With the current trend of the fourth industrial revolution, machine learning technique is increasingly adopted in various water industry fields. In this review paper, recent studies using machine learning to predict flood, water consumption, water quality, and water treatment processes are summarized. In the typical water purification processes such as flocculation, disinfection, and filtration, machine learning was able to present high-accuracy prediction results for complex non-linear mechanisms. Hybrid machine learning methods, combining multiple algorithms, generally outperformed machine learning results using only one algorithm. A more microscopic machine learning approach can provide valuable information to the operators in the water industry.
Methanogenic community shift and comparison were determined by 454 pryosequencing for two different full-scale anaerobic digesters treating municipal sludge. For monitoring long-term of microbial communities, samples were collected for two year at three-monthly basis. The two mesophilic AD bioreactor were operated at similar operating conditions, but different substrate streams. Methanospirillum were identified as the key drivers of methanogenesis in full-scale anaerobic digester treating municipal sludge. In Joongrang (JR) digester, Methanospirillum was dominant (48%±10.3) over almost all period, but the dominant genus move to Methanosaeta and Methanoculleus due to low acetate concentration (0.02 g/L), total ammonia nitrogen concentration, respectively. In Asan digester (AS), Methanospirillum also was dominant (41%±12.6) like JR digester, but methanogenic community shift was examined twice. One of those was from Methanospirillum to Methanophaerula due to pH sharply decrease (<5.5) and second shift was Methanosaeta increase due to low VFAs concentration (0.25 g/L).