Modeling Water Treatment Reactor Hydraulics Using Reactor Networks

Abstract

Reactor hydraulics are integral to water treatment processes such as disinfection and chemical contaminant oxidation. This work uses reactor networksconceptual reactors arranged in parallel and series combinations-to simplify the accurate modeling of residence time in water treatment reactors. Reactor networks were selected for 14 clearwells, ozone contactors, clarifiers, and filters using tracer data sets from literature. For seven of 14 full-scale reactors, two parallel tanks-in-series reactors best balanced accuracy and simplicity. Reactor networks accurately represented tracer data using between two and eight fitting parameters, while segregated flow analysis required 32-164 inputs (two per tracer data point). The modeling work revealed that tracer data sets may have overestimated baffle factors by >10% in 10 of the 14 full-scale reactors, potentially as a result of inaccurate measurement of flow rate or volume. Applications of reactor networks include more accurate approaches to disinfection regulation and modeling the degradation of emerging contaminants.

Assessing flow segregation and mixing by modeling residual disinfectant conversion

“…RN models achieve better fits to nonreactive tracer data as the number of parallel flows increases from one to two and from two to three (Gorzalski et al, ). This can be seen in Figure a, where the parallel TIS RN model achieved a better fit than a single TIS reactor.…”

“…This is consistent with a study by Craik (), who predicted similar inactivation using the SF, TIS, and MM models when the oxidant decay rate approached zero. If a residual oxidant is used, then selecting from PFR t 10 , TIS, SF, or RN models will be governed by the resources available to the modeler, the ultimate use of the model, and potential detrimental consequences of reactor overdesign (Gorzalski et al, ).…”

“…Reactor models for the two clearwells included SF, MM, TIS, and RN models, which are discussed in detail in the SI. A description of these models can also be found in reference textbooks and the literature (Craik, ; Crittenden et al, ; Fogler, ; Gorzalski et al, ; Levenspiel, ; Najm et al, ; Pfeiffer & Barbeau, ). Models were developed from the tracer data in Figure , which shows the normalized tracer concentration ( C / C 0 ) versus θ (normalized run time, θ = t / τ ).…”

“…The number of tanks, n , for TIS models representing clearwells A and B were found to be 5 and 3, respectively, via least‐squares regression. RN models were selected based on fits to nonreactive tracer data using a process described by Gorzalski et al (). The parallel TIS RN model was selected for clearwell A (see diagram in Figure a for fitting parameters).…”

“…The extended t 10 and extended CSTR methods for ozone contactors credit greater Ct by incorporating oxidant decay (USEPA, ) but do not account for micromixing. Alternative reactor models more accurately represent macromixing, including the axial dispersion, tanks‐in‐series (TIS), SF, and reactor network (RN) models (Crittenden et al, ; Ducoste, Carlson, & Bellamy, ; Gorzalski, Harrington, & Coronell, ; Najm, Brown, Gramith, & Hargy, ; Teefy & Singer, ). The TIS, SF, and RN models, which are described in detail in the SI, vary significantly in their assumptions about micromixing (see Table ).…”