‫ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ ﺗﻮﺍﺑﻊ ﻫﺪﻑﻫﺎﻱ ﻣﺨﺘﻠﻒ ﺑﺮﺍﻱ ﭘﻴﺪﺍ ﻛﺮﺩﻥ ﻣﻘﺎﺩﻳﺮ ﻣﺠﻬﻮﻝ ﭘﺎﺭﺍﻣﺘﺮﻫﺎﻱ ﻛﻨﺘﺮﻟﻲ‬ ‫‪۱‬‬

‫ﺩﺭ ﺗﻨﻈﻴﻢ ﭘﺎﺭﺍﻣﺘﺮﻫﺎﻱ ﻛﻨﺘﺮﻟﻲ‪ ،‬ﺭﻭﺵﻫﺎﻱ ﻣﺨﺘﻠﻔﻲ ﺑﺮﺍﻱ ﺍﺭﺯﻳﺎﺑﻲ ﻋﻤﻠﻜﺮﺩ ﺳﻴﺴﺘﻢﻫﺎﻱ ﻛﻨﺘﺮﻟﻲ ﺑﻪ ﻛﺎﺭ ﮔﺮﻓﺘﻪ ﻣﻲﺷﻮﻧﺪ‪ .‬ﺳﻪ‬
‫ﺷﺎﺧﺺ ﺭﺍﻳﺞ ﺑﺮﺍﻱ ﺑﻬﻴﻨﻪﺳﺎﺯﻱ ﻭ ﺍﺭﺯﻳﺎﺑﻲ ﭘﺎﺳﺦ ﺳﻴﺴﺘﻢ‪ ،(ITSE) ،(IAE) ،(ISE) ،‬ﻫﺴﺘﻨﺪ‪ .‬ﺍﻳﻦ ﺳﻪ ﺷﺎﺧﺺ ﺑﻬﻴﻨﻪ‬
‫ﺳﺎﺯﻱ ﺑﻪ ﻋﻨﻮﺍﻥ ﺗﺎﺑﻊ ﻫﺪﻓﻲ ‪ ۱‬ﻛﻪ ﺑﻪ ﺩﻧﺒﺎﻝ ﺣﺪﺍﻗﻞ ﻛﺮﺩﻥ ﺁﻥ ﺑﺮﺍﻱ ﭘﻴﺪﺍ ﻛﺮﺩﻥ ﭘﺎﺭﺍﻣﺘﺮﻫﺎﻱ ﻛﻨﺘﺮﻟﻲ ﻛﻪ ﺷﺎﻣﻞ ﺗﻨﺎﺳﺒﻲ‪،‬‬
‫ﺍﻧﺘﮕﺮﺍﻟﻲ ﺩﺭ ﺁﺯﻣﺎﻳﺶ ﻓﺸﺎﺭ ﻣﺨﺎﺯﻥ ﻣﺘﻮﺍﻟﻲ ﺍﺳﺖ‪.‬‬
‫ﺍﮔﺮ ﺑﻪ ﺩﺳﺘﻮﺭ ﻛﺎﺭ ﻭ ﺧﻮﺍﺳﺘﻪﻫﺎﻱ ﺁﺯﻣﺎﻳﺶ ﻣﺮﺑﻮﻃﻪ ﻣﺮﺍﺟﻌﻪ ﻛﻨﻴﺪ‪ ،‬ﻣﺸﺎﻫﺪﻩ ﻣﻲﻛﻨﻴﺪ ﻛﻪ ﺩﺭ ﭼﻨﺪ ﺑﺨﺶ ﺁﻥ ﺑﻪ ﺍﻳﻦ ﺍﺳﻢﻫﺎ‬
‫ﺍﺷﺎﺭﻩ ﺷﺪﻩ ﺍﺳﺖ ﺩﺭ ﺍﺩﺍﻣﻪ ﺑﻪ ﺍﺧﺘﺼﺎﺭ ﺑﻪ ﻫﺮ ﻛﺪﺍﻡ ﺍﺯ ﺍﻳﻦ ﺳﻪ ﺭﻭﺵ ﻭ ﺯﻣﺎﻧﻲ ﻛﻪ ﺩﺭ ﻣﺤﺎﺳﺒﻪ ﭘﺎﺭﺍﻣﺘﺮﻫﺎﻱ ﻛﻨﺘﺮﻟﻲ ﺁﺯﻣﺎﻳﺶ‬
‫ﻧﻘﺶ ﺧﻮﺩ ﺭﺍ ﺍﻳﻔﺎ ﻣﻲﻛﻨﺪ ﺧﻮﺍﻫﻴﻢ ﭘﺮﺩﺍﺧﺖ‪.‬‬

‫ﺭﻭﺵ )‪ISE (Integral of Squared Error‬‬ ‫‪۱.۱‬‬

‫ﺍﻳﻦ ﺷﺎﺧﺺ ﻣﺠﻤﻮﻉ ﺍﻧﺘﮕﺮﺍﻟﻲ ﻣﺮﺑﻌﺎﺕ ﺧﻄﺎ ﺭﺍ ﻣﺤﺎﺳﺒﻪ ﻣﻲﻛﻨﺪ ﻛﻪ ﻓﺮﻣﻮﻝ ﺁﻥ ﺑﻪ ﺍﻳﻦ ﺻﻮﺭﺕ ﺍﺳﺖ‪:‬‬
‫∞ ∫‬
‫= ‪ISE‬‬ ‫‪e2 (t)dt‬‬ ‫)‪(۱‬‬
‫‪0‬‬

‫ﺑﺎ ﺗﻮﺟﻪ ﺑﻪ ﻓﺮﻣﻮﻝ ﺑﺎﻻ ﻣﻲﺗﻮﺍﻥ ﺑﻪ ﺍﻳﻦ ﺟﻤﻊﺑﻨﺪﻱ ﺭﺳﻴﺪ ﻛﻪ ﺑﻪ ﺩﻟﻴﻞ ﺑﻪ ﺗﻮﺍﻥ ﺩﻭ ﺭﺳﻴﺪﻥ ﻣﻘﺪﺍﺭ ﺧﻄﺎ ﺩﺭ ﻓﺮﻣﻮﻝ ﺗﺄﻛﻴﺪ ﺑﻴﺸﺘﺮﻱ‬
‫ﺑﺮ ﺧﻄﺎﻫﺎﻱ ﺑﺰﺭﮒ ﺩﺍﺭﺩ‪ ،‬ﺯﻳﺮﺍ ﻣﺮﺑﻊ ﺧﻄﺎﻫﺎ ﺩﺭ ﻧﻈﺮ ﮔﺮﻓﺘﻪ ﻣﻲﺷﻮﺩ ﻭ ﻣﻘﺎﺩﻳﺮ ﺧﻄﺎﻫﺎﻱ ﻛﻮﭼﻚ ﺩﺭ ﻣﻘﺎﻳﺴﻪ ﺑﺎ ﺧﻄﺎﻫﺎﻱ ﺑﺰﺭﮒ‬
‫ﺍﺯ ﺍﺛﺮ ﻛﻤﺘﺮﻱ ﺑﺮﺧﻮﺭﺩﺍﺭ ﻫﺴﺘﻨﺪ‪.‬‬

‫ﺭﻭﺵ )‪IAE (Integral of Absolute Error‬‬ ‫‪۲.۱‬‬

‫ﺩﺭ ﺍﻳﻦ ﺭﻭﺵ‪ ،‬ﻣﻘﺪﺍﺭ ﺍﻧﺘﮕﺮﺍﻟﻲ ﻗﺪﺭﻣﻄﻠﻖ ﺧﻄﺎ ﺑﺮﺭﺳﻲ ﻣﻲﺷﻮﺩ ﻛﻪ ﻓﺮﻣﻮﻝ ﺁﻥ ﺑﻪ ﺷﺮﺡ ﺯﻳﺮ ﺍﺳﺖ‪:‬‬

‫∫‬ ‫∞‬
‫= ‪IAE‬‬ ‫‪|e(t)|dt‬‬ ‫)‪(۲‬‬
‫‪0‬‬

‫ﺑﺎ ﺗﻮﺟﻪ ﺑﻪ ﻓﺮﻣﻮﻝ ﺁﻥ ﺩﺭ ﺍﻳﻦ ﺭﻭﺵ ﻣﺤﺎﺳﺒﻪ ﺗﺎﺑﻊ ﻫﺪﻑ ﻣﻘﺎﺩﻳﺮ ﺧﻄﺎ ﺑﺰﺭﮒ ﻭ ﻛﻮﭼﻚ ﺩﺍﺭﺍﻱ ﻭﺯﻥ ﻳﻜﺴﺎﻧﻲ ﻫﺴﺘﻨﺪ‪.‬‬

‫ﺭﻭﺵ )‪ITSE (Integral of Time-weighted Squared Error‬‬ ‫‪۳.۱‬‬

‫ﺩﺭ ﺍﻳﻦ ﺭﻭﺵ‪ ،‬ﺯﻣﺎﻥ ﺑﻪ ﻋﻨﻮﺍﻥ ﺿﺮﻳﺐ ﻭﺯﻧﻲ ﺩﺭ ﺧﻄﺎﻱ ﻣﺮﺑﻌﻲ ﺍﻋﻤﺎﻝ ﻣﻲﺷﻮﺩ‪ ،‬ﻛﻪ ﻓﺮﻣﻮﻝ ﺁﻥ ﺑﻪ ﺷﺮﺡ ﺯﻳﺮ ﻣﻲﺑﺎﺷﺪ‪:‬‬
‫∞ ∫‬
‫= ‪ITSE‬‬ ‫‪te2 (t)dt‬‬ ‫)‪(۳‬‬
‫‪0‬‬

‫ﺍﻳﻦ ﺭﻭﺵ ﺩﺍﺭﺍﻱ ﻣﺎﻫﻴﺖ ﻣﺸﺎﺑﻪ ﺑﺎ ﺭﻭﺵ ‪ ISE‬ﺑﻮﺩﻩ ﻭﻟﻲ ﺗﻔﺎﻭﺕ ﺁﻥ ﺑﻪ ﺿﺮﺏ ﺷﺪﻥ ﻣﻴﺰﺍﻥ ﺯﻣﺎﻥ ﺩﺭ ﻣﻘﺪﺍﺭ ﻣﺮﺑﻊ ﻫﺮ ﺧﻄﺎ‬
‫ﺍﺳﺖ ﺑﻪ ﻫﻤﻴﻦ ﺩﻟﻴﻞ ﺧﻄﺎﻫﺎﻳﻲ ﻛﻪ ﺩﺭ ﻟﺤﻈﻪﻫﺎﻱ ﺟﺪﻳﺪ ﺗﺮﻱ ﺍﺗﻔﺎﻕ ﻣﻲﺍﻓﺘﻨﺪ ﺩﺍﺭﺍﻱ ﻭﺯﻥ ﺑﻴﺸﺘﺮﻱ ﺩﺭ ﻣﻘﺎﻳﺴﻪ ﺑﺎ ﺧﻄﺎﻫﺎﻱ‬
‫ﻟﺤﻈﻪﻫﺎﻱ ﻗﺒﻠﻲ ﻫﺴﺘﻨﺪ ﺑﻪ ﻫﻤﻴﻦ ﺩﻟﻴﻞ ﺍﺛﺮ ﺧﻄﺎﻱ ﻟﺤﻈﺎﺕ ﺟﺪﻳﺪ ﺩﺭ ﻣﻘﺪﺍﺭ ﺗﺎﺑﻊ ﻫﺪﻑ ﺍﺛﺮ ﺑﻴﺸﺘﺮﻱ ﺩﺍﺭﻧﺪ‪.‬‬
‫‪1‬‬
‫‪Objective Function‬‬

‫‪۱‬‬
 ‫ﺍﺭﺗﺒﺎﻁ ﺑﺎ ﺁﺯﻣﺎﻳﺶ ﻓﺸﺎﺭ ﻣﺨﺎﺯﻥ ﻣﺘﻮﺍﻟﻲ‬ ‫‪۲‬‬
‫ﺩﺭ ﺍﻳﻦ ﺁﺯﻣﺎﻳﺶ ﻣﺎ ﺩﺭ ﺩﻭﺑﺨﺶ ﺍﺯ ﺗﻮﺍﺑﻊ ﻫﺪﻑ ﺑﺮﺍﻱ ﭘﻴﺪﺍ ﻛﺮﺩﻥ ﭘﺎﺭﺍﻣﺘﺮﻫﺎﻱ ﻣﻮﺭﺩ ﻧﻴﺎﺯ ﺍﺳﺘﻔﺎﺩﻩ ﻛﺮﺩﻳﻢ‪:‬‬

‫‪ .۱‬ﭘﻴﺪﺍ ﻛﺮﺩﻥ ﭘﺎﺭﺍﻣﺘﺮﻫﺎﻱ ﻻﺯﻡ ﺑﺮﺍﻱ ﺗﺨﻤﻴﻦ ﺭﻓﺘﺎﺭ ﺳﻴﺴﺘﻢ ﺑﻪ ﺻﻮﺭﺕ ﻣﺮﺗﺒﻪ ﻳﻚ ﻭ ﻣﺮﺗﺒﻪ ‪ ۲‬ﻛﻪ ﺑﺎ ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ ﺗﺎﺑﻊ‬
‫‪ tfest‬ﺍﻧﺠﺎﻡ ﺷﺪ‪.‬‬

‫‪ .۲‬ﭘﻴﺪﺍ ﻛﺮﺩﻥ ﭘﺎﺭﺍﻣﺘﺮﻫﺎﻱ ﻛﻨﺘﺮﻟﻲ ﺑﺎ ﺩﺭ ﻧﻈﺮ ﮔﺮﻓﺘﻦ ﺭﻓﺘﺎﺭ ﻣﺮﺗﺒﻪ ﻳﻚ ﻭ ﺩﻭ ﺣﺎﺻﻠﻪ ﺍﺯ ﻣﺮﺣﻠﻪﻱ ﻗﺒﻞ ﻛﻪ ﺑﺎ ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ‬
‫ﺍﻟﮕﻮﺭﻳﺘﻢ ﮊﻧﺘﻴﻚ ﻭ ﺗﺎﺑﻊ ﻫﺪﻑ ‪ ISE‬ﺍﺳﺘﻔﺎﺩﻩ ﺷﺪ‪.‬‬

‫ﺍﮔﺮ ﺑﻪ ﻓﺎﻳﻞﻫﺎ ﻣﺘﻠﺐ ﺁﺯﻣﺎﻳﺶ ﻣﺮﺑﻮﻃﻪ ﻣﺮﺍﺟﻌﻪ ﻛﻨﻴﺪ ﻣﺸﺎﻫﺪﻩ ﻣﻲﻛﻨﻴﺪ ﻛﻪ ﻳﻚ ﻓﺎﻳﻞ ﺑﻪ ﻋﻨﻮﺍﻥ ‪sys_performance‬‬
‫ﺩﺭ ﻣﻴﺎﻥ ﻓﺎﻳﻞﻫﺎ ﻭﺟﻮﺩ ﺩﺍﺭﺩ‪ ،‬ﺩﺭ ﺧﻂ ‪ ۱۰‬ﻣﺸﺎﻫﺪﻩ ﻣﻲﻛﻨﻴﺪ ﻛﻪ ﻣﻲﺗﻮﺍﻥ ﺑﺎ ﺗﻐﻴﻴﺮ ﻧﺎﻡ ‪ ISE_PI‬ﺑﻪ ﻳﻜﻲ ﺍﺯ ﻧﺎﻡﻫﺎﻱ ﺫﻛﺮ‬
‫ﺷﺪﻩ ﺩﺭ ﻛﺎﻣﻨﺖ ﻣﺮﺑﻮﻃﻪ ﻧﻮﻉ ﺗﺎﺑﻊ ﻫﺪﻑ ﻣﺪﻧﻈﺮ ﺭﺍ ﺗﻐﻴﻴﺮ ﺩﺍﺩ!‬

‫ﺷﻜﻞ ‪ :۱‬ﺷﻤﺎﻱ ﻣﺤﻴﻂ ﺑﻼﻙ ﺩﻳﺎﮔﺮﺍﻡ ﺑﺴﺘﻪ ﺷﺪﻩ ﺩﺭ ﺩﺍﺧﻞ ﺳﻴﻤﻮﻟﻴﻨﻚ‪.‬‬

‫ﺣﺎﻝ ﺳﻮﺍﻟﻲ ﻛﻪ ﻣﻤﻜﻦ ﺍﺳﺖ ﭘﻴﺶ ﺑﻴﺎﻳﺪ ﺍﻳﻦ ﻫﺴﺖ ﻛﻪ ﺩﺭ ﺩﺍﺧﻞ ﺳﻴﻤﻮﻟﻴﻨﻜﻲ ﻛﻪ ﺗﺎﺑﻊ ﻫﺪﻑﻫﺎﻱ ﻣﺮﺑﻮﻃﻪ ﻛﻪ ﺩﺭ ﺩﺍﺧﻞ‬
‫ﺍﻟﮕﻮﺭﻳﺘﻢ ﮊﻧﺘﻴﻚ ﻗﺮﺍﺭ ﺩﺍﺩﻩ ﻣﻲﺷﻮﺩ ﺑﻪ ﭼﻪ ﺻﻮﺭﺕ ﺗﻮﻟﻴﺪ ﻣﻲﺷﻮﻧﺪ‪ ،‬ﺑﺮﺍﻱ ﺭﻭﺷﻦ ﺗﺮ ﺷﺪﻥ ﻣﻮﺿﻮﻉ ﺑﻪ ﺗﻮﺟﻪ ﺑﻪ ﺷﻜﻞ ‪ ۱‬ﻣﻲﺗﻮﺍﻥ‬
‫ﻣﺸﺎﻫﺪﻩ ﻛﺮﺩ ﻛﻪ ﻣﻘﺪﺍﺭ ﺧﻄﺎﻱ ﻭﺭﻭﺩﻱ ﺑﻪ ﻛﻨﺘﺮﻟﺮ ﺍﺯ ﻣﺪﺍﺭ ﺑﺴﺘﻪ ‪ ۱‬ﺍﻧﺸﻌﺎﺏ ﮔﺮﻓﺘﻪﻣﻲﺷﻮﺩ ﻭ ﺑﺎ ﺗﻮﺟﻪ ﺑﻪ ﻓﺮﻣﺖ ﺗﻮﺍﺑﻊ ﻫﺪﻑ‬
‫ﻣﺮﺑﻮﻃﻪ ﻳﺎ ﺍﺯ ﺁﻥ ﻗﺪﺭ ﻣﻄﻠﻖ ﮔﺮﻓﺘﻪ ﺷﺪﻩ ﻳﺎ ﺑﻪ ﺗﻮﺍﻥ ﺩﻭ ﻣﻲﺭﺳﺪ‪ ،‬ﺳﭙﺲ ﺑﺴﺘﻪ ﺑﻪ ﺍﻳﻨﻜﻪ ﻧﻮﻉ ﺗﺎﺑﻊ ﻫﺪﻑ ﺑﻪ ﺻﻮﺭﺕ ‪ITSE‬‬
‫ﺑﺎﺷﺪ ﺩﺭ ﻭﺍﺣﺪ ﺯﻣﺎﻥ ﻧﻴﺰ ﺿﺮﺏ ﻣﻲﺷﻮﺩ ﺩﺭ ﻧﻬﺎﻳﺖ ﺑﺎ ﻋﺒﻮﺭ ﺍﺯ ﻳﻚ ﺍﻧﺘﮕﺮﺍﻟﻴﺮ ) ‪ ( 1s‬ﺗﺎﺑﻊ ﻫﺪﻑ ﻣﺪ ﻧﻈﺮ ﺩﺭﺳﺖ ﻣﻲﺷﻮﺩ ﻭ ﺑﺎ‬
‫ﺍﺳﺘﻔﺎﻩ ﺍﺯ ﺑﻼﻙ ‪ ToWorkSpace‬ﻛﻪ ﻧﺎﻡ ﺁﻥ ﻫﻢ ﺍﺳﻢ ﻳﻜﻲ ﺍﺯ ﺗﻮﺍﺑﻊ ﻫﺪﻓﻲ ﺍﺳﺖ ﺑﻪ ﻣﺘﻠﺐ ﺑﺮﮔﺮﺩﺍﻧﺪﻩ ﻣﻲﺷﻮﺩ ﻛﻪ ﺩﺭ‬
‫ﺩﺍﺧﻞ ﻓﺎﻳﻞ ‪ sys_performance‬ﻧﻮﻉ ﺗﺎﺑﻊ ﻫﺪﻓﻲ ﻛﻪ ﺩﺭ ﺩﺍﺧﻞ ﻭﺭﻙ ﺍﺳﭙﻴﺲ ﻣﺘﻠﺐ ﺍﺳﺖ ﺑﻪ ﻋﻨﻮﺍﻥ ﻣﻘﺪﺍﺭ ﺗﺎﺑﻊ ﻫﺪﻑ‬
‫ﻗﺮﺍﺭ ﺩﺍﺩﻩ ﻣﻲﺷﻮﺩ‪.‬‬
‫‪1‬‬
‫‪Close Loop‬‬

‫‪۲‬‬
 ‫ﻧﺘﻴﺠﻪﮔﻴﺮﻱ‬ ‫‪۳‬‬
‫ﺍﻧﺘﺨﺎﺏ ﺭﻭﺵ ﻣﻨﺎﺳﺐ ﺑﻪ ﻣﺎﻫﻴﺖ ﺳﻴﺴﺘﻢ ﻭ ﺍﻫﺪﺍﻑ ﻛﻨﺘﺮﻟﻲ ﺑﺴﺘﮕﻲ ﺩﺍﺭﺩ‪ .‬ﺍﮔﺮ ﻛﺎﻫﺶ ﺧﻄﺎﻱ ﮔﺬﺭﺍ ﺍﻫﻤﻴﺖ ﺑﻴﺸﺘﺮﻱ ﺩﺍﺷﺘﻪ‬
‫ﺑﺎﺷﺪ‪ ISE ،‬ﮔﺰﻳﻨﻪ ﺑﻬﺘﺮﻱ ﺍﺳﺖ‪ .‬ﺩﺭ ﺣﺎﻟﻲ ﻛﻪ ﺍﮔﺮ ﻛﺎﻫﺶ ﻧﻮﺳﺎﻧﺎﺕ ﻭ ﺭﻓﺘﺎﺭ ﻧﺮﻡ ﻣﻮﺭﺩ ﻧﻴﺎﺯ ﺑﺎﺷﺪ‪ IAE ،‬ﻣﻨﺎﺳﺐﺗﺮ ﺍﺳﺖ‪.‬‬
‫ﺍﻣﺎ ﺍﮔﺮ ﺑﻪ ﺩﻧﺒﺎﻝ ﺑﻬﺒﻮﺩ ﻋﻤﻠﻜﺮﺩ ﺑﻠﻨﺪﻣﺪﺕ ﻭ ﻛﺎﻫﺶ ﺧﻄﺎﻱ ﻣﺎﻧﺪﮔﺎﺭ ﺑﺎﺷﻴﻢ‪ ITSE ،‬ﮔﺰﻳﻨﻪ ﺑﻬﺘﺮﻱ ﺧﻮﺍﻫﺪ ﺑﻮﺩ‪.‬‬

‫‪۳‬‬