣﻨﻈﻮﺭ ﭘﻴﺶﺑﻴﻨﻲ ﺟﺎﻣﻊ ﻭ ﺩﻗﻴﻖﺗﺮ‪ ،‬ﺍﺯ ﺩﻳﮕﺮ ﭘﺎﺭﺍﻣﺘﺮﻫﺎﻱ ﻣﺆﺛﺮ ﺩﺭ ﺷﮑﺴﺖ ﻟﻮﻟﻪﻫـﺎ ﻧﺎﺷـ ﻲ ﺍﺯ‬
‫ﺷﺮﺍﻳﻂ ﻣﺤﻴﻄﻲ ﻭ ﺍﻗﻠﻴﻤﻲ‪ ،‬ﺑﺎﺭﮔﺬﺍﺭﻱ ﺗﺮﺍﻓﻴﻜﻲ ﻭ ﺷﺮﺍﻳﻂ ﺑﻬﺮﻩﺑﺮﺩﺍﺭﻱ ﻣﺎﻧﻨﺪ ﻋﻤﻖ ﻧﺼﺐ ﻟﻮﻟﻪﻫﺎ ﻭ ﻓﺸﺎﺭ ﺁﺏ ﺩﺭ ﺷﺒﮑﻪ ﺑﻬﺮﻩ ﮔﺮﻓﺘﻪ ﺷﺪ ﻭ ﺑﺎ ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ ﺭﻭﺷﻬﺎﻱ‬
‫ﺭﮔﺮﺳﻴﻮﻥ ﭼﻨﺪ ﻣﺘﻐﻴﺮﻩ‪ ،‬ﻫﻮﺵ ﻣﺼﻨﻮﻋﻲ )ﻋﺼﺒﻲ ﻭ ﻋﺼﺒ ﻲ ‪ -‬ﻓﺎﺯﻱ( ﻭ ﻫﻤﭽﻨﻴﻦ ﺭﻭﺵ ﺩﺍﺩﻩ ﮐﺎﻭﻱ ﺭﮔﺮﺳﻴﻮﻥ ﭼﻨﺪ ﺟﻤﻠﻪﺍﻱ ﺗﮑﺎﻣﻠﻲ )ﺑﺮﻧﺎﻣﻪ ﻧﻮﻳﺴﻲ ﮊﻧﺘﻴﮏ( ﺑـﻪ‬
‫ﭘﻴﺶ ﺑﻴﻨﻲ ﻧﺮﺥ ﺷﻜﺴﺖ ﻟﻮﻟﻪ ﭘﺮﺩﺍﺧﺘﻪ ﺷﺪ‪ .‬ﺑﻪﻣﻨﻈﻮﺭ ﺍﺭﺯﻳﺎﺑﻲ ﺭﻭﺷﻬﺎﻱ ﺍﺭﺍﺋﻪ ﺷﺪﻩ ﺑﺎ ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ ﺩﺍﺩﻩﻫﺎﻱ ﺟﻤﻊﺁﻭﺭﻱ ﺷﺪﻩ ﺣﻮﺍﺩﺙ ﻟﻮﻟﻪﻫﺎ ﺩﺭ ﺷﺒﻜﻪ ﺗﻮﺯﻳﻊ ﺁﺏ‬
‫ﻳﻜﻲ ﺍﺯ ﺷﻬﺮﻫﺎﻱ ﻛﺸﻮﺭ‪ ،‬ﻣﻄﺎﻟﻌﻪ ﻣﻮﺭﺩﻱ ﺍﻧﺠﺎﻡ ﮔﺮﺩﻳﺪ ﻭ ﻧﺘﺎﻳﺞ ﺍﺧﺬ ﺷﺪﻩ ﻣﻮﺭﺩ ﺑﺮﺭﺳﻲ ﻭ ﻣﻘﺎﻳﺴﻪ ﻗﺮﺍﺭ ﮔﺮﻓﺖ‪ .‬ﻧﺘﺎﻳﺞ ﺣﺎﺻﻞ ﺍﺯ ﻣﺪﻝ ﺷﺒﻜﻪ ﻋﺼﺒﻲ ﻭ ﺭﮔﺮﺳﻴﻮﻥ‬
‫ﭼﻨﺪ ﺟﻤﻠﻪﺍﻱ ﺗﮑﺎﻣﻠﻲ ﺗﺮﮐﻴﺒﻲ ﻣﺆﻳﺪ ﺍﻳﻦ ﻗﺎﺑﻠﻴﺖ ﺑﻮﺩ ﮐﻪ ﺍﻳﻦ ﺭﻭﺷﻬﺎ ﺑﺎ ﺗﺮﮐﻴﺐ ﺍﺛﺮ ﻧﻮﻓﻪﻫﺎﻱ ﭘﺎﺭﺍﻣﺘﺮﻫﺎﻱ ﻏﻴﺮﺩﺍﺋﻤﻲ ﻭ ﺭﻭﻧﺪ ﻳﻜﻨﻮﺍﺧﺖ ﺷﮑﺴﺘﻲ ﮐﻪ ﭘﺎﺭﺍﻣﺘﺮﻫﺎﻱ‬
‫ﺩﺍﺋﻤﻲ ﭘﺪﻳﺪ ﻣﻲﺁﻭﺭﻧﺪ ﻭ ﻫﻤﭽﻨﻴﻦ ﻗﺮﺍﺭ ﺩﺍﺩﻥ ﺗﻌﺪﺍﺩ ﺑﻴﺸﺘﺮﻱ ﺍﺯ ﭘﺎﺭﺍﻣﺘﺮﻫﺎﻱ ﻣﺆﺛﺮ‪ ،‬ﺩﺭ ﭘﻴﺶ ﺑﻴﻨﻲ ﻧﺮﺥ ﺷﻜﺴﺖ ﻟﻮﻟﻪﻫﺎ ﺗﻮﺍﻧﻤﻨﺪﺗﺮ ﻋﻤﻞ ﻣﻲ ﻛﻨﻨﺪ‪.‬‬

‫ﻭﺍﮊﻩ ﻫﺎﻱ ﻛﻠﻴﺪﻱ‪ :‬ﺷﺒﮑﻪ ﺗ ﻮﺯﻳﻊ ﺁﺏ‪ ،‬ﺍﻧﺰﻭﺍﻝ ﻭ ﺷﮑﺴﺖ ﻟﻮﻟﻪﻫﺎ‪ ،‬ﻣـﺪﻳﺮﻳﺖ ﺑﻬﻴﻨـﻪ ﺳـﺎﺯﻱ‪ ،‬ﭘـﻴﺶﺑﻴﻨـﻲ ‪ ،‬ﺭﮔﺮﺳـﻴﻮﻥ‪ ،‬ﺷـﺒﮑﻪ ﻋـﺼﺒﻲ‬
‫ﻣﺼﻨﻮﻋﻲ‪ ،‬ﺳﻴﺴﺘﻢ ﻋﺼﺒﻲ ‪ -‬ﻓﺎﺯﻱ‬

‫‪Study (Prediction) of Main Pipes Break Rates in Water Distribution‬‬

‫‪Systems Using Intelligent and Regression Methods‬‬

‫‪Massoud Tabesh 1‬‬ ‫‪Arash Aghaei 2‬‬ ‫‪Jaber Soltani 3‬‬

‫‪(Received Apr. 7, 2009‬‬ ‫)‪Accepted May 7, 2010‬‬

‫‪Abstract‬‬
‫‪Optimum operation of water distribution networks is one of the priorities of sustainable development of water‬‬
‫‪resources, considering the issues of increasing efficiency and decreasing the water losses. One of the key‬‬
‫‪subjects in optimum operational management of water distribution systems is preparing rehabilitation and‬‬
‫‪replacement schemes, prediction of pipes break rate and evaluation of their reliability. Several approaches have‬‬
‫‪been presented in recent years regarding prediction of pipe failure rates which each one requires especial data‬‬
‫‪sets. Deterministic models based on age and deterministic multi variables and stochastic group modeling are‬‬
‫‪examples of the solutions which relate pipe break rates to parameters like age, material and diameters. In this‬‬
‫‪1.Assoc. Prof., Center of Excellence for Eng. and Management of‬‬ ‫‪ - ۱‬ﺩﺍﻧﺸﻴﺎﺭ ﻭ ﻋﻀﻮ ﻗﻄﺐ ﻋﻠﻤﻲ ﻣﻬﻨﺪﺳﻲ ﻭ ﻣﺪﻳﺮﻳﺖ ﺯﻳﺮ ﺳﺎﺧﺘﻬﺎ‪ ،‬ﺩﺍﻧﺸﻜﺪﻩ ﻣﻬﻨﺪﺳﻲ ﻋﻤﺮﺍﻥ‪،‬‬
‫‪Infrastructures, School of Civil Eng. Tehran University, Tehran, Iran‬‬ ‫ﭘﺮﺩﻳﺲ ﺩﺍﻧﺸﻜﺪﻩﻫﺎﻱ ﻓﻨﻲ‪ ،‬ﺩﺍﻧﺸﮕﺎﻩ ﺗﻬﺮﺍﻥ‬
‫‪2. Instructor, Dept. of Civil Eng., Taft Branch, Islamic Azad University‬‬ ‫‪ - ۲‬ﻣﺪﺭﺱ ﺩﺍﻧﺸﮕﺎﻩ ﺁﺯﺍﺩ ﺍﺳﻼﻣﻲ‪ ،‬ﻭﺍﺣﺪ ﺗﻔﺖ‪ ،‬ﮔﺮﻭﻩ ﻣﻬﻨﺪﺳﻲ ﻋﻤﺮﺍﻥ‪ ،‬ﺗﻔﺖ )ﻧﻮﻳﺴﻨﺪﻩ ﻣﺴﺌﻮﻝ(‬
‫‪(Correponding Author) (+98 351) 8258858 aghaei@iauyazd.ac.ir‬‬ ‫‪aghaei@iauyazd.ac.ir (۰۳۵۱) ۸۲۵۸۸۵۸‬‬
‫‪3. Assist. Prof. of Civil Eng., Zabol University, Zabol‬‬
‫‪ - ۳‬ﺍﺳﺘﺎﺩﻳﺎﺭ ﺩﺍﻧﺸﻜﺪﻩ ﻣﻬﻨﺪﺳﻲ ﻋﻤﺮﺍﻥ‪ ،‬ﺩﺍﻧﺸﮕﺎﻩ ﺯﺍﺑﻞ‪ ،‬ﺯﺍﺑﻞ‬

‫ﺷﻤﺎﺭﻩ ‪ ٢‬ﺳﺎﻝ ‪١٣٩٠‬‬ ‫ﺁﺏ ﻭ ﻓﺎﺿﻼﺏ‬ ‫‪‬‬

 ‫‪paper besides the mentioned parameters, more factors such as pipe depth and hydraulic pressures are considered‬‬
‫‪as well. Then using multi variable regression method, intelligent approaches (Artificial neural network and neuro‬‬
‫‪fuzzy models) and Evolutionary polynomial Regression method (EPR) pipe burst rate are predicted. To evaluate‬‬
‫‪the results of different approaches, a case study is carried out in a part of Mashhad water distribution network.‬‬
‫‪The results show the capability and advantages of ANN and EPR methods to predict pipe break rates, in‬‬
‫‪comparison with neuro fuzzy and multi-variable regression methods.‬‬

‫‪Keywords: Water Distribution Networks, Pipe Break, Optimization, Prediction, Regression, ANN,‬‬
‫‪Neuro-Fuzzy Systems.‬‬

‫ﺗﻌﻤﻴﺮ ﻭ ﻧﮕﻬﺪﺍﺭﻱ ﭘﻴﺸﮕﻴﺮﺍﻧﻪ ﺩﺭ ﺷـﺮﻛﺘﻬﺎﻱ ﺁﺏ ﻭ ﻓﺎﺿـﻼﺏ ﻛـﻪ ﺑـﺮ‬
‫ﻣﺒﻨﺎﻱ ﺑﺮﻧﺎﻣﻪﺭﻳﺰﻱ ﻣﺪﻭﻥ ﺗﻌﻤﻴﺮ ﻭ ﺳـﺮﻭﻳﺲ ﺩﺳـﺘﮕﺎﻫﻬﺎ ﻭ ﺗﺠﻬﻴـﺰﺍﺕ‬
‫ﻗﺒﻞ ﺍﺯ ﺧﺮﺍﺑﻲ ﺍﺳﺘﻮﺍﺭ ﺍﺳﺖ‪ ،‬ﺍﺯ ﺳﺎﻝ ‪ ۲۲۳) ۱۳۸۲‬ﻣﻴﻠﻴﺎﺭﺩ ﺭﻳـﺎﻝ( ﺑـﻪ‬
‫ﺑﻌﺪ‪ ،‬ﻫﺰﻳﻨﻪﻫﺎ ﻛﺎﻫﺶ ﻳﺎﻓﺘﻪ ﺍﺳﺖ‪.[۱] .‬‬
‫ﺑﻪﻃﻮﺭ ﮐﻠﻲ ﺷﮑﺴﺖ ﻣﮑﺎﻧﻴﮑﻲ ﻟﻮﻟـﻪﻫـﺎ ﺗﺤـﺖ ﺷـﺮﺍﻳﻂ ﻋـﺎﺩﻱ )ﻳﻌﻨـﻲ‬
‫ﻓﺮﺳﻮﺩﮔﻲ ﻃﺒﻴﻌﻲ ﺁﻧﻬﺎ( ﺑﻪﻋﻨﻮﺍﻥ ﻣﺆﻟﻔﻪ ﺍﺻﻠﻲ ﺷـﺒﮑﻪﻫـﺎﻱ ﺗﻮﺯﻳـﻊ ﺁﺏ‬
‫ﺷﻬﺮﻱ‪ ،‬ﻫﻨﮕﺎﻣﻲ ﺭﺥ ﻣﻲﺩﻫﺪ ﮐﻪ ﺗﻨﺶﻫﺎﻱ ﺣﺎﺻﻞ ﺍﺯ ﺷﺮﺍﻳﻂ ﻣﺤﻴﻄﻲ ﻭ‬
‫ﻋﻤﻠﻴﺎﺗﻲ‪ ،‬ﻣﺘﺠﺎﻭﺯ ﺍﺯ ﻗﺎﺑﻠﻴﺖ ﺍﺭﺗﺠﺎﻋﻲ ﻣﺤـﺪﻭﺩ ﺳـﺎﺯﻩﺍﻱ ﻟﻮﻟـﻪ ﮔـﺮﺩﺩ‪.‬‬
‫ﺍﻧﺰﻭﺍﻝ ﻟﻮﻟﻪﻫﺎ ﻣﺘﺄﺛﺮ ﺍﺯﻋﻮﺍﻣـﻞ ﻣﺘﻌـﺪﺩﻱ ﻣﺎﻧﻨـﺪ ﻗـﺪﻣﺖ )ﺳـﻦ(‪ ،‬ﻗﻄـﺮ‪،‬‬
‫ﺟـﻨﺲ‪ ،‬ﭘﺎﻳـﺪﺍﺭﻱ ﻭ ﺧـﻮﺭﺩﮔﻲ ﺧــﺎﻙ ﻭ ﺁﺏ‪ ،‬ﺷـﺮﺍﻳﻂ ﻧـﺼﺐ‪ ،‬ﺍﺟــﺮﺍ ﻭ‬
‫ﺑﻬﺮﻩ ﺑﺮﺩﺍﺭﻱ ﺍﺳﺖ ﻛﻪ ﻣﻲﺗﻮﺍﻧﺪ ﺑﻪﻋﻨـﻮﺍﻥ ﺷـﺮﺍﻳﻂ ﻣﺤﻴﻄـﻲ ﻭ ﻋﻤﻠﻴـﺎﺗﻲ‬
‫ﺍﺳﺘﺎﺗﻴﮑﻲ ﻭ ﻳﺎ ﻣﺘﻐﻴﺮ ﺩﺭ ﺑﺎﺯﻩ ﻃﻮﻝ ﻟﻮﻟﻪ‪ ،‬ﺑﺮ ﺣﺴﺐ ﺯﻣﺎﻥ ﺗﻌﺮﻳﻒ ﺷﻮﺩ‪.‬‬
‫ﺑــﻪ ﺳــﺎﺩﮔﻲ ﻣــﺸﺨﺺ ﺍﺳــﺖ ﻛــﻪ ﭘﺎﺭﺍﻣﺘﺮﻫــﺎﻱ ﺍﻳــﺴﺘﺎ ﺩﺭ ﺯﻣــﺎﻥ‪ ،‬ﺑــﻪ‬
‫ﺧﺼﻮﺻﻴﺎﺕ ﻟﻮﻟﻪ ﻭ ﺷﻴﻮﻩ ﻧـﺼﺐ ﺁﻧﻬـﺎ ﻣـﺮﺗﺒﻂ ﻫـﺴﺘﻨﺪ‪ .‬ﺩﺭ ﺣـﺎﻟﻲ ﮐـﻪ‬
‫ﭘﺎﺭﺍﻣﺘﺮﻫﺎﻱ ﺩﻳﻨﺎﻣﻴﮑﻲ‪ ،‬ﺁﻧﻬـﺎﻳﻲ ﻫـﺴﺘﻨﺪ ﮐـﻪ ﭘﻴﺮﺍﻣـﻮﻥ ﻟﻮﻟـﻪ ﻭ ﻣﺤـﻴﻂ‬
‫ﺍﻃﺮﺍﻑ ﺁﻥ ﻗﺮﺍﺭ ﺩﺍﺭﻧﺪ‪ .‬ﻧﮑﺘﻪ ﺍﺳﺎﺳﻲ‪ ،‬ﺗﻮﺟﻪ ﺑـﻪ ﺍﻳـﻦ ﻣﻄﻠـﺐ ﺍﺳـﺖ ﮐـﻪ‬
‫ﺷﮑﺴﺖ ﺩﺭ ﻫﺮ ﻣﮑﺎﻥ‪ ،‬ﺷﺮﺍﻳﻂ ﺑﻪﺧﺼﻮﺻﻲ ﺩﺍﺭﺩ ﻭ ﻳﮏ ﻣﺪﻝ ﭘﻴﺶﺑﻴﻨﻲ‬
‫ﺷﮑﺴﺖ‪ ،‬ﻗﺎﺑﻠﻴﺖ ﺗﻌﻤﻴﻢ ﺑـﺮﺍﻱ ﻋﻤـﻮﻡ ﺭﺍ ﻧـﺪﺍﺭﺩ؛ ﻟـﺬﺍ ﺁﻧـﺎﻟﻴﺰ ﺩﺍﺩﻩﻫـﺎﻱ‬
‫ﺷﮑﺴﺖ ﻣﻲﺗﻮﺍﻧﺪ ﺑﻪ ﻣـﺸﺨﺺ ﻧﻤـﻮﺩﻥ ﺩﻻﻳـﻞ ﺍﺻـﻠﻲ ﺷﮑـﺴﺖﻫـﺎ ﺩﺭ‬
‫ﻣﺤﺪﻭﺩﻩﻫﺎﻱ ﻭﻳﮋﻩﺍﻱ ﺍﺯ ﺷﺒﮑﻪ ﻳﺎ ﺩﺭ ﻟﻮﻟﻪﺍﻱ ﺧﺎﺹ ﮐﻤﮏ ﻧﻤﻮﺩﻩ ﻭ ﮐﻠﻴﻪ‬
‫ﺗﻐﻴﻴﺮﺍﺕ ﺩﺭ ﻃﺮﺍﺣﻲ ﻭ ﺳﻴﺎﺳﺖﻫﺎﻱ ﺯﻳﺮ ﺳﺎﺧﺖ ﺭﺍ ﻣﺸﺨﺺ ﻧﻤﺎﻳﺪ‪.‬‬
‫ﺑﺮﺍﻱ ﺁﻧﺎﻟﻴﺰ ﺩﺍﺩﻩﻫﺎ ﻭ ﺷﻜﺴﺖ ﻟﻮﻟﻪﻫﺎ‪ ،‬ﺭﻭﺷـﻬﺎﻱ ﻣﺨﺘﻠﻔـﻲ ﻭﺟـﻮﺩ ﺩﺍﺭﺩ‬
‫]‪ ۲‬ﻭ ‪ .[۳‬ﺗﺤﻘﻴﻘﺎﺕ ﺯﻳﺎﺩﻱ ﺩﺭ ﺯﻣﻴﻨﻪ ﺁﻧﺎﻟﻴﺰ ﺷﮑﺴﺖ ﻭ ﭘﻴﺶﺑﻴﻨﻲ ﻧﺮﺥ‬
‫ﺷﻜﺴﺖ ﻟﻮﻟﻪﻫﺎ ﺍﻧﺠﺎﻡ ﺷﺪﻩ ﺍﺳﺖ ﺍﻣﺎ ﻫﻨﻮﺯ ﺍﮐﺜﺮﻳﺖ ﻣﺪﻝﻫﺎﻱ ﭘﻴﺶﺑﻴﻨﻲ‬
‫ﺷﮑﺴﺖ‪ ،‬ﻣﻨﺤـﺼﺮﺍﹰ ﺗﻌﻤـﻴﻢ ﻳﺎﻓﺘـﻪ ﺍﺯ ﻓﺎﮐﺘﻮﺭﻫـﺎﻱ ﺍﺳـﺘﺎﺗﻴﮑﻲ ﻭ ﺗﻌـﺪﺍﺩ‬
‫ﻣﺤﺪﻭﺩﻱ ﺍﺯ ﭘﺎﺭﺍﻣﺘﺮﻫﺎﻱ ﻣﺆﺛﺮ ﺩﺭ ﺷﮑﺴﺖ ﻫـﺴﺘﻨﺪ ﻭ ﺟﺎﻣﻌﻴـﺖ ﮐـﺎﻓﻲ‬
‫ﻧﺪﺍﺭﻧﺪ ]‪ ۴‬ﻭ ‪ .[۵‬ﻣﺪﻝ ﻣﺘﺪﺍﻭﻝ ﺷﺎﻣﻴﺮ‪ ١‬ﻭ ﻫﺎﻭﺍﺭﺩ‪ ٢‬ﻳـﻚ ﻣـﺪﻝ ﻧﻤـﺎﻳﻲ‬
‫ﺑﺮﺍﻱ ﭘﻴﺶﺑﻴﻨﻲ ﻧﺮﺥ ﺷﻜﺴﺖ ﻟﻮﻟﻪﻫﺎ ﺑﺮ ﺣﺴﺐ ﺯﻣﺎﻥ ﺑـﻪﺻـﻮﺭﺕ ﺯﻳـﺮ‬
‫ﺍﺳﺖ ]‪[۶‬‬
‫‪BR  BR ( t )  e‬‬
‫‪0‬‬
‫) ‪K ( tt‬‬
‫‪0‬‬
‫)‪(۱‬‬
‫ﺩﺭ ﺳــﺎﻝ ‪ ۱۳۷۸‬ﺑــﻪ ‪ ۲۳۹‬ﻣﻴﻠﻴــﺎﺭﺩ ﺭﻳــﺎﻝ ﺩﺭ ﺳــﺎﻝ ‪ ،۱۳۸۱‬ﺍﻓــﺰﺍﻳﺶ‬
‫‪1‬‬
‫‪2‬‬
‫‪Shamir‬‬
‫‪Howard‬‬
‫ﭼﺸﻤﮕﻴﺮﻱ ﻧﺸﺎﻥ ﻣﻲﺩﻫﺪ‪ .‬ﺩﺭﺣﺎﻟﻲ ﮐـﻪ ﺑـﺎ ﺷـﺮﻭﻉ ﻧﮕـﺮﺵ ﻭ ﺍﺳـﺘﻘﺮﺍﺭ‬
‫‪ -١‬ﻣﻘﺪﻣﻪ‬
‫ﻣﺤﺪﻭﺩﻳﺖ ﺣﺠﻢ ﻣﻨﺎﺑﻊ ﺁﺏ ﺷﻴﺮﻳﻦ ﻭ ﺍﻓﺰﺍﻳﺶ ﺗﻘﺎﺿﺎ ﺑﺮﺍﻱ ﺍﻳﻦ ﮐﺎﻻﻱ‬
‫ﮐﻤﻴﺎﺏ ﻭ ﺍﺭﺯﺷﻤﻨﺪ ﺍﺯ ﻳﮏ ﺳﻮ ﻭ ﭘﺮﺍﮐﻨﺪﮔﻲ ﺯﻣﺎﻧﻲ ﻭ ﻣﮑﺎﻧﻲ ﻧـﺎﻫﻤﮕﻮﻥ‬
‫ﻣﻴﺎﻥ ﻣﻨﺎﺑﻊ ﻭ ﻧﻴﺎﺯ ﺁﺑﻲ ﺍﺯ ﺳﻮﻱ ﺩﻳﮕﺮ‪ ،‬ﻣﻮﺿﻮﻉ ﺁﺏ ﻭ ﻣﺪﻳﺮﻳﺖ ﺑﺮ ﺁﻥ‬
‫ﺭﺍ ﺍﺯ ﺣﻴﻄﻪﻫﺎﻱ ﻣﻬﻨﺪﺳﻲ ﮔﺬﺭﺍﻧﺪﻩ ﻭ ﺟﺎﻳﮕﺎﻩ ﻭﻳﮋﻩﺍﻱ ﺭﺍ ﺩﺭ ﮔﺴﺘﺮﻩﻫـﺎﻱ‬
‫ﺍﻗﺘﺼﺎﺩ ﻭ ﻣﺪﻳﺮﻳﺘﻲ‪ ،‬ﻓﺮﻫﻨﮕﻲ ﻭ ﮔﺎﻩ ﺣﺘﻲ ﺳﻴﺎﺳـﻲ ﻭ ﺍﻣﻨﻴﺘـﻲ ﺑـﺮﺍﻱ ﺁﻥ‬
‫ﺭﻗﻢ ﺯﺩﻩ ﺍﺳﺖ‪ .‬ﺑﺎﻳﺪ ﺗﻮﺟﻪ ﺩﺍﺷﺖ ﺩﺭ ﺩﻧﻴﺎﻱ ﻣﺪﺭﻥ ﺍﻣﺮﻭﺯﻱ ﺑﺮﺍﻱ ﺗﻐﻴﻴﺮ‬
‫ﺩﺭ ﻣﻔﺎﻫﻴﻢ‪ ،‬ﺍﺩﺑﻴﺎﺕ ﻭ ﻣﺪﻳﺮﻳﺖ ﺧﺮﺩﻣﻨﺪﺍﻧﻪ ﺁﺏ‪ ،‬ﺑﺮﻧﺎﻣﻪﺭﻳﺰﻱ‪ ،‬ﺗﻮﺳﻌﻪ ﻭ‬
‫ﺑﻬﺮﻩﺑﺮﺩﺍﺭﻱ ﺑﻬﻴﻨﻪ ﺍﺯ ﺯﻳﺮﺳﺎﺧﺘﻬﺎﻱ ﺷﺒﻜﻪﻫـﺎﻱ ﺗﻮﺯﻳـﻊ ﺁﺏ‪ ،‬ﺑـﻪﻋﻨـﻮﺍﻥ‬
‫ﮔﺴﺘﺮﺩﻩﺗﺮﻳﻦ ﻭ ﭘﻴﭽﻴﺪﻩ ﺗـﺮﻳﻦ ﺑﺨـﺶ ﻳـﻚ ﺳﻴـﺴﺘﻢ ﺁﺑﺮﺳـﺎﻧﻲ ﻣﻄـﺮﺡ‬
‫ﻣﻲ ﮔﺮﺩﺩ‪ .‬ﺗﺄﻣﻴﻦ ﻭ ﺭﺳﺎﻧﺪﻥ ﺁﺏ ﻣﻮﺭﺩ ﻧﻴﺎﺯ ﻣﺸﺘﺮﻛﺎﻥ ﺧﺎﻧﮕﻲ‪ ،‬ﻣﺼﺎﺭﻑ‬
‫ﻋﻤﻮﻣﻲ‪ ،‬ﺗﺠﺎﺭﻱ ﻭ ﺻﻨﻌﺘﻲ ﺑﺎ ﻛﻤﻴ‪‬ﺖ‪ ،‬ﻛﻴﻔﻴﺖ ﻭ ﻓﺸﺎﺭ ﻣﻨﺎﺳﺐ ﺑﺮ ﺍﺳﺎﺱ‬
‫ﺍﺳﺘﺎﻧﺪﺍﺭﺩﻫﺎﻱ ﺑﻴﻦﺍﻟﻤﻠﻠﻲ ﻣﻮﺟﻮﺩ ﺍﺯ ﻣﺤﻞ ﺗﻮﻟﻴﺪ ﺑـﻪ ﻣـﺼﺮﻑ ﺩﺭ ﻃـﻮﻝ‬
‫ﺩﻭﺭﻩ ﻃﺮﺡ ﻭ ﻋﻤﺮ ﻣﻔﻴﺪ ﺳﻴﺴﺘﻢ ﺍﺯ ﺍﻫﻢ ﻭﻇﺎﻳﻒ ﺍﻳﻦ ﺯﻳﺮﺳﺎﺧﺖ ﺍﺳـﺖ‪.‬‬
‫ﺑﺎ ﺗﻮﺟﻪ ﺑﻪ ﺍﻳﻦ ﻣﻬﻢ‪ ،‬ﺣﻔﻆ ﻭ ﺍﺣﻴﺎﻱ ﺍﻳﻦ ﺷﺮﻳﺎﻥﻫـﺎ ﻫﻤـﻮﺍﺭﻩ ﺑـﻪﻋﻨـﻮﺍﻥ‬
‫ﻳﻜﻲ ﺍﺯ ﺑﺮﻧﺎﻣﻪﻫﺎﻱ ﻋﻤﺪﻩ ﺩﺳﺘﮕﺎﻫﻬﺎﻱ ﺍﺟﺮﺍﻳﻲ ﻣﺮﺑﻮﻃﻪ ﺍﺳﺖ‪.‬‬
‫ﻋﻤﻮﻣﺎﹰ ﺑﺮﻧﺎﻣﻪﺭﻳﺰﻱﻫﺎ ﺑﻪ ﺩﻭ ﮔﻮﻧﻪ ﮐﻮﺗﺎﻩ ﻣﺪﺕ ﺑﺮﺍﻱ ﻣﻘﺎﺻﺪ ﻋﻤﻠﻴـﺎﺗﻲ ﻭ‬
‫ﺑﻠﻨﺪ ﻣﺪﺕ ﺑﺮﺍﻱ ﺍﻫﺪﺍﻑ ﺳﺮﻣﺎﻳﻪﮔﺬﺍﺭﻱ ﻣﻄﺮﺡ ﻣﻲﮔﺮﺩﺩ‪ .‬ﺑﺮﻧﺎﻣـﻪﺭﻳـﺰﻱ‬
‫ﮐﻮﺗﺎﻩ ﻣﺪﺕ ﺑﺮﺍﻱ ﺗﺄﻣﻴﻦ ﻣﻨﺎﺑﻊ ﮐﺎﻓﻲ ﻭ ﻧﻴﺎﺯﻫـﺎﻱ ﺁﺗـﻲ ﺑـﻪ ﭘـﻴﺶﺑﻴﻨـﻲ‬
‫ﺗﻌﺪﺍﺩ ﺷﮑﺴﺖﻫﺎ ﺩﺭ ﻟﻮﻟﻪﻫﺎﻱ ﺍﺻﻠﻲ ﺁﺑﺮﺳﺎﻧﻲ ﮐﻪ ﺍﺣﺘﻤﺎﻻﹰ ﺩﺭ ﻃﻲ ﭼﻨـﺪ‬
‫ﺳﺎﻝ ﺁﺗﻲ ﺭﺥ ﺧﻮﺍﻫﺪ ﺩﺍﺩ‪ ،‬ﻧﻴﺎﺯ ﺩﺍﺭﺩ ]‪ .[۱‬ﺩﺭ ﺑﺮﻧﺎﻣﻪﺭﻳﺰﻱ ﺑﻠﻨﺪ ﻣـﺪﺕ‪،‬‬
‫ﻫﺪﻑ‪ ،‬ﺷﻨﺎﺳﺎﻳﻲ ﺻﺤﻴﺢ ﻭ ﺩﻗﻴﻖ ﺍﺯ ﻧﺮﺥ ﻓﺮﺳﻮﺩﮔﻲ ﻟﻮﻟـﻪﻫـﺎﻱ ﻣـﺪﻓﻮﻥ‬
‫ﺑﺮﺍﻱ ﻳﺎﻓﺘﻦ ﺯﻣﺎﻥ ﺑﻬﻴﻨﻪ ﺗﻌﻮﻳﺾ ﺑﻪﺟﺎﻱ ﺗﻌﻤﻴﺮ ﻟﻮﻟﻪ ﺍﺳﺖ‪ .‬ﺑﺪﻳﻬﻲ ﺍﺳﺖ‬
‫ﮐﻪ ﺩﺭ ﺗﻤﺎﻡ ﺍﻳﻦ ﺑﺮﻧﺎﻣـﻪﺭﻳـﺰﻱﻫـﺎ‪ ،‬ﻳـﺎﻓﺘﻦ ﺷـﺎﺧﺺ ﻧـﺮﺥ ﺷﮑـﺴﺖ‪ ،‬ﺍﺯ‬
‫ﮐﻠﻴﺪﻫﺎﻱ ﺭﺍﻫﮕﺸﺎﻱ ﻣﺴﺌﻠﻪ ﻣﺤﺴﻮﺏ ﻣﻲﺷﻮﺩ‪ .‬ﺳـﻨﺎﺭﻳﻮﻱ ﭘـﻴﺶﺑﻴﻨـﻲ‬
‫ﺗﻌﺪﺍﺩ ﺣﻮﺍﺩﺙ ﺩﺭ ﺷﺒﮑﻪ‪ ،‬ﻫﻤﮕﺎﻡ ﺑﺎ ﺑﺮﻧﺎﻣﻪﺭﻳﺰﻱ ﺑﻬﺮﻩﺑـﺮﺩﺍﺭﻱ ﺑﻬﻴﻨـﻪ ﺍﺯ‬
‫ﺁﻧﻬﺎ ﻋﻼﻭﻩ ﺑﺮ ﺻﺮﻓﻪﺟﻮﻳﻲ ﺩﺭ ﻫﺰﻳﻨﻪﻫﺎ ﻭ ﺍﻓﺰﺍﻳﺶ ﺩﺭﺁﻣـﺪﻫﺎ‪ ،‬ﺍﻓـﺰﺍﻳﺶ‬
‫ﻣﻴﺰﺍﻥ ﺳﺮﻭﻳﺲﺩﻫﻲ‪ ،‬ﻃﻮﻝ ﻋﻤﺮ ﻣﻔﻴﺪ ﺷﺒﻜﻪ ﻭ ﺍﻃﻤﻴﻨﺎﻥﭘﺬﻳﺮﻱ ﺳﻴـﺴﺘﻢ‬
‫ﺭﺍ ﺳﺒﺐ ﺧﻮﺍﻫﺪ ﺷﺪ‪.‬‬
‫ﺩﺭ ﺗﺄﮐﻴﺪ ﺍﻫﻤﻴﺖ ﻣﻮﺿﻮﻉ ﺑﺮﺍﺳﺎﺱ ﺑﺮﺭﺳـﻲﻫـﺎﻱ ﺍﻧﺠـﺎﻡ ﺷـﺪﻩ‪ ،‬ﻣﻴـﺰﺍﻥ‬
‫ﻫﺰﻳﻨﻪﻫﺎﻱ ﻧﮕﻬﺪﺍﺭﻱ ﻭ ﺗﻌﻤﻴﺮﺍﺕ ﺳﻨﺘﻲ ﺑﺨﺶ ﺁﺏ ﺍﺯ ‪ ۹۴‬ﻣﻴﻠﻴﺎﺭﺩ ﺭﻳﺎﻝ‬

‫‪‬‬ ‫ﺁﺏ ﻭ ﻓﺎﺿﻼﺏ‬ ‫ﺷﻤﺎﺭﻩ ‪ ٢‬ﺳﺎﻝ ‪١٣٩٠‬‬

‫]‪ .[۲‬ﺍﺯ ﺑﺮﺭﺳﻲ ﻭ ﺁﻧﺎﻟﻴﺰ ﺩﺍﺩﻩﻫﺎﻱ ﺷﻬﺮ ﮐﺮﻣﺎﻥ ﺑـﺮﺍﻱ ﺗﻐﻴﻴـﺮﺍﺕ ﺗﻌـﺪﺍﺩ‬ ‫ﮐﻪ ﺩﺭ ﺁﻥ‬
‫ﺣﻮﺍﺩﺙ ﺳﺎﻻﻧﻪ )‪ (NB‬ﺩﺭ ﺑﺮﺍﺑﺮ ﻗﻄﺮ ﻟﻮﻟﻪ ﺁﺯﺑﺴﺖ ﺭﺍﺑﻄﻪ ﺯﻳﺮ ﺭﺍ ﺍﺭﺍﺋـﻪ‬ ‫‪ BR‬ﻣﺸﺨﺼﻪ ﻧﺮﺥ ﺷﻜﺴﺖ ﺩﺭ ﺳﺎﻝ ‪) t‬ﺑﺮ ﺣﺴﺐ ‪ / yr / 1000ft‬ﺗﻌﺪﺍﺩ‬
‫ﮐﺮﺩﻩﺍﻧﺪ ]‪[۲‬‬ ‫ﺷﻜﺴﺖ( ‪ t o ،‬ﺳﺎﻝ ﭘﺎﻳﻪ ﺁﻧﺎﻟﻴﺰ‪ BR(t0) ،‬ﻧﺮﺥ ﺷﻜﺴﺖ ﺩﺭ ﺳـﺎﻝ ‪ t o‬ﻭ ‪K‬‬
‫‪NB  9.38e‬‬ ‫‪0.1653 De‬‬ ‫‪2‬‬
‫‪r  0.9194‬‬ ‫)‪(۵‬‬ ‫ﻧﺮﺥ ﺭﺷﺪ )‪ (1/ yr‬ﺍﺳﺖ‪ .‬ﺿﺮﻳﺐ ‪ K‬ﺑﺮﺣﺴﺐ ﺟﻨﺲ ﻭ ﻗﻄﺮ ﻟﻮﻟﻪﻫﺎ ﺑﻴﻦ‬
‫ﺁﻧﻬﺎ ﻫﻤﭽﻨﻴﻦ ﺑﺎ ﺁﻧﺎﻟﻴﺰ ﺩﺍﺩﻩﻫﺎ ﺑﺮﺍﻱ ﺗﻐﻴﻴﺮﺍﺕ ﻧﺮﺥ ﺷﮑـﺴﺖ ﺩﺭ ﻣﻘﺎﺑـﻞ‬ ‫‪ ۰/۰۵‬ﺗﺎ ‪ ۰/۱۵‬ﻣﺘﻐﻴﺮ ﺍﺳﺖ‪.‬‬
‫ﻗﻄﺮ ﻟﻮﻟﻪﻫﺎﻱ ﺁﺯﺑﺴﺖ ﻧﻴﺰ ﺭﺍﺑﻄﻪ ﺯﻳﺮ ﺭﺍ ﺑﻪﺩﺳﺖ ﺁﻭﺭﺩﻧﺪ‬ ‫ﮐﺘﻠﺮ‪ ١‬ﻭ ﮔﻮﻟﺘﺮ‪ ،٢‬ﻧﻴﺰ ﺑﺎ ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ ﺷـﺎﺧﺺ ﺗﻌـﺪﺍﺩ ﮐـﻞ ﺷﮑـﺴﺖﻫـﺎﻱ‬
‫ﺳﺎﻻﻧﻪ‪ ،‬ﻣﻄﺎﻟﻌﺎﺗﻲ ﺭﺍ ﺑﺮﺍﻱ ﺑﻴﺎﻥ ﺗﻐﻴﻴـﺮﺍﺕ ﺗﻌـﺪﺍﺩ ﺣـﻮﺍﺩﺙ ﺳـﺎﻻﻧﻪ ﺩﺭ‬
‫‪09092‬‬ ‫ﺑﺮﺍﺑﺮ ﺳﻦ ﻭ ﻗﻄﺮ ﻟﻮﻟﻪﻫﺎﻱ ﭼﺪﻧﻲ ﻭ ﺁﺯﺑﺴﺖ ﺩﺭ ﺷﻬﺮ ﻭﻳﻨﻴﭙـﮓ‪ ٣‬ﮐﺎﻧـﺎﺩﺍ‬
‫‪BR  0.01 ‬‬ ‫‪r 2  0.729‬‬ ‫)‪(۶‬‬
‫‪De 2‬‬ ‫ﺍﻧﺠﺎﻡ ﺩﺍﺩﻩ ﻭ ﺭﺍﺑﻄﻪﺍﻱ ﺑﻪﺷﺮﺡ ﺫﻳﻞ ﺍﺭﺍﺋﻪ ﺩﺍﺩﻧﺪ ]‪[۷‬‬
‫‪NB  K 0  Ae‬‬ ‫)‪( ۲‬‬
‫ﻫﻤﺎﻧﮕﻮﻧﻪ ﮐﻪ ﺍﺯ ﺭﻭﺍﺑﻂ ﺑﺎﻻ ﺑﺮ ﻣﻲﺁﻳﺪ‪ ،‬ﺍﺯ ﻧﺮﺥ ﺷﮑﺴﺖ ﻳﺎ ﺷﮑﺴﺖ‪ ،‬ﺗﻨﻬﺎ‬ ‫ﻛﻪ ﺩﺭ ﺍﻳﻦ ﺭﺍﺑﻄﻪ‬
‫ﺍﺭﺗﺒﺎﻃﺎﺕ ﻭﺍﺣﺪﻱ ﺑﺎ ﻳﮑﻲ ﺍﺯ ﭘﺎﺭﺍﻣﺘﺮﻫﺎﻱ ﻣﺴﺘﻘﻞ ﺍﺯ ﺯﻣﺎﻥ ﺑﻴﺎﻥ ﮔـﺸﺘﻪ‬ ‫‪ NB‬ﺗﻌﺪﺍﺩ ﺷﮑﺴﺖ ﺩﺭ ﺳﺎﻝ‪ K0 ،‬ﭘﺎﺭﺍﻣﺘﺮ ﺭﮔﺮﺳﻴﻮﻧﻲ ﻭ ‪ Ae‬ﺳﻦ ﻟﻮﻟﻪ ﺑﺮ‬
‫ﻛﻪ ﺍﻳﻦ ﻣﺤﺪﻭﺩﻳﺘﻬﺎ ﺑﺎﻋﺚ ﻣﻲﺷـﻮﺩ ﮐـﻪ ﻧـﺮﺥ ﺷﻜـﺴﺖ ﻳـﺎ ﺷﻜـﺴﺖ ﺍﺯ‬ ‫ﺣﺴﺐ ﺳﺎﻝ ﺍﺳﺖ‪ .‬ﻃﺒﻖ ﺍﻳﻦ ﺗﺤﻘﻴﻖ ﺭﻭﻧﺪ ﺗﻐﻴﻴﺮﺍﺕ ﺳﻦ ﺑـﺮﺍﻱ ﻫـﺮ ﺩﻭ‬
‫ﺟﺎﻣﻌﻴﺖ ﮐﻤﺘﺮﻱ ﺑﺮﺧﻮﺭﺩﺍﺭ ﺑﺎﺷﺪ‪ .‬ﺑﺮﺍﻱ ﺣﻞ ﺍﻳﻦ ﻣﺤﺪﻭﺩﻳﺘﻬﺎ‪ ،‬ﺁﻗﺎﻳﻲ ﻭ‬ ‫ﻣﻮﺭﺩ ﺧﻄﻲ ﺑﻮﺩﻩ ﻭ ﻧﺸﺎﻥ ﺩﻫﻨﺪﻩ ﺍﻓﺰﺍﻳﺶ ﺛﺎﺑﺖ ﺗﻌﺪﺍﺩ ﺷﮑـﺴﺖ ﺩﺭ ﻫـﺮ‬
‫ﻫﻤﮑــﺎﺭﺍﻥ ﺑــﺎ ﺍﺳــﺘﻔﺎﺩﻩ ﺍﺯ ﺍﻟﮕــﻮﻱ ﺑﺮﻧﺎﻣــﻪﻧﻮﻳــﺴﻲ ﮊﻧﺘﻴــﮏ ﻭ ﺗﺮﮐﻴــﺐ‬
‫ﺳﺎﻝ ﺍﺳﺖ ﮐﻪ ﻣﻌﻤﻮﻻﹰ ﺩﺭ ﻭﺍﻗﻌﻴﺖ ﺑﻪ ﺍﻳﻦ ﺻﻮﺭﺕ ﻧﺨﻮﺍﻫﺪ ﺑﻮﺩ‪ ،‬ﺯﻳﺮﺍ ﺑﺎ‬
‫ﭘﺎﺭﺍﻣﺘﺮﻫﺎﻱ ﺩﺍﺋﻤـﻲ ﻭ ﻏﻴﺮﺩﺍﺋﻤـﻲ ﺩﺭ ﻳﮑـﻲ ﺍﺯ ﺷـﺒﮑﻪﻫـﺎﻱ ﺗﻮﺯﻳـﻊ ﺁﺏ‬
‫ﮔﺬﺷﺖ ﺯﻣﺎﻥ‪ ،‬ﻟﻮﻟﻪﻫﺎ ﻓﺮﺳﻮﺩﻩ ﺷﺪﻩ ﻭ ﻣﻘﺎﻭﻣﺖ ﺁﻧﻬﺎ ﮐـﺎﻫﺶ ﻣـﻲﻳﺎﺑـﺪ ﻭ‬
‫ﺷﻬﺮﻱ ﮐﺸﻮﺭ‪ ،‬ﻓﺮﻣﻮﻝﻫﺎﻱ ﭘﻴﺶ ﺑﻴﻨﻲ ﺗﻌﺪﺍﺩ ﺷﮑﺴﺖ ﺩﺭ ﻟﻮﻟﻪﻫﺎﻱ ﺍﺻﻠﻲ‬
‫ﺁﺳﻴﺐ ﭘﺬﻳﺮﻱ ﺁﻧﻬﺎ ﺑﻴﺸﺘﺮ ﻣﻲ ﺷﻮﺩ‪ .‬ﺩﺭ ﻧﺘﻴﺠﻪ ﺑـﺎ ﻃـﻮﻻﻧﻲ ﺷـﺪﻥ ﻋﻤـﺮ‬
‫ﺁﺑﺮﺳﺎﻧﻲ ﺭﺍ ﺑﺎ ﻣﺘﻐﻴﺮﻫﺎﻱ ﻗﺎﺑﻞ ﺗﺸﺨﻴﺺ ﺻﺮﻳﺢ‪ ،‬ﺑﻬﺒﻮﺩ ﺩﺍﺩﻧﺪ‪ ،‬ﻫﺮ ﭼﻨﺪ‬ ‫ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ ﻟﻮﻟﻪﻫﺎ‪ ،‬ﺗﻌﺪﺍﺩ ﺷﮑﺴﺖﻫﺎﻱ ﺳﺎﻻﻧﻪ ﺍﻓﺰﺍﻳﺶ ﺑﻴﺸﺘﺮﻱ ﺧﻮﺍﻫﺪ‬
‫ﮐﻪ ﭘﺎﺭﺍﻣﺘﺮ ﻓﺸﺎﺭ ﺩﺭ ﺁﻥ ﺑﻪﻃﻮﺭ ﻣﺴﺘﻘﻴﻢ ﺩﺭ ﻧﻈﺮ ﮔﺮﻓﺘﻪ ﻧﺸﺪﻩ ﺑﻮﺩ ]‪.[۱۰‬‬ ‫ﻳﺎﻓﺖ‪ .‬ﺁﻧﻬﺎ ﻫﻤﭽﻨﻴﻦ ﺑﺮﺍﻱ ﻧﺮﺥ ﺷﻜـﺴﺖ )‪ (BR‬ﺭﺍﺑﻄـﻪ ﺯﻳـﺮ ﺭﺍ ﺍﺭﺍﺋـﻪ‬
‫ﺩﺭ ﺍﻳﻦ ﻣﻘﺎﻟﻪ ﺑـﺎ ﺍﺳـﺘﻔﺎﺩﻩ ﺍﺯ ﭼﻬـﺎﺭ ﺭﻭﺵ ﭘﻴـﺸﺮﻓﺘﻪ ﻣﺤﺎﺳـﺒﺎﺗﻲ ﻳﻌﻨـﻲ‬ ‫ﻛﺮﺩﻧﺪ‬
‫ﺭﻭﺷﻬﺎﻱ ﺷﺒﮑﻪﻫﺎﻱ ﻋﺼﺒﻲ‪ ،‬ﺳﻴﺴﺘﻢﻫﺎﻱ ﻋﺼﺒﻲ‪ -‬ﻓﺎﺯﻱ‪ ،‬ﺑﺮﻧﺎﻣﻪﻧﻮﻳﺴﻲ‬ ‫)‪BR  (2.002  0.0064De‬‬ ‫)‪(۳‬‬
‫ﮊﻧﺘﻴﮏ ﺗﮑﺎﻣﻠﻲ ﻭ ﺭﮔﺮﺳﻴﻮﻥ ﭼﻨﺪ ﻣﺘﻐﻴـﺮﻩ‪ ،‬ﺑـﺎ ﺑﻬـﺮﻩ ﮔﻴـﺮﻱ ﺍﺯ ﻣﺠﻤـﻮﻉ‬
‫ﻛﻪ ﺩﺭ ﺍﻳﻦ ﺭﺍﺑﻄﻪ‬
‫ﭘﺎﺭﺍﻣﺘﺮﻫﺎﻱ ﻗﺎﺑﻞ ﭘﻴﺶ ﺑﻴﻨﻲ ﺻﺮﻳﺢ ﺩﺭ ﻣﺪﻝﺳﺎﺯﻱ ﮐﻪ ﺍﺯ ﺩﺍﺩﻩﻫﺎﻱ ﺛﺒﺖ‬
‫‪ BR‬ﺑﻴﺎﻧﮕﺮ ﺗﻌﺪﺍﺩ ﺣﻮﺍﺩﺙ ﺳﺎﻟﻴﺎﻧﻪ ﺩﺭ ﻫﺮ ﮐﻴﻠﻮﻣﺘﺮ ﻟﻮﻟﻪ ‪ l‬ﺍﻡ ﻭ ‪ De‬ﻗﻄﺮ‬
‫ﺣﻮﺍﺩﺙ ﻗﺎﺑﻞ ﺍﺳﺘﺨﺮﺍﺝ ﺍﺳﺖ )ﻋﻤﻖ‪ ،‬ﻃﻮﻝ‪ ،‬ﻗﻄـﺮ‪ ،‬ﺳـﻦ ﻭ ﻓـﺸﺎﺭ(‪ ،‬ﺑـﻪ‬
‫ﻟﻮﻟﻪ ﺑﺮ ﺣﺴﺐ ﻣﻴﻠﻲﻣﺘﺮ ﺍﺳﺖ‪ .‬ﺭﺍﺑﻄﻪ ﺑﺎﻻ ﺩﺭ ﻟﻮﻟﻪﻫـﺎﻱ ﭼـﺪﻧﻲ ﺍﺯ ﻗﻄـﺮ‬
‫ﭘﻴﺶ ﺑﻴﻨﻲ ﻧﺮﺥ ﺣﻮﺍﺩﺙ ﺩﺭ ﺷﺒﻜﻪﻫﺎﻱ ﺗﻮﺯﻳﻊ ﺁﺏ )ﻟﻮﻟـﻪ ﻫـﺎ( ﭘﺮﺩﺍﺧﺘـﻪ‬
‫‪ ۱۰۰‬ﺗﺎ ‪ ۳۰۰‬ﻣﻴﻠﻲﻣﺘﺮ ﺗﻌﺮﻳﻒ ﺷﺪﻩ ﺍﺳﺖ‪.‬‬
‫ﺷﺪ‪ .‬ﻫﻤﭽﻨﻴﻦ ﺑﺎ ﺗﻮﺟﻪ ﺑﻪ ﻛﻴﻔﻴﺖ ﻭ ﻛﻤﻴ‪‬ﺖ ﺩﺍﺩﻩﻫﺎﻱ ﺩﺭ ﺩﺳﺘﺮﺱ‪ ،‬ﻧﻘﺎﻁ‬ ‫ﺳﻮ ﻭ ﻫﻤﮑﺎﺭﺍﻥ‪ ٤‬ﺑﺮ ﺍﺳﺎﺱ ﺩﺍﺩﻩﻫﺎﻱ ﺷﮑﺴﺖ ﻟﻮﻟﻪﻫﺎﻱ ﺷﺒﮑﻪ ﺗﻮﺯﻳﻊ ﺁﺏ‬
‫ﺿﻌﻒ ﻭ ﻗﻮﺕ ﺍﻳـﻦ ﺭﻭﺷـﻬﺎ ﺑﺮﺭﺳـﻲ ﮔﺮﺩﻳـﺪ‪ .‬ﻧﺘﻴﺠـﻪ ﺍﻳـﻦ ﺗﺤﻘﻴـﻖ ﺩﺭ‬ ‫ﺷﻬﺮ ﺳﻨﺖ ﻟﻮﺋﻴﺲ ‪ ٥‬ﺭﺍﺑﻄﻪ ﺯﻳﺮ ﺭﺍ ﺑﺮﺍﻱ ﻧﺮﺥ ﺷﮑﺴﺖ ﻟﻮﻟﻪ ‪ l‬ﺍﻡ )ﺗﻌـﺪﺍﺩ‬
‫ﻣﺪﻳﺮﻳﺖ ﺑﻬﻴﻨﻪ ﺑﻬﺮﻩﺑﺮﺩﺍﺭﻱ ﻭ ﺍﻃﻤﻴﻨﺎﻥﭘﺬﻳﺮﻱ ﺯﻳﺮﺳـﺎﺧﺖ ﺷـﺒﻜﻪﻫـﺎﻱ‬ ‫ﺷﮑﺴﺖ ﺩﺭ ﻫﺮ ﻣﺎﻳﻞ ﻟﻮﻟﻪ ‪ l‬ﺍﻡ( ﺍﺳﺘﺨﺮﺍﺝ ﮐﺮﺩﻧﺪ ]‪[۸‬‬
‫ﺗﻮﺯﻳﻊ ﺁﺏ ﺷﻬﺮﻱ ﮐﺎﺭﺑﺮﺩ ﻣﺴﺘﻘﻴﻢ ﺩﺍﺷﺘﻪ ﻭ ﺑﺎﻋـﺚ ﺑﻬﺒـﻮﺩ ﻧﺘـﺎﻳﺞ ﺁﻧﻬـﺎ‬ ‫‪0.6858 2.7158 2.7658‬‬
‫ﺧﻮﺍﻫﺪ ﮔﺮﺩﻳﺪ‪.‬‬ ‫‪BR ‬‬ ‫‪‬‬ ‫‪‬‬ ‫‪ 0.42‬‬ ‫)‪(۴‬‬
‫‪De3.28 De1.3131 De 3.5792‬‬
‫‪٦‬‬
‫ﺩﺭ ﭘﮋﻭﻫﺸﻲ ﺩﻳﮕﺮ ﺳﺎﮐﻠﻮﺗﻲ ﺑﺎ ﺑﻬﺮﻩﮔﻴﺮﻱ ﺍﺯ ﺷﺒﮑﻪ ﻋﺼﺒﻲ ﻣﺼﻨﻮﻋﻲ‪،‬‬
‫‪ -٢‬ﺭﻭﺵ ﺗﺤﻘﻴﻖ‬ ‫ﭘﻴﺶ ﺑﻴﻨﻲ ﻧﺮﺥ ﺷﮑﺴﺖ ﻟﻮﻟﻪﻫﺎ ﺩﺭ ﺷﺒﮑﻪ ﺗﻮﺯﻳﻊ ﺁﺏ ﻳﮑﻲ ﺍﺯ ﺷـﻬﺮﻫﺎﻱ‬
‫‪٧‬‬
‫‪ -١-٢‬ﺷﺒﮑﻪﻫﺎﻱ ﻋﺼﺒﻲ ﻣﺼﻨﻮﻋﻲ‬ ‫ﮐﺎﻧﺎﺩﺍ ﺭﺍ ﺑﺎ ﺗﻮﺟﻪ ﺑﻪ ﺗﺄﺛﻴﺮ ﻋﻤﺪﻩ ﺷﺮﺍﻳﻂ ﺍﻗﻠﻴﻤﻲ ﻭ ﻳﺨﺒﻨﺪﺍﻥ ﺩﺭ ﺷﮑﺴﺖ‬
‫ﺷﺒﻜﻪﻫﺎﻱ ﻋﺼﺒﻲ‪ ،‬ﺳﻴـﺴﺘﻢﻫـﺎﻱ ﺩﻳﻨـﺎﻣﻴﻜﻲ ﻫـﺴﺘﻨﺪ ﻛـﻪ ﺑـﺎ ﭘـﺮﺩﺍﺯﺵ‬ ‫ﻟﻮﻟﻪﻫﺎ ﻣﻮﺭﺩ ﺑﺮﺭﺳﻲ ﻗﺮﺍﺭ ﺩﺍﺩﻩ ﺍﺳﺖ ]‪ .[۹‬ﺩﺭ ﻃـﻲ ﺳـﺎﻟﻬﺎﻱ ﺍﺧﻴـﺮ ﻧﻴـﺰ‬
‫ﺩﺍﺩﻩﻫﺎﻱ ﺗﺠﺮﺑﻲ‪ ،‬ﺩﺍﻧﺶ ﻭ ﻳﺎ ﻗﺎﻧﻮﻥ ﻧﻬﻔﺘﻪ ﺩﺭ ﺩﺍﺩﻩﻫﺎ ﺭﺍ ﺑﻪ ﺳﺎﺧﺘﺎﺭ ﺷﺒﻜﻪ‬ ‫ﺗﺎﺑﺶ ﻭ ﻋﺎﺑﺪﻳﻨﻲ ﺑﺎ ﺗﻮﺟﻪ ﺑﻪ ﺁﻣﺎﺭ ﺣﻮﺍﺩﺙ ﺑﺮﺧﻲ ﺷﺒﮑﻪﻫﺎﻱ ﺁﺑﺮﺳـﺎﻧﻲ‬
‫ﻣﻨﺘﻘﻞ ﻣﻲﻛﻨﻨﺪ‪ .‬ﺍﻳﻦ ﺍﻟﮕﻮﺭﻳﺘﻢﻫﺎ ﺑﺎ ﺳﺎﺧﺘﺎﺭ ﭘﺮﺳﭙﺘﺮﻭﻥ‪ ،‬ﺍﺯ ﭼﻨﺪﻳﻦ ﻻﻳﻪ‬ ‫ﺷﻬﺮﻱ ﮐﺸﻮﺭ‪ ،‬ﺗﺤﻘﻴﻘﺎﺗﻲ ﺑﺮ ﺭﻭﻱ ﻧﺮﺥ ﺷﮑﺴﺖ ﻟﻮﻟﻪﻫﺎ ﺍﻧﺠـﺎﻡ ﺩﺍﺩﻩﺍﻧـﺪ‬
‫)ﻃﺒﻖ ﺗﺤﻘﻴﻘﺎﺕ ﻗﺒﻠﻲ ﻣﻌﻤﻮﻻﹰ ﺳﻪ ﻻﻳﻪ( ﻭ ﻫﺮ ﻻﻳﻪ ﺍﺯ ﺗﻌـﺪﺍﺩﻱ ﻭﺍﺣـﺪ‬
‫ﭘﺮﺩﺍﺯﺷــﮕﺮ ﺑــﻪ ﻧــﺎﻡ ﻧــﺮﻭﻥ ﺗــﺸﻜﻴﻞ ﺷــﺪﻩ ﺍﺳــﺖ ]‪ ۱۱‬ﻭ ‪ .[۱۲‬ﻧــﺮﻭﻥ‬ ‫‪1‬‬
‫‪2‬‬
‫‪Kettler‬‬
‫‪Goulter‬‬
‫ﻣﺼﻨﻮﻋﻲ ﻣﻲﺗﻮﺍﻧﺪ ﻳﻚ ﺗﺎﺑﻊ ﺭﻳﺎﺿﻲ ﻏﻴﺮ ﺧﻄﻲ ﺑﺎﺷﺪ ﻭ ﺑﻪﻃﻮﺭ ﻣﺴﺘﻘﻞ‬ ‫‪3‬‬
‫‪Winnipeg‬‬
‫‪4‬‬
‫‪Su et al.‬‬
‫‪5‬‬
‫‪St. Louis‬‬
‫‪7‬‬ ‫‪6‬‬
‫)‪Artifical Neural Network (ANN‬‬ ‫‪Sacluti‬‬

‫ﺷﻤﺎﺭﻩ ‪ ٢‬ﺳﺎﻝ ‪١٣٩٠‬‬ ‫ﺁﺏ ﻭ ﻓﺎﺿﻼﺏ‬ ‫‪‬‬

‫ﻳﮏ ﻣﺠﻤﻮﻋﻪ ﻓﺎﺯﻱ ﺍﺳﺖ ﻭ ﺧﺮﻭﺟﻲ ﻫﺮ ﮔـﺮﻩ ﺑﺮﺍﺑـﺮ ﺩﺭﺟـﻪ ﻋـﻀﻮﻳﺖ‬
‫ﻣﺘﻐﻴﺮ ﻭﺭﻭﺩﻱ ﺩﺭ ﺍﻳﻦ ﻣﺠﻤﻮﻋﻪ ﻓﺎﺯﻱ ﺍﺳﺖ‪ .‬ﺩﺭ ﻻﻳﻪ ﺩﻭﻡ ﻫﺮ ﮔﺮﻩ ﺩﺭﺟﻪ‬
‫ﻓﻌﺎﻟﻴﺖ ﻳﮏ ﻗﺎﻧﻮﻥ ﺭﺍ ﻣﺤﺎﺳﺒﻪ ﻣﻲﮐﻨﺪ‪ .‬ﻻﻳﻪ ﺳﻮﻡ ﻻﻳﻪ ﻋﻤﻠﮕﺮ ﺍﺳﺖ ﮐﻪ‬
‫ﻣﻴﺎﻧﮕﻴﻦ ﻭﺯﻥ ﻫﺮ ﻗﺎﻧﻮﻥ ﺭﺍ ﻣﺤﺎﺳﺒﻪ ﻣﻲﮐﻨﺪ‪ .‬ﻻﻳﻪ ﭼﻬﺎﺭﻡ ﺧﺮﻭﺟـﻲ ﻫـﺮ‬
‫ﮔﺮﻩ ﺩﺭ ﺍﻳﻦ ﻻﻳﻪ ﺍﺳﺖ ﻭ ﺩﺭ ﻻﻳﻪ ﭘﻨﺠﻢ‪ ،‬ﻫﺮ ﮔﺮﻩ ﻣﻘﺪﺍﺭ ﺧﺮﻭﺟﻲ ﻧﻬـﺎﻳﻲ‬
‫ﺍﺯ ‪ ANFIS‬ﺭﺍ ﻣﺤﺎﺳﺒﻪ ﻣﻲﮐﻨﺪ‪.‬‬
‫ﻋﻤﻞ ﻣﻲ ﻛﻨﺪ‪ .‬ﺩﺭ ﻧﺘﻴﺠﻪ ﻳﻚ ﺷﺒﻜﻪ ﻋﺼﺒﻲ ﻛﻪ ﺍﺯ ﺍﺟﺘﻤﺎﻉ ﺍﻳﻦ ﻧـﺮﻭﻥﻫـﺎ‬
‫ﺗﺸﻜﻴﻞ ﻣﻲﺷﻮﺩ ﻧﻴﺰ ﻣﻲ ﺗﻮﺍﻧﺪ ﻳﻚ ﺳﻴﺴﺘﻢ ﻛﺎﻣﻼﹰ ﭘﻴﭽﻴـﺪﻩ ﻭ ﻏﻴﺮﺧﻄـﻲ‬
‫ﺑﺎﺷﺪ‪.‬‬
‫ﺩﺭ ﺍﻳﻦ ﺗﺤﻘﻴﻖ ﺑﻪﻋﻨﻮﺍﻥ ﻳﮏ ﺭﻭﺵ ﻫﻮﺷﻤﻨﺪ‪ ،‬ﻣﺪﻟﻲ ﺑﺮﺍﻱ ﺑﺮﺭﺳﻲ ﻧﺮﺥ‬
‫ﺷﮑﺴﺖ ﻣﮑﺎﻧﻴﮑﻲ ﻟﻮﻟﻪﻫﺎ ﺍﺯ ﻃﺮﻳﻖ ﺟﻤﻊﺁﻭﺭﻱ ﺩﺍﺩﻩﻫﺎﻱ ﻭﺍﻗﻌﻲ ﻭ ﻣﺪﻝ‬
‫ﮐﺮﺩﻥ ﺗﻌﺪﺍﺩ ﺯﻳﺎﺩﻱ ﺍﺯ ﭘﺎﺭﺍﻣﺘﺮﻫﺎﻱ ﻣﺆﺛﺮ ﺩﺭ ﺣﻮﺍﺩﺙ ﺑﻪﻣﻨﻈﻮﺭ ﻣﺪﻳﺮﻳﺖ‬
‫ﺑﻬﻴﻨﻪ ﺑﻬﺮﻩﺑﺮﺩﺍﺭﻱ ﺍﺯ ﺷﺒﮑﻪﻫﺎ ﺍﺭﺍﺋﻪ ﮔﺮﺩﻳﺪ‪.‬‬

‫‪١‬‬
‫‪ -٢-٢‬ﺳﻴﺴﺘﻢ ﺍﺳﺘﻨﺘﺎﺝ ﻓﺎﺯﻱ ‪ -‬ﻋﺼﺒﻲ ﺗﻄﺎﺑﻘﻲ‬
‫ﺳﻴﺴﺘﻢﻫﺎﻱ ﺍﺳﺘﻨﺘﺎﺝ ﻓﺎﺯﻱ‪ -‬ﻋﺼﺒﻲ ﺗﻄﺎﺑﻘﻲ ﺑﺎ ﺗﺮﮐﻴﺐ ﻋﻤﻠﮑﺮﺩ ﻣﻨﻄﻘﻲ‬
‫ﺳﻴﺴﺘﻢﻫﺎﻱ ﻓﺎﺯﻱ ﻭ ﺷﺒﮑﻪﻫﺎﻱ ﻋﺼﺒﻲ ﻭ ﺑـﺎ ﺍﺗﮑـﺎ ﺑـﻪ ﺗﺮﮐﻴـﺐ ﻗـﺪﺭﺕ‬
‫ﻳـﺎﺩﮔﻴﺮﻱ ﻭ ﺗـﻮﺍﻥ ﺁﻣــﻮﺯﺵﭘـﺬﻳﺮﻱ ﺷـﺒﮑﻪﻫــﺎﻱ ﻋـﺼﺒﻲ‪ ،‬ﺭﺍﺑﻄـﻪ ﺑــﻴﻦ‬
‫)‪ -۱‬ﺍﻟﻒ(‬ ‫ﻣﺘﻐﻴﺮﻫﺎﻱ ﻣﺨﺘﻠﻒ ﺭﺍ ﺷﻨﺎﺳﺎﻳﻲ ﻣﻲﻧﻤﺎﻳﻨﺪ ﻭ ﺩﺭ ﺍﻳﻨﺠﺎ ﺑﻪ ﭘﻴﺶﺑﻴﻨﻲ ﻧﺮﺥ‬
‫ﺷﮑﺴﺖ ﻟﻮﻟﻪﻫﺎ ﻣﻲ ﭘﺮﺩﺍﺯﺩ‪ .‬ﺳﻴـﺴﺘﻢ ﻓـﺎﺯﻱ ﺍﺯ ﺟﻤﻠـﻪ ﺭﻭﺷـﻬﺎﻱ ﻫـﻮﺵ‬
‫ﻣﺼﻨﻮﻋﻲ ﻣﺒﺘﻨﻲ ﺑﺮ ﻗﻮﺍﻋﺪ ﻣﻨﻄﻘﻲ ﺍﮔﺮ‪ -‬ﺁﻧﮕﺎﻩ ﺍﺳﺖ‪ .‬ﻧﻘﻄﻪ ﺷﺮﻭﻉ ﻳـﮏ‬
‫ﺳﻴﺴﺘﻢ ﻓﺎﺯﻱ ﺑﻪﺩﺳﺖ ﺁﻭﺭﺩﻥ ﭘﺎﻳﮕﺎﻩ ﻗﻮﺍﻋـﺪ ﺁﻥ ﺍﺳـﺖ‪ .‬ﺍﻳـﻦ ﭘﺎﻳﮕـﺎﻩ‪،‬‬
‫ﻣﺠﻤﻮﻋﻪﺍﻱ ﺍﺯ ﻗﻮﺍﻧﻴﻦ ﻣﻨﻄﻘﻲ ﺍﮔﺮ ﺁﻧﮕﺎﻩ ﻓﺎﺯﻱ ﺍﺯ ﺩﺍﻧﺶ ﻓﺮﺩ ﺧﺒـﺮﻩ ﻳـﺎ‬
‫ﺩﺍﻧﺶ ﺣﻮﺯﻩ ﻣﻮﺭﺩ ﻧﻈﺮ ﺍﺳﺖ ﮐﻪ ﺭﺍﺑﻄﻪ ﺑﻴﻦ ﻣﺘﻐﻴﺮﻫـﺎﻱ ﻓـﺎﺯﻱ ﺭﺍ ﺑﻴـﺎﻥ‬
‫ﻣﻲﺩﺍﺭﻧﺪ‪ .‬ﺑﺮﺍﻱ ﭘﻴﺎﺩﻩﺳـﺎﺯﻱ ﻳـﮏ ﺳﻴـﺴﺘﻢ ﻓـﺎﺯﻱ ﺗﻮﺳـﻂ ﺷـﺒﮑﻪﻫـﺎﻱ‬
‫)‪ -۱‬ﺏ(‬ ‫ﻋــﺼﺒﻲ ﺳــﺎﺧﺘﺎﺭﻫﺎﻱ ﻣﺨﺘﻠﻔــﻲ ﭘﻴــﺸﻨﻬﺎﺩ ﺷــﺪﻩ ﺍﺳــﺖ ﮐــﻪ ﻳﮑــﻲ ﺍﺯ‬
‫ﺷﮑﻞ ‪ -۱‬ﺳﻴﺴﺘﻢ ﺍﺳﺘﺘﻨﺎﺝ ﻓﺎﺯﻱ ﻧﻮﻉ ﺳﻮﮔﻨﻮ ﺑﻪ ﺻﻮﺭﺕ ﻳﮏ ﺳﺎﺧﺘﺎﺭ‬ ‫ﻗﺪﺭﺗﻤﻨﺪﺗﺮﻳﻦ ﺍﻳﻦ ﺳﺎﺧﺘﺎﺭﻫﺎ‪ ،‬ﺳﺎﺧﺘﺎﺭ ﺷﺒﮑﻪ ﻓﺎﺯﻱ‪ -‬ﻋﺼﺒﻲ ﻣﻮﺳﻮﻡ ﺑﻪ‬
‫ﺷﺒﮑﻪ ﺟﻠﻮ ﺭﻭﻧﺪﻩ ﻭ ﻣﻌﻤﺎﺭﻱ ﺷﺒﮑﻪ ‪ ANFIS‬ﺁﻥ‬ ‫‪ ANFIS‬ﺍﺳــﺖ‪ .‬ﺍﺯ ﻗﺎﺑﻠﻴﺘﻬــﺎﻱ ﻣﻬــﻢ ﺍﻳــﻦ ﺳــﺎﺧﺘﺎﺭ ﺍﺳــﺘﻔﺎﺩﻩ ﺁﻥ ﺩﺭ‬
‫ﻣﺪﻝ ﺳﺎﺯﻱ ﻭ ﮐﻨﺘﺮﻝ ﺳﻴـﺴﺘﻢﻫـﺎﻱ ﭘﻴﭽﻴـﺪﻩ ﺍﺳـﺖ‪ .‬ﺑـﻪﻃـﻮﺭ ﮐﻠـﻲ ﻫـﺮ‬
‫ﺍﻟﮕــﻮﺭﻳﺘﻢ ﻳــﺎﺩﮔﻴﺮﻱ ﺗﺮﮐﻴﺒــﻲ ﺩﺭ ﻣــﺪﻝ ‪ ANFIS‬ﺍﺟــﺎﺯﻩ ﻣــﻲﺩﻫــﺪ ﮐــﻪ‬ ‫ﭘﺪﻳﺪﻩﺍﻱ ﮐﻪ ﺑﺘﻮﺍﻥ ﺁﻥ ﺭﺍ ﺑﺎ ﻣﺠﻤﻮﻋﻪﺍﻱ ﺍﺯ ﻣﺸﺎﻫﺪﺍﺕ ﺭﻓﺘﺎﺭ ﺛﺒﺖ ﮐﺮﺩ‪،‬‬
‫ﺳﻴﺴﺘﻢﻫﺎﻱ ﻓﺎﺯﻱ ﺩﺭ ﻣﺒﺎﺣﺚ ﺁﻣﻮﺯﺵ ﭘﺎﺭﺍﻣﺘﺮﻫﺎ ﺍﺯ ﺍﻟﮕﻮﺭﻳﺘﻢ ﺁﻣﻮﺯﺵ‬ ‫ﺑﺪﻭﻥ ﻧﻴﺎﺯ ﺑﻪ ﻓﺮﻣﻮﻝ ﻳﺎ ﻣﺪﻝ ﺧﺎﺹ ﺭﻳﺎﺿﻲ ﺑﺮﺍﻱ ﻃﺮﺍﺣﻲ‪ ،‬ﺑﻪﮐﻤﮏ ﺍﻳﻦ‬
‫ﮔﺴﺘﺮﺵ ﺑﺮﮔﺸﺘﻲ ﺗﻄﺎﺑﻘﻲ ﺍﺳﺘﻔﺎﺩﻩ ﻧﻤﺎﻳﻨﺪ‪ .‬ﺩﺭ ﺍﻳﻦ ﺭﻭﺵ ﺑﺎ ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ‬ ‫ﺭﻭﺵ ﻗﺎﺑﻞ ﺷﺒﻴﻪﺳﺎﺯﻱ ﺍﺳﺖ‪ .‬ﻭﻳﮋﮔﻲ ﺑﺴﻴﺎﺭ ﻣﻬﻢ ﺍﻳﻦ ﻧﻮﻉ ﺳﻴـﺴﺘﻢﻫـﺎ‬
‫ﺍﻟﮕﻮﺭﻳﺘﻢ ﺷﻴﺐ ﻧﺰﻭﻟﻲ ﺧﻄﺎ‪ ،‬ﻣﻘﺪﺍﺭ ﺧﻄﺎ ﺑﻪﺳﻤﺖ ﻭﺭﻭﺩﻱﻫﺎ ﭘﺨﺶ ﺷﺪﻩ‬ ‫ﺍﻳﻦ ﺍﺳﺖ ﮐﻪ ﻧﻴﺎﺯﻱ ﺑﻪ ﻓﺮﻣﻮﻝ ﻳﺎ ﻣﺪﻝ ﺭﻳﺎﺿﻲ ﺑﺮﺍﻱ ﻃﺮﺍﺣـﻲ ﻧﺪﺍﺭﻧـﺪ‪.‬‬
‫ﻭ ﭘﺎﺭﺍﻣﺘﺮﻫﺎ ﺗﺼﺤﻴﺢ ﻣﻲﺷﻮﻧﺪ‪ .‬ﺍﻳﻦ ﺭﻭﺵ ﺁﻣﻮﺯﺵ ﺩﻗﻴﻘﺎﹰ ﻫﻤﺎﻧﻨﺪ ﺭﻭﺵ‬ ‫ﻟﺬﺍ ﺩﺭ ﻣﻮﺭﺩ ﻃﺮﺍﺣﻲ ﺳﻴﺴﺘﻢﻫﺎﻳﻲ ﮐﻪ ﻧﺘﻮﺍﻥ ﻋﻤﻠﮑﺮﺩ ﺁﻧﻬﺎ ﺭﺍ ﺑﻪﺻـﻮﺭﺕ‬
‫ﭘﺲﺍﻧﺘﺸﺎﺭ ﺧﻄـﺎﻱ ﻣـﻮﺭﺩ ﺍﺳـﺘﻔﺎﺩﻩ ﺩﺭ ﺷـﺒﮑﻪﻫـﺎﻱ ﻋـﺼﺒﻲ ﻣـﺼﻨﻮﻋﻲ‬ ‫ﻣﺪﻝ ﺭﻳﺎﺿﻲ ﺑﻴﺎﻥ ﮐﺮﺩ‪ ،‬ﺑﺴﻴﺎﺭ ﻣﻔﻴﺪ ﻭ ﻣﻨﺎﺳﺐ ﺑﻪﻧﻈﺮ ﻣﻲﺭﺳﻨﺪ‪ .‬ﺷﮑﻞ ‪۱‬‬
‫)‪ (ANN‬ﺍﺳﺖ‪ .‬ﻃﺒﻖ ﺳﺎﺧﺘﺎﺭ ﺷﺒﮑﻪ ‪ ANFIS‬ﮐﻪ ﺩﺭ ﺷـﮑﻞ ‪ ۱‬ﻣﻼﺣﻈـﻪ‬ ‫ﺳﺎﺧﺘﺎﺭ ﻳﮏ ﺳﻴﺴﺘﻢ ﺍﺳﺘﺘﻨﺎﺝ ﻓـﺎﺯﻱ‪ -‬ﻋـﺼﺒﻲ ‪ ۵‬ﻻﻳـﻪ‪ ،‬ﺑـﺎ ﺩﻭ ﻣﺘﻐﻴـﺮ‬
‫ﻣﻲﮔـﺮﺩﺩ‪ ،‬ﻣـﻲ ﺗـﻮﺍﻥ ﺧﺮﻭﺟـﻲ ﮐـﻞ )‪ (f‬ﺭﺍ ﺑـﻪﺻـﻮﺭﺕ ﺗﺮﮐﻴـﺐ ﺧﻄـﻲ‬ ‫ﻭﺭﻭﺩﻱ ﮐـﻪ ﻫـﺮ ﻭﺭﻭﺩﻱ ﺩﺍﺭﺍﻱ ﻳـﮏ ﺗــﺎﺑﻊ ﻋـﻀﻮﻳﺖ ﺍﺳـﺖ ﺭﺍ ﻧـﺸﺎﻥ‬
‫ﭘﺎﺭﺍﻣﺘﺮﻫﺎﻱ ﻧﺘﻴﺠﻪ ﻧﻮﺷﺖ‬ ‫ﻣﻲﺩﻫﺪ‪ .‬ﺩﺭ ﺷﮑﻞ ‪ -۱‬ﺍﻟﻒ ﺳﻴﺴﺘﻢ ﻓﺎﺯﻱ ﺳﻮﮔﻨﻮ ﺑﺎ ﺩﻭ ﻭﺭﻭﺩﻱ )‪ (x,y‬ﻭ‬
‫‪f  w 1f1  w 2f 2  ( w 1x ) p1  ( w 1 y)q1  ( w 1 ) r1‬‬ ‫ﻳـــــﮏ ﺧﺮﻭﺟـــــﻲ )‪ (f‬ﺑـــــﻪﻫﻤـــــﺮﺍﻩ ﺩﻭ ﻗـــــﺎﻧﻮﻥ ﻭ ﺩﺭ ﺷـــــﮑﻞ‬
‫‪ ( w 2 x ) p1  ( w 2 y )q 2  ( w 2 ) r2‬‬ ‫)‪(۷‬‬ ‫‪ -۱‬ﺏ ﻣﻌﻤﺎﺭﻱ ﺳﻴﺴﺘﻢ ‪ ANFIS‬ﻣﻌﺎﺩﻝ ﺁﻥ ﺗﺮﺳﻴﻢ ﺷﺪﻩ ﺍﺳﺖ‪.‬‬
‫ﮐﻪ ﺩﺭ ﺍﻳﻦ ﺭﺍﺑﻄﻪ‬ ‫ﭼﻨﺎﻧﭽﻪ ﺧﺮﻭﺟﻲ ﻫﺮ ﻻﻳﻪ ﺷﺒﮑﻪ ﺑﻪ ﺻﻮﺭﺕ ‪ ) Oi,j‬ﺧﺮﻭﺟﻲ ‪ i‬ﺍﻣﻴﻦ ﮔﺮﻩ‬
‫‪ w i‬ﺩﺭﺟﻪ ﻓﻌﺎﻟﻴـﺖ ﻧﺮﻣـﺎﻝ ﺷـﺪﻩ ﻗـﺎﻧﻮﻥ ‪i‬ﺍﻡ ﻭ ‪ pi ،ri‬ﻭ ‪ qi‬ﭘﺎﺭﺍﻣﺘﺮﻫـﺎﻱ‬ ‫ﺩﺭ ‪ j‬ﺍﻣﻴﻦ ﻻﻳﻪ( ﻧﺸﺎﻥ ﺩﺍﺩﻩ ﺷﻮﺩ‪ ،‬ﺁﻧﮕﺎﻩ ﻣـﻲﺗـﻮﺍﻥ ﻋﻤﻠﮑـﺮﺩ ﻻﻳـﻪﻫـﺎﻱ‬
‫ﺗﻄﺒﻴﻘﻲ ﺍﻳﻦ ﻻﻳﻪ ﻫﺴﺘﻨﺪ ﮐﻪ ﭘﺎﺭﺍﻣﺘﺮﻫﺎﻱ ﻧﺘﻴﺠﻪ ﻧﻴﺰ ﺧﻮﺍﻧﺪﻩ ﻣﻲﺷﻮﻧﺪ‪ .‬ﺑﻪ‬ ‫ﻣﺨﺘﻠﻒ ﺷﮑﻞ ‪ ۱‬ﺭﺍ ﺑﻪﺻﻮﺭﺕ ﺯﻳﺮ ﺑﻴﺎﻥ ﻧﻤﻮﺩ‪:‬‬
‫ﺍﻳﻦ ﺗﺮﺗﻴﺐ ﭘﻴﺎﺩﻩﺳﺎﺯﻱ ﻳﮏ ﺳﻴﺴﺘﻢ ﻓﺎﺯﻱ ﺑﻪ ﮔﻮﻧﻪﺍﻱ ﺍﻧﺠﺎﻡ ﻣﻲﺷﻮﺩ ﮐﻪ‬ ‫ﻻﻳﻪ ﺍﻭﻝ ﻻﻳﻪ ﻭﺭﻭﺩﻱ ﺍﺳﺖ ﺑﻪﻃﻮﺭﻱ ﮐﻪ ﻫﺮ ﮔﺮﻩ ﺩﺭ ﺍﻳﻦ ﻻﻳﻪ‪ ،‬ﻣﻌـﺎﺩﻝ‬
‫ﻗﺎﺑﻠﻴﺖ ﻳﺎﺩﮔﻴﺮﻱ ﺩﺍﺷﺘﻪ ﺑﺎﺷﺪ‪ .‬ﺑﻨﺎﺑﺮﺍﻳﻦ ﻣﻘﺎﺩﻳﺮ ﭘﺎﺭﺍﻣﺘﺮﻫﺎﻱ ﻧﺘﻴﺠـﻪ ﺑـﺎ‬
‫ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ ﺭﻭﺵ ﮐﻤﺘﺮﻳﻦ ﻣﺠﻤﻮﻉ ﻣﺮﺑﻌﺎﺕ ﺧﻄﺎ ﺑﻪﺩﺳﺖ ﻣﻲﺁﻳﻨـﺪ‪ .‬ﺑـﺎ‬
‫‪1‬‬
‫)‪Adaptive Neuro-Fuzzy Inference System (ANFIS‬‬

‫‪‬‬ ‫ﺁﺏ ﻭ ﻓﺎﺿﻼﺏ‬ ‫ﺷﻤﺎﺭﻩ ‪ ٢‬ﺳﺎﻝ ‪١٣٩٠‬‬

 ‫‪ x 11 x 12‬‬ ‫‪x13 ... x1K‬‬ ‫‪‬‬
‫ﺗﺮﮐﻴﺐ ﺍﻳﻦ ﺭﻭﺵ ﻭ ﺭﻭﺵ ﭘـﺲﺍﻧﺘـﺸﺎﺭ ﺧﻄـﺎ‪ ،‬ﻳـﮏ ﺭﻭﺵ ﺁﻣﻮﺯﺷـﻲ‬
‫‪‬‬ ‫‪‬‬ ‫ﺗﺮﮐﻴﺒﻲ ﺍﻳﺠﺎﺩ ﻣﻲﺷﻮﺩ ﮐﻪ ﺑﻪﺻـﻮﺭﺕ ﺯﻳـﺮ ﻋﻤـﻞ ﻣـﻲﮐﻨـﺪ‪ :‬ﺩﺭ ﻫـﺮ ﺩﻭ‬
‫‪ x 21 x 22‬‬ ‫‪x 23 ... x 2 K ‬‬
‫‪X   x 31 x 32‬‬ ‫‪x 33 ... x 3K   X1 X 2 X 3 ..... X K ‬‬ ‫)‪(۹‬‬ ‫ﺁﻣﻮﺯﺵ ﻫﻨﮕﺎﻡ ﺣﺮﮐﺖ ﺭﻭ ﺑﻪ ﺟﻠﻮ‪ ،‬ﺧﺮﻭﺟﻲ ﮔﺮﻩﻫﺎ ﺑﻪﺻﻮﺭﺕ ﻋﺎﺩﻱ ﺗﺎ‬
‫‪‬‬ ‫‪‬‬
‫‪... ...‬‬ ‫‪... ........ ‬‬ ‫ﻻﻳﻪ ﭼﻬﺎﺭﻡ ﻣﺤﺎﺳﺒﻪ ﻣـﻲﺷـﻮﻧﺪ ﻭ ﺳـﭙﺲ ﭘﺎﺭﺍﻣﺘﺮﻫـﺎﻱ ﻧﺘﻴﺠـﻪ ﺗﻮﺳـﻂ‬
‫‪x x‬‬ ‫‪x N 3 ... x NK ‬‬
‫‪ N1 N 2‬‬ ‫ﺭﻭﺷﻬﺎﻱ ﻣﺤﺎﺳﺒﻪ ﺧﻄﺎ ﻧﻈﻴﺮ ﮐﻤﺘﺮﻳﻦ ﻣﺠﻤﻮﻉ ﻣﺮﺑﻌـﺎﺕ ﺧﻄـﺎ ﻣﺤﺎﺳـﺒﻪ‬
‫ﺟﻤﻠﻪ ‪ j‬ﺍﻡ ﺭﺍﺑﻄﻪ ‪ ۸‬ﻣﻲﺗﻮﺍﻧﺪ ﺑﻪﺻﻮﺭﺕ ﺯﻳﺮ ﻧﻮﺷﺘﻪ ﺷﻮﺩ‬ ‫ﻣﻲﺷﻮﻧﺪ‪ .‬ﺩﺭ ﺍﺩﺍﻣﻪ ﭘﺲ ﺍﺯ ﻣﺤﺎﺳـﺒﻪ ﺧﻄـﺎ ﺩﺭ ﺑﺎﺯﮔـﺸﺖ ﺭﻭ ﺑـﻪ ﻋﻘـﺐ‪،‬‬
‫‪‬‬
‫‪z Jn1  X1 ‬‬
‫‪ES J ,1‬‬
‫‪.X 2 ‬‬
‫‪ESJ , 2‬‬
‫‪.X 3 ‬‬
‫‪ES J , 3‬‬
‫‪............. X K ‬‬
‫‪ES J ,K ‬‬
‫‪‬‬ ‫‪j  2....d‬‬ ‫ﻧﺴﺒﺖ ﺧﻄﺎ ﺑﺮ ﺭﻭﻱ ﭘﺎﺭﺍﻣﺘﺮﻫﺎﻱ ﺷﺮﻁ‪ ،‬ﭘﺨﺶ ﺷـﺪﻩ ﻭ ﺑـﺎ ﺍﺳـﺘﻔﺎﺩﻩ ﺍﺯ‬
‫)‪(۱۰‬‬ ‫ﺭﻭﺵ ﺷﻴﺐ ﻧﺰﻭﻟﻲ ﺧﻄﺎ‪ ،‬ﻣﻘﺪﺍﺭ ﺁﻧﻬﺎ ﺗﺼﺤﻴﺢ ﻣﻲﺷﻮﺩ‪ .‬ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ ﺍﻳـﻦ‬
‫ﺭﻭﺵ ﺩﺭ ﺣﻞ ﻣﺴﺎﺋﻞ ﺷﺒﮑﻪﻫﺎﻱ ﺗﻮﺯﻳﻊ ﺁّﺏ ﺩﺭ ﺳﺎﻟﻬﺎﻱ ﺍﺧﻴﺮ ﺩﺭ ﺣـﺎﻝ‬
‫ﻛﻪ ﺩﺭ ﺁﻥ‬ ‫ﮔﺴﺘﺮﺵ ﺍﺳﺖ ]‪ ۱۳‬ﻭ ‪.[۱۴‬‬
‫‪ j ، zJ‬ﺍﻣﻴﻦ ﺑﺮﺩﺍﺭ ﺳﺘﻮﻧﻲ ﺣﺎﺻﻞ ﻭﺭﻭﺩﻱﻫﺎﻱ ﻛﺎﻧﺪﻳﺪ ﺷـﺪﻩ )‪ (xi‬ﻭ ‪ES‬‬
‫ﻣــﺎﺗﺮﻳﺲ ﺗــﻮﺍﻥﻫﺎﺳــﺖ‪ .‬ﺑﻨــﺎﺑﺮﺍﻳﻦ ﺑﺮﻣﺒﻨــﺎﻱ ﻣــﺎﺗﺮﻳﺲ ﺍﺻــﻠﻲ ﺭﻭﺵ‬ ‫‪ -٣-٢‬ﺭﻭﺵ ﺑﺮﻧﺎﻣﻪﻧﻮﻳﺴﻲ ﮊﻧﺘﻴﮏ ﺍﻟﮕـﻮﻳﻲ ﺩﺭ ﺭﮔﺮﺳـﻴﻮﻥ ﺗﺮﮐﻴﺒـﻲ‪-‬‬
‫ﺑﺮﻧﺎﻣﻪﻧﻮﻳﺴﻲ ﮊﻧﺘﻴﻚ ﻗﺎﻋﺪﻩﺍﻱ ﻣﻲﺗﻮﺍﻥ ﻧﻮﺷﺖ‬ ‫ﺗﮑﺎﻣﻠﻲ‬
‫‪Y  a0  a1  Z1  a2  Z2  a3  Z3  a4  Z4 ‬‬ ‫ﺗﻠﻔﻴﻖ ﺭﮔﺮﺳﻴﻮﻥ ﻋﺪﺩﻱ ﻭ ﻧﻤـﺎﺩﻳﻦ ﺑـﻪ ﻳـﻚ ﺭﮔﺮﺳـﻴﻮﻥ ﭼﻨﺪﺟﻤﻠـﻪﺍﻱ‬
‫‪٢‬‬
‫‪a 0  a1  X1  a 2  X1X2  a3  X3 X1  a 4  X2 X3‬‬ ‫ﺗﺮﮐﻴﺒﻲ‪ -‬ﺗﻜﺎﻣﻠﻲ ﺑﺮﺍﻱ ﻧﺨﺴﺘﻴﻦ ﺑﺎﺭ ﺗﻮﺳﻂ ﮔﻴﻮﺳﺘﻮ ﻟﻴﺴﻲ‪ ١‬ﻭ ﺳﺎﻭﻳﭻ‬
‫)‪(۱۱‬‬ ‫ﺍﺭﺍﺋﻪ ﮔﺮﺩﻳﺪ ]‪ .[۱۵‬ﺩﺭ ﺍﻳﻦ ﺭﻭﺵ ﺑﺎ ﺍﻟﮕﻮﻱ ﺑﺮﻧﺎﻣـﻪﻧﻮﻳـﺴﻲ ﮊﻧﺘﻴـﻚ‪،٣‬‬
‫ﺑﺎ ﺗﻮﺟﻪ ﺑـﻪ ﺁﻧﭽـﻪ ﺩﺭ ﻣـﻮﺭﺩ ﺳـﺎﺧﺘﺎﺭ ﻭ ﭘﺮﻭﺳـﻪ ﺭﮔﺮﺳـﻴﻮﻥ ﺗﺮﮐﻴﺒـﻲ‪-‬‬ ‫ﺑﺮﻧﺎﻣﻪﻫﺎﻱ ﻛﺎﻣﭙﻴﻮﺗﺮﻱ ﺑﺮﺍﻱ ﺣﻞ ﻣﺴﺌﻠﻪ ﺩﺍﺩﻩ ﺷﺪﻩ‪ ،‬ﺍﺳﺘﻨﺘﺎﺝ ﻣـﻲﺷـﻮﺩ‪.‬‬
‫ﺗﮑﺎﻣﻠﻲ ﺑﻴﺎﻥ ﺷﺪ‪ ،‬ﺑﺮﺍﻱ ﺗﻮﺍﺑﻊ ﺯﻳﺮ ﺳـﺎﺧﺖ ﻣـﺪﻝ‪ ،‬ﺍﺯ ﺟﻌﺒـﻪ ﺍﺑـﺰﺍﺭ ‪GA‬‬ ‫ﺑﻪﺩﻟﻴﻞ ﺍﺳﺘﺮﺍﺗﮋﻱ ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ ﺳﺎﺧﺘﺎﺭﻫﺎﻱ ﭼﻨﺪ ﺟﻤﻠﻪﺍﻱ ﻣﺸﺎﺑﻪ ﻗﻮﺍﻋـﺪ‬
‫ﻣﻮﺟﻮﺩ ﺩﺭ ﻧﺮﻡﺍﻓﺰﺍﺭ ‪ Matlab 6.5‬ﺑﻬﺮﻩ ﮔﺮﻓﺘﻪ ﺷﺪ ﻭ ﭘﺮﻭﺳﻪ ﺑﻪﺻـﻮﺭﺕ‬ ‫‪ ،GP‬ﻋﻼﻭﻩ ﺑﺮ ﺍﻳﻨﮑﻪ ﺩﺭ ﺍﻳﻦ ﺭﻭﺵ ﻧﻴﺎﺯﻱ ﺑﻪ ﺗﻌﻴﻴﻦ ﻣﺪﻝ ﺭﮔﺮﺳﻴﻮﻥ ﺍﺯ‬
‫ﺭﺍﺑﻂ ﮔﺮﺍﻓﻴﮑﻲ ﺩﺭ ﻣﺤﻴﻂ ﻧﺮﻡ ﺍﻓﺰﺍﺭ ﺍﺟﺮﺍ ﮔﺮﺩﻳﺪ‪.‬‬ ‫ﻗﺒﻞ ﻧﻴﺴﺖ‪ ،‬ﭘﻴﭽﻴﺪﮔﻲ ﻋﺒﺎﺭﺗﻬﺎﻱ ﺑﺮﺁﻭﺭﺩ ﺷﺪﻩ ﻭ ﻫﻤﭽﻨـﻴﻦ ﻣـﺸﻜﻼﺕ‬
‫ﻣﺮﺗﺒﻂ ﺑﺎ ‪ GP‬ﻛﻼﺳﻴﻚ ﻧﻴﺰ ﻛﺎﻫﺶ ﻣﻲﻳﺎﺑﺪ‪ .‬ﺗﻤﺎﻡ ﻋﺒﺎﺭﺗﻬﺎﻱ ﺣﺎﺻﻠﻪ ﺑﺎ‬
‫‪٤‬‬
‫‪ -٤-٢‬ﺭﮔﺮﺳﻴﻮﻥ ﻏﻴﺮ ﺧﻄﻲ‬ ‫ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ ﻓﺮﺍﻳﻨﺪ ﺗﻜـﺎﻣﻠﻲ ﺑـﻪﺻـﻮﺭﺕ ﻃـﺮﻑ ﺭﺍﺳـﺖ ﺭﺍﺑﻄـﻪ ﺯﻳـﺮ ﺩﺭ‬
‫ﺭﮔﺮﺳﻴﻮﻥ ﭼﻨﺪ ﻣﺘﻐﻴﺮﻩ ﺭﻭﺷﻲ ﺑﺮﺍﻱ ﻣﻄﺎﻟﻌﻪ ﺳﻬﻢ ﭼﻨﺪ ﻣﺘﻐﻴﺮ ﻣﺴﺘﻘﻞ ﺩﺭ‬ ‫ﻣﻲﺁﻳﻨﺪ‬
‫ﭘﻴﺶ ﺑﻴﻨﻲ ﻣﺘﻐﻴﺮ ﻭﺍﺑﺴﺘﻪ ﺍﺳﺖ‪ .‬ﻫﺪﻑ ﺍﺯ ﮐﺎﺭﺑﺮﺩ ﺭﻭﺵ ﺭﮔﺮﺳﻴﻮﻥ ﭼﻨـﺪ‬ ‫‪m‬‬
‫‪y   a j .Z j  a 0‬‬ ‫)‪(۸‬‬
‫ﻣﺘﻐﻴﺮﻩ ﺑﺮﺍﻱ ﭘﻴﺶ ﺑﻴﻨﻲ ﻧﺮﺥ ﺷﮑـﺴﺖ ﻟﻮﻟـﻪﻫـﺎ ﺩﺭ ﺷـﺒﮑﻪ ﺗﻮﺯﻳـﻊ ﺁﺏ‪،‬‬ ‫‪j1‬‬

‫ﻟﺤﺎﻅ ﮐﺮﺩﻥ ﺗﻌﺪﺍﺩ ﻋﻮﺍﻣﻞ ﺑﻴﺸﺘﺮﻱ ﺍﺳﺖ ﮐﻪ ﺩﺭ ﺷﮑﺴﺖ ﻟﻮﻟﻪﻫـﺎ ﻣـﺆﺛﺮ‬

‫ﻫﺴﺘﻨﺪ ﮐﻪ ﺗﺎ ﺣﺪ ﺍﻣﮑﺎﻥ ﺑﺘﻮﺍﻥ ﻣﺪﻝ ﺟﺎﻣﻊﺗﺮﻱ ﺩﺭ ﻣﻘﺎﻳﺴﻪ ﺑﺎ ﺗﺤﻘﻴﻘﺎﺕ‬ ‫ﮐﻪ ﺩﺭ ﺍﻳﻦ ﺭﺍﺑﻄﻪ‬
‫ﻗﺒﻠﻲ ﺑﻪﻣﻨﻈﻮﺭ ﭘﻴﺶ ﺑﻴﻨﻲﻫﺎﻱ ﺩﻗﻴﻖﺗﺮ ﺍﻧﺠﺎﻡ ﺩﺍﺩ‪ .‬ﺭﮔﺮﺳﻴﻮﻥ ﭼﻨﺪ ﻣﺘﻐﻴﺮﻩ‬ ‫‪ y‬ﺗﺨﻤﻴﻦ ﺣﺪﺍﻗﻞ ﻣﺮﺑﻌﺎﺕ ﻣﻘﺎﺩﻳﺮ ﻫﺪﻑ‪ aj ،‬ﻳﻚ ﭘﺎﺭﺍﻣﺘﺮ ﻗﺎﺑـﻞ ﺗﻨﻈـﻴﻢ‬
‫ﺩﺭ ﺍﻳﻦ ﺗﺤﻘﻴـﻖ ﺍﺯ ﻃﺮﻳـﻖ ﻧﻮﺷـﺘﻦ ﮐـﺪ ﺩﺭ ﻧـﺮﻡ ﺍﻓـﺰﺍﺭ ‪Matlab 7.04‬‬
‫ﺑﺮﺍﻱ ﺟﻤﻠﻪﻱ ‪ j‬ﺍﻡ‪ a0 ،‬ﻳﻚ ﺑﺎﻳﺎﺱ )ﺛﺎﺑﺖ( ﺍﻧﺘﺨﺎﺑﻲ‪ m ،‬ﺗﻌﺪﺍﺩ ﺗﺮﻡﻫﺎﻱ‬
‫ﺍﺳﺘﻔﺎﺩﻩ ﺷﺪ‪ .‬ﺩﺭ ﺍﻳﻦ ﻣﻄﺎﻟﻌﻪ ﻧﺮﺥ ﺷﮑﺴﺖ ﺑـﻪﻋﻨـﻮﺍﻥ ﻣﺘﻐﻴـﺮ ﻭﺍﺑـﺴﺘﻪ ﻭ‬ ‫ﻋﺒﺎﺭﺕ ﻭ ‪ Zj‬ﻳـﻚ ﻣﺘﻐﻴـﺮ ﺗﻐﻴﻴـﺮ ﺷـﻜﻞ ﻳﺎﻓﺘـﻪ ﺍﺳـﺖ‪ .‬ﺛﺎﺑـﺖ ‪ j‬ﭘﻴﺎﻣـﺪ‬
‫ﻋﻮﺍﻣﻞ ﻣﺆﺛﺮ ﺩﺭ ﺷﮑﺴﺖ ﻟﻮﻟﻪ ﻫﺎ ﻧﻈﻴﺮ ﻗﻄﺮ‪ ،‬ﻃﻮﻝ ‪ ،‬ﺳﻦ‪ ،‬ﻋﻤﻖ ﻧﺼﺐ ﻭ‬ ‫ﻣﺘﻐﻴﺮﻫﺎﻱ ﻣﺴﺘﻘﻞ ﭘﻴﺶﮔﻮﻳﻲ ﻛﻨﻨﺪﻩ ﻭﺭﻭﺩﻱ ﺍﺳﺖ )‪ ( x1 x2 … xk‬ﻛﻪ‬
‫ﻓﺸﺎﺭ ﻫﻴﺪﺭﻭﻟﻴﮑﻲ ﺑﻪﻋﻨﻮﺍﻥ ﻣﺘﻐﻴﺮﻫﺎﻱ ﻣﺴﺘﻘﻞ ﺩﺭ ﻧﻈﺮ ﮔﺮﻓﺘﻪ ﺷﺪﻧﺪ‪.‬‬ ‫‪ k‬ﺗﻌﺪﺍﺩ ﻣﺘﻐﻴﺮﻫﺎﻱ ﻣـﺴﺘﻘﻞ ﭘـﻴﺶﮔـﻮﻳﻲ ﻛﻨﻨـﺪﻩ ﺍﺳـﺖ‪ .‬ﻣﺒﻨـﺎﻱ ﺭﻭﺵ‬
‫ﺭﮔﺮﺳﻴﻮﻥ ﭼﻨﺪﺟﻤﻠﻪﺍﻱ ﺗﺮﮐﻴﺒﻲ‪ -‬ﺗﻜﺎﻣﻠﻲ‪ ،‬ﺍﻧﺠﺎﻡ ﻳﻚ ﺟﺴﺘﺠﻮﻱ ﭘﻮﻳـﺎ‬
‫‪ -۳‬ﻣﻌﻴﺎﺭﻫﺎﻱ ﺍﺭﺯﻳﺎﺑﻲ‬ ‫ﺍﺳﺖ ﻛﻪ ﺍﺯ ﻃﺮﻳﻖ ﻗﻴﺎﺱ ﻣﺮﺣﻠﻪ ﺑﻪ ﻣﺮﺣﻠﻪﻱ ﺭﮔﺮﺳـﻴﻮﻥﻫـﺎ ﺑـﻪ ﺗﻜﺎﻣـﻞ‬
‫ﺍﺯ ﺟﻤـﻠﻪ ﺗﻮﺍﺑـﻊ ﻫﺪﻑ ﻭ ﻣﻌﻴﺎﺭﻫﺎﻱ ﺍﺭﺯﻳﺎﺑﻲ ﺷﺒﮑﻪ ﻋﺼﺒﻲ ﻣﺼﻨﻮﻋﻲ ﮐﻪ‬ ‫ﻣﻲﺭﺳﺪ‪ .‬ﺑﻪﻋﻨﻮﺍﻥ ﻣﺜﺎﻝ ﭼﻨﺎﻧﭽﻪ ﻣﺎﺗﺮﻳﺲ ﻭﺭﻭﺩﻱ ‪ X‬ﺑـﻪﺻـﻮﺭﺕ ﺯﻳـﺮ‬
‫ﺩﺭ ﺍﻳﻦ ﭘﮋﻭﻫﺶ ﺑﻪ ﮐﺎﺭ ﺭﻓﺘﻪ‪ ،‬ﻣﻴﺎﻧﮕﻴﻦ ﻣﺠﺬﻭﺭﺍﺕ ﺧﻄﺎ‪ ،٥‬ﺟﺬﺭ ﻣﻴﺎﻧﮕﻴﻦ‬ ‫ﺑﺎﺷﺪ‬
‫ﻣﺠﺬﻭﺭﺍﺕ ﺧﻄﺎ‪ ٦‬ﻭ ﺷﺎﺧﺺ ﺗﻄﺎﺑﻖ‪ ٧‬ﺍﺳﺖ‪ .‬ﺭﻭﺍﺑﻂ ﺑـﻪﮐـﺎﺭ ﺭﻓﺘـﻪ ﺷـﺪﻩ‬

‫‪4‬‬
‫)‪Non-Linear Regression (NLR‬‬ ‫‪1‬‬
‫‪5‬‬
‫)‪Mean Squared Error (MSE‬‬ ‫‪Giustolisi‬‬
‫‪6‬‬ ‫‪2‬‬
‫)‪Root Mean Squared Error (RMSE‬‬ ‫‪Savic‬‬
‫‪7‬‬ ‫‪3‬‬
‫)‪Index of Accordance (IOA‬‬ ‫)‪Genetic Programming (GP‬‬

‫ﺷﻤﺎﺭﻩ ‪ ٢‬ﺳﺎﻝ ‪١٣٩٠‬‬ ‫ﺁﺏ ﻭ ﻓﺎﺿﻼﺏ‬ ‫‪‬‬

‫ﺁﻧﺎﻟﻴﺰ‪ ،‬ﺑﺮﺭﺳﻲ ﺷـﺪﻧﺪ ﻛـﻪ ﺩﺭ ﻧﻬﺎﻳـﺖ ﺷـﺎﺧﺺ ﻧـﺮﺥ ﺷﻜـﺴﺖ )ﺗﻌـﺪﺍﺩ‬
‫ﺣﻮﺍﺩﺙ ﺳﺎﻻﻧﻪ ﺩﺭ ﻭﺍﺣﺪ ﻃﻮﻝ( ﮐـﻪ ﺑـﻪ ﺧـﻮﺑﻲ ﺑﻴـﺎﻧﮕﺮ ﺗﻤـﺎﻡ ﺟﻮﺍﻧـﺐ‬
‫‪MSE   ( y actual,i  y prediction,i ) 2‬‬
‫ﻣﺴﺌﻠﻪ ﺍﺳﺖ ﺑﻪﻋﻨـﻮﺍﻥ ﻳـﮏ ﭘـﺎﺭﺍﻣﺘﺮ ﺷـﺎﺧﺺ ﺍﻧﺘﺨـﺎﺏ ﺷـﺪ‪ .‬ﻭﻳﮋﮔـﻲ‬
‫ﺷﺎﺧﺺ ﺍﻳﻦ ﺍﺳﺖ ﻛﻪ ﻣﻲﺗﻮﺍﻥ ﺑﺮ ﺍﺳﺎﺱ ﺁﻥ ﻗﻀﺎﻭﺕ ﻳﺎ ﻣﻘﺎﻳـﺴﻪﻫـﺎﻱ‬
‫ﻻﺯﻡ ﺭﺍ ﺍﻧﺠﺎﻡ ﺩﺍﺩ ]‪.[۲‬‬
‫ﺑﺮﺍﻱ ﺍﺭﺯﻳﺎﺑﻲ ﺭﻭﺵ ﺍﺭﺍﺋـﻪ ﺷـﺪﻩ ﻭ ﺑﺮﺭﺳـﻲ ﺷﺎﺧـﺼﻬﺎﻱ ﻣـﺆﺛﺮ ﺩﺭ‬
‫ﺷﻜﺴﺖ ﻟﻮﻟﻪﻫﺎ‪ ،‬ﻋﻼﻭﻩ ﺑﺮ ﻛﻴﻔﻴﺖ ﺩﺳﺘﺮﺳﻲ ﺩﺍﺩﻩﻫـﺎ‪ ،‬ﻣﻴـﺰﺍﻥ ﺻـﺤﺖ ﻭ‬
‫ﺩﻗﺖ ﺁﻧﻬﺎ ﻧﻴﺰ ﻣﺪ ﻧﻈﺮ ﺍﺳـﺖ‪ .‬ﺑـﺮﺍﻱ ﺍﻳـﻦ ﻣﻨﻈـﻮﺭ ﺍﺯ ﭘﺎﻳﮕـﺎﻩ ﺩﺍﺩﻩﻫـﺎﻱ‬
‫ﺧﻄﻮﻁ ﺍﺻﻠﻲ ﺷـﺒﻜﻪ ﺗﻮﺯﻳـﻊ ﺁﺏ ﺩﺭ ﻧﺎﺣﻴـﻪ ﭼﻬـﺎﺭ ﺷـﻬﺮ ﻣـﺸﻬﺪ ﮐـﻪ ﺍﺯ‬
‫ﻟﻮﻟﻪﻫﺎﻱ ﺑﺎ ﻗﻄﺮﻫﺎﻱ ‪ ٧٠٠ ،٨٠٠‬ﻭ ‪ ٦٠٠‬ﻣﻴﻠﻲﻣﺘـﺮ ﺷـﺮﻭﻉ ﺷـﺪﻩ ﻭ ﺑـﻪ‬
‫ﻗﻄﺮﻫﺎﻱ ‪ ١٠٠ ،١٥٠ ،٢٠٠ ،٢٥٠ ،٣٠٠ ،٤٠٠‬ﻭ ‪ ٨٠‬ﻣﻴﻠﻲﻣﺘـﺮ ﺧـﺘﻢ‬
‫ﻣﻲﺷﺪ‪ ،‬ﺍﺳﺘﻔﺎﺩﻩ ﮔﺮﺩﻳﺪ‪ .‬ﺑﺮﺍﻱ ﭘﺮﻫﻴﺰ ﺍﺯ ﻣﺤﺎﺳﺒﺎﺕ ﻃﻮﻻﻧﻲ ﻭ ﺗﻜﺮﺍﺭﻱ ﻭ‬
‫ﺑﻪﺩﻟﻴﻞ ﺩﺭ ﺍﺧﺘﻴﺎﺭ ﻧﺪﺍﺷﺘﻦ ﺍﻃﻼﻋـﺎﺕ ﺩﻗﻴـﻖ ﻭ ﻣﻨﺎﺳـﺐ ﺍﺯ ﺍﻧـﺸﻌﺎﺑﺎﺕ‬
‫ﻓﺮﻋﻲ‪ ،‬ﺗﻨﻬﺎ ﺑﺮﺍﻱ ﺷﺒﻜﻪ ﻟﻮﻟﻪﻫﺎﻱ ﺍﺻﻠﻲ ﺍﺯ ﻗﻄﺮ ‪ ٨٠‬ﺗﺎ ‪ ٣٠٠‬ﻣﻴﻠـﻲﻣﺘـﺮ‬
‫ﻛﻪ ﺭﻭﻧﺪ ﺷﻜﺴﺖ ﻭ ﺧﺮﺍﺑﻲ ﺩﺭ ﺁﻧﻬﺎ ﺑﻪﺻﻮﺭﺕ ﻣﺤﺴﻮﺳﻲ ﺗﺤـﺖ ﺗـﺄﺛﻴﺮ‬
‫ﻋﻮﺍﻣــﻞ ﮔﻔﺘــﻪ ﺷــﺪﻩ ﺑــﻮﺩ‪ ،‬ﭘــﻴﺶﺑﻴﻨــﻲ ﻧــﺮﺥ ﺷﻜــﺴﺖ ﺍﻧﺠــﺎﻡ ﺷــﺪ‪ .‬ﺩﺭ‬
‫ﻣﺪﻝﺳﺎﺯﻱ‪ ،‬ﺍﺯ ﻟﻮﻟﻪﻫﺎﻱ ﺁﺯﺑـﺴﺖ ﻗﻄـﺮ ‪ ٨٠‬ﺗـﺎ ‪ ٣٠٠‬ﻣﻴﻠـﻲﻣﺘـﺮ‪ ،‬ﺁﻣـﺎﺭ‬
‫ﺣﻮﺍﺩﺙ ﺭﻭﺯﺍﻧﻪ ﺛﺒﺖ ﺷﺪﻩ ﺩﺭ ﺳـﺎﻟﻬﺎﻱ ‪ ١٣٨٤‬ﺗـﺎ ‪ ٢٤٣٨) ٨٥‬ﺭﮐـﻮﺭﺩ‬
‫ﺑﺮﺭﺳﻲ ﺷﺪﻩ( ﻭ ﺍﻃﻼﻋﺎﺕ ﻓﺸﺎﺭ ﻃﻲ ﺍﻳﻦ ﺳﺎﻟﻬﺎ ﺍﺳﺘﻔﺎﺩﻩ ﺷـﺪ‪ .‬ﻻﺯﻡ ﺑـﻪ‬
‫ﺫﻛﺮ ﺍﺳـﺖ ﻛـﻪ ﺑﻴـﺸﺘﺮﻳﻦ ﻧـﺮﺥ ﺣـﻮﺍﺩﺙ ﺩﺭ ﺷـﺒﻜﻪ ﺧﻄـﻮﻁ ﺍﺻـﻠﻲ ﺩﺭ‬
‫ﻣﺤﺪﻭﺩﻩ ﻗﻄﺮﻱ ﻣﺬﻛﻮﺭ ﺑـﻮﺩﻩ ﺍﺳـﺖ‪ .‬ﻫﻤﭽﻨـﻴﻦ ﺩﺭ ﺯﻣـﺎﻥ ﺍﻳـﻦ ﻣﻄﺎﻟﻌـﻪ‬
‫ﺑﺮﺍﺳﺎﺱ ﻓﺸﺎﺭﻫﺎﻱ ﺛﺒﺖ ﺷﺪﻩ ﺩﺭ ﻭﺍﺣﺪ ﺣﻮﺍﺩﺙ ﻧﺎﺣﻴـﻪ ﭼﻬـﺎﺭ‪ ،‬ﺑﻴـﺸﻴﻨﻪ‬
‫ﻓﺸﺎﺭ ﺩﺭ ﻧﺎﺣﻴﻪ ‪ ٨/٥‬ﺍﺗﻤﺴﻔﺮ ﻭ ﺣﺪﺍﻗﻞ ﺁﻥ ﻧﻴﺰ ﺩﺭ ﺑﻌﻀﻲ ﺳﺎﻋﺎﺕ ‪١/٢‬‬
‫ﺍﺗﻤﺴﻔﺮ ﺑﻮﺩ‪ .‬ﻫﻤﭽﻨﻴﻦ ﺑﺎ ﺗﻮﺟﻪ ﺑﻪ ﺍﻳﻦ ﻣﻄﻠﺐ ﮐﻪ ﻟﻮﻟﻪﻫﺎ ﺑﺎ ﻣﺸﺨـﺼﺎﺕ‬
‫ﺳﺎﺧﺘﺎﺭﻱ ﻳﮑﺴﺎﻥ ﺍﺯ ﻧﻈﺮ ﻗﺪﻣﺖ‪ ،‬ﻗﻄـﺮ‪ ،‬ﺟـﻨﺲ ‪ ،‬ﺑﺎﻳـﺪ ﻧـﺮﺥ ﺷﮑـﺴﺖ‬
‫ﻳﮑﺴﺎﻧﻲ ﺩﺍﺷﺘﻪ ﺑﺎﺷﻨﺪ‪ ،‬ﺑﻪﻣﻨﻈـﻮﺭ ﺍﺳـﺘﺨﺮﺍﺝ ﭘﺎﺭﺍﻣﺘﺮﻫـﺎﻱ ﺍﺻـﻠﻲ ﻣـﺪﻝ‬
‫ﺷﻜﺴﺖ ﺍﺯ ﮐﻞ ﻟﻮﻟﻪﻫﺎﻱ ﻣﻨﻄﻘﻪ ﻣﻄﺎﻟﻌﺎﺗﻲ ﺧﻮﺍﻩ ﺩﭼﺎﺭ ﺣﺎﺩﺛﻪ ﺷﺪﻩ ﺑﺎﺷﻨﺪ‬
‫ﻳﺎ ﺧﻴﺮ‪ ،‬ﺍﺳﺘﻔﺎﺩﻩ ﮔﺮﺩﻳﺪ‪ .‬ﮔﺴﺘﺮﻩ ﻣﺤﺪﻭﺩﻩ ﻣﻄﺎﻟﻌﺎﺗﻲ ﺩﺭ ﺷﮑﻞ ‪ ٢‬ﻧﻤﺎﻳـﺎﻥ‬
‫ﺍﺳﺖ ]‪.[۱۰‬‬
‫‪ -۲‬ﺟﻤﻊﺁﻭﺭﻱ ﺣﺘﻲﺍﻟﻤﻘﺪﻭﺭ ﺗﻤﺎﻡ ﭘﺎﺭﺍﻣﺘﺮﻫﺎﻱ ﻣﺆﺛﺮ ﻭ ﺩﺧﻴﻞ ﺩﺭ ﻣﺴﺌﻠﻪ‬
‫‪ -۵‬ﻧﺘﺎﻳﺞ ﻣﺪﻝﺳﺎﺯﻱ ﭘﻴﺶﺑﻴﻨـﻲ ﻧـﺮﺥ ﺷﻜـﺴﺖ ﻟﻮﻟـﻪﻫـﺎﻱ‬
‫ﺷﺒﮑﻪ ﺗﻮﺯﻳﻊ ﺁﺏ‬
‫ﺍﻧﺠﺎﻡ ﻣﺪﻝﺳﺎﺯﻱ ﭘﻴﺶﺑﻴﻨﻲ ﻧﺮﺥ ﺷﻜﺴﺖ ﻟﻮﻟﻪﻫﺎﻱ ﺷـﺒﮑﻪ ﺗﻮﺯﻳـﻊ ﺁﺏ‬
‫ﻧﻴﺰ ﻣﺎﻧﻨﺪ ﺗﻤﺎﻡ ﺑﺮﻧﺎﻣـﻪﻫـﺎﻱ ﻣـﺪﻝﺳـﺎﺯﻱ ﻧﻴﺎﺯﻣﻨـﺪ ﻣﺠﻤﻮﻋـﻪ ﺩﺍﺩﻩﻫـﺎ ﻭ‬
‫ﺍﻃﻼﻋﺎﺗﻲ ﺍﺳﺖ ﻛـﻪ ﭘﺎﺭﺍﻣﺘﺮﻫـﺎ ﻭ ﺿـﺮﺍﻳﺐ ﻣـﺪﻝ ﺗﻮﺳـﻂ ﺁﻧﻬـﺎ ﺷـﻜﻞ‬
‫ﻣﻲﮔﻴﺮﻧﺪ‪ .‬ﻫﻤﺎﻧﻄﻮﺭ ﻛـﻪ ﻗـﺒﻼﹰ ﻧﻴـﺰ ﺍﺷـﺎﺭﻩ ﺷـﺪ ﺩﺭ ﻣـﺪﻝﻫـﺎﻱ ﺣﺎﺿـﺮ‪،‬‬
‫ﻭﺭﻭﺩﻱﻫﺎ ﺷﺎﻣﻞ ﺗﻌﺪﺍﺩ ﺯﻳﺎﺩﻱ ﺍﺯ ﭘﺎﺭﺍﻣﺘﺮﻫﺎﻱ ﻣﺆﺛﺮ ﺩﺭ ﺷﮑﺴﺖ ﻟﻮﻟﻪﻫﺎ‬
‫ﻣﺎﻧﻨﺪ ﻗﻄﺮ‪ ،‬ﻃﻮﻝ‪ ،‬ﺳﻦ‪ ،‬ﻋﻤـﻖ ﻧـﺼﺐ ﻭ ﻓـﺸﺎﺭ ﻫﻴـﺪﺭﻭﻟﻴﮑﻲ ﻟﻮﻟـﻪﻫـﺎﻱ‬
‫ﺁﺯﺑﺴﺖ ﺍﺯ ﻗﻄﺮ ‪ ۸۰‬ﺗﺎ‪ ۳۰۰‬ﻣﻴﻠﻲﻣﺘﺮ ﻭ ﺧﺮﻭﺟﻲ‪ ،‬ﻧﺮﺥ ﺷﮑﺴﺖ ﺩﺭ ﻫـﺮ‬
‫ﻟﻮﻟﻪ ﺍﺳﺖ‪.‬‬
‫‪‬‬ ‫ﺁﺏ ﻭ ﻓﺎﺿﻼﺏ‬
‫ﺑﻪﺻﻮﺭﺕ ﺯﻳﺮ ﺍﺳﺖ‬
‫‪1 n‬‬
‫)‪(۱۲‬‬
‫‪n i 1‬‬
‫‪1 n‬‬
‫‪RMSE ‬‬ ‫‪ ( y actual,i  y prediction,i )2‬‬ ‫)‪(۱۳‬‬
‫‪n i 1‬‬
‫‪n‬‬ ‫‪2‬‬
‫‪ yprediction,i  yactual,i ‬‬
‫‪IOA 1‬‬ ‫‪ i1‬‬ ‫‪‬‬ ‫)‪(۱۴‬‬
‫‪n‬‬ ‫‪2‬‬
‫‪( yprediction,i  ya varege,i  yactual,i  ya varege,i ) ‬‬
‫‪ i1‬‬ ‫‪‬‬
‫ﮐﻪ ﺩﺭ ﺁﻥ‬
‫‪ y actual‬ﺩﺍﺩﻩﻫــﺎﻱ ﻭﺍﻗﻌــﻲ )ﻣــﺸﺎﻫﺪﻩ ﺷــﺪﻩ(‪ y prediction ،‬ﺩﺍﺩﻩﻫــﺎﻱ‬
‫ﭘﻴﺶ ﺑﻴﻨﻲﺷﺪﻩ‪ y average ،‬ﻣﺘﻮﺳﻂ ﺩﺍﺩﻩﻫﺎﻱ ﻭﺍﻗﻌﻲ ﻭ ‪ n‬ﺗﻌﺪﺍﺩ ﻣﺸﺎﻫﺪﺍﺕ‬
‫ﺍﺳﺖ‪.‬‬
‫‪ -۴‬ﺍﺭﺯﻳﺎﺑﻲ ﺭﻭﺵ‬
‫ﻫﻤﺎﻧﮕﻮﻧﻪ ﮐﻪ ﺑﻴﺎﻥ ﮔﺮﺩﻳﺪ‪ ،‬ﻫﺪﻑ ﺍﺻﻠﻲ ﺍﻳﻦ ﺗﺤﻘﻴـﻖ ﭘـﻴﺶﺑﻴﻨـﻲ ﻧـﺮﺥ‬
‫ﺷﮑﺴﺖ ﻟﻮﻟﻪﻫﺎ ﺩﺭ ﭼﺎﺭﭼﻮﺏ ﻳﮏ ﺳﻴﺎﺳﺖ ﺑﻬﻴﻨﻪ ﻣﺪﻳﺮﻳﺖ ﺑﻬﺮﻩ ﺑﺮﺩﺍﺭﻱ‬
‫ﺑﺎ ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ ﺍﻧﻮﺍﻉ ﺭﻭﺷﻬﺎﻱ ﻫﻮﺵ ﻣﺼﻨﻮﻋﻲ ﻭ ﺭﮔﺮﺳﻴﻮﻥ ﭼﻨﺪ ﻣﺘﻐﻴـﺮﻩ‬
‫ﻭ ﻣﻘﺎﻳﺴﻪ ﺗﻮﺍﻧﻤﻨﺪﻱﻫﺎﻱ ﻫـﺮ ﻳـﮏ ﺍﺯ ﺁﻧﻬـﺎ ﺑـﻮﺩ‪ .‬ﺑـﺮﺍﻱ ﭘﻴـﺸﺒﺮﺩ ﺍﻳـﻦ‬
‫ﻣﻨﻈﻮﺭ‪ ،‬ﺍﺳﺘﺮﺍﺗﮋﻱﻫﺎﻱ ﺯﻳﺮ ﺍﻧﺠﺎﻡ ﭘﺬﻳﺮﻓﺖ‪:‬‬
‫‪ -۱‬ﺑﺮﺭﺳــﻲ ﻭ ﺁﻧــﺎﻟﻴﺰ ﺣــﻮﺍﺩﺙ ﺩﺭ ﺷــﺒﻜﻪﻫــﺎ‪ :‬ﻫــﺪﻑ ﺍﺯ ﺍﻧﺠــﺎﻡ ﺁﻧــﺎﻟﻴﺰ‬
‫ﺣﻮﺍﺩﺙ‪ ،‬ﻳﺎﻓﺘﻦ ﺍﺭﺗﺒﺎﻃﺎﺕ ﻗﺎﺑﻞ ﻓﻬـﻢ ﻭ ﺁﺷـﻜﺎﺭ ﺑـﻴﻦ ﻧـﺮﺥ ﺷﻜـﺴﺖ ﻭ‬
‫ﻣﺸﺨﺼﺎﺗﻲ ﺍﺯ ﻟﻮﻟﻪ ﺍﺳﺖ ﻛﻪ ﺑﻴﺸﺘﺮ ﻗﺎﺑﻞ ﺩﺳﺘﺮﺳـﻲ ﺍﺳـﺖ ﺯﻳـﺮﺍ ﺭﻓﺘـﺎﺭ‬
‫ﺣﻮﺍﺩﺙ ﺩﺭ ﻟﻮﻟﻪﻫﺎ ﺗﺤﺖ ﺗﺄﺛﻴﺮ ﭘﺎﺭﺍﻣﺘﺮﻫﺎﻱ ﻏﻴﺮ ﻣﺸﺨﺺ‪ ،‬ﻣـﺸﺘﺮﻙ ﻭ‬
‫ﭘﻴﭽﻴﺪﻩ ﺍﺳﺖ‪ .‬ﭼﻪ ﭘﺎﺭﺍﻣﺘﺮﻫﺎﻳﻲ ﮐﻪ ﺑﻪ ﺧﺼﻮﺻﻴﺎﺕ ﻟﻮﻟﻪ ﻭ ﻛﻴﻔﻴﺖ ﻧﺼﺐ‬
‫ﺁﻧﻬﺎ ﻣﺮﺗﺒﻂ ﻫﺴﺘﻨﺪ ﻭ ﭼﻪ ﺁﻧﻬﺎﻳﻲ ﮐﻪ ﺑﻪ ﻣﺤﻴﻂ ﭘﻴﺮﺍﻣﻮﻥ ﻟﻮﻟﻪ ﻭ ﺷـﺮﺍﻳﻂ‬
‫ﻋﻤﻠﻴﺎﺗﻲ ﺷﺒﮑﻪ ﻟﻮﻟﻪﻫﺎ ﻭﺍﺑﺴﺘﻪﺍﻧﺪ‪.‬‬
‫ﺩﺭ ﻳﻚ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ‪ :‬ﺩﺭ ﺭﺍﺳﺘﺎﻱ ﺍﻧﺠﺎﻡ ﺁﻧـﺎﻟﻴﺰ ﺟـﺎﻣﻊ‪ ،‬ﻻﺯﻡ ﺍﺳـﺖ ﻛـﻪ‬
‫ﻣﺸﺨﺼﺎﺕ ﻟﻮﻟﻪﻫﺎ ﻣﺎﻧﻨﺪ ﺟﻨﺲ‪ ،‬ﻗﻄﺮ‪ ،‬ﻃﻮﻝ‪ ،‬ﺳﻦ‪ ،‬ﻋﻤﻖ ﻧـﺼﺐ ﻟﻮﻟـﻪ ﻭ‬
‫ﺷﺮﺍﻳﻂ ﺑﻬﺮﻩﺑﺮﺩﺍﺭﻱ )ﻓﺸﺎﺭ ﻫﻴﺪﺭﻭﻟﻴﮑﻲ( ﮐﻪ ﺍﻃﻼﻋﺎﺕ ﺁﻥ ﺩﺭ ﺷـﺮﺍﻳﻂ‬
‫ﻣﻮﺟﻮﺩ ﮐﺸﻮﺭ ﻗﺎﺑـﻞ ﺩﺳـﺘﺮﺱ ﺑﺎﺷـﺪ ﺭﺍ ﺩﺭ ﻧﻈـﺮ ﮔﺮﻓـﺖ ﻭ ﻧـﺴﺒﺖ ﺑـﻪ‬
‫ﺟﻤﻊﺁﻭﺭﻱ ﻭ ﺫﺧﻴﺮﻩ ﺁﻥ ﺩﺭ ﻳﻚ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﺍﻗﺪﺍﻡ ﻧﻤﻮﺩ‪ .‬ﺩﺭ ﺍﻳﻦ ﺭﺍﺳﺘﺎ‬
‫ﺑﻪ ﺑﺮﺭﺳﻲ ﺗﻌﺪﺍﺩ ﺯﻳﺎﺩﻱ ﺍﺯ ﺷﻬﺮﻫﺎﻱ ﮐـﺸﻮﺭ ﮐـﻪ ﺍﺣﺘﻤـﺎﻝ ﻭﺟـﻮﺩ ﺁﻣـﺎﺭ‬
‫ﻣﻨﺎﺳﺐ ﺩﺭ ﺁﻧﻬﺎ ﻭﺟﻮﺩ ﺩﺍﺷﺖ‪ ،‬ﺍﻗﺪﺍﻡ ﮔﺮﺩﻳﺪ ﻭ ﺩﺭ ﻧﻬﺎﻳـﺖ ﻧﺎﺣﻴـﻪ ﭼﻬـﺎﺭ‬
‫ﺁﺑﻔﺎﻱ ﻣﺸﻬﺪ ﺑﻪﻋﻨﻮﺍﻥ ﻣﺤﺪﻭﺩﻩ ﻣﻄﺎﻟﻌﺎﺗﻲ ﺩﺭ ﺍﻳﻦ ﻣﺮﺣﻠﻪ ﺍﻧﺘﺨﺎﺏ ﺷﺪ‪.‬‬
‫‪ -۳‬ﺍﻧﺘﺨﺎﺏ ﺷﺎﺧﺺ ﻣﺒﻨﺎﻱ ﺁﻧﺎﻟﻴﺰ‪ :‬ﺍﺯ ﺑﻴﻦ ﺩﺍﺩﻩﻫﺎﻱ ﺟﻤﻊﺁﻭﺭﻱ ﺷﺪﻩ ﻭ‬
‫ﺑﺎ ﺗﻮﺟﻪ ﺑﻪ ﺳﻮﺍﺑﻖ ﻣﻄﺎﻟﻌـﺎﺕ‪ ،‬ﺷﺎﺧـﺼﻬﺎﻱ ﻣﺨﺘﻠﻔـﻲ ﺑـﻪﻋﻨـﻮﺍﻥ ﻣﺒﻨـﺎﻱ‬
‫ﺷﻤﺎﺭﻩ ‪ ٢‬ﺳﺎﻝ ‪١٣٩٠‬‬
 ‫‪ ۱۰۰۰۰‬ﮔﺎﻡ‪ ،‬ﺗﺤﺖ ﺁﻣﻮﺯﺵ ﻗﺮﺍﺭ ﮔﺮﻓﺘﻨﺪ‪.‬‬ ‫‪ -۱-۵‬ﻣﺪﻝ ﺷﺒﮑﻪ ﻋﺼﺒﻲ ﻣﺼﻨﻮﻋﻲ‬
‫ﺟﺪﻭﻝ ‪ ،۱‬ﺟﺰﺋﻴﺎﺕ ﺑﻌﻀﻲ ﺍﺯ ﻣﻬﻢﺗﺮﻳﻦ ﺳﺎﺧﺘﺎﺭ ﻣﺪﻝﻫﺎ ﺭﺍ ﺩﺭ ﻧﻪ ﺣﺎﻟﺖ‬ ‫ﺩﺭ ﺍﻳﻦ ﻣﻘﺎﻟﻪ ﺍﺯ ﺷﺒﻜﻪ ﻋﺼﺒﻲ ﻣﺼﻨﻮﻋﻲ ﺑﺎ ﺳﺎﺧﺘﺎﺭ ﭘﺮﺳﭙﺘﺮﻭﻥ ﭼﻨﺪ ﻻﻳﻪ‬
‫ﻧﺸﺎﻥ ﻣﻲﺩﻫﺪ‪ .‬ﺩﺭ ﻧﻬﺎﻳﺖ ﺑﺎ ﺗﻮﺟﻪ ﺑﻪ ﻣﻌﻴﺎﺭﻫﺎﻱ ﻣﻴـﺰﺍﻥ ﺧﻄـﺎ ﻭ ﻣـﺪﺕ‬ ‫ﭘﻴﺶﺧﻮﺭ ﺑﺎ ﺍﻧﺘﺸﺎﺭ ﺑﻪ ﻋﻘـﺐ ﺧﻄﺎﻫـﺎ ﺍﺳـﺘﻔﺎﺩﻩ ﺷـﺪ‪ .‬ﺗﻌـﺪﺍﺩ ﺩﺍﺩﻩﻫـﺎﻱ‬
‫ﺯﻣﺎﻥ ﺍﺟﺮﺍﻱ ﻣﺪﻝ‪ ،‬ﮔﺰﻳﻨـﻪ ‪ ۵‬ﺑـﺎ ‪ ۲۰۰۰‬ﻓﺮﺍﻳﻨـﺪ ﺁﻣـﻮﺯﺵ ﺑـﻪﻫﻤـﺮﺍﻩ ‪۵‬‬ ‫ﺟﻤﻊﺁﻭﺭﻱ ﺷﺪﻩ ﺑﺮﺍﻱ ﻟﻮﻟﻪﻫﺎﻱ ﺁﺯﺑﺴﺖ ﺩﺭ ﻣﻨﻄﻘﻪ ﻣﻮﺭﺩ ﻣﻄﺎﻟﻌـﻪ ‪٣٣٨‬‬
‫ﭘﺎﺭﺍﻣﺘﺮ ﻭﺭﻭﺩﻱ ﻭ ﻳﻚ ﭘﺎﺭﺍﻣﺘﺮ ﺧﺮﻭﺟﻲ ﺑﺮﮔﺰﻳﺪﻩ ﺷﺪ ﮐﻪ ﺩﺍﺭﺍﻱ ﺩﻭ ﻻﻳﻪ‬ ‫ﺭﮐﻮﺭﺩ ﺑﻮﺩﻩ ﮐﻪ ﺍﺯ ﺍﻳﻦ ﺗﻌـﺪﺍﺩ‪ ٨٠ ،‬ﺩﺭﺻـﺪ ﺑـﺮﺍﻱ ﺁﻣـﻮﺯﺵ ﺷـﺒﮑﻪ‪١٥ ،‬‬
‫ﭘﻨﻬﺎﻥ ﺑﻪﺗﺮﺗﻴﺐ ﺑﺎ ‪ ۵‬ﻭ ‪ ۱۰‬ﻧﺮﻭﻥ ﺑﻮﺩ ﻭ ﺗﻮﺍﺑﻊ ﻣﺤﺮﻙ ﻻﻳﻪ ﺍﻭﻝ ﻭ ﺩﻭﻡ‬ ‫ﺩﺭﺻﺪ ﺑﺮﺍﻱ ﺗﺴﺖ ﻭ ‪ ٥‬ﺩﺭﺻﺪ ﺑﺮﺍﻱ ﺻﺤﺖﺳﻨﺠﻲ ﻧﺘﺎﻳﺞ ﭘـﻴﺶ ﺑﻴﻨـﻲ‪،‬‬
‫ﺁﻥ ﺗﺎﻧﮋﺍﻧﺖ ﺳﻴﮕﻤﻮﺋﻴﺪ ﻭ ﺗﺎﺑﻊ ﻣﺤﺮﻙ ﻻﻳﻪ ﺧﺮﻭﺟﻲ‪ ،‬ﺧﻄﻲ ﺑﻮﺩ‪.‬‬ ‫ﻣﻮﺭﺩ ﺍﺳﺘﻔﺎﺩﻩ ﻗﺮﺍﺭ ﮔﺮﻓﺖ‪.‬‬
‫ﭘﺲ ﺍﺯ ﺗﻌﻴﻴﻦ ﺑﻬﺘﺮﻳﻦ ﺳﺎﺧﺘﺎﺭ ﻣﻤﮑـﻦ ﺑـﺮﺍﻱ ﺷـﺒﮑﻪ ﻋـﺼﺒﻲ ﺑﺎﻳـﺪ‬ ‫ﺑﺮﺍﻱ ﺍﻧﺘﺨﺎﺏ ﻣﺪﻝ ﻣﻨﺎﺳﺐ ﺑﺮﺍﻱ ﭘﻴﺶﺑﻴﻨﻲ ﻧﺮﺥ ﺷﮑﺴﺖ ﻟﻮﻟـﻪﻫـﺎ ﺑـﺮ‬
‫ﺻﺤﺖ ﻭ ﻫﻤﭽﻨﻴﻦ ﮐﺎﺭﺍﻳﻲ ﺁﻥ ﻧﻴﺰ ﺑﺮﺭﺳﻲ ﺷﻮﺩ‪ .‬ﺑﺮﺍﻱ ﺍﻳﻦ ﻣﻨﻈﻮﺭ ﻧﺘﺎﻳﺞ‬ ‫ﺍﺳﺎﺱ ﺷﺒﮑﻪ ﻋﺼﺒﻲ ﻣﺼﻨﻮﻋﻲ ﭼﻨﺪﻳﻦ ﻣﺪﻝ ﻣﺨﺘﻠﻒ ﺷـﺒﻜﻪ ﻋـﺼﺒﻲ ﺑـﺎ‬
‫ﺣـﺎﺻﻞ ﺍﺯ ﺗﺴﺖ ﻭ ﺻﺤﺖﺳﻨﺠﻲ ﺷﺒﮑﻪ ﻣﻨﺘﺨﺐ ﺩﺭ ﻣﻘﺎﻳﺴﻪ ﺑﺎ ﺩﺍﺩﻩﻫﺎﻱ‬ ‫ﺗﻌﺪﺍﺩ ﻻﻳﻪﻫﺎ‪ ،‬ﻧﺮﻭﻥﻫﺎ ﻭ ﺗﻮﺍﺑﻊ ﻣﺤﺮﻙ ﻣﺨﺘﻠـﻒ ﺍﻳﺠـﺎﺩ ﺷـﺪﻩ ﻭ ﻣـﻮﺭﺩ‬
‫ﺁﺯﻣﺎﻳﺶ ﻗﺮﺍﺭ ﮔﺮﻓﺘﻨﺪ‪ .‬ﻫﻤﭽﻨـﻴﻦ ﺑـﺮﺍﻱ ﺭﺳـﻴﺪﻥ ﺑـﻪ ﺑﻬﺘـﺮﻳﻦ ﺟـﻮﺍﺏ‪،‬‬
‫ﺷﺒﮑﻪﻫﺎﻱ ﻋﺼﺒﻲ ﻣﻨﺘﺨﺐ ﺑـﺎ ﺗﻌـﺪﺍﺩ ﮔﺎﻣﻬـﺎﻱ ﻣﺨﺘﻠـﻒ ﺑـﻴﻦ ‪ ۵۰۰‬ﺗـﺎ‬

‫ﺷﮑﻞ ‪ -٢‬ﮐﺮﻭﮐﻲ ﻣﺤﺪﻭﺩﻩﻫﺎ ﻭ ﺧﻄﻮﻁ ﻫﻢ ﻓﺸﺎﺭ ﺩﺭ ﻣﻨﻄﻘﻪ ﻣﻄﺎﻟﻌﺎﺗﻲ ]‪[١۶‬‬

‫ﺟﺪﻭﻝ ‪ -۱‬ﻣﺸﺨﺼﺎﺕ ﺳﺎﺧﺘﺎﺭﻫﺎﻱ ﻣﺨﺘﻠﻒ ﻣﺪﻝﻫﺎﻱ ﺷﺒﮑﻪ ﻋﺼﺒﻲ‬

‫ﺗﻌﺪﺍﺩ ﻧﺮﻭﻥﻫﺎ ﺩﺭ ﻫﺮ ﻳﮏ ﺍﺯ ﻻﻳﻪﻫﺎ‬ ‫ﺗﺎﺑﻊ ﻣﺤﺮﮎ ﺩﺭ ﻫﺮ ﻳﮏ ﺍﺯ ﻻﻳﻪ ﻫﺎ‬ ‫ﺷﻤﺎﺭﻩ‬
‫ﻻﻳﻪ ﺍﻭﻝ ﻻﻳﻪ ﺩﻭﻡ ﻻﻳﻪ ﺧﺮﻭﺟﻲ‬ ‫ﻻﻳﻪ ﺧﺮﻭﺟﻲ‬ ‫ﻻﻳﻪ ﺩﻭﻡ‬ ‫ﻻﻳﻪ ﺍﻭﻝ‬ ‫ﺣﺎﻟﺖ‬
‫‪١‬‬ ‫‪-‬‬ ‫‪٥‬‬ ‫ﺧﻄﻲ‬ ‫‪-‬‬ ‫ﺗﺎﻧﮋﺍﻧﺖ ﺳﻴﮕﻤﻮﺋﻴﺪ‬ ‫‪١‬‬
‫‪١‬‬ ‫‪-‬‬ ‫‪١٠‬‬ ‫ﺧﻄﻲ‬ ‫‪-‬‬ ‫ﺗﺎﻧﮋﺍﻧﺖ ﺳﻴﮕﻤﻮﺋﻴﺪ‬ ‫‪٢‬‬
‫‪١‬‬ ‫‪-‬‬ ‫‪١٥‬‬ ‫ﺧﻄﻲ‬ ‫‪-‬‬ ‫ﺗﺎﻧﮋﺍﻧﺖ ﺳﻴﮕﻤﻮﺋﻴﺪ‬ ‫‪٣‬‬
‫‪١‬‬ ‫‪٥‬‬ ‫‪٥‬‬ ‫ﺧﻄﻲ‬ ‫ﺗﺎﻧﮋﺍﻧﺖ ﺳﻴﮕﻤﻮﺋﻴﺪ ﺗﺎﻧﮋﺍﻧﺖ ﺳﻴﮕﻤﻮﺋﻴﺪ‬ ‫‪٤‬‬
‫‪١‬‬ ‫‪١٠‬‬ ‫‪٥‬‬ ‫ﺧﻄﻲ‬ ‫ﺗﺎﻧﮋﺍﻧﺖ ﺳﻴﮕﻤﻮﺋﻴﺪ ﺗﺎﻧﮋﺍﻧﺖ ﺳﻴﮕﻤﻮﺋﻴﺪ‬ ‫‪٥‬‬
‫‪١‬‬ ‫‪١٥‬‬ ‫‪٥‬‬ ‫ﺧﻄﻲ‬ ‫ﺗﺎﻧﮋﺍﻧﺖ ﺳﻴﮕﻤﻮﺋﻴﺪ ﺗﺎﻧﮋﺍﻧﺖ ﺳﻴﮕﻤﻮﺋﻴﺪ‬ ‫‪٦‬‬
‫‪١‬‬ ‫‪٥‬‬ ‫‪١٠‬‬ ‫ﺧﻄﻲ‬ ‫ﺗﺎﻧﮋﺍﻧﺖ ﺳﻴﮕﻤﻮﺋﻴﺪ ﺗﺎﻧﮋﺍﻧﺖ ﺳﻴﮕﻤﻮﺋﻴﺪ‬ ‫‪٧‬‬
‫‪١‬‬ ‫‪١٠‬‬ ‫‪١٠‬‬ ‫ﺧﻄﻲ‬ ‫ﺗﺎﻧﮋﺍﻧﺖ ﺳﻴﮕﻤﻮﺋﻴﺪ ﺗﺎﻧﮋﺍﻧﺖ ﺳﻴﮕﻤﻮﺋﻴﺪ‬ ‫‪٨‬‬
‫‪١‬‬ ‫‪١٥‬‬ ‫‪١٠‬‬ ‫ﺧﻄﻲ‬ ‫ﺗﺎﻧﮋﺍﻧﺖ ﺳﻴﮕﻤﻮﺋﻴﺪ ﺗﺎﻧﮋﺍﻧﺖ ﺳﻴﮕﻤﻮﺋﻴﺪ‬ ‫‪٩‬‬

‫ﺷﻤﺎﺭﻩ ‪ ٢‬ﺳﺎﻝ ‪١٣٩٠‬‬ ‫ﺁﺏ ﻭ ﻓﺎﺿﻼﺏ‬ ‫‪‬‬

 ‫ﺷﮑﻞ ‪ -٣‬ﻧﺘﺎﻳﺞ ﺷﺒﮑﻪ ﻋﺼﺒﻲ ﻣﺼﻨﻮﻋﻲ ﺑﺮﺍﻱ ﻣﻘﺎﺩﻳﺮ ﺷﺒﻴﻪﺳﺎﺯﻱ ﺷﺪﻩ ﻭ ﻭﺍﻗﻌﻲ ﺩﺭ ﺳﻪ ﻣﺮﺣﻠﻪ ﺁﻣﻮﺯﺵ‪ ،‬ﺗﺴﺖ ﻭ ﺻﺤﺖ ﺳﻨﺠﻲ‬

‫ﺷﮑﻞ ‪ -٤‬ﻣﻘﺎﻳﺴﻪ ﻧﺮﺥ ﺷﮑﺴﺖ ﭘﻴﺶﺑﻴﻨﻲ ﺷﺪﻩ ﺗﻮﺳﻂ ﺷﺒﮑﻪ ﻋﺼﺒﻲ ﻭ ﺩﺍﺩﻩﻫﺎﻱ ﻭﺍﻗﻌﻲ ]‪[١۶‬‬

‫ﻣـﻮﺭﺩ ﻣﻄﺎﻟﻌـﻪ‪ ،‬ﺑـﺮ ﺍﺳــﺎﺱ ﻗﻄـﺮ‪ ،‬ﻃـﻮﻝ‪ ،‬ﺳــﻦ‪ ،‬ﻋﻤـﻖ ﻧـﺼﺐ ﻭ ﻓــﺸﺎﺭ‬ ‫ﻣﺸﺎﻫﺪﻩﺍﻱ ﺑﺮﺍﻱ ﻣﺤﺎﺳﺒﻪ ﻧﺮﺥ ﺷﮑﺴﺖ ﻟﻮﻟﻪﻫﺎ ﺩﺭ ﺷﮑﻞ ‪ ٣‬ﻧﻤﺎﻳﺶ ﺩﺍﺩﻩ‬
‫ﻫﻴﺪﺭﻭﻟﻴﮑﻲ ﻣﻄﺎﺑﻖ ﺭﺍﺑﻄﻪ ﺯﻳﺮ ﺑﻪﺩﺳﺖ ﺁﻣﺪ‬ ‫ﺷﺪ‪ .‬ﺑﺮﺭﺳﻲﻫﺎ ﮐﺎﻣﻼﹰ ﻣﺸﺨﺺ ﻣﻲﮐﻨﺪ ﮐـﻪ ﻧﺘـﺎﻳﺞ‪ ،‬ﺑـﺴﻴﺎﺭ ﻧﺰﺩﻳـﮏ ﺑـﻪ‬
‫ﺩﺍﺩﻩﻫﺎﻱ ﻣﺸﺎﻫﺪﻩﺍﻱ ﺑﻮﺩﻩ ﻭ ﺩﺍﺭﺍﻱ ﺧﻄﺎﻫﺎﻱ ﺑﺴﻴﺎﺭ ﺍﻧﺪﮎ ﻭ ﻗﺎﺑﻞ ﻗﺒﻮﻟﻲ‬
‫‪ ‬‬
‫‪ ‬‬ ‫‪ ‬‬
‫‪ ‬‬ ‫‪‬‬
‫‪‬‬
‫‪ De Demax ‬‬ ‫‪‬‬
‫‪‬‬
‫‪‬‬
‫‪‬‬
‫‪‬‬
‫‪‬‬ ‫ﺑﻮﺩ‪ .‬ﺍﻳﻦ ﻣﻮﺿﻮﻉ‪ ،‬ﺁﻣﻮﺯﺵ ﻣﻨﺎﺳﺐ ﻭ ﺗﻮﺍﻧﺎﻳﻲ ﺷﺒﮑﻪ ﺭﺍ ﻧﺸﺎﻥ ﻣﻲ ﺩﻫﺪ‪.‬‬
‫‪ Lt Lt‬‬ ‫‪‬‬
‫ﺑــﻪﻣﻨﻈــﻮﺭ ﻣﻘﺎﻳــﺴﻪ ﻧﺘــﺎﻳﺞ ﺷــﺒﻴﻪﺳــﺎﺯﻱ ﺷــﺪﻩ ﻭ ﻣــﺸﺎﻫﺪﻩ ﺷــﺪﻩ‪ ،‬ﻧــﺮﺥ‬
‫‪ ‬‬ ‫‪ ‬‬ ‫‪‬‬ ‫‪‬‬ ‫‪‬‬ ‫‪‬‬ ‫‪‬‬ ‫‪‬‬
‫‪ ‬‬ ‫‪ ‬‬ ‫‪‬‬ ‫‪max ‬‬
‫‪‬‬ ‫‪‬‬ ‫‪‬‬
‫‪‬‬ ‫‪‬‬
‫‪BRl  Purelin w 3  tan sig w 2  tan sig w1   Pr Prmax   b1    b2    b3   BR max‬‬
‫‪ ‬‬
‫‪ ‬‬
‫‪ ‬‬
‫‪ ‬‬
‫‪‬‬
‫‪‬‬
‫‪ Hi Hi max ‬‬
‫‪‬‬
‫‪‬‬ ‫‪‬‬
‫‪‬‬
‫‪‬‬
‫‪‬‬
‫ﺷﮑﺴﺖﻫﺎﻱ ﺷﺒﻴﻪﺳﺎﺯﻱ ﺷﺪﻩ ﺑﺮﺍﻱ ﺷﺒﮑﻪ ﻋﺼﺒﻲ ﻣﻨﺘﺨـﺐ ﻭ ﺩﺍﺩﻩﻫـﺎﻱ‬
‫‪‬‬ ‫‪ ‬‬
‫‪ ‬‬
‫‪ ‬‬
‫‪ ‬‬
‫‪ ‬‬ ‫‪‬‬
‫‪ Ae Ae ‬‬
‫‪‬‬ ‫‪max ‬‬ ‫‪‬‬
‫‪‬‬ ‫‪‬‬
‫‪‬‬
‫‪‬‬ ‫‪‬‬
‫‪‬‬ ‫ﻣﺸﺎﻫﺪﻩﺍﻱ ﺑﺎ ﺷﻴﺐ ﺧـﻂ ‪ ١:١‬ﺩﺭ ﺷـﻜﻞ ‪ ٤‬ﺗﺮﺳـﻴﻢ ﺷـﺪ‪ .‬ﺑـﺎ ﺗﻮﺟـﻪ ﺑـﻪ‬
‫)‪(١٥‬‬ ‫ﺿﺮﻳﺐ ﻫﻤﺒﺴﺘﮕﻲ ﺑﺎﻻﻱ ﺩﺍﺩﻩﻫﺎ‪ ،‬ﻣﻲﺗﻮﺍﻥ ﮔﻔﺖ ﮐﻪ ﺷﺒﮑﻪ ﺑـﻪﺧـﻮﺑﻲ ﺍﺯ‬
‫ﻛﻪ ﺩﺭ ﺍﻳﻦ ﺭﺍﺑﻄﻪ‬ ‫ﭘﺲ ﻳﺎﺩﮔﻴﺮﻱ ﺑﺮﺁﻣﺪﻩ ﻭ ﻣﻲﺗﻮﺍﻧﺪ ﺧﻮﺩ ﺭﺍ ﺑﻪ ﺣﺎﻟﺘﻬﺎﻱ ﻣﺨﺘﻠـﻒ ﺗﻌﻤـﻴﻢ‬
‫ﻣﻘﺪﺍﺭ ‪ BRL‬ﻧﺮﺥ ﺷﮑـﺴﺖ ﻟﻮﻟـﻪ ‪ L‬ﺍﻡ‪ De ،‬ﻗﻄـﺮ‪ Lt ،‬ﻃـﻮﻝ‪ Pr ،‬ﻓـﺸﺎﺭ‬ ‫ﺩﻫﺪ‪ .‬ﺩﺭ ﻧﻬﺎﻳﺖ ﻣﻌﺎﺩﻟﻪ ﻧﺮﺥ ﺷﮑﺴﺖ ﻟﻮﻟﻪﻫﺎﻱ ﺁﺯﺑـﺴﺖ ﺑـﺮﺍﻱ ﻣﻨﻄﻘـﻪ‬
‫ﻫﻴـــــﺪﺭﻭﻟﻴﮑﻲ‪ Hi ،‬ﻋﻤـــــﻖ ﻧـــــﺼﺐ‪ Ae ،‬ﺳـــــﻦ ﻟﻮﻟـــــﻪ ‪ L‬ﺍﻡ ﻭ‬

‫‪‬‬ ‫ﺁﺏ ﻭ ﻓﺎﺿﻼﺏ‬ ‫ﺷﻤﺎﺭﻩ ‪ ٢‬ﺳﺎﻝ ‪١٣٩٠‬‬

‫‪ ۵‬ﭘــﺎﺭﺍﻣﺘﺮ ﻭﺭﻭﺩﻱ‪ ،‬ﻳــﮏ ﭘــﺎﺭﺍﻣﺘﺮ ﺧﺮﻭﺟــﻲ‪ ۲۷۰ ،‬ﺟﻔــﺖ ﺍﻃﻼﻋــﺎﺕ‬
‫ﻭﺭﻭﺩﻱ ﺑﺮﺍﻱ ﺁﻣﻮﺯﺵ‪ ۵۱ ،‬ﺟﻔﺖ ﺍﻃﻼﻋﺎﺕ ﻭﺭﻭﺩﻱ ﺑﺮﺍﻱ ﺗﺴﺖ‪۱۷ ،‬‬
‫ﺟﻔﺖ ﺍﻃﻼﻋﺎﺕ ﻭﺭﻭﺩﻱ ﺑﺮﺍﻱ ﺻـﺤﺖﺳـﻨﺠﻲ ﺍﺯ ﻧـﻮﻉ ‪،genfis1 FIS‬‬
‫ﺍﻧﺪﺍﺯﻩ ﮔﺎﻣﻬﺎ ﺑﻴﻦ ‪ ۱۰‬ﺗﺎ ‪ ،۱۵۰‬ﻧﻮﻉ ﺭﻭﺵ ﺑﻬﻴﻨﻪﺳﺎﺯﻱ‪ ،‬ﺗـﺎﺑﻊ ﻋـﻀﻮﻳﺖ‬
‫ﻫﻴﺒﺮﻳــﺪ ﻭ ﺍﻧــﻮﺍﻉ ﺗﻮﺍﺑــﻊ ﻋــﻀﻮﻳﺖ ﻣــﻮﺭﺩ ﺑﺮﺭﺳــﻲ ﺷــﺎﻣﻞ ‪gbellmf,‬‬
‫‪ . trimf, trapmf, gaussmf, gauss2mf‬ﻣﺸﺎﺑﻪ ﺭﻭﺵ ﻋﺼﺒﻲ ﺑـﺮﺍﻱ‬
‫ﺑﺮﺭﺳﻲ ﺟﻮﺍﺑﻬﺎ ﺍﺯ ﺭﻭﺍﺑﻂ ‪ ۱۲‬ﺗﺎ ‪ ۱۴‬ﺍﺳﺘﻔﺎﺩﻩ ﺷﺪ‪ .‬ﺑﺮﺍﻱ ﺍﻧﺘﺨـﺎﺏ ﻣـﺪﻝ‬
‫ﻣﻨﺎﺳﺐ ﻧﻴﺰ ﺑﺎ ﺗﻮﺟﻪ ﺑـﻪ ﺭﻭﺷـﻬﺎﻱ ﻣﺨﺘﻠـﻒ ﺍﺷـﺎﺭﻩ ﺷـﺪﻩ‪ ،‬ﻣﻘـﺪﺍﺭ ﺧﻄـﺎ‬
‫ﻣﺤﺎﺳﺒﻪ ﻭ ﺑﺎ ﻳﮑﺪﻳﮕﺮ ﻣﻘﺎﻳﺴﻪ ﮔﺮﺩﻳﺪ‪ .‬ﻧﺘـﺎﻳﺞ ﻣﺤﺎﺳـﺒﺎﺕ ﺩﺭ ﺟـﺪﻭﻝ ‪۲‬‬
‫ﺍﺭﺍﺋﻪ ﺷﺪﻩ ﺍﺳﺖ‪.‬‬
‫ﻫﻤﺎﻥﻃﻮﺭ ﮐﻪ ﺍﺯ ﺟﺪﻭﻝ ﻣﻼﺣﻈﻪ ﻣﻲﺷﻮﺩ ﺑﺎ ﻫـﺮ ﻳـﮏ ﺍﺯ ﺭﻭﺷـﻬﺎﻱ‬
‫ﻣﺤﺎﺳﺒﻪ ﺧﻄﺎ‪ ،‬ﺗﺎﺑﻊ ﻋﻀﻮﻳﺖ ‪ Gaussmf‬ﺑﻬﺘﺮﻳﻦ ﺟﻮﺍﺏ ﺭﺍ ﺩﺍﺩﻩ ﺍﺳـﺖ‪.‬‬
‫ﺩﺭ ﺍﺩﺍﻣﻪ ﺑﺮﺍﻱ ﺻﺤﺖ ﮐﺎﺭﺑﺮﺩ ﻭ ﮐﺎﺭﺍﻳﻲ‪ ،‬ﻣﺪﻝ ﻣﻮﺭﺩ ﺻﺤﺖﺳﻨﺠﻲ ﻗﺮﺍﺭ‬
‫ﮔـﺮﻓﺖ ﮐﻪ ﻧﺘـﺎﻳﺞ ﻫﺮ ﺳﻪ ﻣﺮﺣﻠـﻪ ﺩﺭ ﺷﮑﻞ ‪ ٥‬ﺍﺭﺍﺋﻪ ﺷﺪﻩ ﺍﺳﺖ‪ .‬ﺑﺮﺭﺳـﻲ‬
‫ﺳﺎﺧﺘﺎﺭ ﻣﺪﻝ ﺩﺭ ﻧﻈﺮ ﮔﺮﻓﺘﻪ ﺷﺪﻩ ﺑﺮﺍﻱ ﺍﻳﻦ ﺗﺤﻘﻴﻖ ﻋﺒﺎﺭﺕ ﺑﻮﺩ ﺍﺯ ﺗﻌﺪﺍﺩ‬
‫ﻣﺸﺨـﺼﺎﺕ ‪BRmax ، Aemax ، Hi max ، Prmax ، Lt max ، De max‬‬
‫ﻣﻘﺎﺩﻳﺮ ﺣﺪﺍﮐﺜﺮ ﻣﺮﺑﻮﻁ ﺑﻪ ﻟﻮﻟﻪﻫﺎﻱ ﺁﺯﺑـﺴﺖ ﺩﺭ ﺷـﺒﮑﻪ ﻣـﻮﺭﺩ ﻣﻄﺎﻟﻌـﻪ‬
‫ﺍﺳﺖ‪ .‬ﻫﻤﭽﻨـﻴﻦ ‪ w 3 ، w 2 ، w 1‬ﻣـﺎﺗﺮﻳﺲﻫـﺎﻱ ﻭﺯﻥ ﻭ ‪b 3 ، b 2 ، b 1‬‬
‫ﻣﺎﺗﺮﻳﺲ ﻣﻘﺎﺩﻳﺮ ﺛﺎﺑﺖ ﺑﻪﺩﺳﺖ ﺁﻣﺪﻩ ﺍﺯ ﺷﺒﮑﻪ ﻋﺼﺒﻲ ﺑﺮﺍﻱ ﻻﻳﻪ ﻣﺨﺘﻠﻒ‬
‫ﺷﺒﮑﻪ ﻋﺼﺒﻲ ﻣﻨﺘﺨﺐ ﻫﺴﺘﻨﺪ ]‪.[١۱‬‬
‫‪ -۲-۵‬ﻣﺪﻝ ﺳﻴﺴﺘﻢ ﻋﺼﺒﻲ‪ -‬ﻓﺎﺯﻱ‬
‫ﻣﻄﺎﺑﻖ ﺭﻭﺵ ﺷﺒﮑﻪ ﻋﺼﺒﻲ‪ ،‬ﺳﻴﺴﺘﻢ ‪ ANFIS‬ﻃﺮﺍﺣﻲ ﺷﺪﻩ ﻧﻴﺰ ﻣﺘﺸﮑﻞ‬
‫ﺍﺯ ﭘﻨﺞ ﻭﺭﻭﺩﻱ ﺑﻪ ﺷﺒﮑﻪ ﺷﺎﻣﻞ ﻗﻄﺮ‪ ،‬ﻃﻮﻝ‪ ،‬ﺳﻦ‪ ،‬ﻋﻤﻖ ﻧـﺼﺐ ﻭ ﻓـﺸﺎﺭ‬
‫ﻫﻴﺪﺭﻭﻟﻴﮑﻲ ﺍﺯ ﻟﻮﻟﻪﻫـﺎﻱ ﺁﺯﺑـﺴﺖ ﺑـﺎ ﻗﻄـﺮ ‪ ۸۰‬ﺗـﺎ ‪ ۳۰۰‬ﻣﻴﻠـﻲﻣﺘـﺮ ﻭ‬
‫ﺧﺮﻭﺟــﻲ ﺷــﺎﻣﻞ ﻧــﺮﺥ ﺷﮑــﺴﺖ ﺩﺭ ﻫــﺮ ﻟﻮﻟــﻪ ﺑــﻮﺩ‪ .‬ﺗﻌــﺪﺍﺩ ﺩﺍﺩﻩﻫــﺎﻱ‬
‫ﺟﻤﻊﺁﻭﺭﻱ ﺷﺪﻩ ﻫﻤﺎﻥ ﺩﺍﺩﻩﻫﺎ ﺑﺮﺍﻱ ﺭﻭﺵ ﭘﻴﺸﻴﻦ ﺑﻮﺩ ﮐﻪ ‪ ۸۰‬ﺩﺭﺻﺪ ﺍﺯ‬
‫ﺁﻧﻬﺎ ﺑﺮﺍﻱ ﺁﻣﻮﺯﺵ ﺷﺒﮑﻪ‪۱۵ ،‬ﺩﺭﺻﺪ ﺁﻥ ﺑﺮﺍﻱ ﺗﺴﺖ ﻭ ‪ ۵‬ﺩﺭﺻﺪ ﺑـﺮﺍﻱ‬
‫ﺻﺤﺖﺳﻨﺠﻲ ﻧﺘﺎﻳﺞ ﭘﻴﺶﺑﻴﻨﻲ ﻣﻮﺭﺩ ﺍﺳﺘﻔﺎﺩﻩ ﻗﺮﺍﺭ ﮔﺮﻓﺖ‪ .‬ﻣﺸﺨـﺼﺎﺕ‬

‫ﺟﺪﻭﻝ ‪ -٢‬ﻣﻘﺎﺩﻳﺮ ﻣﺘﻮﺳﻂ ﻣﺮﺑﻌﺎﺕ ﺧﻄﺎﻱ ﺁﻣﻮﺯﺵ‪ ،‬ﺗﺴﺖ ﻭ ﺻﺤﺖ ﺳﻨﺠﻲ ﻣﺪﻝ ﺷﺒﮑﻪ ﻋﺼﺒﻲ‪ -‬ﻓﺎﺯﻱ ﺑﺎ ﺍﻧﻮﺍﻉ ﺗﻮﺍﺑﻊ ﻋﻀﻮﻳﺖ‬
‫ﻣﺮﺣﻠﻪ ﺻﺤﺖ ﺳﻨﺠﻲ ﺷﺒﮑﻪ‬ ‫ﻣﺮﺣﻠﻪ ﺗﺴﺖ ﺷﺒﮑﻪ‬ ‫ﻣﺮﺣﻠﻪ ﺁﻣﻮﺯﺵ ﺷﺒﮑﻪ‬ ‫ﻧﻮﻉ ﺗﺎﺑﻊ‬
‫‪RMSE‬‬ ‫‪MSE‬‬ ‫‪IOA‬‬ ‫‪RMSE‬‬ ‫‪MSE‬‬ ‫‪IOA‬‬ ‫‪RMSE‬‬ ‫‪MSE‬‬ ‫‪IOA‬‬ ‫ﻋﻀﻮﻳﺖ‬
‫‪٠/٠٥٠٢٨٨‬‬ ‫‪٠/٠٠٢٥٢٩ ٠/٩٨٨‬‬ ‫‪٠/٠٢٣٠٥١‬‬ ‫‪٠/٠٠٠٥٣١‬‬ ‫‪٠/٩٩٧٩٩ ٠/٠٢٤٦٤٢‬‬ ‫‪٠/٠٠٠٦٠٧‬‬ ‫‪٠/٩٩٦٨٢ Gaussmf‬‬
‫‪٠/٠٣٢١٤٨‬‬ ‫‪٠/٠٠١٠٣٤ ٠/٩٩١٣١‬‬ ‫‪٠/٠٢٤١٨١‬‬ ‫‪٠/٠٠٠٥٨٥‬‬ ‫‪٠/٩٩٧٠٤ ٠/٠٢٩٢٩٧‬‬ ‫‪٠/٠٠٠٨٥٨‬‬ ‫‪٠/٩٩٧٦٤ Gauss2mf‬‬
‫‪٠/٠٤٦٨٥٨‬‬ ‫‪٠/٠٠٢١٩٦ ٠/٩٨٧١٩‬‬ ‫‪٠/٠٣٦٥٣٣‬‬ ‫‪٠/٠٠١٣٣٥‬‬ ‫‪٠/٨٨٤٩١ ٠/٠٢٦٣٥‬‬ ‫‪٠/٠٠٠٦٩٤‬‬ ‫‪٠/٩٩٧٣٢ Gbellmf‬‬
‫‪٠/٠٥٧٨٣٣‬‬ ‫‪٠/٠٠٣٣٤٥ /٩٩٥٥١‬‬ ‫‪٠/٠٢٠٧٨٥‬‬ ‫‪٠/٠٠٠٤٣٢‬‬ ‫‪٠/٩٩٨٠٢ ٠/٠٢٥٦٧٣‬‬ ‫‪٠/٠٠٠٦٥٩‬‬ ‫‪٠/٩٩٧٥٥‬‬ ‫‪Trimf‬‬

‫ﺷﻜﻞ ‪ -۵‬ﻧﺘﺎﻳﺞ ﺷﺒﮑﻪ ﻋﺼﺒﻲ ‪ -‬ﻓﺎﺯﻱ ﺑﺮﺍﻱ ﻣﻘﺎﺩﻳﺮ ﺷﺒﻴﻪﺳﺎﺯﻱ ﺷﺪﻩ ﻭ ﻭﺍﻗﻌﻲ ﺩﺭ ﺳﻪ ﻣﺮﺣﻠﻪ ﺁﻣﻮﺯﺵ‪ ،‬ﺗﺴﺖ ﻭ ﺻﺤﺖ ﺳﻨﺠﻲ‬

‫ﺷﻤﺎﺭﻩ ‪ ٢‬ﺳﺎﻝ ‪١٣٩٠‬‬ ‫ﺁﺏ ﻭ ﻓﺎﺿﻼﺏ‬ ‫‪‬‬

 ‫ﺷﮑﻞ ‪ -٦‬ﻧﺘﺎﻳﺞ ﺷﺒﮑﻪ ﻋﺼﺒﻲ‪ -‬ﻓﺎﺯﻱ ﺑﺎ ﺭﻭﺷﻬﺎﻱ ﻣﺨﺘﻠﻒ ﻣﺤﺎﺳﺒﻪ ﺧﻄﺎ‬

‫‪ NB‬ﺗﻌﺪﺍﺩ ﺣﻮﺍﺩﺙ ﺩﺭ ﺳﺎﻝ ﺁﻳﻨﺪﻩ ﻭ ﺗﺮﻡﻫﺎﻱ ﺳـﻤﺒﻠﻴﻚ ‪،Ae ،De ،Lt‬‬ ‫ﺷﻜﻠﻬﺎ ﻭ ﻣﻘﺎﻳﺴﻪ ﺑﻴﻦ ﻣﻘﺎﺩﻳﺮ ﻣﺸﺎﻫﺪﻩ ﺷﺪﻩ ﻭ ﺷﺒﻴﻪﺳﺎﺯﻱ ﺷﺪﻩ ﺩﺭ ﻫﺮ ﺳﻪ‬
‫‪ Pr‬ﻭ ‪ Hi‬ﺑﻪﺗﺮﺗﻴﺐ ﺷﺎﺧﺼﻬﺎﻱ ﻃﻮﻝ ﮐﻠﻲ‪ ،‬ﻗﻄﺮ‪ ،‬ﺳﻦ‪ ،‬ﻓﺸﺎﺭ ﻣﺘﻮﺳـﻂ ﻭ‬ ‫ﻣﺮﺣﻠﻪ ﺁﻣﻮﺯﺵ‪ ،‬ﺗﺴﺖ ﻭ ﺻﺤﺖﺳﻨﺠﻲ ﺷﺒﮑﻪ ﻧﺸﺎﻥ ﺍﺯ ﻣﻨﺎﺳـﺐ ﺑـﻮﺩﻥ‬
‫ﻋﻤﻖ ﻧﺼﺐ ﻟﻮﻟﻪ ﻣﻌﺎﺩﻝ ﻫﺮ ﮐﻼﺱ ﻫﺴﺘﻨﺪ‪ .‬ﻭﺭﻭﺩﻱﻫﺎﻱ ﺍﻧﺘﺨﺎﺏ ﺷـﺪﻩ‬ ‫ﻣــﺪﻝ ﺍﻧﺘﺨــﺎﺑﻲ ﺑــﺮﺍﻱ ﻣﺤــﺪﻭﺩﻩ ﻣــﻮﺭﺩ ﻧﻈــﺮ ﺩﺍﺭﺩ‪ .‬ﻫﻤﭽﻨــﻴﻦ ﻧﺘــﺎﻳﺞ‬
‫ﺩﺭﺳﺎﺧﺘﺎﺭ ﺍﻳﻦ ﭼﻨﺪ ﺟﻤﻠﻪﺍﻱ ﺩﺍﺭﺍﻱ ﻳﻚ ﺗﻮﺍﻥ ﺻﺤﻴﺢ ﻫـﺴﺘﻨﺪ‪ .‬ﺑـﺮﺍﻱ‬ ‫ﺑﻪﮐﺎﺭﮔﻴﺮﻱ ﺷﺒﮑﻪﻫﺎﻱ ﻋﺼﺒﻲ‪ -‬ﻓﺎﺯﻱ ﻣﻌﺮﻓﻲ ﺷﺪﻩ ﺩﺭ ﭘﻴﺶﺑﻴﻨـﻲ ﻧـﺮﺥ‬
‫ﭘﺬﻳﺮﺵ ﻣﺪﻝﻫﺎﻱ ﭘﻴﺶ ﺑﻴﻨـﻲ ﺍﺯ ﺑـﻴﻦ ‪ ٢٤٠‬ﻣـﺪﻝ ﺗﻮﻟﻴـﺪ ﺷـﺪﻩ‪ ،‬ﻣﻌﻴـﺎﺭ‬ ‫ﺷﮑﺴﺖ ﻟﻮﻟﻪﻫﺎﻱ ﺷـﺒﮑﻪ ﺁﺑﺮﺳـﺎﻧﻲ‪ ،‬ﺑـﺎ ﺍﺳـﺘﻔﺎﺩﻩ ﺍﺯ ﺭﻭﺷـﻬﺎﻱ ﻣﺨﺘﻠـﻒ‬
‫ﺑﻴﺸﻴﻨﻪ ﻭ ﻣﻴﺎﻧﮕﻴﻦ ﺧﻄﺎ‪ ،‬ﺩﺭ ﻛﻨﺎﺭ ﺷﺎﺧﺺ ﺁﻣﺎﺭﻱ ﺿﺮﻳﺐ ﺗﻌﻴﻴﻦ‪ ٢‬ﺑﻪﻛﺎﺭ‬ ‫ﻣﺤﺎﺳﺒﻪ ﺧﻄﺎ ﺑﺎ ﻳﮑﺪﻳﮕﺮ ﻣﻘﺎﻳﺴﻪ ﺷﺪ )ﺷﻜﻞ ‪ .(٦‬ﺍﺯ ﺑﺮﺭﺳـﻲ ﻫـﺎ ﮐـﺎﻣﻼﹰ‬
‫ﮔﺮﻓﺘﻪ ﻣﻲﺷـﻮﺩ ﺗـﺎ ﻛـﺎﺭﺍﻳﻲ ﻫـﺮ ﻣـﺪﻝ ﺍﺭﺯﻳـﺎﺑﻲ ﮔـﺮﺩﺩ‪ .‬ﺿـﺮﻳﺐ ‪CoD‬‬ ‫ﻣﺸﺨﺺ ﺍﺳﺖ ﮐﻪ ﻧﺘﺎﻳﺞ‪ ،‬ﺑـﺴﻴﺎﺭ ﻧﺰﺩﻳـﮏ ﺑـﻪ ﺩﺍﺩﻩﻫـﺎﻱ ﻣـﺸﺎﻫﺪﺍﺗﻲ ﻭ‬
‫ﺑﻪﻋﻨﻮﺍﻥ ﻋﻤﺪﻩﺗﺮﻳﻦ ﺷﺎﺧﺺ ﮐﺎﺭﺍﻳﻲ ﻣﺪﻝ ﺑﻪﺻﻮﺭﺕ ﺭﺍﺑﻄﻪ ‪ ١٧‬ﻣﺤﺎﺳﺒﻪ‬ ‫ﺩﺍﺭﺍﻱ ﺧﻄﺎﻫﺎﻱ ﺑﺴﻴﺎﺭ ﺍﻧﺪﮎ ﻭ ﻗﺎﺑﻞ ﻗﺒﻮﻝ ﻭ ﻣﻄﻠﻮﺑﻲ ﺑﻮﺩﻧﺪ‪.‬‬
‫ﻣﻲﺷﻮﺩ‬
‫‪١‬‬
‫‪)2‬‬ ‫‪ -۳-۵‬ﻣﺪﻝ ﺭﮔﺮﺳﻴﻮﻧﻲ ﭼﻨﺪ ﺟﻤﻠﻪﺍﻱ ﺗﮑﺎﻣﻠﻲ‬
‫‪N 1 ‬‬
‫‪( NB  NB‬‬ ‫‪exp‬‬
‫‪CoD  1 ‬‬ ‫‪‬‬ ‫‪N‬‬
‫*‬ ‫‪2‬‬
‫)‪(١٧‬‬ ‫ﺑﺎ ﭘﻬﻨﻪ ﺑﻨﺪﻱ ﻣﻨﻄﻘﻪ ﻣﻄﺎﻟﻌﺎﺗﻲ‪ ،‬ﻛﻞ ﻣﺠﻤﻮﻋﻪ ﺩﺍﺩﻩﻫﺎﻱ ﺍﺳﺘﺨﺮﺍﺝ ﺷﺪﻩ ﺍﺯ‬
‫‪N  ( NB NB‬‬ ‫)‬
‫‪N‬‬
‫‪exp‬‬ ‫‪exp‬‬
‫ﻣﺸﺨﺼﺎﺕ ﺷﺒﮑﻪ ﻭ ﺁﻣﺎﺭ ﺣﻮﺍﺩﺙ ﺑﻪ ﭼﻨﺪﻳﻦ ﻣﺤﺪﻭﺩﻩ ﺗﻘﺴﻴﻢ ﻣﻲﺷﻮﺩ‪.‬‬
‫ﺩﺭ ﻫﻤﻴﻦ ﺭﺍﺳـﺘﺎ ﻫـﺮ ﻳـﮏ ﺍﺯ ﮐـﻼﺱﻫـﺎﻱ ﻗﻄـﺮﻱ ﺍﺻـﻠﻲ ﺷـﺶﮔﺎﻧـﻪ‬
‫ﻛﻪ ﺩﺭ ﺁﻥ‬
‫)‪ ٨٠‬ﺗﺎ ‪ ٣٠٠‬ﻣﻴﻠﻲﻣﺘﺮ( ﺩﺭ ﭘﻬﻨﻪﺑﻨﺪﻱﻫﺎ ﺑﺎ ﻟﻮﻟـﻪﻫـﺎﻱ ﻓﺮﺿـﻲ ﻣﻌـﺎﺩﻝ‪،‬‬
‫‪ N‬ﺗﻌﺪﺍﺩ ﭘﺮﺩﺍﺯﺵ ﺩﺍﺩﻩﻫﺎ‪ NB ،‬ﺗﻌﺪﺍﺩ ﺷﮑـﺴﺖﻫـﺎﻱ ﭘـﻴﺶ ﺑﻴﻨـﻲ ﺷـﺪﻩ‬
‫ﻣﺪﻝ ﺷﺪﻧﺪ‪ .‬ﺑـﺎ ﺗﻮﺟـﻪ ﺑـﻪ ﭘﺎﺭﺍﻣﺘﺮﻫـﺎﻱ ﺍﻧﺘﺨـﺎﺏ ﺷـﺪﻩ‪ ،‬ﺳـﺎﺧﺘﺎﺭ ﻛﻠـﻲ‬
‫ﺗﻮﺳﻂ ﻣﺪﻝ‪ NBexp ،‬ﺗﻌﺪﺍﺩ ﺷﮑﺴﺖﻫﺎﻱ ﺛﺒﺖ ﺷﺪﻩ ﺑـﺮﺍﻱ ﻫـﺮ ﮐـﻼﺱ‬
‫ﭼﻨﺪﺟﻤﻠﻪﺍﻱ ﺯﻳﺮ ﺑﺎ ﺍﻟﮕﻮﻱ ﺭﮔﺮﺳﻴﻮﻧﻲ ﺗﮑﺎﻣﻠﻲ‪ ،‬ﺍﺭﺍﺋﻪ ﺷﺪ‬
‫ﻟﻮﻟﻪ ﻭ ‪ NB*exp‬ﻣﻴﺎﻧﮕﻴﻦ ﺗﻌﺪﺍﺩ ﺷﮑـﺴﺖﻫـﺎﻱ ﺛﺒـﺖ ﺷـﺪﻩ ﺑـﺮﺍﻱ ﻫـﺮ‬
‫‪‬‬ ‫‪‬‬ ‫‪‬‬ ‫‪ ‬‬
‫ﮐﻼﺱ ﺍﺳﺖ‪.‬‬ ‫‪NB  Ae  Lt  De  Pr .Hi  a 0‬‬ ‫)‪( ١٦‬‬
‫ﺍﺯ ﺑﻴـﻦ ﺗﻤـﺎﻡ ﻣـﺪﻝﻫﺎﻱ ﭘﻴﺶﺑﻴﻨـﻲ ﺷﻜﺴﺖ ﻛﻪ ﺑــﺮﺍﻱ ﻣﺤـﺪﻭﺩﻩﻫﺎﻱ‬ ‫ﻛﻪ ﺩﺭ ﺍﻳﻦ ﺭﺍﺑﻄﻪ‬

‫‪2‬‬ ‫‪1‬‬
‫‪Coefficient of Determination‬‬ ‫)‪Evolutionary Polynomial Regression (EPR‬‬

‫‪‬‬ ‫ﺁﺏ ﻭ ﻓﺎﺿﻼﺏ‬ ‫ﺷﻤﺎﺭﻩ ‪ ٢‬ﺳﺎﻝ ‪١٣٩٠‬‬

‫ﻋﺼﺒﻲ‪ -‬ﻓﺎﺯﻱ ﺩﺭ ﺷﮑﻞ ‪ ٩‬ﺍﺭﺍﺋﻪ ﺷﺪﻩ ﺍﺳﺖ‪ .‬ﺑـﺎ ﺗﻮﺟـﻪ ﺑـﻪ ﻧﻤﻮﺩﺍﺭﻫـﺎﻱ‬ ‫ﻣﻨﻄﻘﻪ ﻣﻄﺎﻟﻌﺎﺗﻲ ﺗﺄﻳﻴﺪ ﮔﺮﺩﻳﺪﻩ ﺑـﻮﺩ‪ ،‬ﻣـﺪﻝ ﭘـﻴﺶﺑﻴﻨـﻲ ﺷﻜـﺴﺖ ﺯﻳـﺮ‬
‫ﺷﮑﻞ ‪ ٩‬ﻭ ﻣﻨﺤﻨﻲ ﻧﺮﺥ ﺷﮑﺴﺖ ﻭﺍﻗﻌﻲ‪ ،‬ﻣﻮﺍﺭﺩ ﺯﻳﺮ ﻣﺸﺎﻫﺪﻩ ﻣﻲﺷﻮﺩ‪:‬‬ ‫ﺍﻧﺘﺨﺎﺏ ﮔﺮﺩﻳﺪ‪ .‬ﺳﭙﺲ ﺑﺎ ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ ﻃﻮﻝ ﻣﺠﻤﻮﻉ ﻟﻮﻟﻪﻫﺎﻱ ﻫﺮ ﮐﻼﺱ‬
‫ﻗﻄﺮﻱ‪ ،‬ﻧﺮﺥ ﺷﮑﺴﺖ ﻟﻮﻟﻪﻫﺎﻱ ﻣﻨﻄﻘﻪ ﻣﺤﺎﺳﺒﻪ ﮔﺮﺩﻳﺪ‬
‫)‪(۱۸‬‬
‫) ‪BR  round ( 0.097846  De 1  Lt  0.0012545  Ae 2 .Pr 2‬‬

‫ﻛﻪ ﺩﺭ ﺍﻳﻦ ﺭﺍﺑﻄﻪ‬

‫‪ Ae‬ﺳﻦ ﻟﻮﻟﻪ‪ Lt ،‬ﻃﻮﻝ ﻟﻮﻟﻪ‪ De ،‬ﻗﻄﺮ ﻟﻮﻟﻪ ﻭ ‪ Pr‬ﻓﺸﺎﺭ ﺍﺳﺖ‪.‬‬
‫ﺑﺮﺍﺳﺎﺱ ﺭﺍﺑﻄﻪ ‪ ،۱۸‬ﺳﻦ ﻟﻮﻟﻪ ﺗﻨﻬﺎ ﻓﺎﻛﺘﻮﺭﻱ ﻧﻴﺴﺖ ﻛﻪ ﺑﻪﻃﻮﺭ ﻋﻤﺪﻩ ﺑﺮ‬
‫ﺧﺮﺍﺑﻲ ﻭ ﺯﻭﺍﻝ ﻟﻮﻟﻪﻫﺎ ﺗﺄﺛﻴﺮ ﻣﻲﮔﺬﺍﺭﺩ‪ ،‬ﺑﻠﻜﻪ ﻃﻮﻝ ﻟﻮﻟﻪ‪ ،‬ﻗﻄﺮ ﻭ ﻓﺸﺎﺭ ﻧﻴﺰ‬
‫ﺩﺭ ﺗﻌﻴﻴﻦ ﺷﻜﺴﺖﻫﺎﻱ ﻳﻚ ﻟﻮﻟﻪ ﻣـﺆﺛﺮ ﻫـﺴﺘﻨﺪ ﻭ ﺩﻳﮕـﺮ ﭘﺎﺭﺍﻣﺘﺮﻫـﺎﻱ‬
‫ﺛﺒﺖ ﺷﺪﻩ‪ ،‬ﺩﺭ ﺷﮑﺴﺖ ﻟﻮﻟﻪﻫـﺎﻱ ﻣﻨﻄﻘـﻪ ﺑـﺪﻭﻥ ﺗﺄﺛﻴﺮﺍﻧـﺪ‪ .‬ﺩﺭ ﺷـﮑﻞ ‪۷‬‬
‫ﺩﺍﺩﻩﻫﺎﻱ ﭘﻴﺶﺑﻴﻨﻲ ﺷـﺪﻩ ﺗﻮﺳـﻂ ﻣـﺪﻝ ﺑﺮﮔﺰﻳـﺪﻩ ﺩﺭ ﻣﻘﺎﺑـﻞ ﺩﺍﺩﻩﻫـﺎﻱ‬
‫ﺷﮑﻞ ‪ -٧‬ﻧﻤﺎﻳﺶ ﭘﻴﺶﺑﻴﻨﻲ ﺷﻜﺴﺖ ﺭﮔﺮﺳﻴﻮﻥ ﭼﻨﺪ ﺟﻤﻠﻪﺍﻱ ﺗﻜﺎﻣﻠﻲ‬ ‫ﻭﺍﻗﻌﻲ ﺛﺒﺖ ﺷﺪﻩ ﺷﮑﺴﺖ‪ ،‬ﺗﺮﺳﻴﻢ ﺷﺪﻩﺍﻧﺪ‪ .‬ﺩﺭ ﻋﺒﺎﺭﺗﻬﺎﻱ ﭼﻨﺪ ﺟﻤﻠﻪﺍﻱ‪،‬‬
‫ﺩﺭ ﻣﻘﺎﺑﻞ ﺷﻜﺴﺖﻫﺎﻱ ﺛﺒﺖ ﺷﺪﻩ‬ ‫ﻓﺮﻣﻮﻝﻫﺎﻳﻲ ﮐﻪ ﭘﺎﺭﺍﻣﺘﺮﻫﺎﻳﻲ ﺑﺎ ﺗﺄﺛﻴﺮ ﺷﻨﺎﺧﺘﻪ ﺷﺪﻩ ﺩﺍﺭﻧﺪ ﻭ ﺍﺯ ﺿـﺮﻳﺐ‬
‫‪ CoD‬ﻣﻨﺎﺳـﺐ ) ﺑــﺎﻻﺗﺮ ﺍﺯ ‪ (۰/۸۵‬ﺑﺮﺧﻮﺭﺩﺍﺭﻧـﺪ‪ ،‬ﺑﺮﮔﺰﻳــﺪﻩ ﻣـﻲﺷــﻮﻧﺪ‬
‫]‪ .[۱۲‬ﺷﻜﻞ ‪ ۸‬ﻣﻘﺪﺍﺭ ﻣﺤﺎﺳﺒﻪ ﺷـﺪﻩ ﺿـﺮﻳﺐ ‪ COD‬ﺩﺭ ﺑﺮﺍﺑـﺮ ﺷـﻤﺎﺭﻩ‬
‫ﺭﻭﻧﺪ ﺗﻮﻟﻴﺪ ﺷﺪﻩ ﺭﺍ ﻧﺸﺎﻥ ﻣﻲﺩﻫﺪ‪.‬‬
‫‪COD‬‬

‫‪ -۴-۵‬ﻣﺪﻝ ﺭﮔﺮﺳﻴﻮﻥ ﭼﻨﺪ ﻣﺘﻐﻴﺮﻩ ﻏﻴﺮﺧﻄﻲ‬

‫ﻣﺪﻝﺳﺎﺯﻱ ﺭﮔﺮﺳﻴﻮﻧﻲ ﭼﻨﺪ ﻣﺘﻐﻴـﺮﻩ ﺑـﺮﺍﻱ ﭘـﻴﺶ ﺑﻴﻨـﻲ ﻧـﺮﺥ ﺷﻜـﺴﺖ‬
‫ﻟﻮﻟﻪﻫﺎﻱ ﺷﺒﮑﻪ ﺗﻮﺯﻳﻊ ﺁﺏ ﺩﺭ ﻣﻨﻄﻘﻪ ﻣﻮﺭﺩ ﻣﻄﺎﻟﻌﻪ ﻭ ﺑﺮﺍﺳﺎﺱ ﺩﺍﺩﻩﻫﺎﻱ‬
‫ﺟﻤﻊﺁﻭﺭﻱ ﺷﺪﻩ ﺍﺯ ﻃﺮﻳﻖ ﻧﻮﺷﺘﻦ ﮐﺪ ‪ M-file‬ﺩﺭ ﻧـﺮﻡ ﺍﻓـﺰﺍﺭ ‪Matlab‬‬
‫‪ 7.04‬ﺍﻧﺠﺎﻡ ﮔﺮﻓﺖ‪ .‬ﺩﺭ ﻧﻬﺎﻳﺖ ﺭﺍﺑﻄﻪ ﻧﺮﺥ ﺷﮑﺴﺖ ﻟﻮﻟﻪﻫـﺎﻱ ﺁﺯﺑـﺴﺖ‬
‫ﺷﻜﻞ ‪ -٨‬ﺷﺎﺧﺺ ﺑﺮﺍﺯﺵ ‪ COD‬ﺩﺭ ﻃﻲ ﻓﺮ ﺍﻳﻨﺪ ﺗﻮﻟﻴﺪ‬
‫ﺑﺮﺍﻱ ﻣﻨﻄﻘﻪ ﻣﻮﺭﺩ ﻣﻄﺎﻟﻌﻪ ﺑﻪﺻﻮﺭﺕ ﺯﻳﺮ ﺑﻪﺩﺳﺖ ﺁﻣﺪ‬
‫) ‪BR   1 .1775 ( De 0 .16287 )  0 .0076 ( Lt 0.2668 )  2 .7178 (Pr 0.169‬‬
‫‪ -۱‬ﺟﻮﺍﺑﻬﺎﻱ ﺭﻭﺵ ‪ ANFIS‬ﺑﺮﺍﻱ ﺳﺮﻱ ﺩﺍﺩﻩﻫﺎﻱ ﺷـﻬﺮ ﻣـﺸﻬﺪ ﺩﺍﺭﺍﻱ‬
‫) ‪ 2.271( Hi 0.0196 )  0.1381( Ae 0.00028‬‬
‫ﻧﻮﺳﺎﻧﺎﺕ ﭘﻲ ﺩﺭ ﭘﻲ ﺍﺳﺖ ﻭ ﺗﻔﺎﻭﺕ ﻣﻘﺪﺍﺭ ﻧﺮﺥ ﺷﮑﺴﺖ ﻣﺤﺎﺳﺒﻪ ﺷـﺪﻩ‬ ‫)‪(١٩‬‬
‫ﺩﺭ ﻣﺪﻝ ‪ ANFIS‬ﺑﺎ ﻣﻘﺪﺍﺭ ﺣﻘﻴﻘﻲ ﺩﺭ ﺑﺴﻴﺎﺭﻱ ﻣﻮﺍﺭﺩ ﻣﻌﻨﻲﺩﺍﺭ ﺍﺳـﺖ‪،‬‬
‫ﻟﺬﺍ ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ ﺍﻳﻦ ﮔﺰﻳﻨﻪ ﭘﻴﺸﻨﻬﺎﺩ ﻧﻤﻲﺷﻮﺩ‪.‬‬ ‫ﺟﺪﻭﻝ ‪ ٣‬ﻣﻘﺎﻳﺴﻪ ﺟﻮﺍﺑﻬﺎﻱ ﮔﺮﻓﺘﻪ ﺷﺪﻩ ﺑﺎ ﺩﺍﺩﻩﻫﺎﻱ ﻣﺸﺎﻫﺪﺍﺗﻲ ﺑﺮﺍﻱ‬
‫‪ -۲‬ﺗﻔﺎﻭﺕ ﻣﻌﻨﻲﺩﺍﺭﻱ ﺑﻴﻦ ﻧﺮﺥ ﺷﮑﺴﺖ ﻣﺤﺎﺳﺒﻪ ﺷﺪﻩ ﺩﺭ ﻣﺪﻝ ‪ANN‬‬ ‫ﻣﺤﺎﺳﺒﻪ ﻧﺮﺥ ﺷﮑﺴﺖ ﻟﻮﻟﻪﻫﺎ ﺑﺎ ﻣﻌﻴﺎﺭﻫﺎﻱ ﺍﺭﺯﻳﺎﺑﻲ ﻣﺨﺘﻠﻒ ﺭﺍ ﻧﻤﺎﻳﺶ‬
‫ﻭ ﻣﺪﻝ ‪ EPR‬ﺑﺎ ﻣﻘﺪﺍﺭ ﺣﻘﻴﻘﻲ ﻭﺟﻮﺩ ﻧﺪﺍﺭﺩ‪ ،‬ﺑﻨـﺎﺑﺮﺍﻳﻦ ﺍﻳـﻦ ﺭﻭﺷـﻬﺎ ﺩﺭ‬ ‫ﺩﺍﺩﻩ ﺍﺳﺖ‪.‬‬
‫ﺍﺭﺟﺤﻴﺖ ﺑﻴﺸﺘﺮﻱ ﻧﺴﺒﺖ ﺑﻪ ﺩﻳﮕﺮ ﺭﻭﺷـﻬﺎﻱ ﻣﻌﻤـﻮﻝ ﺩﺭ ﻣـﺪﻝﺳـﺎﺯﻱ‬ ‫ﺟﺪﻭﻝ ‪ -٣‬ﻣﺘﻮﺳﻂ ﻣﺮﺑﻌﺎﺕ ﺧﻄﺎﻱ ﻣﺪﻝ ﺭﮔﺮﺳﻴﻮﻥ ﭼﻨﺪ ﻣﺘﻐﻴﺮﻩ‬
‫ﺷﮑﺴﺖ ﺩﺭ ﻟﻮﻟﻪﻫﺎ ﻗﺮﺍﺭ ﻣﻲﮔﻴﺮﻧﺪ‪.‬‬ ‫ﻣﻌﻴﺎﺭﻫﺎﻱ ﺧﻄﺎﻱ ﻣﺪﻝ‬ ‫ﻧﻮﻉ‬
‫‪ -۳‬ﺭﻭﻧﺪ ﺗﻐﻴﻴﺮﺍﺕ ﻧﺮﺥ ﺷﮑﺴﺖ ﺩﺭ ﻣﺪﻝﺳـﺎﺯﻱ ‪ NLR‬ﺗﻘﺮﻳﺒـﺎﹰ ﻣـﺸﺎﺑﻪ‬ ‫‪RMSE‬‬ ‫‪MSE‬‬ ‫‪IOA‬‬ ‫ﺗﺎﺑﻊ‬
‫ﺣﺎﻟﺖ ﻭﺍﻗﻌﻲ ﺍﺳﺖ ﻭﻟﻲ ﻣﺤﺪﻭﺩﻩ ﺗﻐﻴﻴﺮﺍﺕ ﺩﺭ ﺁﻥ ﺑﻴـﺸﺘﺮ ﺍﺯ ﺩﻭ ﺭﻭﺵ‬ ‫‪٠/٢٧٧٨٩ ٠/٠٧٧٢٢٢٨٥٤ ٠/٩٣٨٣٣‬‬ ‫‪NLR‬‬
‫‪ EPR‬ﻭ ‪ ANN‬ﺑﻮﺩﻩ ﻭ ﻧﺰﺩﻳﮏ ﺑﻪ ‪ ANN‬ﺍﺳﺖ‪.‬‬
‫‪ -٤‬ﻧــﺮﺥ ﺷﮑــﺴﺖ ﻣﺤﺎﺳــﺒﻪ ﺷــﺪﻩ ﺩﺭ ﻣــﺪﻝ ‪ ANN‬ﻭ ﻣــﺪﻝ ‪ EPR‬ﺩﺭ‬ ‫‪ -۶‬ﻧﺘﺎﻳﺞ ﻭ ﺑﺤﺚ‬
‫ﻗﻄﺮﻫﺎﻱ ﮐﻮﭼﮏ ﻳﻌﻨﻲ ﮐﻤﺘﺮ ﺍﺯ ‪ ٢٠٠‬ﻣﻴﻠﻲﻣﺘﺮ‪ ،‬ﺍﺯ ﺗﻄﺎﺑﻖ ﺑﺴﻴﺎﺭ ﺧﻮﺑﻲ‬ ‫ﻧﺘﺎﻳﺞ ﺑﻪ ﺩﺳﺖ ﺁﻣﺪﻩ ﺍﺯ ﻣﺪﻝﻫﺎﻱ ﺑﺮﻧﺎﻣﻪﻧﻮﻳـﺴﻲ ﮊﻧﺘﻴـﮏ ﻭ ﺭﮔﺮﺳـﻴﻮﻥ‬
‫ﺑﺎ ﻣﻴـﺰﺍﻥ ﺣﻘﻴﻘـﻲ ﺑـﺮﺧـﻮﺭﺩﺍﺭ ﺍﺳﺖ ﮐﻪ ﺍﻳﻦ ﻣﻮﺿـﻮﻉ ﺑﻬﺮﻩﮔﻴﺮﻱ ﺍﺯ ﺍﻳﻦ‬ ‫ﻏﻴﺮﺧﻄﻲ ﻭ ﺭﻭﺍﺑﻂ ﺍﺳﺘﺨﺮﺍﺟﻲ ﺍﺯ ﻣﺪﻝﻫﺎﻱ ﺷﺒﮑﻪ ﻋـﺼﺒﻲ ﻣـﺼﻨﻮﻋﻲ ﻭ‬

‫ﺷﻤﺎﺭﻩ ‪ ٢‬ﺳﺎﻝ ‪١٣٩٠‬‬ ‫ﺁﺏ ﻭ ﻓﺎﺿﻼﺏ‬ ‫‪‬‬

 ‫ﺭﻭﺷﻬﺎ ﺭﺍ ﻧﺴﺒﺖ ﺑﻪ ﺳﺎﻳﺮ ﺭﻭﺷﻬﺎ ﻣﻤﺘﺎﺯ ﻣﻲﮐﻨﺪ‪.‬‬
‫‪2.50‬‬

‫ﻻﺯﻡ ﺑﻪﺫﮐﺮ ﺍﺳﺖ ﮐﻪ ﺗﻌـﺪﺍﺩ ﻭ ﻧـﻮﻉ ﭘﺎﺭﺍﻣﺘﺮﻫـﺎﻱ ﻣـﻮﺛﺮ ﺷـﺒﮑﻪ ﺩﺭ‬

‫‪2.00‬‬

‫ﭘﻴﺶ ﺑﻴﻨﻲ ﻧﺮﺥ ﺷﮑﺴﺖ ﺑﺮﺍﻱ ﻫﻤﻪ ﻣﺪﻝﻫﺎ ﻳﮑﺴﺎﻥ ﺑﻮﺩ‪.‬‬ ‫‪1.50‬‬

‫ﻧﺮﺥ ﺷﻜﺴﺖ‬
‫‪1.00‬‬

‫‪ -٧‬ﻧﺘﻴﺠﻪﮔﻴﺮﻱ‬ ‫‪0.50‬‬

‫ﻣﺪﻓﻮﻥ ﺑﻮﺩﻥ ﺷﺒﻜﻪﻫﺎﻱ ﺗﻮﺯﻳﻊ ﺁﺏ ﺩﺭ ﺯﻳـﺮ ﺧـﺎﻙ ﺳـﺒﺐ ﺷـﺪﻩ ﮐـﻪ ﺍﺯ‬ ‫‪0.00‬‬

‫ﻟﺤﺎﻅ ﺑﺼﺮﻱ ﻗﺎﺑﻞ ﭘﺎﻳﺶ ﻧﺒﺎﺷﻨﺪ‪ .‬ﻫﻤﭽﻨﻴﻦ ﻋﻼﻭﻩ ﺑﺮ ﻣﺸﻜﻼﺕ ﺣﻔﻆ‬

‫ﺳﻼﻣﺖ ﻭ ﻧﮕﻬﺪﺍﺭﻱ ﺁﻧﻬﺎ‪ ،‬ﻣﻌﺎﻳﺐﺷﺎﻥ ﻧﻴﺰ ﺗﺎ ﺣﺪ ﻣـﺰﻣﻦ ﻭ ﺑﺤﺮﺍﻧـﻲ ﺍﺯ‬ ‫‪2.50‬‬

‫ﺩﻳﺪ ﻣﺪﻳﺮﺍﻥ ﻭ ﺑﻬﺮﻩ ﺑﺮﺩﺍﺭﺍﻥ ﭘﻨﻬـﺎﻥ ﻣـﻲﻣﺎﻧـﺪ ﻭ ﺑـﺮﻭﺯ ﻣـﺸﻜﻼﺕ ﻓﻨـﻲ‬ ‫‪2.00‬‬

‫ﭘﻴﭽﻴﺪﻩ ﺩﺭ ﺷـﺒﻜﻪ ﺭﺍ ﺑﺎﻋـﺚ ﻣـﻲﺷـﻮﺩ‪ .‬ﭘﺪﻳـﺪﻩ ﺷﮑـﺴﺖ ﺩﺭ ﻟﻮﻟـﻪﻫـﺎﻱ‬

‫ﺁﺑﺮﺳﺎﻧﻲ ﺩﺭ ﺳﺎﻟﻴﺎﻥ ﮔﺬﺷﺘﻪ ﺗﻮﺳـﻂ ﻣـﺪﻝﻫـﺎﻱ ﺁﻣـﺎﺭﻱ ﺯﻳـﺎﺩﻱ ﻣـﻮﺭﺩ‬
‫‪1.50‬‬

‫ﻧﺮﺥ ﺷﻜﺴﺖ‬
‫ﻣﻄﺎﻟﻌﻪ ﻗـﺮﺍﺭ ﮔﺮﻓﺘـﻪ ﮐـﻪ ﺗﻨﻬـﺎ ﺗﻌـﺪﺍﺩ ﻣﺤـﺪﻭﺩﻱ ﺍﺯ ﺍﻳـﻦ ﻣـﺪﻝﻫـﺎ ﮐـﻪ‬ ‫‪1.00‬‬

‫ﺍﻃﻤﻴﻨﺎﻥﭘﺬﻳﺮﻱ ﻣﻜﺎﻧﻴﻜﻲ ﺭﺍ ﺗﺸﺮﻳﺢ ﻣﻲﮐﺮﺩﻧﺪ‪ ،‬ﺗﻮﺳـﻌﻪ ﺩﺍﺩﻩ ﺷـﺪﻩﺍﻧـﺪ‪.‬‬ ‫‪0.50‬‬

‫ﻭﻟﻲ ﻧﮑﺘﻪ ﻗﺎﺑﻞ ﺗﺄﻣﻞ ﺩﺭ ﺍﻳﻦ ﺟﺎ ﺍﺳﺖ ﮐﻪ ﻫﺮ ﻳﮏ ﺍﺯ ﺍﻳﻦ ﻣﺪﻝﻫﺎ ﻓﻘـﻂ‬ ‫‪0.00‬‬

‫ﺑﺎ ﺗﮏ ﭘﺎﺭﺍﻣﺘﺮﻫﺎﻱ ﻣﺆﺛﺮ ﺩﺭ ﺷﮑﺴﺖ ﺑﻴﺎﻥ ﻣﻲﺷﻮﻧﺪ‪ ،‬ﻟـﺬﺍ ﺍﺯ ﺟﺎﻣﻌﻴـﺖ‬

‫ﮐﻤﺘﺮﻱ ﺑﺮﺧﻮﺭﺩﺍﺭ ﺑﻮﺩﻩ ﻭ ﻗﺎﺑﻠﻴﺖ ﺗﻌﻤﻴﻢ ﺑﻪ ﺗﻤﺎﻣﻲ ﻣﺴﺎﺋﻞ ﻣﻄﺮﺡ ﺷـﺪﻩ‬
‫ﺩﺭ ﺑﻬﻴﻨﻪﺳﺎﺯﻱ ﻣﻬﻨﺪﺳﻲ ﺳﻴﺴﺘﻢﻫﺎﻱ ﺁﺑﺮﺳﺎﻧﻲ ﺭﺍ ﺩﺍﺭﺍ ﻧﻴﺴﺘﻨﺪ‪.‬‬ ‫‪2.50‬‬

‫ﺩﺭ ﺍﻳﻦ ﻣﻘﺎﻟﻪ ﺑـﻪﻋﻨـﻮﺍﻥ ﻳـﮏ ﮐـﺎﺭ ﺟﺪﻳـﺪ‪ ،‬ﺗﻌﻴـﻴﻦ ﻧـﺮﺥ ﺷﮑـﺴﺖ‬ ‫‪2.00‬‬

‫ﻣﮑﺎﻧﻴﮑﻲ ﺍﺯ ﻃﺮﻳﻖ ﺍﺭﺍﺋﻪ ﻣﺪﻝ ﺷﺒﮑﻪ ﻋـﺼﺒﻲ ﻣـﺼﻨﻮﻋﻲ‪ ،‬ﻣـﺪﻝ ﺳﻴـﺴﺘﻢ‬ ‫‪1.50‬‬
‫ﻧﺮﺥ ﺷﻜﺴﺖ‬

‫ﺍﺳﺘﻨﺘﺎﺝ ﻓﺎﺯﻱ‪ ،‬ﺑﺮﻧﺎﻣـﻪﻧﻮﻳـﺴﻲ ﮊﻧﺘﻴـﮏ ﻗﺎﻋـﺪﻩﺍﻱ ﻭ ﺭﮔﺮﺳـﻴﻮﻥ ﭼﻨـﺪ‬ ‫‪1.00‬‬

‫ﻣﺘﻐﻴﺮﻩ ﻏﻴﺮﺧﻄﻲ ﻣﻄﺮﺡ ﮔﺮﺩﻳﺪ‪ .‬ﺑـﺮﺍﻱ ﺍﻳـﻦ ﻣﻨﻈـﻮﺭ ﺑـﺎ ﺑﻬـﺮﻩﮔﻴـﺮﻱ ﺍﺯ‬ ‫‪0.50‬‬

‫ﻣﺠﻤﻮﻋـﻪ ﻣﺘﻐﻴﺮﻫــﺎﻱ ﻗﺎﺑـﻞ ﺍﻧــﺪﺍﺯﻩﮔﻴــﺮﻱ ﺩﺭ ﻳـﮏ ﺳﻴــﺴﺘﻢ ﺁﺑﺮﺳــﺎﻧﻲ‬

‫)ﺟﻨﺲ‪ ،‬ﺳﻦ‪ ،‬ﻃﻮﻝ‪ ،‬ﻗﻄﺮ‪ ،‬ﻋﻤﻖ ﻧـﺼﺐ ﻭ ﻓـﺸﺎﺭ( ﺑـﻪﻣﻨﻈـﻮﺭ ﻣـﺪﻳﺮﻳﺖ‬
‫‪0.00‬‬

‫ﺑﻬﻴﻨﻪ ﺑﻬﺮﻩﺑﺮﺩﺍﺭﻱ ﺍﺯ ﺷـﺒﮑﻪ ﮐـﻪ ﻻﺯﻣـﻪ ﺍﺻـﻠﻲ ﺁﻥ ﺗﺤﻠﻴـﻞ ﻭ ﺩﺭﻳﺎﻓـﺖ‬

‫ﺻﺤﻴﺢ ﺍﺯ ﻧﺮﺥ ﺣﻮﺍﺩﺙ ﺩﺭ ﺷـﺒﮑﻪ ﺁﺑﺮﺳـﺎﻧﻲ ﺧـﺼﻮﺻﺎﹰ ﻟﻮﻟـﻪﻫـﺎ ﺍﺳـﺖ‪،‬‬
‫‪2.50‬‬

‫ﺭﻭﺍﺑﻂ ﻭ ﻣﺪﻝﻫﺎﻳﻲ ﺍﺳﺘﺨﺮﺍﺝ ﺷﺪ‪ .‬ﻫﻤﭽﻨﻴﻦ ﺑﺎ ﺗﻮﺟﻪ ﺑـﻪ ﺑﺮﺭﺳـﻲ ﻫـﺎﻱ‬

‫‪2.00‬‬

‫ﺍﻧﺠﺎﻡ ﺷﺪﻩ ﺍﺯ ﻛﻴﻔﻴﺖ ﺧﺮﻭﺟﻲ ﻣﺪﻝﻫﺎ‪ ،‬ﻗﺎﺑﻠﻴﺖ ﺭﻭﺷﻬﺎﻱ ﺷﺒﮑﻪ ﻋـﺼﺒﻲ‬ ‫‪1.50‬‬
‫ﻧﺮﺥ ﺷﻜﺴﺖ‬

‫ﻣﺼﻨﻮﻋﻲ ﻭ ﺑﺮﻧﺎﻣﻪ ﻧﻮﻳﺴﻲ ﮊﻧﺘﻴﮏ ﺩﺭ ﻣﺪﻝﺳﺎﺯﻱ ﺷﮑﺴﺖ ﻟﻮﻟﻪﻫﺎ ﻣﻮﺭﺩ‬ ‫‪1.00‬‬

‫ﺗﺄﻳﻴــﺪ ﻗــﺮﺍﺭ ﮔﺮﻓــﺖ ﮐــﻪ ﺍﻳــﻦ ﺭﻭﺷــﻬﺎ ﺑــﺮﺍﻱ ﺍﺳــﺘﻔﺎﺩﻩ ﺩﺭ ﻣﺤﺎﺳــﺒﺎﺕ‬ ‫‪0.50‬‬

‫ﺍﻃﻤﻴﻨﺎﻥ ﭘﺬﻳﺮﻱ ﻭ ﺍﻣﮑﺎﻥ ﮐﺎﺭﺑﺮﻱ ﺩﺭ ﺷـﺒﻜﻪﻫـﺎﻱ ﺗﻮﺯﻳـﻊ ﺁﺏ ﻣﻌﺮﻓـﻲ‬ ‫‪0.00‬‬

‫ﻣﻲﺷﻮﻧﺪ‪.‬‬

‫ﺷﻜﻞ ‪ -٩‬ﻧﻤﻮﺩﺍﺭﻫﺎﻱ ﻧﺮﺥ ﺷﮑﺴﺖ ﭘﻴﺶﺑﻴﻨﻲ ﺷﺪﻩ ﺑﺎ ﺭﻭﺷﻬﺎﻱ ﻣﺘﻔﺎﻭﺕ‬

‫‪ -۸‬ﻣﺮﺍﺟﻊ‬
‫‪1- Elahi Panah, N. A. (1998). “Subsequent water distribution networks in the country until 1400.” J. of Water‬‬
‫)‪and Environment, 28, 4-15. (In Persian‬‬
‫‪2- Tabesh, M., and Abedini, A.A. (2005). “Analysis of pipe failure in water distribution network.” J. of Iran‬‬
‫)‪Water Resources Research, 1(1), 78-79. (In Persian‬‬
‫‪3- Tabesh, M., Jafari, H., and Delavari, M. R. (2005). “Management model of accidents and breaks in water‬‬
‫)‪distribution system using GIS.” J. of Waterand Wastewater, 70, 2-15. (In Persian‬‬

‫‪‬‬ ‫ﺁﺏ ﻭ ﻓﺎﺿﻼﺏ‬ ‫ﺷﻤﺎﺭﻩ ‪ ٢‬ﺳﺎﻝ ‪١٣٩٠‬‬

4- Newport, R. (1981). “Factors influencing the occurrence of bursts in iron water mains.” Water Supply and
Management, 3, 274-278.
5- Walski, T. M., and Pelliccia, A. (1982). “Economic analysis of water main breaks.” J. of AWWA, 74, 140-
147.
6- Shamir, U., and Howard, C.D.D. (1979). “An analytical approach to scheduling pipe replacement.” J. of
AWWA., 71, 248-258.
7- Kettler, A.J., and Goulter, L.C. (1985). “An analysis of pipe breakage in urban water distribution networks.”
Can. J. Civ. Eng., 12(2), 286-293.
8- Su, Y.C., Mays, L.W., Duan, N., and Lansey, K.E. (1987), “Reliability based optimization model for water
distribution systems.” J. Hydr. Eng., 114(12), 1539-1556.
9- Sacluti, F. R. (1999). “Modeling water distribution pipe failures using artificial neural networks.” M.Sc.
Thesis, Department of Civil and Envir. Eng., University of Alberta, Canada.
10-Tabesh, M., Aghaei, A., and Abrishami, J. (2008). “Investigation of the effects of influential parameters on
pipe burst in water distribution systems using evolutionary polynomial regression method.” J. of Faculty of
Engineering (Special issue on: Civil Engineering), 42(6), 691-703. (In Persian)
11- Tabesh, M., Soltani, J., Farmani, R., and Savic, D. (2009). “Assessing pipe failure rate and mechanical
reliability of water distribution networks using data driven modeling.” J. of Hydroinformatics, 11(1), 1-17.
12- Farmani, R., Walters, G.A., and Savic, D. (2005). “Trade off between total cost and reliability for anytown
water distribution network.” J. Water Resour. Plng. and Mgmt., 131(3), 161-171.
13- Tabesh, M., and Dini, M. (2009). “Fuzzy and neuro- fuzzy models for short-term water demand forecasting in
Tehran.” Iranian J. of Science and Technology, Transaction B: Engineering, 33(B1), 61-77.
14- Karamoz, M., Tabesh, M., Nazif, S., and Moridi, A. (2005). “Predict the water pressure in water supply
networks and fuzzy inference.” J. of Water and Wastewater, 53, 3-14. (In Persian)
15- Giustolosi, O., and Savic, D.A. (2004). “A novel genetic programming strategy: Evolutionary polynomial
Regression.” 6th International Conference on Hydro Informatics, Liong, Phoon and Babovic (eds.),
Singapore, 1, 787-794.
16- Tabesh, M.., and Soltani, J. (2007). “Optimal operational management of water sypply systems by
evaluation of mechanical reliability using artificial neural networks (ANN).” Proc. of the 2nd National
Conference on Iran Water Resources, (2WRM 2007), Faculty of Engineering, Isfahan University of
Technology, Isfahan, Iran, 23-24. (In Persian)

١٣٩٠ ‫ ﺳﺎﻝ‬٢ ‫ﺷﻤﺎﺭﻩ‬ ‫ ﺁﺏ ﻭ ﻓﺎﺿﻼﺏ‬