Experiments and adaptive-network-based fuzzy inference system modelling in a hybrid up-flow anaerobic sludge blanket reactor to assess industrial azadirachtin effluent quality.

Experimental investigations were carried out for the treatment of industrial azadirachtin effluent in a hybrid up-flow anaerobic sludge blanket (HUASB) reactor continuously for 115 days in three stages at mesophilic temperature (30 - 35˚C). An adaptive-network-based fuzzy inference system (ANFIS) modelling and statistical regression analysis were applied with the raw data. In the ANFIS modelling as well as in the statistical regression analysis, the operating parameters such as initial pH, influent COD, effluent COD and biogas generation (X1, X2, X3 and X4) were taken as variables and effluent BOD values as a response (Y). The average percentage error (APE) values of ANFIS modelling were 2.18, 12.29, and 0.01%, for stage-I, II and III respectively. These values indicated that ANFIS modelling performed well in all the three stages and provided more accurate results.