Date of Graduation

5-2021

Document Type

Dissertation

Degree Name

Doctor of Philosophy in Engineering (PhD)

Degree Level

Graduate

Department

Chemical Engineering

Advisor/Mentor

Jamie A. Hestekin and Christa N. Hestekin

Committee Member

David M. Ford

Second Committee Member

Wen Zhang

Third Committee Member

Julian Fairey

Keywords

Electrodeionization, Microalgae, Supervised Learning, Wastewater treatment

Abstract

Wastewater has a serious impact on environment and public health due to its high concentration of nutrients and toxic contaminants. Without proper treatment, excess nutrients discharged in wastewater can cause a damage to the ecosystem such as undesirable pH shifts, cyanotoxin production, and low dissolved oxygen concentrations.

Main objectives of this dissertation work were to investigate i) the biofuel potential of P. cruentum when grown in swine wastewater, ii) the influence of four most commonly used ion exchange resins on the system efficiency and selectivity for the removal of sodium, calcium, and magnesium ions, and iii) the modeling of wafer-enhanced electrodeionization with data science and machine learning techniques.

The growth and lipid production of the microalgae Porphyridium (P.) cruentum grown in swine wastewater (ultra-filtered and raw) were examined as compared with control media (L−1, modified f/2) at two different salt concentrations (seawater and saltwater). The cultivation of P. cruentum in the treated swine wastewater media (seawater = 5.18 ± 2.3 mgl−1day−1, saltwater = 3.32 ± 1.93 mgl−1day−1) resulted in a statistically similar biomass productivity compared to the control medium (seawater = 2.61 ± 2.47 mgl−1day−1, saltwater = 6.53 ± 0.81 mgl−1day−1) at the corresponding salt concentration. Furthermore, no major differences between the fatty acid compositions of microalgae in the treated swine wastewater medium and the control medium were observed.

The performance comparison of four commonly used cation exchange resins (Amberlite IR120 Na+, Amberlite IRP 69, Dowex MAC 3 H+, and Amberlite CG 50) and their influence on the current efficiency and selectivity for the removal of cations from a highly concentrated salt stream were also reported in this work. The current efficiencies were high for all the resin types studied. Results also revealed that weak cation exchange resins favor the transport of the monovalent ion (Na+) while strong cation exchange resins either had no strong preference or preferred to transport the divalent ions (Ca2+ and Mg2+). Moreover, the strong cation exchange resins in powder form generally performed better in wafers than those in the bead form for the selective removal of divalent ions (selectivity > 1). To further understand the impact of particle size, resins in the bead form were ground into a powder. After grinding the strong cation resins displayed similar behavior (more consistent current efficiency and preference for transporting divalent ions) to the strong cation resins in powder form. This indicates the importance of resin size in the performance of wafers.

Through this research, the modeling of wafer-enhanced electrodeionization with high concentration multi-ion solution has been accomplished. This paper is the first study that uses data science and machine learning techniques for the modeling of wafer-enhanced electrodeionization with high concentration multi-ion solutions. With the use of data science and machine learning, the sodium, calcium, and magnesium ion concentrations were predicted with multioutput regression and neural networks multilayer perceptron (NN-MLP), and the observed effects of different resin wafers were confirmed using both multioutput and single output regression as well as leave-one-out cross validation and NN-MLP.

Share

COinS