Modeling of decolorisation dyes by ozonation techniques using Levenberg-Marquardt neural network

Acid Black 1 (AB1) and Acid Yellow 19 (AY19) are synthetic dyes widely used in industries such as textiles and cosmetics, which have complex structures. Special techniques are needed to degrade the dye before it goes out into the environment. One technique that has been done is to use Advance Oxidation Process (AOPs) using ozonation process. The process requires a combination of values of 4 parameters to obtain the optimum decolorization percentage. The parameters are dye concentration (mg / L), ozone concentration (mg / L), pH, and temperature (oC). In laboratory experiments to obtain a combination of parameters that yield the optimum decolorization percentage, requires relatively high time and cost. This research proposes modeling using Levenberg-Marquardt neural network (LM-NN) to obtain combination of parameters that can yield optimum decolorization percentage. The experimental results with the ozonation method obtained consecutive decolorization percentage values AB1 and AY19 were 93% and 98%. The percentage of decolorization was produced from a combination of 4 parameters, with values of 740 mg/L, 13 mg/L, 6.20, and 380C, while for the AY19 was 830 mg/L, 15 mg/L, 6.20, and 390C. The result of prediction of AB1 and AY19 by using LM-NN model in sequence is 95,25% and 99,99%. The condition was generated from the combination of the 4 parameters, for AB1 the values of each parameter were 707.50 mg/L, 12.70 mg/L, 7.50 and 380C, while the AY19 was 587.50 mg/L, 15.50 mg/L, 7,65, and 330C. The result of experimental and predicted using the LM-NN model, referring to statistical test results with 95% confidence level, indicating that the percentage of decolorization did not occur of difference significantly (p-value> 0.05), both for AB1 and AY19. It shows that modeling with LM-NN can be used to predict the optimum decolorization percentage with predetermined parameters.