Performance and membrane fouling behaviors analysis with SVR-LibSVM model in a submerged anaerobic membrane bioreactor treating low-strength domestic sewage

Abstract

Abstract A submerged anaerobic membrane bioreactor (SAnMBR) was operated for low-strength domestic sewage treatment over a period of 190 days. Average chemical oxygen demand (COD) removal efficiency ranged from 64.41 to 83.49%, and the removal of ammonia nitrogen (NH 4 + -N) and total phosphorus (TP) were relatively poor. No acidification occurred in the reactor during operation. The performance and membrane fouling behaviors of SAnMBR were investigated and support vector regression model with LibSVM package (SVR-LibSVM) was established to predict performance and analyze the effect of extracellular polymeric substances (EPSs) on membrane fouling. The effect of organic loading rate (OLR) on COD removal efficiency at each stage was also studied according to the model. Both transmembrane pressure (TMP) and total membrane resistance ( R t ) have a significant decline when membrane flushing was carried out. However, physical flushing could alleviate rather than eliminate membrane fouling.