Fouling is an inevitable problem in membrane water treatment plant. It can be measured by trans-membrane pressure (TMP) in the constant flux operation, and chemical cleaning is carried out when TMP reaches a critical value. An early fouilng alarm is defined as warning the critical TMP value appearance in advance. The alarming method was developed using one of machine learning algorithms, decision tree, and applied to a ceramic microfiltration (MF) pilot plant. First, the decision tree model that classifies the normal/abnormal state of the filtration cycle of the ceramic MF pilot plant was developed and it was then used to make the early fouling alarm method. The accuracy of the classification model was up to 96.2% and the time for the early warning was when abnormal cycles occurred three times in a row. The early fouling alram can expect reaching a limit TMP in advance (e.g., 15-174 hours). By adopting TMP increasing rate and backwash efficiency as machine learning variables, the model accuracy and the reliability of the early fouling alarm method were increased, respectively.

The production process of ultra-pure water (UPW) involves dozens of unit processes such as reverse osmosis (RO), pretreatments, membrane degasifier, and several ion exchange processes. Recently, continuous electrodeionization (CEDI) has replaced the 2-bed and 3-tower (2B3T) ion exchange process. As a result, the majority of wastewater in UPW production now comes from the RO concentrate. The important of RO in UPW production is to produce high-quality water with a low ion concentration (around 1 mg/L) for CEDI feed water. Minimizing RO concentrate is essential to reduce the wastewater produced in the UPW production process. This can be achieved by maximizing the recovery of the RO system. However, increasing the recovery is limited by the water quality of the RO permeate. To ensure high-quality permeate water, the RO system is designed with a two-pass configuration. The recovery of each pass in the RO system is limited (e.g., < 85%) due to the expected increase in permeate water concentration at higher RO feed water concentrations. Interestingly, tests using 4-inch RO modules with low concentration feed water (≤ 35 mg/L as NaCl) revealed that the permeate concentration remains almost constant regardless of the feed water concentration. This implies that the recovery of the first RO pass can be increased as long as the average feed/concentrate concentration of the second RO pass is less than 35 mg/L. According to this design criterion for the RO system, the recovery of the first and second RO pass, with a feed water concentration of 250 mg/L as NaCl, can be increased up to 94.8% and 96.0%, respectively. Compared to the conventional RO system design (e.g., 70% and 80% for the first and second RO pass) for UPW production, this maximum recovery design reduces the volume of RO feed and concentrate by up to 38.4% and 89.2%, respectively.

Forward osmosis (FO) process is a chemical potential driven process, where highly concentrated draw solution (DS) is used to take water through semi-permeable membrane from feed solution (FS) with lower concentration. Recently, commercial FO membrane modules have been developed so that full-scale FO process can be applied to seawater desalination or water reuse. In order to design a real-scale FO plant, the performance prediction of FO membrane modules installed in the plant is essential. Especially, the flux prediction is the most important task because the amount of diluted draw solution and concentrate solution flowing out of FO modules can be expected from the flux. Through a previous study, a theoretical based FO module model to predict flux was developed. However it needs an intensive numerical calculation work and a fitting process to reflect a complex module geometry. The idea of this work is to introduce deep learning to predict flux of FO membrane modules using 116 experimental data set, which include six input variables (flow rate, pressure, and ion concentration of DS and FS) and one output variable (flux). The procedure of optimizing a deep learning model to minimize prediction error and overfitting problem was developed and tested. The optimized deep learning model (error of 3.87%) was found to predict flux better than the theoretical based FO module model (error of 10.13%) in the data set which were not used in machine learning.

High-pressure membrane system like nanofiltration(NF) and reverse osmosis(RO) was investigated as a part of water treatment processes to produce high quality potable water with low organic matter concentration through membrane module tests and design simulation. River water and sand filtration permeate in Busan D water treatment plant were selected as feed water, and NE4040-90 and RE4040-Fen(Toray Chemical Korea) were used as NF and RO membranes, respectively. Total organic carbon(TOC) concentrations of NF and RO permeates were mostly below 0.5 mg/l and the average TOC removal rates of NF and RO membranes were 93.99% and 94.28%, respectively, which means NF used in this study is competitive with RO in terms of organic matter removal ability. Different from ions rejection tendency, the TOC removal rate increases at higher recovery rates, which is because the portion of higher molecular weight materials in the concentrated raw water with increasing recovery rate increases. Discharge of NF/RO concentrates to rivers may not be acceptable because the increased TDS concentration of the concentrates can harm the river eco-system. Thus, the idea of using NF/RO concentrate as the raw water for industrial water production was introduced. The design simulation results with feed water and membranes used in this work reveal that the raw water guideline can be satisfied if the recovery rate of NF/RO system is designed below 80%.