Comments regarding “Seismic damage analysis due to near‐fault multipulse ground motion” by Guan Chen, Jiashu Yang, Ruohan Wang, Kaiqi Li, Yong Liu, Michael Beer; Earthquake Engineering & Structural Dynamics, 2023

This discussion is based on the paper by Chen et al. in 2023 (hereafter referred to as “the original paper/authors”). In their study, the original authors conducted a series of analyses using nonpulse, single‐pulse, and multipulse ground motion records, evaluating their impact on a frame structure, a slope, and a gravity dam. Their key finding suggests that multipulse ground motion leads to more severe structural damage compared to nonpulse and single‐pulse ground motions. However, it is important to note that the seismic damage analysis of the gravity dam in this paper does not adhere to state‐of‐the‐practice recommendations. Consequently, drawing a definitive conclusion regarding the influence and importance of multipulse ground motion records on the seismic response of concrete dams requires further justification. This necessitates the incorporation of high‐fidelity numerical models and probabilistic performance evaluation. We will discuss the significance of modeling assumptions, specifically addressing the dam–rock dynamic interaction in crack propagation and failure in dams.


INTRODUCTION
With the continuous advancement of computational tools in structural engineering and mechanics, there is an increasing expectation for the refinement of response quantities to enable more accurate risk assessment.The seismic response of complex engineering structures, such as dams and nuclear structures, is influenced by several factors. 1 Among these factors, the modeling strategy and the applied stressor are pivotal.Utilizing low-fidelity numerical models to represent physical phenomena, coupled with deterministic stressors that may not encompass the full spectrum of potential scenarios, can lead to unforeseen failure modes that deviate significantly from reality.When these factors are not appropriately considered within the context of probabilistic simulations, it can introduce bias and result in greater dispersion in response quantities. 2 The reference paper by Chen et al. 3 investigates seismic damage in three types of engineering structures: frame structures, slopes, and concrete dams, utilizing near-fault multipulse ground motions.The study uses a total of 21 ground motion records, divided into three groups of seven records each.Irrespective of whether the ground motions exhibit pulse or nonpulse characteristics, all records are normalized to a peak ground acceleration (PGA) of 0.3 g.These PGA-scaled records are subsequently employed for seismic simulations across all three types of structures.As a primary conclusion, the authors of the original paper assert that, "the seismic damage evaluation shows that multipulse ground motions are prone to cause more severe damage to various structural systems than non-and single-pulse ground motions." While this paper makes a valuable contribution to the field of seismic assessment for structures and infrastructures, it is evident that the experimental design, particularly in the case study involving concrete dams, lacks a systematic approach.Consequently, the general conclusion may require further validation.

Seismic response of the dam
The third case study in the reference paper by Chen et al. 3 centers around modeling a concrete gravity dam, with Koyna Dam serving as a representative example.[6][7][8][9][10][11][12][13][14][15][16][17][18][19] However, in the original paper, the complex dam-water-foundation coupled system is simplified to a trapezoidal-type cantilever column problem.The dynamic interaction between the reservoir and the dam is replaced with the outdated Westergaard assumption, which is not accurate for time-history analysis of dams. 20,21Moreover, a rigid base is assumed for this dam, thereby eliminating considerations for foundation flexibility, dynamic interaction with the dam, and seismic wave propagation in the foundation rock domain.Consequently, the overall conclusion regarding the importance of multipulse ground motion in concrete dams is rendered questionable.
Notably, the authors quantify that the mean damage index resulting from multipulse ground motions is approximately 1.4 times that of single-pulse ground motions and 2.2 times that of nonpulse ground motions.3][24][25][26][27][28][29][30][31][32][33] These studies underscore the critical role of engineering demand parameters (EDPs) in comparing pulse-like and nonpulse motions.For instance, Yazdani and Alembagheri 27 demonstrated that fragility curves based on crest displacement and neck damage index are very close for pulse-like and nonpulse motions, with notable distinctions observed in the case of the base damage index.Contrary to this, fig.14 in the original paper conveys a different message.Furthermore, the forward-directivity characteristics of near-fault records significantly contribute to the intensity and variability of EDPs. 26Consequently, attempting to draw generalized conclusions based solely on the seven records included in the original paper, without a comprehensive exploration of their seismological attributes, presents challenges.
It is important to note that the most recent ICOLD (International Commission on Large Dams) workshop on the nonlinear seismic analysis of concrete gravity dams recommends using nonreflecting boundary conditions for the foundation and accounting for the free-field motions.Such considerations substantially alter the dam's response, including its damage behavior. 34Consequently, the quantified results presented in fig.15 of the original paper may not be valid for concrete gravity dams in general and this specific case study, if it is modeled with state-of-the-practice considerations.
Additionally, a few minor modeling and processing issues should be addressed: (1) for gravity dams, a plain strain model is typically recommended, and (2) the applied formula for quantifying the damage index (i.e., the ratio of damaged elements) can be misleading due to variations in element sizes at different locations, with some elements being 10 times larger than others.A more suitable damage index, as recommended by ICOLD participants, can be found in Salamon et al, 34 with a state-of-the-art damage index available in Hariri-Ardebili and Saouma. 15

Optimal intensity measure
Beyond the concerns related to the modeling of concrete dams and their realistic response, the approach taken to scale (rather than select) the ground motion records and compare them does not entirely align with the overall conclusion presented in the original paper.Specifically, all ground motions are scaled to a random PGA of 0.3 g.However, it remains unclear how such a simplistic scaling impacts other intensity measure (IM) parameters within the ground motion, notably the peak ground velocity (PGV) and spectral velocity (  ).In tab. 1 of the original paper, the authors present  35 the peak spectral acceleration of the scaled records and their corresponding periods.In Figure 1 of this discussion, we have visualized the {  , } pairs alongside the fundamental period of the dam (excluding the fluid and rock portions), represented by a dashed line.
As depicted in Figure 1, it becomes apparent that the fundamental period of the dam differs significantly from the majority of nonpulse ground motions, while it aligns more closely (relatively) with multipulse records.Furthermore, damage to this cantilever-type brittle column (i.e., dam example) elongates the period to approximately twice the elastic period, approximately 0.66 s.Consequently, it is not surprising that the structural response (and damage index) resulting from nonpulse motions is lower than that of the other two categories.
It is essential to note that PGA-based normalization (and comparison) is an outdated method, as numerous studies have shown that PGA is not the optimal IM parameter and can lead to significant dispersion even in SDOF systems.The optimal IM parameter for concrete gravity dams has been extensively discussed in. 35,36In particular, 27 examined the optimal IM for a large dataset comprising 75 pulse-like and 60 nonpulse ground motions.Their findings suggested that the pulse nature tends to yield more varied response results.This study also recommended velocity and displacement response spectra at the fundamental period of the dam as optimal IMs for near-fault records.
All these studies collectively argue that the most robust approach to comparing pulse-like and nonpulse motions involves establishing a Probabilistic Seismic Demand Model (PSDM) that simultaneously accounts for both IM and EDP relationship.In the original paper, such a relationship is not provided, nor is the optimal IM parameter utilized.Consequently, the validity of the results presented in fig.15 (in the original paper) is questionable.
Another crucial aspect in ground motion analysis is the energy imposed to the structure.Various energy-related metrics exist, with Arias intensity (  ) being the most renowned.Fig. 3a in the original paper illustrates normalized   (scaled to 100%), offering limited insight into the actual energy imposed on the structure by different ground motion records.This figure merely indicates that the energy input to the system occurs between 30 and 70 s for all records.To provide further clarification, we utilized 100 ground motion records from Hariri-Ardebili and Saouma 35 for a gravity dam to develop both normalized and non-normalized Arias intensity plots, as displayed in Figure 2. Comparatively, the normalization process demonstrates a similar dispersion among the curves as seen in the original paper (fig.3a).However, in reality, there is a fundamentally distinct nature and variability when considering non-normalized curves.

FINAL RECOMMENDATION
The primary aim of the original authors was to demonstrate that multipulse ground motions induce higher structural responses and more significant damage in various structures, especially concrete gravity dams.In light of this, the discussers propose that the comparison between multipulse, single-pulse, and nonpulse ground motions should follow a systematic approach, such as Probabilistic Seismic Demand Modeling (PSDM).The quantification of damage indices using the PGA-scaled method in the original paper may or may not be valid if an optimal IM is identified and employed.Furthermore, the discussers do not recommend adopting the quantified damage index for the concrete gravity dams presented in the original paper.This hesitation arises from the oversimplifications made by the authors in their approach.The case study, as presented by the authors, does not align with state-of-the-practice recommendations for nonlinear seismic analyses of dams.Constructing a model that accounts for proper foundation-dam dynamic interaction and simulates wave propagation in the rock medium would likely yield substantially different results.
In summary, while the original paper offers valuable insights into the impact of multipulse ground motions on structural responses and damage, the discussers urge caution in drawing generalized conclusions.A more systematic approach, considering optimal IMs and adhering to best practices in seismic analysis, should be employed to ensure the robustness of such findings in the field of earthquake engineering and structural dynamics.

A C K N O W L E D G M E N T S
The authors have nothing to report.

D ATA AVA I L A B I L I T Y S TAT E M E N T
None.

F I G U R E 1 2
−  relationship for the scaled ground motions in three categories (collected from tab. 1 in the original paper).Comparison of normalized versus original   -time curves with data from Hariri-Ardebili and Saouma.