• No se han encontrado resultados

Contenidos desarrollados

In document Carrera por Internet Fase I (página 63-73)

CAPÍTULO 3 3 ANALISIS DE COSTOS

4.3 Diseño del Sitio Web

4.3.5 Contenidos desarrollados

The preceding sections provided examples of how predicted LRV for a given Da and required Da for a target LRV are dependent on the reactor model used to represent the RTD. A similar analysis was performed for all 35 reactors, and this section discusses the corresponding results. Figure 4.4 shows plots of log reduction versus Da for the PFR, PFR t10, TIS, SF, and RN models for three example reactors. Plots for all 35 reactors are included in Figure C.2 through Figure C.6 of APPENDIX C. As expected, the PFR model predicts the highest log reduction for any given Da. PFR t10 predicts the lowest log reduction up to a given Da that varies from reactor to reactor. At higher Da, the PFR t10 model predicts more removal than the TIS, SF, and RN models.

121

Figure 4.4. Log reduction versus Da plots for three example reactors. Selected reactors provide examples where (A) SF model produces errant predictions due to nonmonotonic tracer data, (B) TIS, SF, and RN models yield similar predictions, and (C) TIS model predictions differ

significantly from SF and RN models. APPENDIX C contains these plots for all 35 reactors, and panels (A), (B), and (C) correspond to reactors 3H, 3F, and 5B shown in Figure C.3 in

122

For some reactors, the LRV predicted by the SF approached infinity when Da exceeded a certain threshold (e.g., Figure 4.4A). This is an artifact resulting from nonmonotonic tracer data. When the observed tracer concentration decreases, particularly early in the tracer run, the SF model predicts negative contaminant concentrations for that particular PFR, which when

summed with the other hypothetical PFRs produces an errantly high prediction of log reduction. Nonmonotonic tracer curves could result from fluctuating background concentrations of the tracer or from inaccuracies in measuring tracer concentrations. This should be considered a limitation of the SF model: it should not be used for nonmonotonic tracer data due to erroneous predictions that become increasingly significant above a threshold Da or log reduction.

Predictions of log reduction for TIS, SF, and RN models were similar at lower log reductions, with differences in predictions occurring at higher log reductions. This indicates that the higher the target LRV, the more important the selection of reactor model. Depending on the observed RTD and the corresponding fit of reactor models, the TIS may yield similar predictions to the more complex RN and SF models as shown in Figure 4.4B. However, there were also reactors where the TIS model did not have the same flexibility to represent the observed RTD, resulting in predictions that differed significantly from the SF and RN models (e.g., Figure 4.4C).

Box and whisker plots were created to visualize the range of Da required for the 35 reactors to achieve a target log reduction using different models. These results, shown in Figure 4.5, excluded the PFR and SF models. The PFR model was excluded because it underestimates required Da (see Figure 4.2 through Figure 4.4), and the SF model was excluded due to the issue with nonmonotonic tracer data discussed earlier (Figure 4.4A and corresponding discussion). The results shown in Figure 4.5 are applicable to any first-order reaction where oxidant concentration

123

is constant. At low log reductions such as 0.5-log, higher Da would be required if the PFR t10 model was used than if the TIS or RN models were used. This supports the use of the PFR t10 model in disinfection regulation when conservative reactor design and operation is desirable for LRVs of 0.5.

At higher log reductions, the TIS and RN models require higher Da than the PFR t10 model. For example, the median values of Da for the TIS and RN models were approximately twice as high as for the PFR t10 model when targeting 6-log reduction. Because the TIS and RN models more accurately represent RTD than the PFR t10 model, especially at the lower residence time portions of the RTD curve (Chapter 2), the TIS and RN models are more appropriate for processes targeting 6-log reduction. This finding should inform the selection of RTD model in the design and regulation of disinfection, particularly in water reuse applications where higher log reductions are required.

The PFR t10, TIS, and RN models required similar Da at 3.0-log, but significantly different Da at 0.5-log or 6.0-log (see Figure 4.5). The log reduction versus Da figures in APPENDIX C show that predictions from the PFR t10, TIS, and RN models often intersect around 2.0- to 4.0-log. The PFR t10 model predictions are likely most accurate in this range. The following section will examine at what log reduction the PFR t10 model ceases to be

conservative, which is also the point at which it is most accurate (i.e., produces the same prediction as the RN model).

124

Figure 4.5. Required Da to achieve different log reductions using PFR t10, TIS and RN models. Note that of 35 reactors, one at the 6-log treatment target would require values of Da greater than 1000 using the RN model, and thus is not shown on the plot. This box and whiskers plot provides the 25th percentile, median, and 75th percentile as horizontal lines. The vertical lines represent observations that are within 1.5 times the interquartile range. Individual data points are outliers that are outside the range represented by the vertical lines.

In document Carrera por Internet Fase I (página 63-73)

Documento similar