As the authors have gained more experience in assisting with the transient analysis of combined sewer overflow storage tunnels, a recurring issue has been the choice of design conditions to use in the analysis. The default approach has typically been to select some extreme rainfall event under the assumption that since that provides the most severe limitation to storing the inflow, so it must also provide the worst-case situation with respect to transients. Especially as the knowledge of the importance of trapped air during the filling process has expanded, it seems appropriate to revisit this fundamental assumption to determine whether that expectation is reasonable. Prior to the recognition that air entrapment may play an important role in transient events, the main concern with respect to numerical modeling for design was with surge in tunnel ventilation or access shafts. It seems that even with only this consideration, the choice of a limiting design condition is not straightforward and the additional consideration of air-water interactions makes the choice even more difficult.
A numerical model that can simulate the filling process, albeit with an approximate handling of the air dynamics, was used to investigate this issue in more depth for an array of inflow conditions for a hypothetical tunnel with an idealized geometry. The simulations varied some aspects of the tunnel geometry as well as the nature of the inflow hydrograph. The model outputs included the maximum surge in vertical shafts connected to the tunnel, air demand through the shafts during the transient, volume of entrapped air pockets and several other potentially important parameters. The simulation results are analyzed to define when worst-case conditions may develop. For example, it is observed that the magnitude of computed surges is directly correlated with the inflow rate at the instant that the tunnel becomes completely filled. Although this is not necessarily an unexpected finding, this has not previously been used as a basis for analyzing potential designs. We also observe that trapped air volume is not strongly correlated with tunnel inflow conditions and that large variations can occur. This latter finding suggests that one should be careful to investigate a number of filling scenarios to ensure that the range of system response is evaluated. A number of general design recommendations are presented.