This study explored the usefulness and implications of the Bayesian hyperparameter optimization in developing species distribution models (SDMs). A variety of machine learning (ML) algorithms, namely, support vector machine (SVM), random forest (RF), boosted regression tree (BRT), XGBoost (XGB), and Multilayer perceptron (MLP) were used for predicting the occurrence of four benthic macroinvertebrate species. The Bayesian optimization method successfully tuned model hyperparameters, with all ML models resulting an area under the curve (AUC) > 0.7. Also, hyperparameter search ranges that generally clustered around the optimal values suggest the efficiency of the Bayesian optimization in finding optimal sets of hyperparameters. Tree based ensemble algorithms (BRT, RF, and XGB) tended to show higher performances than SVM and MLP. Important hyperparameters and optimal values differed by species and ML model, indicating the necessity of hyperparameter tuning for improving individual model performances. The optimization results demonstrate that for all macroinvertebrate species SVM and RF required fewer numbers of trials until obtaining optimal hyperparameter sets, leading to reduced computational cost compared to other ML algorithms. The results of this study suggest that the Bayesian optimization is an efficient method for hyperparameter optimization of machine learning algorithms.
The ballasted flocculation effects of the mill scale and magnetite on activated sludge were investigated. Both ballasted flocculants (BF) could remarkably improve the sludge settleability in terms of zone settling velocity (ZSV) and sludge volume index (SVI). With the BF dosage of 0.2 to 2.0 g-BF/g-SS, the magnetite particles showed better efficiency on improving settling behavior of activated sludge than the mill scale due to higher surface area and hydrophobic property. The efficiency of SVI30 with magnetite injection was 2.5 to 11.3% higher than mill scale injection and that of the ZSV appreciated from 23.7% to 44.4% for magnetite injection. Averaged floc size of the BF sludge with magnetite dosage (0.5 g-BF/g-SS) was 2.3 times higher than that of the control sludge. Dewaterability of the sludge was also greatly improved by addition of the BF. The specific resistance to filtration (SRF) was reduced exponentially with increasing the dosage of BF. However, the BF’s particle size effect on the SRF looks to be marginal. Consequently, for improving the dewaterability, the BF played a physical role to remove the pore water of the biological flocs by intrusive attachment and a chemical role to induce aggregation of the flocs by charge neutralization.
Osmotic power is to produce electric power by using the chemical potential of two flows with the difference of salinity. Water permeates through a semipermeable membrane from a low concentration feed solution to a high concentration draw solution due to osmotic pressure. In a pressure retarded osmosis (PRO) process, river water and wastewater are commonly used as low salinity feed solution, whereas seawater and brine from the SWRO plant are employed as draw solution. During the PRO process using wastewater effluent as feed solution, PRO membrane fouling is usually caused by the convective or diffusive transport of PRO which is the most critical step of PRO membrane in order to prevent membrane fouling. The main objective of this study is to assess the PRO membrane fouling reduction by pretreatment to remove organic matter using coagulation-UF membrane process. The experimental results obtained from the pretreatment test showed that the optimum ferric chloride and PAC dosage for removal of organic matter applied for the coagulation and adsorption process was 50 mg/L as FeCl3 (optimum pH 5.5). Coagulation-UF pretreatment process was higher removal efficiency of organic matter, as also resulting in the substantial improvement of water flux of PRO membrane.
Recently, various researches have been studied, such as water treatment, water reuse, and seawater desalination using CDI (Capacitive deionization) technology. Also, applications like MCDI (Membrane capacitive deionization), FCDI (Flow-capacitive deionization), and hybrid CDI have been actively studied. This study tried to investigate various factors by an experiment on the TDS (Total dissolved solids) removal characteristics using MCDI module in aqueous solution. As a result of the TDS concentration of feed water from 500 to 2,000 mg/L, the MCDI cell broke through faster when the higher TDS concentration. In the case of TDS concentration according to the various flow rate, 100 mL/min was stable. In addition, there was no significant difference in the desorption efficiency according to the TDS concentration and method of backwash water used for desorption. As a result of using concentrated water for desorption, stable adsorption efficiency was shown. In the case of the MCDI module, the ions of the bulk solution which is escaped from the MCDI cell to the spacer during the desorption process are more important than the concentration of ions during desorption. Therefore, the MCDI process can get a larger amount of treated water than the CDI process. Also, prepare a plan that can be operated insensitive to the TDS concentration of backwash water for desorption.
Chitosan, natural organic polymer, has been applied in water treatment as adsorbent due to non-toxic for human being. The amino group as functional group, can interacts with cation and anion at the same time. The prepared chitosan bead (HCB) was crosslinked to increase chemical stability (HCB-G) and both HCB and HCB-G were prepared to increase physical strength by drying referred to DCB and DCB-G, respectively. The adsorption effect for crosslinking and drying for four types of chitosan bead was tested using pseudo fist order (PFO), pseudo second order (PSO), and intraparticle diffusion model (ID). Regardless of PFO and PSO, the order of K, rate constant, is as followed: HCB > HCB-G > DCB > DCB-G for Cu(II) and phosphate. Drying leading to contraction of bead significantly reduced adsorption rate due to reduce the porosity of chitosan. In addition, crosslingking also negatively effect on adsorption rate. When compared with Cu(II) using hydrogel bead, phosphate showed higher value than Cu(II) for PFO and PSO. The application of ID showed that both hydrogel beads (HCB and HCB-G) obtained a very low R2 ranging to 0.37 to 0.81, while R2 can be obtained to over 0.9 for DCB and DCB-G, indicting ID is appropriate for low adsorption rate.