Predviđanje separacionih karakteristika kompleksirajuće-mikrofiltracionog procesa primenom veštačkih neuronskih mreža

Abstract

Zagađenost životne sredine, pre svega vode, teškim metalima predstavlja ekološki problem širom sveta. Povećana koncentracija teških metala u komunalnim i industrijskim otpadnim vodama predstavlja ozbiljnu pretnju s obzirom da se metali ne mogu razgraditi u prirodi i da neki mogu imati toksične efekte na biljke i životinje kao i na čoveka. Da bi se njihova koncentracija smanjila na adekvatan nivo definisan zakonskom regulativom, neophodno je primeniti metode prečišćavanja pre ispuštanja u recipijente. Membranski procesi se sve više koriste u oblasti zaštite životne sredine kao i za pripremu i preradu vode za potrebe prehrambene i farmaceutske industrije, petrohemije i dr. Glavne prednosti ovih procesa su mala energetska potrošnja, velika efikasnost, pouzdanost i mala količina otpada. Kompleksirajuće-mikrofiltracioni proces je hibridna membranska separaciona metoda koja se primenjuje za uklanjanje jona teških metala iz vode. Zasnovana je na konceptu da se mali joni metala, koji bi prolazili kroz pore mikrofiltracione membrane, ukrupne kompleksiranjem/vezivanjem sa makromolekulima. Na ovaj način nastale čestice postaju veće od pora na membrani i bivaju zadržane na površini i na taj način uklonjene iz vode. Da bi ovaj proces bio primenljiv u praksi potrebno je ostvariti visok fluks i protok prečišćene vode i visok koeficijent zadržavanja jona metala. Stoga bi mogućnost predviđanja separacionih karakteristika u sistemu bila od izuzetnog značaja za primenu kompleksirajuće-mikrofiltracionog procesa. Za predviđanje vrednosti fluksa razvijeni su određeni matematički modeli kao što su model procesa kontrolisanog pritiskom, model teorije gela, model osmotskog pritiska, model otpora. Ipak, nijedan od ovih modela nije zadovoljavajući i ne može opisati sve oblasti u kojima se odvija proces. Uspešnost kompleksiranja kao i sprečavanje pada fluksa zavisi od uslova procesa i različitih radnih parametara. Međusobne veze u sistemu jon metala – makromolekul – membrana su nelinearne i nedovoljno definisane. Veštačke neuronske mreže privlače sve više pažnje kada matematički linearni modeli nisu primenljivi, jer se mogu koristiti ulazni podaci sa nelinearnim odnosima umesto fizički zavisnih odnosa ulaznih vrednosti. Glavni cilj ove disertacije bio je razvoj i optimizacija modela primenom veštačkih neuronskih mreža za predviđanje separacionih karakteristika kompleksirajuće-mikrofiltracionog procesa uklanjanja teških metala iz vode. Istraživanja u okviru ove disertacije bila su podeljena u dva segmenta. U prvom delu, eksperimentalno su ispitani parametri koji mogu uticati na koeficijent zadržavanja i fluks permeata kao što su radni pritisak, pH vrednost rastvora, početna koncentracija jona teških metala, koncentracija agensa za kompleksiranje i prisustvo jedinjenja na bazi aminokiselina kao koliganda. Utvrđeno je da najveći uticaj na proces imaju pritisak, pH vrednost i koncentracija kompleksirajućeg agensa...

Environmental pollution, primarily water, with heavy metals is an environmental problem around the world. The increased concentration of heavy metals in municipal and industrial wastewater represent a serious threat as metals cannot be degraded in nature and some can have toxic effects on plants and animals as well as humans. In order to reduce their concentration to an adequate level defined by legislation, it is necessary to apply purification methods before discharge into recipients. Membrane processes are increasingly used in the field of environmental protection as well as for the preparation and processing of water for the food and pharmaceutical industry, petrochemistry, etc. The main advantages of these processes are low energy consumption, high efficiency, reliability and small amount of waste. The complexing-microfiltration process is a hybrid membrane separation method used to remove heavy metal ions from water. It is based on the concept that small metal ions, which would pass through the pores of the microfiltration membrane, are enlarged by complexation / binding with macromolecules. The particles formed in this way become larger than the pores on the membrane and are retained on the surface and thus removed from the water. In order for this process to be applicable in practice, it is necessary to achieve a high flux and flow of purified water and a high retention coefficient of metal ions. Therefore, the ability to predict the separation characteristics in the system would be extremely important for the application of the complexing-microfiltration process. To predict the flux value, certain mathematical models have been developed, such as the model of the process controlled by pressure, the model of gel theory, the model of osmotic pressure, the model of resistance. However, none of these models is satisfactory and cannot describe all the areas in which the process takes place. The success of complexation as well as the prevention of flux drop depends on the process conditions and various operating parameters. The interactions in the ion metal - macromolecule - membrane system are nonlinear and insufficiently defined. Artificial neural network (ANN) models are attracting increasing attention for use in situations where mathematical linear models are not applicable because they may have nonlinear relationships between variables instead of physical relationships of input values. The main goal of this dissertation was to develop and optimize a model of an artificial neural network to predict the separation characteristics of the complexing-microfiltration process of removing heavy metals from water. The research within this dissertation was divided into two segments. In the first part, parameters that may affect the retention coefficient and permeate flux such as working pressure, pH value of the solution, initial concentration of heavy metal ions, concentration of complexing agents and the presence of compounds based on amino acid as coligand. The greatest influence on the process was found to have the pressure, pH value and concentration of the complexing agent...