Skip to Main Content
 

Global Search Box

 
 
 
 

ETD Abstract Container

Abstract Header

MULTI-OBJECTIVE OPTIMIZATION OF CORE WATER TREATMENT PROCESSES INCORPORATING ULTRAVIOLET AND FLUORESCENCE SPECTROSCOPIC MEASUREMENTS

Arjmandi, Parvin, Arjmandi

Abstract Details

2017, Doctor of Philosophy, University of Akron, Civil Engineering.
This study presents a multi-objective optimization (MO) evaluation and comparison of the trade-off between important water quality treatment parameters and the treatment cost for the Akron Water Supply (AWS) in Akron, Ohio. The well-known NSGA_II algorithm was chosen for the MO solution search for three objective functions and multiple MO cases were evaluated. Model functions were developed and categorized based on fluorescence peaks and UV254 absorbance measurements as inputs. Specific model functions included a Settled Cr model, DOC-TI (dissolved organic carbon treatment index), settled turbidity models, and chemical cost. The dissolved organic carbon and particulate functions were built using the multi-layer perceptron (MLP) method with genetic algorithm (GA) and simulated annealing (SA) search algorithms and the chemical cost was calculated directly as a function of chemical dose for four chemicals (potassium permanganate, chlorine dioxide, alum, and powdered activated carbon). Fluorescence peaks for 554 days were obtained via a preprocessing tool generated in MATLAB and were utilized in fluorescence-based MLP models. Five MLP models were developed and mean squared error (MSE), R-Squared, and prediction interval averaged weight (PIAW) of each model were evaluated as model performance factors. Parametric analysis was also done to examine model behavior. Only the models that satisfied both statistical performance and “reasonable” physical behavior were evaluated further. The final accepted models could predict unseen data with 51% and 91% R-squared, for the settled Cr and the fluorescence-based settled turbidity, and 85% and 23% R-squared, for the settled UV254 and the UV-based settled turbidity respectively. The NSGA-II optimized the objective functions by adjusting the chemical dose for alum, PAC, KMnO4, and ClO2. The number of solutions was shown to have high sensitivity to constraints and the lower and upper bounds. Although, the constraints were defined based on the simulated values of the objective functions and their descriptive statistics, about 98% of 554 days could be optimized such that the cost saving percent of the total cost of these days had a median 14.40 percent and 14.96 percent in a fixed PAC dose condition, and 16.68 percent and 17.03 percent for a varied PAC dose condition for fluorescence-based and UV-based cases, respectively. The obtained cost saving of all 554 days was the result of the high PAC dose saving during the optimization process. The MO process was done for two different PAC dose initializations- one, for a fixed PAC dose and the other one, for a varied PAC dose- that both could increase the cost saving. The NSGA-II performance for 554 days was done by an 2D hypervolume (HV) method as a known metric. The HV of the fluorescence-based and the UV-based MO Pareto fronts for each day were plotted and the bounded area of each front and a defined reference point was calculated. Because, the reference point was chosen to cover both Pareto fronts of a single day, it brought some limitation in HV calculation such that the it could be evaluated for the days that their Pareto fronts had an overlap. The HV values for 554 days showed a small difference for fluorescence-based and UV-based cases. The fluorescence-based case obtained 1.88 and 3.74, and the UV-based MO case showed 2.16 and 4.53 for two different PAC dose conditions. The NSGA-II could also optimize the alum dose on average by 7.53 mg/L and 6.29 mg/L for the fluorescence-based and the UV-based MO cases in a fixed PAC dose condition, and 11.06 mg/L and 6.33 mg/L when PAC dose was varied. The results of this study demonstrate that fluorescence and UV-based inputs for treatment process models for particulate and dissolved organic carbon removal could provide cost-savings of more than 10% daily. Validation of this approach for other water sources could be beneficial.
Christopher Miller, Dr (Advisor)
Stephan Duirk, Dr (Committee Member)

Recommended Citations

Citations

  • Arjmandi, Arjmandi, P. (2017). MULTI-OBJECTIVE OPTIMIZATION OF CORE WATER TREATMENT PROCESSES INCORPORATING ULTRAVIOLET AND FLUORESCENCE SPECTROSCOPIC MEASUREMENTS [Doctoral dissertation, University of Akron]. OhioLINK Electronic Theses and Dissertations Center. http://rave.ohiolink.edu/etdc/view?acc_num=akron1501345847002254

    APA Style (7th edition)

  • Arjmandi, Arjmandi, Parvin. MULTI-OBJECTIVE OPTIMIZATION OF CORE WATER TREATMENT PROCESSES INCORPORATING ULTRAVIOLET AND FLUORESCENCE SPECTROSCOPIC MEASUREMENTS. 2017. University of Akron, Doctoral dissertation. OhioLINK Electronic Theses and Dissertations Center, http://rave.ohiolink.edu/etdc/view?acc_num=akron1501345847002254.

    MLA Style (8th edition)

  • Arjmandi, Arjmandi, Parvin. "MULTI-OBJECTIVE OPTIMIZATION OF CORE WATER TREATMENT PROCESSES INCORPORATING ULTRAVIOLET AND FLUORESCENCE SPECTROSCOPIC MEASUREMENTS." Doctoral dissertation, University of Akron, 2017. http://rave.ohiolink.edu/etdc/view?acc_num=akron1501345847002254

    Chicago Manual of Style (17th edition)