COMPARISON, ARTIFICIAL NEURAL NETWORK MODELING AND GENETIC ALGORITHM OPTIMIZATION OF THE RESINOID AND POTASSIUM YIELDS FROM WHITE LADY'S BEDSTRAW (Galium mollugo L.) BY CONVENTIONAL, REFLUX AND ULTRASOUND-ASSISTED AQUEOUS--ETHANOLIC EXTRACTION.

In this work, the yields of resinoid and potassium were obtained from aerial parts of white lady's bedstraw (Galium mollugo L.) by maceration, reflux extraction and ultrasound-assisted extraction using aqueous ethanol solutions as solvents. The main goal was to define the influence of the extraction technique and the ethanol concentration on the resinoid and potassium yields. The resinoid and potassium yields were determined by the solvent evaporation from the liquid extracts to constant weight and the AAS emission method, respectively. The dependence of resinoid and potassium yields on the ethanol concentration was described by linear and quadratic polynomial models, respectively. The best potassium extraction selectivity of 0.077 g K/g of dry extract was achieved by maceration at the ethanol concentrations of 10 g/100 g. The artificial neural network (ANN) was successfully applied to estimate the resinoid and potassium yields based on the ethanol concentration in the extracting solvent and the time duration for all three extraction techniques employed. The response surface methodology was also used to present the dependence of ANN results on the operating factors. The extraction process was optimized using the ANN model coupled with genetic algorithm. The maximum predicted resinoid and potassium yields of 30.4 and 1.67 g/100 g of dry plant were obtained by the ultrasonic extraction (80 min) using the 10 g/100 g aqueous ethanol solution. [ABSTRACT FROM AUTHOR]

U radu je prikazan prinos rezinoida i kalijuma dobijenim iz nadzemnih delova belog ivanjskog cveća (Galium mollugo L.) konvencionalnom (maceracijom), refluks i ultrazvučnom vodeno-etanolnom ekstrakcijom. Glavni cilj je bio da se definiŠe uticaj ekstrakcionih tehnika i koncentracije etanola na prinos rezinoida i kalijuma. Prinos rezinoida, odnosno kalijuma određen je isparavanjem rastvarača iz tečnih ekstrakata, odnosno metodom atomske apsorpcione spektrometrije. Zavisnost prinosa rezinoida i kalijuma, kao i selektivnosti kalijuma od koncentracije etanola matematički je opisano linearnim, kvadratnim polinomnim modelom. Najveća selektivnost kalijuma postiže se maceracijom i iznosi 0,077 g K/g suvog ekstrakta. VeŠtačka neuronska mreža (VNM) je uspeŠno primenjena za predviđanje prinosa rezinoida i kalijuma na osnovu koncentracije etanola pri ekstrakciji rastvaračem i vremenskog trajaja za sve tri primenjene tehnike ekstrakcije. Metodom odziva povrŠine predstavljena je zavisnost rezultata VNM od operativnih uslova. Proces ekstrakcije je optimizovan primenom genetičkog algoritma (GA) na rezultate VNM. Određeno je da optimalna koncentracija etanola iznosi 10 g/100 g, pri kojoj se postiže maksimalni prinos rezinoida od 30.4 g/100 g i kalijuma od 1.67 g/100 g ultrazvučnom ekstrakcijom od 80 min. Relativno procentno odstupanje optimalnih prinosa od eksperimentalno dobijenih prinosa za isto vreme iznosi ±9.3% ,odnosno ±0.3%. [ABSTRACT FROM AUTHOR]