Stochastic Simulations of Process Performance in UV Disinfection Systems
Yousra Mohamed Hamdy Ahmed1, Xing Li1, Angela Ortiz1, Chengyue Shen2, O. Karl Scheible2, Daphne Chiu3, Geeta Rijal4, Keith Mahoney5, Bruno Ferran6, XXX7, Ernest R. Blatchley III1,8
1Lyles School of Civil Engineering, Purdue University, West Lafayette, IN
2HDR Engineering, Mahwah, NJ
3Citizens Energy Group, Indianapolis, IN
4Metropolitan Water Reclamation District of Greater Chicago, Chicago, IL
5New York City Department of Environmental Protection, New York, NY
6Ozonia North America, Leonia, New Jersey
7LIT
8Division of Environmental & Ecological Engineering, Purdue University, West Lafayette, IN
Open channel configurations are widely used in UV disinfection systems for wastewater. Due to the size of the open channel UV disinfection systems, evaluating their performance experimentally at the design stage is often not economically feasible, especially for design optimizations.
The central hypothesis of this work is that CFD-I models (which are essentially deterministic, by nature) can be applied via a stochastic approach to simulate process performance, including variability, by allowing appropriate variations in input variables. As such, this approach has the potential to yield numerical simulation results that can be used to refine reactor design and tailor operating conditions, so as to improve reactor performance and efficiency. Figure 1 illustrates variability in the concentration of viable E. coli that entered the open channel UV disinfection system at the Belmont WWTP in Indianapolis, IN (N0), as well as the UV254 dose-response behavior of these same organisms.CFD simulations are being conducted on full-scale and pilot-scale, open-channel UV disinfection systems using Fluent (Ansys Software). I-field simulations are being conducted following a ray-tracing approach using Photopia (LTI Optics, Westminster, CO). Dose distribution estimates for each operating condition are developed via a Lagrangian approach, by mapping of simulated particle trajectories through the simulated I-field of each reactor. Process performance is simulated by integration of the dose distribution with target microorganism dose-response behavior following the segregated-flow model.The input parameters of the systems were measured using standardized experimental methods (e.g. Ambient Biodosimetry and Collimated beam experiments) and on-site measuring devices.
Monte Carlo simulations have been conducted to compare the measured behavior of the UV disinfection system from biodosimetry (AB) at Belmont with the predictions of the stochastic model. As a first cut, the conditions corresponding to the 14 AB experiments conducted at the Belmont were simulated. Variability of dose-response behavior and N0 of E. coli, as defined in Figure 1, were used as input parameters, along with measured values of flow rate, transmittance, and lamp usage for each test. A comparison of these simulation results with measurements from the AB tests (as mean values) is provided in Figure 2 and MC simulation of these tests is shown in Figure 3. In general, the simulations and measured behavior were in good agreement across these tests. It is also important to note that for all 14 conditions tested, the concentration of viable E. coli was well below the regulatory limits that apply to the Belmont facility. These simulations are being expanded to incorporate other sources of variability, and to extend to the other full-scale UV disinfection systems that are included in this research.
The results of this project illustrate the potential for stochastic simulations to account for and accurately describe the performance of UV disinfection systems. Incorporation of these methods offers the potential to substantially reduce the capital and operating costs, as well as energy usage associated with UV disinfection systems by providing more accurate and comprehensive descriptions of their performance. Moreover, because these methods provide a detailed accounting of the dose distribution delivered by a UV reactor, it is possible to accurately predict the performance of the reactor relative to new or emerging treatment endpoints.