Closure to “skeletonizing Pipes in Series Within Urban Water Distribution Systems Using a Transient-Based Method” by Yuan Huang, Feifei Zheng, Huan-Feng Duan, Tuqiao Zhang, Xinlei Guo, and Qingzhou Zhang
Yuan Huang,Feifei Zheng,Huan‐Feng Duan,Qingzhou Zhang
DOI: https://doi.org/10.1061/(asce)hy.1943-7900.0001727
IF: 2.4
2020-01-01
Journal of Hydraulic Engineering
Abstract:It was interesting to see the topic of skeletonization arising again, and the authors have obviously given some serious thought to it. The original paper proposes a modification to earlier published transient-based approaches (e.g., Wylie and Streeter 1993), striving to ensure that skeletonized systems can reasonably capture the overall transient properties of the original pipe system. The method focused on the skeletonization of pipes in series having no branch lines and no intermediate demands. In the original paper, in order to capture the essential wave propagation dynamics, both a transient phase criterion [i.e., Eq. (3) or (6)] and a transient amplitude criterion [i.e., Eq. (4) or (11)] are combined with traditional hydraulic equivalence theory based on steady-state analysis [i.e., Eq. (5)]. The result is then incorporated into the skeletonization process for pipes in series. Three assessment metrics [errmax in Eq. (14), errmin in Eq. (15), and errcum in Eq. (16)] are proposed to quantitatively measure the impacts of model skeletonization on the system’s transient dynamics. The discussers believe that model skeletonization is a must-have process after collecting information from a geographic information system, since a large number of dummy nodes are almost always created. Normally, such dummy nodes do not contribute to the accuracy of transient analysis and thus should be removed as part of the traditional skeletonization process. The point of this discussion is not to devalue the importance of model skeletonization in general, but rather to ask if the “special treatment” of skeletonization for transient analysis is necessary, valuable, and safe in practice. Thus, in summary, for sequences of pipes that are purely in series and have no internal demand, the discussers wonder if the equivalent pipe with artificially adjusted diameter, wave speed, and friction factor provides a sufficient benefit to the efficiency of the hydraulic calculation. The challenge is that, really, any transientbased skeletonization method is designed to omit specific interactions and wave reflections and transmissions that occur in the original series pipes, a replacement that must in someway alter their wave characteristics. The traditional skeletonization is to reduce large water distribution systemmodels while maintaining, through the lens of steady-state analysis, a nearly identical hydraulic behavior of the larger original model; however, the transient-based skeletonization method unavoidably modifies transient behavior. The level of modification, represented as errmax, errmin, and errcum can be minimized with the reduced Etol in the authors’ Eq. (4), but the reduced Etol inevitably tends to minimize the level of model skeletonization, as presented in authors’ Table 3 and Fig. 8. The numerical and conceptual benefits of model skeletonization are clearly proportional to the level of skeletonization with the goal of significantly reducing analysis costs. However, if the same process jeopardizes the accuracy of the analysis (e.g., the case of Etol 1⁄4 1 in Fig. 8), the discussers do not recommend transient-based model skeletonization, particularly since modern computers are so much faster than the pioneering ones used in the past. Also, it is pointless to apply the transient-based skeletonization to simply eliminate several single internal nodes [e.g., Fig. 7(a)] for better accuracy while it still cannot maintain the same hydraulic behavior of the original model. In addition, the computational accuracy and efficiency of the transient-based skeletonization are significantly influenced by computational time step, number of internal reaches, various Courant number constraints, and whatever interpolation or wave speed adjustment approaches are adopted. For example, the results shown in the authors’ Table 3 and Figs. 5 and 6 will likely vary considerably depending on the settings of the computational parameters. Also, the method of characteristics applied to the transient analysis has the inherent requirement of calculating internal reaches depending on time step and reach size. The equivalent—skeletonized—long pipe can be easily divided into small pipes for the internal reach calculation, so the computational benefit would not be achieved in these cases despite the distortion of implementing transient-based model skeletonization. There is another, perhaps even more, concerning point that is of great importance in transient analysis. Skeletonization inevitably loses at least some elevation information at internal nodes, so the simplified system cannot capture the transient pressure heads at these internal points. In extreme cases, then, the simplified system may ignore the potential risk of cavitation or water column separation in the internal nodes, which may well influence the interpretation of the safety of components both inside and outside the scope of the skeletonized portion. The practical analyses and interpretation of the interaction of transient waves and events in complex systems are inevitably difficult to assess. Perhaps, in the end, the authors again raise the traditional and unresolvable tension between model simplicity and model accuracy. Certainly, they have made a bold and reasoned attempt toward a better understanding of model skeletonization, including transient dynamics. However, taking into account the unavoidable modifications to the system’s transient behavior, some important questions arise. First, what level of model skeletonization is recommended considering the distortion of transient dynamics? Second, what is the implication of skeletonization for other numerical factors such as computational time step, number of internal