The new seismic hazard model ESHM20 of Europe: Investigating the implications to the seismic design and risk assessment of major industrial facilities across Europe

Abstract

The seismic performance and safety of major European industrial facilities have a global interest for the whole of Europe and its citizens. However, the seismic design of these facilities is based on national, sometimes outdated seismic hazard analyses. The results of the commonly developed, fully harmonized newly released European Seismic Hazard Model ESHM20 provide a pertinent reference for seismic hazard at European scale and have been officially adopted as an “acceptable representation of the seismic hazard in Europe” in the ongoing revision of Eurocode 8. This study presents a large-scale investigation of the impact of the potential adoption of ESHM20 on the design of new industrial facilities as well as on the potential seismic risk of existing facilities at European level with respect to the current seismic codes. The horizontal elastic response spectra using ESHM20 in combination with the revised Eurocode 8 for selected industrial sites are compared with the respective response spectra of the national regulations for return periods of 475, 2500 and 5000 years. In addition, a single containment LNG tank is analysed for an industrial site in Germany using the design approach for liquid filled tanks according to the revised Eurocode 8. Furthermore, a deterministic and probabilistic seismic risk assessment of a vessel installed in a five-storey frame is performed for industrial sites in Greece and Germany. The examples show that a consistent procedure all over Europe would be desirable and a benefit for engineers in terms of comparability and achievement of the same safety targets. Based on the conducted illustrative studies, the consequences of a potential adoption of the revised Eurocode 8 and ESHM20 hazard maps are discussed and summarized with respect to the standard harmonisation process in Europe.

1 Introduction

In a globalized world the seismic performance and safety of major European industrial facilities, including essential petrochemical and energy production facilities, have a global interest for the whole of Europe and its citizens. A potential major disaster that may occur in a critical facility somewhere in Europe could affect several countries, probably far beyond the country where it is located. It is well known that severe seismic damage to industrial facilities with loss of life and injuries, damage to natural and built environment, as well as significant economic losses have been reported after many earthquake events (Campedel 2008; Cruz and Valdivia 2011; Erdik und Uckan 2013; Krausmann et al. 2011). One recent example is the one from the 2012 Emilia (Italy) earthquake sequence, which struck a highly industrialized area, causing severe damages and even partial or total collapses to many industrial prefabricated buildings which had not been designed for seismic loads (Minghini et al. 2016; Savoia et al. 2017). Seismic design and safety assessment of facilities all over Europe is practically based on national, often outdated seismic hazard analyses. To this respect, the results of the commonly developed, fully harmonized newly released European Seismic Hazard Model ESHM20 (Danciu et al. 2021), available at http://www.efehr.org, provide a pertinent reference for seismic hazard at European scale. ESHM20 was developed within the EU funded project “Seismology and Earthquake Engineering Research Infrastructure Alliance for Europe" (SERA, http://www.sera-eu.org). The main objectives of the project were to (a) develop an updated, fully harmonized probabilistic seismic hazard model for the Euro-Mediterranean region, (b) facilitate the development of a European seismic risk model (Crowley et al. 2021) and (c) interact with CEN/TC250/SC8 committee to ensure that the ESHM20 output will serve additional engineering requirements as part of the update and revision of Eurocode 8. As a result of this last objective, specific outputs of ESHM20 have been adopted as a reference for the definition of seismic action (elastic response spectra) in the upcoming revision of Eurocode 8 (prEN 1998-1-1 2022). ESHM20 is therefore the most up-to-date homogeneous European seismic hazard model and could be used for European-scale seismic risk investigations.

One of the main particularities of typical industrial facilities is that their production units consist of the primary load-carrying structure and various process engineering components. In addition, supply structures such as free-standing tanks and silos are required for each plant to ensure the supply of material and product storage. Thus, design and construction rules are needed for the primary, secondary, and supply structures. Within the framework of these rules, interactions of primary and secondary structures must also be considered (Butenweg et al. 2020). So far, in most of the national codes in European countries, importance factors are applied in seismic design to scale the hazard values of the return period of 475 years, used for the design of ordinary buildings, so that the higher risk potential of the plant compared to conventional building structures is sufficiently considered. However, the importance factors given in the current version of EN 1998-1 (2004) cannot be generally applied to industrial facilities because they refer solely to standard buildings with their corresponding damage potential. Furthermore, the importance factors given in EN 1998-4 (2006) are limited to silos, tanks or pipelines. Suitable importance factors for industrial facilities should consider the impact on human, health and safety, environment, and lifeline functionalities within the plants. Such a comprehensive approach is presently not available in national or international standards. A reasonable approach is described in the VCI Guidelines “The seismic load case in plant construction” (VCI-Guideline 2022a, b) which complements the rules of EN 1998-4 (2006) for industrial facilities. Similar guidelines were developed in France, i.e. “Guide méthodologique pour la conception, l’installation et le diagnostic des équipements en zone sismique” (AFPS 2011). The values of the importance factors given in these documents were defined upon the basis of interdisciplinary expertise and have been checked against international standards.

Usually, importance factors applying a simple linear scaling of response spectra to a reference value are proposed in standards for higher return periods. However, while they represent a relative safety level with respect to the reference value, they cannot be directly compared to the values resulting for specific higher return periods. Such a linear scaling can thus only constitute a rough approximation, since, for different return periods, both spectral shapes and the boundaries of seismic zones are different. It is certainly more meaningful to adopt a seismic hazard map for a higher return period instead of applying importance factors. Such are available in some national codes, e.g. the Italian seismic code NCT (2018) provides seismic hazard for return periods of 975 and 2475 years. At European scale ESHM20 (Danciu et al. 2021) provides such seismic hazard map at pan-European scale, i.e. for the return periods of 975, 2500 and 5000 years. The use of the ESHM20 maps for different return periods eliminates the errors related to simple scaling of the code spectra and provides a reliable basis for probabilistic and performance-based design, which allows an individually tailored design of facilities to minimize damage and costs due to loss of functionality for different exceedance probabilities. Furthermore, their introduction could correct the problem with non-seismic zones in seismic codes, which implies that there will be no seismic action, no matter how high the adopted importance factor may be, while it is clear that for higher return periods parts of non-seismic zones may indeed acquire some degree of seismic hazard. New hazard maps based on ESHM20 will also provide an excellent basis for the assessment of existing facilities, which, as has been said, are often designed with outdated seismic codes or even without any consideration of seismic actions. The safety verification of existing facilities using the ESHM20 seismic input will allow a fully utilization of the load-bearing reserves by means of nonlinear analysis (Meskouris et al. 2019).

ESHM20 can also be integrated in modern systems for the protection and safety control of industrial facilities. These new and innovative systems consist of four basic components: (1) the seismic sensor network, (2) the local monitoring systems within the facility, (3) the communication infrastructure and (4) the integration of the sensor data in digital models of the monitored facility. ESHM20 can support the optimal arrangement of the regional seismic network in the regarded area with stochastic simulations of scenario earthquakes. Fundamental element of these novel systems is the automated interaction of smart sensors or sensor systems, which are capable not only to record motions or strains, but also to process the recordings decentrally and to forward the results of the assessment to decision makers in a compact form (see e.g. Parolai et al. 2015, 2017; Bindi et al. 2016; Boxberger et al. 2017)). Guidelines for the seismic risk assessment of major-hazard industrial plants and its mitigation through smart sensors were already developed by Ciucci et al. (2019). In Germany, the novel early warning and rapid response system for earthquake events in the Lower Rhine region “ROBUST” is under development with financial support from the German Federal Ministry for Economic Affairs and Energy (BMWI) and applied to a chemical production unit close to Cologne (Butenweg 2021). Furthermore, similar approaches were developed in the project SYNER-G (Pitilakis et al. 2016) and for school buildings in Greece (Pitilakis et al. 2022). ESHM20 will also allow the planning and installation of systems regardless of national borders and contribute to the harmonization process of seismic risk assessment of facilities in Europe. Hence, considering that ESHM20 is an attractive and powerful new tool for seismic hazard across Europe, it is necessary to thoroughly validate it both for residential buildings and industrial facilities across Europe.

In the present paper we investigate the impact of the potential adoption of ESHM20 hazard maps in combination with the provisions of the revised EC8 (prEN 1998-1-1 2022) on the operation, performance and safety requirements of industrial facilities at European level with respect to the current practice. To this regard, 79 industrial sites in nine European countries were selected to provide a representative subset of facilities distributed throughout Europe. One should keep in mind, however, that for the local-scale seismic hazard and risk assessment of a specific industrial facility or strategic infrastructure, in general, the resolution of a pan-European model such as ESHM20 might not be adequate and a more detailed and site-specific approach should be preferred, which however is beyond the scope of this work. For each of the 79 selected industrial sites, horizontal elastic response spectra based on the upcoming revision of Eurocode 8 (prEN 1998-1-1 2022) for three return periods, i.e., 475 years, 2500 years, and 5000 years¹ are calculated. These spectra are compared with the respective response spectra used for the design of the selected facilities according to the national regulations, including the current code version EN 1998-1 (2004). Considering that the seismic hazard in ESHM20 is estimated for rock site conditions (V_s,30 > 800 m/s), the site amplification at each site is estimated based on estimations of the local site conditions using mainly the proxy of topography and geology data, as proposed in the site response model of ESRM20 (Weatherill et al. 2021), as well on other more detailed data where available. These approximations may be considered as acceptable for a large-scale application as the one presented herein; however, the obtained seismic hazard for individual sites should be treated with caution as site-specific ground response analyses are usually recommended for the actual design of an industrial facility and should be considered as more pertinent.

The consequences of the application of the compiled spectra on the design is investigated for two typical industrial structures. At first, a single containment LNG tank that stores the liquid in an inner cylindrical steel tank is analysed using a deterministic design approach. The product vapour is contained by an outer steel tank with dome roof and integrated suspended insulated ceiling. The inner and outer tank are analysed for a return period of 5000 years for one of the industrial sites in Germany using the new design approach for liquid filled tanks according to prEN 1998-4 (2022). Furthermore, a vessel installed in a typical five-storey moment resisting steel frame with technical installations is analysed. A deterministic design and a simplified reliability-based probabilistic verification of the storage vessel according to prEN 1998-1-1 (2022) are performed for industrial sites in Greece and Germany.

The goal of this work, which is ongoing within Working Group 13 of the European Association for Earthquake Engineering, (EAEE) (www.eaee.org), is to identify potential inadequacies in the design of existing industrial facilities in terms of considered seismic demand and to highlight expected impact of the adoption of the new European Seismic Hazard Model on the design of new industrial facilities, as well as on the potential seismic risk of existing facilities.

2 Seismic demand for industrial facilities in the Eurocodes

In the current Eurocode 8, the European Standard for the seismic design of structures, the design of industrial facilities is covered for silos, tanks and pipelines in EN 1998-4 (2006). In this part recommended values for importance factors, γ_i for four importance classes I-IV are provided. The recommended values range from 0.8 for importance class I (associated with low risk for loss of life and negligible risk for the social and economic consequences of failures) up to 1.6 for class IV (for situations with exceptional risk of life and extreme economic and social consequences of failure). However, the importance factors are national determined parameters that can be more precisely defined in the National Annexes of the member states and may differentiate for the various seismic areas of the country. The general part EN 1998-1 (2004) for the definition of seismic action further suggests that the importance factors may be defined for a return period considered as more appropriate for the design of a specific structure, T_L, when the seismic hazard is available for a different reference return period T_LR (usually 475 years), as (T_LR/T_L)^−1/k, where k is an exponent depending on seismicity, generally in the order of 3.

In prEN 1998-1-1 (2022), importance factors are replaced by so called performance factors γ_LS,CC, based on the adopted Limit State (LS) and Consequence Class (CC) as given in Table 1. The seismic performance for systems in industrial facilities is addressed by prEN 1998-4, 4.3 (2022) corresponding to the Significant Damage (SD) limit state, in which the structural system and its ancillary elements are significantly damaged, but both retain their structural integrity with controlled leakage of contents. Depending on the characteristics and the function of the specific structure (silo, tank, pipeline, ancillary element) considered, the LS of Damage Limitation (DL) or Fully Operational LS (OP) may be checked additionally. The LS of DL allows a limited extent and amount of damage to the structural system with ancillary elements, that allows a fast recovery of the system capacity up to a predefined level of operation after a damage check and control. Furthermore, liquid-filled systems should remain leak-proof. The LS of OP is defined as the one in which the structural system with ancillary elements remains fully serviceable. Liquid filled systems should remain leak-proof. The seismic action is specified in terms of its return period with respect to the specified limit states and consequence class of the structural system with ancillary elements. The values of the performance factors are up to 1.8 for the Significant Damage (SD) limit state which is recommended for structures covered by prEN 1998-4 (2022) and can be used as a proxy for a return period of 2500 years, unless otherwise specified by the National Annex or relevant Authorities. Alternatively, the performance requirements may be achieved by appropriate return periods T_LS,CC, depending on the limit states and consequence class of the structure and its ancillary elements. In the following, the principles of limit state SD according to pr EN 1998-4 (2022) are applied, which already include requirements for structural damage and leak tightness of systems. Specific damage criteria for OP and DL are not considered. The definition and check of OP and DL limit states is task of the structural engineer, together with the process engineers of the plant operator and relevant authorities. Finally, it should be mentioned that the principles of calculation and design given in prEN 1998-4 (2022) can also be applied for structures with consequence class CC4 (e.g. LNG tanks).

Table 1 Recommended performance factors γ_LS,CC according to prEN 1998-4 (2022)

3 Study area and selected industrial sites

The main criteria of selection of the industrial sites were the facility type, size and amount of material handling. A total of 79 industrial facilities in 9 countries (Belgium, France, Italy, Netherlands, Spain, Switzerland, Germany, Greece, Cyprus) were finally selected to provide a representative subset of facilities distributed throughout Europe as a basis for the studies. The spatial distribution of the industrial sites in Europe is shown in Fig. 1. Table 2 provides the identifier, geographic information and name of the facilities.

Fig. 1

Location of the selected industrial sites in Europe

Table 2 Description of the selected industrial sites in Europe

4 ESHM20 seismic hazard output at major industrial sites in Europe

The results of ESHM20, available at www.efehr.org, are given across Europe for a computational grid with 97,920 points, equally spaced at 0.1–0.1 degrees. ESHM20 includes hazard maps for specified intensity measure types (PGA, spectral acceleration with 5% dam** at predominant periods in the range of 0.05–5 s) and six mean return periods (i.e. 50, 100, 475, 975, 2500 and 5000 years), hazard curves at every computational site, depicting the mean, median (50th) and four quantiles (5th, 16th, 84th and 95th) for all intensity measure types and Uniform Hazard Spectra (UHS) for the mean, median and four quantiles and the six return periods.

In addition to the above, and because of the interaction of ESHM20 with CEN/TC250/SC8, two additional hazard maps were developed for the two seismic hazard parameters adopted by prEN 1998-1-1 (2022) to anchor the horizontal elastic response spectra for rock conditions, namely S_α,475 and S_β,475. S_α,475 is the maximum response spectral acceleration (5% dam**) corresponding to the constant acceleration range of the horizontal elastic response spectrum, while S_β,475 is the spectral acceleration (5% dam**) at the vibration period T_β = 1 s of the horizontal elastic response spectrum, both specified on site category A (rock site conditions) and for a return period of 475 years. For each point of the ESHM20 grid, S_α,475 and S_β,475 were computed from the respective UHS. For the computation of S_α,475, for each grid point of ESHM20 the spectral period T_peak that corresponds to the peak of the median UHS is first identified and then, S_α,475 is calculated as the average spectral value over the range of periods between 0.5 T_peak and 1.5 T_peak (Labbe and Paolucci 2022). For example, if T_peak equals 0.2 s, S_α,475 is calculated as the average spectral value over periods between 0.10 s and 0.30 s. S_β,475 is directly the UHS value for a spectral period of 1 s. The two maps of S_α,475 and S_β,475 will be included in an Annex of prEN 1998-1-1 (2022), as an “acceptable representation of the seismic hazard in Europe for the return period of 475 years”.

Figure 2 shows the indicative seismic hazard curves from ESHM20 in terms of PGA at selected important industrial sites in Italy, Greece, and Belgium, as obtained from the EFEHR platform (http://hazard.efehr.org/), along with the respective maps with their locations. It is reminded that the seismic hazard curves refer to rock conditions (site category A). The probabilities of exceedance in 50 years corresponding to the return periods of 475, 2500 and 5000 years (equal to 10%, 2% and 1% respectively) are shown in Fig. 2. As expected, the seismic hazard in Italy and Greece, two of the most earthquake-prone countries in Europe, is much higher compared to seismic hazard of the low seismicity country of Belgium. Similar seismic hazard curves are available from ESHM20 for the spectral accelerations as well.

Fig. 2

Location of the selected industrial sites and site-specific seismic hazard curves for rock conditions obtained from ESHM20 (Danciu et al. 2021) in a Italy, b Greece and c Belgium

Figure 3 compares the ESHM20 site-specific hazard curves on rock with those obtained through linear scaling of the ESHM20 475-year hazard using the performance factors given in Table 1 for the sites GR-3, BE-1, IT-1 and DE-1 to investigate whether the proposed linear scaling over- or underestimates the seismic hazard for higher return periods. A very good match can be observed for GR-3 site located in Thessaloniki, Greece, while for the other three sites the linear scaling results in a significant underestimate of PGA for return periods higher than 475 years, and an overestimate for return periods less than 475 years. It is reminded that the prEN 1998-1-1 (2022) provides the option of using directly the hazard parameters for the selected return periods when available, which would eliminate the observed insufficiency of the code provisions for higher return periods. The derivation of hazard maps with S_α and S_β for higher return periods such as 2500 years or 5000 years from ESHM20 would certainly support towards this direction. Nevertheless, it should be pointed out that the two alternatives represent two different approaches for the designer. Considering the return periods as a measure allows to use hazard values resulting from a probabilistic hazard assessment, considering the frequency of occurrence of earthquakes, as defined in the earthquake catalogues. On the other hand, the scaling of hazard values starting with 475 years only represents a relative safety level with respect to this reference value and cannot be directly compared to the values resulting from a specific return period. Due to the non-linear nature of the seismic hazard curve, a linear scaling can only be an approximation and e.g. satisfy the design target to be n-times safer than the 475-years design.

Fig. 3

Comparison of the ESHM20 site specific hazard curves on rock with the linear scaling of the ESHM20 475-year hazard using the performance factors of prEN 1998-4 (2022) as given in Table 1 for the industrial sites GR-3, BE-1, IT-1 and DE-1

5 Impact of the adoption of ESHM20 on the seismic design of industrial facilities

5.1 Methodology

To investigate the impact of the adoption of ESHM20 on the seismic design of industrial facilities, elastic response spectra according to prEN 1998-1-1 (2022) are compared with the respective elastic response spectra of the codes currently (or previously) in force in the countries where the industrial sites are located. These codes are either national codes or National Annexes of the current EN 1998-1 (2004). The elastic response spectra according to prEN 1998-1-1 (2022) are constructed with the two seismic hazard parameters for rock conditions, S_α,475 and S_β,475 proposed by ESHM20.

The influence of local site conditions on the elastic response spectra is considered based on the respective provisions of each code, using the proxy of topography- and geology-based European site response of ESRM20 (Weatherill et al. 2021), complemented with more detailed data where available. Elastic response spectra are initially calculated for a return period of 475 years, and then appropriate importance factors (for the national codes) or performance factors (for prEN 1998-4, 2022), commonly used for the design of industrial facilities as described in Sect. 2, are applied to calculate the elastic response spectra as a proxy for higher return periods. Details on the calculation of the different elastic response spectra are provided hereinafter.

5.2 Spectra according to prEN 1998-1-1 (2022)

The elastic response spectrum for a specific return period (RP), instead of being anchored with only one parameter, namely the Peak Ground Acceleration (PGA), is anchored with two parameters, S_α, which is the spectrum value at the plateau spanning over a certain spectral period range, and S_β, which is the spectrum value at T_β = 1 s (Fig. 4). The S_a and S_β parameters are obtained from the respective values at rock conditions (site category A), S_a,RP and S_β,RP, multiplied by the short-period (F_α) and intermediate period (F_β) site-amplification factors, respectively. Site-amplification factors are dependent on the site category, which is based on two parameters, namely H₈₀₀ (depth of the conventional bedrock formation, identified by a shear wave velocity at least equal to 800 m/s) and v_s,H (time-averaged shear wave velocity of the top 30 m or down to H₈₀₀ if H₈₀₀ < 30 m). The site categorization scheme introduced in the prEN 1998-1-1(2022) is given in Table 3. Based on the site category, prEN 1998-1-1 (2022) provides appropriate formulas for the computation of F_α and F_β, introducing a dependency not only on the site class (through v_s,H), but also on the level of seismic ground motion (through hazard parameters S_a,RP and S_β,RP). In this way, the proposed amplification factors are continuous with v_s,H, smoothening the strong jumps that are existent in EN 1998-1 (2004) between adjacent soil categories (Paolucci et al. 2021), as well as intensity-dependent, incorporating in this way the effect of soil nonlinearity for strong ground shaking. Alternatively, default intensity-dependent values of F_α and F_β are introduced for the cases where a simplified identification of the site category is adopted because of insufficient geotechnical information. The proposed analytical and default amplification factors are given in Table 4. In addition to the site amplification factors, a topographic amplification factor may be applied, similarly to EN 1998-1 (2004) (Labbé and Paolucci 2022). However, each country may adopt different standards for site categorization and estimation of site amplification, as has been recently proposed for example by Pitilakis et al. (2019, 2020).

Fig. 4

Horizontal elastic response spectrum with horizontal log scale according to prEN 1998-1-1 (2022)

Table 3 Site categorization according to prEN 1998–1-1 (2022)

Table 4 Site amplification factors according to prEn 1998–1-1 (2022)

For the current application, to classify the European sites with major industrial facilities of Table 2 into one of the site categories of Table 3, each industrial site is represented by a single point with respect to its spatial extension. This is a rough simplification, as in fact industrial facilities span over large areas, exceeding sometimes tens of square kilometres, i.e., the differentiation of site conditions across the industrial site might be non-negligible. Thereafter for each site a single V_s,30 is determined either from the slope-based site model of ESRM20 or from available local information and data (e.g., in Greece). With respect to soil depth, available information for specific sites is used (e.g., for Germany obtained from the subsoil class map included in DIN EN 1998-1/NA (2021), for Greece from available local information and data), while for the other sites an intermediate depth within the range of 30–100 m is assumed. After classification of all sites into one of the categories of Table 3, the amplification factors of F_α and F_β with both the analytical and the default formulas are calculated. All these approximate approaches, including the inference of V_s,30 from slope and geology, the assumption of uniform site conditions over the whole industrial area and the adoption of code-based site amplification factors, inevitably increase the uncertainties related to the assessment of site amplification, which for the case of industrial facilities certainly needs to be as rigorous as possible. However, given the large scale of the application and the fact that the scope of the investigation is to compare the results obtained with the new and the current coded, these assumptions are within acceptable margins.

Regarding the determination of S_a,RP and S_β,RP (for site category A), it is a national choice whether they will be obtained from spectral acceleration maps for a specific return period, or by using performance factors γ_LS,CC, based on the adopted Limit State (LS) and Consequence Class (CC). Details on the values proposed prEN 1998-4 (2022) are given in Sect. 2.

In the current application, the elastic response spectra are evaluated for three return periods, i.e. (a) 475 years using directly the S_α,475 and S_β,475 values for site category A provided by ESHM20, (b) 2500 years using S_α,475 and S_β,475 increased by performance factors equal to 1.8 considering CC3-b consequence class and significant damage (SD) limit state and (c) 5000 years applying to S_α,475 and S_β,475 a performance factor equal to 2.2 considering CC3-b consequence class and near collapse (NC) limit state as given in Table 1. In all cases the elastic response spectra are evaluated for rock conditions (site category A) and taking local site conditions into account. For the results shown in the next section, the code name C0 is used for prEN 1998-1-1 (2022).

5.3 National codes

The National Codes are either stand-alone codes (e.g., SIA 261 in Switzerland, EAK 2003 in Greece and DIN4149 for Germany) or National Annexes of the current EN 1998-1 (2004) (Table 5). For the results shown in the next section, the code name C1 is used for the national codes currently in force, while C2 is used for older seismic codes, applicable only for Germany and Greece as shown in Table 5.

Table 5 National Codes applied for industrial sites

Following the same rationale as for the prEN 1998-1-1 (2022), the elastic response spectra of the national codes are evaluated for three return periods, i.e. (a) 475 years using the seismic hazard parameters (mainly PGA) from the national seismic hazard maps, (b) 2500 years applying an importance factor equal to 1.6, the highest value given in EN 1998-4 (2006) for importance class IV, and (c) 5000 years applying an importance factor equal to 2.19 (Sect. 5.2), obtained from the formula given in the EN 1998-1 (2004). In all cases the elastic response spectra are calculated for rock conditions (site category A) and taking local site conditions into account according to the relevant clauses of the national codes.

Although a detailed description of the particularities of each national code is beyond the scope of this paper, we should note that the national codes/annexes of specific countries, such as Greece and Cyprus, fully endorse the clauses of the EN 1998-1 (2004) regarding soil categorization, importance factors, control periods and soil amplification factors, while others differentiate substantially. For example, the site classes in the categorization schemes of the national annexes of Germany and Belgium are defined with different V_s,30 boundaries compared EN 1998-1 (2004), while the site amplification factors in the German, Italian, Spanish and Swiss codes are dependent not only on the site class but also on the level of ground acceleration. We should also highlight that the Italian Code is the only one which provides values for PGA at rock a_g, spectral amplification F₀ and the control-period T_C* for different return periods between 30 and 2475 years for a grid of 10,751 points. This is the reason why for the Italian National Annex the elastic response spectra for the 5000-years return period are not evaluated, as the performance factor cannot be generalized for increasing other parameters than a_g. On the other hand, the code of the Netherlands provides a seismic hazard map which refers to a return period of 5000 years, which was scaled down to the return period of 475 years using the formula given in EN 1998-1 (2004).

5.4 Indicative results

For the results shown in this and the next section, we use the code name C0 to refer to the revised EC8 (prEN 1998-1-1 2022), C1 to refer to the national code currently in force in each country (National Code 1 in Table 5) and C2 to refer to a previous national code (National Code 2 in Table 5).

The elastic response spectra computed with prEN 1998-1-1 (2022) and the national codes for two example sites, one located in Germany (DE-1, Chempark Leverkusen) and one located in Greece (GR-3, HELPE Thessaloniki) are compared in Figs. 5 and 6 respectively, for return periods of 475, 2500 and 5000 years, considering both rock and estimated local site conditions. In the case of the elastic response spectra for rock, the respective UHS obtained from ESHM20 are provided to evaluate the adequacy of the scaling of the 475 years spectrum with the performance factors of prEN 1998-1-1 (2022) in capturing the seismic hazard for higher return periods. It is reminded that the 475-year elastic response spectrum for rock according to prEN 1998-1-1 (2022) is essentially obtained from the ESHM20 UHS for 475 years, as the parameters S_a,475 and S_β,475 needed to anchor the spectrum are obtained from the ESHM20 UHS, so a good match between the prEN1998-1-1 (2022) spectrum and the ESHM20 UHS for 475 years and for rock conditions should be anticipated. For the consideration of local site conditions with prEN 1998-1-1 (2022), the default values of amplification factors F_α and F_β have been adopted (Table 4).

Fig. 5

Horizontal elastic response spectra for the Chempark Leverkusen industrial site (DE-1)

Fig. 6

Horizontal elastic response spectrum for the HELPE Thessaloniki industrial site (GR-3)

Regarding site Chempark Leverkusen in Germany (DE-1), for the 475-year return period, the discrepancies between the elastic response spectra obtained with the three codes are very limited when considering rock conditions (Fig. 5a), with C1 resulting in a slightly higher plateau value compared to C0 and C2. The differences are amplified when taking local site conditions into consideration (Fig. 5b), due to the variation in the values of amplification factors with the different codes. More specifically, amplification factor F_α for the plateau of the spectrum of pr EN 1998-1-1 (2022) is equal to 1.55 for site class C, while the respective amplification factors for national codes C1 and C2 are equal to 1.25, which results in an elevation of the plateau of code C0 above C1 and C2. Moving to the higher return periods, the discrepancy between the plateau of C0 and C1/C2 is further amplified for the case of 2500 years due to the higher performance factor adopted by C0 (1.8 versus 1.6 for C1/C2) (Fig. 5d) and is maintained for the 5000 years where the performance factors for all codes are similar (2.2 for C0 and 2.19 for C1/C2) (Fig. 5e). On the contrary, the descending branches of the spectra when considering local site effects are in good agreement for all examined return periods (Figs. 5b, d, e), meaning that there is no significant impact of the adopted regulation for spectral periods higher than about 0.5 s. This is due to the relatively smaller T_C value for the spectrum of prEN 1998-1-1 (2022), calculated as S_β/S_α, resulting in a narrower plateau for C0 and, thus, a descending branch starting at lower spectral periods. Regarding the comparison with the ESHM20 UHS for rock, we observe that the scaling of the C0 response spectrum with the performance factors of prEN 1998-4 (2022) is inadequate for the specific site, as the C0 rock spectra for 2500 and 5000 years are well below the respective UHS (Fig. 5c and e), meaning that C0 underestimates seismic hazard at this site compared to ESHM20. It is reminded that similar conclusions were drawn based on the comparisons of the ESHM20 hazard curve with its linear approximation with the performance factors of prEN 1998-4 (2022) in Fig. 3.

As for the HELPE site in Thessaloniki, Greece (GR-3), the discrepancies between C0, C1 and C2 are very high when considering either rock or local site conditions, in the first place due to the much higher S_α,RP value for rock obtained from ESHM20 (0.57 g for C0 compared to 0.4 for C0/C1, Fig. 6a). In addition, the fact that the provisions of national code C2 do not account for soil amplification is the cause for the extreme differences at the spectral values at the plateau between codes C0 and C2 when considering local site conditions for all return periods (Figs. 6b, d, e). However, the narrower plateau for C1 due to the lower value for corner period T_C, eliminates the discrepancies between C0 and C1 at the descending branches of the spectra. This is not the case between codes C0 and C2, as the descending branch of C2 spectra is steeper compared to C0 (and C1) for long periods because code C2 has a single descending branch, analogous to (1/T), while codes C0 and C1 have two descending branches, the first proportional to (1/T) and the second proportional to (1/T²). The agreement between the ESHM20 UHS for 2500 and 5000-years with the respective spectra of the code C0 is remarkable, meaning that the linear approximation with the performance factors works well as noted in Sect. 4.

5.5 Discussion on the impact of the revised Eurocode 8

To get an overview of how a potential adoption of prEN 1998-1-1 (2022) combined with ESHM20 S_a,475 and S_β,475 hazard parameters would influence the seismic action used for the design of industrial facilities in terms of elastic acceleration response spectra across Europe compared to the national codes, we focus on two spectral values, i.e., the value at the plateau of the spectrum, max S_e and spectral acceleration at 1.0 s, S_e(T = 1.0 s). We thus computed for all sites and for the three return periods (475, 2500 and 500 years) the ratios of (a) max S_e, using the C0 code to maximum spectral acceleration using C1 or C2 codes and (b) S_e(T = 1.0 s) using the C0 code to spectral acceleration at 1.0 s using the C1 or C2 codes (Figs. 7a, 8a for rock and Figs. 7b, 8b for soil conditions). Sites with zero hazard for 475 years either from ESHM20 or national codes are not presented for any return period, as the scaling of the 475-years hazard results in zero hazard also for higher return periods, implying that there will be no seismic action no matter how high the adopted importance factor may be, an important problem which was highlighted also in the Introduction. The spectral accelerations have been obtained applying the default amplification factors for prEN 1998-1-1 (2022) and should be approached with caution, as in most of the cases the categorization of the sites has been based on approximations (e.g., slope-based V_s,30 values), while the assignment of a uniform V_s,30 value to a whole industrial area is probably simplified. Although site-specific investigations are generally more pertinent for critical infrastructures, the large scale of the herein presented analysis and the scope of the study, which is to compare results between the new and the current codes, justifies such simplifications. The use of the analytical functions instead of default values would decrease the estimated max S_e and S_e (T = 1.0 s) ratios in most cases, as the default amplification factors tend to be more conservative compared to their analytical counterparts (Paolucci et al. 2021). However, the application of the default values can be regarded as the usual case in practise.

Fig. 7

Ratios of maximum spectral acceleration, max S_e, using the revised Eurocode 8 (C0) to max S_e using the national codes C1 (filled symbols) or C2 (line symbols) for the return periods of 475y (green), 2500y (blue) and 5000y (red), considering a rock conditions and b the effect of local site conditions

Fig. 8

Ratios of spectral acceleration at 1.0 s, S_e(T = 1.0 s), using the revised Eurocode 8 (C0) to S_e(T = 1.0 s) using the national codes C1 (filled symbols) or C2 (line symbols) for the return periods of 475y (green), 2500y (blue) and 5000y (red), considering a rock conditions and b the effect of local site conditions

Regarding the effect of the return period on the values of the ratios, it can be observed that max S_e and S_e (T = 1.0 s) ratios for a given site and a given national code C1 or C2 are practically constant (or with very limited variations) for different return periods. This is explained if one considers that for the 2500 and 5000 years return periods the ratios are equal to the respective ratios for the 475-years return period, multiplied by the quantity γ_LS,CC/γ_I, defined as the ratio of the performance factor of prEN 1998-4 (2022), γ_LS,CC, to the importance factor γ_i, of EN 1998-4 (2006), equal to 1.8/1.6 = 1.125 for 2500 years and 2.2/2.19 ≈ 1 for 5000 years. Similarly, the max S_e and S_e(T = 1.0 s) ratios do not seem to be significantly affected by the consideration or not of the local site conditions, indicating that, with the adopted site models, the site-amplification factors of prEN-1998-1-1 (2022) are not far from the current practice, except for the C0/C2 ratios in Greece, which are much higher when site effects are considered due to the lack of site-amplification factors in code C2.

For the industrial sites of Belgium, France and Switzerland, the ratios of both max S_e and S_e(T = 1.0 s) are in most cases below 1.0 regardless of the consideration or not of the effect of local site conditions, meaning that a potential adoption of the revised Eurocode 8 concept would result in a decrease of the design seismic action in these countries compared to what is currently provided for by the national codes. A similar observation can be made for Italy, one of the most earthquake-prone countries in Europe, where max S_e and S_e(T = 1.0 s) ratios are below 1.0 for four out of six sites. The highest decrease occurs in France where the spectral values in consideration are reduced more than 50% compared to the current practice.

On the other hand, Fig. 7a shows an increase of maximum spectral acceleration max S_e with C0 at all sites in Greece compared to both C1 and C2, with max S_e ratios above 1.4 for most of the sites and even higher than 2.0 at selected sites (e.g., GR-8) for rock conditions. The C0/C2 max S_e ratios are even higher if local site conditions are considered, since, as noted previously, C2 does not account for site amplification depending on soil class.

With respect to spectral acceleration at 1.0 s, C0/C1 S_e (T = 1.0 s) ratios are below or around 1.0 for most sites in Greece, regardless of the consideration of site-effects. This is justified by the higher, yet relatively narrower plateau of pr EN 1998-1-1 (2022), which results in a lower value of corner period T_c, i.e., the descending branch for code C0 starts at lower periods and in this way the discrepancies between codes C0 and C1 at longer periods are eliminated, as was also noted for the HELPE site in Fig. 6. On the contrary, C0/C2 S_e (T = 1.0 s) ratios are much higher than 1.0, with values higher than 3 for some sites, due to the descending branch of C2 spectra which is steeper compared to the one of C0.

In Germany, which is the country with the most industrial sites in this study, the max S_e ratios range between 0.35 and 1.7. Most of the sites exhibit ratios closer to 1.0. Some extreme values of the C0/C2 ratios, around the value of 3 for max S_e and even higher for S_e (T = 1.0 s) are observed at some sites (DE-2, DE-3, DE-4), due to the very low amplification factor of 0.75 proposed by national code C2 for the soil class C-S.

A significant increase in max S_e is observed also in the Netherlands, however this originates mostly from the very low seismic hazard of NEN 6702, which is not even mandatory, potentially leading to high ratios for spectral accelerations. Finally, the hazard of the one site that we examined in Cyprus does not seem to be affected by the revised Eurocode 8 concept, with max S_e and S_e (T = 1.0 s) ratios around one, while for the few studied sites in Spain no specific pattern is identified.

The total percentages of the sites with C0/C1 ratios higher or less than 1.0 for both max S_e as well as S_e (T = 1.0 s) are given in Table 6 for all return periods. For the 475-year return period, a potential adoption of revised Eurocode 8 with ESHM20 S_α,475 and S_β,475 values was found to result in an increase of max S_e (respective ratios > 1.0) for half of the investigated sites and in a decrease in the other half. One should keep in mind, however, that the investigated sites are not uniformly distributed across Europe, so the herein obtained results refer only to the specific sites and might change if different sites are selected. Similar results are obtained for the 5000-year return period due to the similar values of importance / performance factors adopted in C0 (2.2), and C1-C2 (2.19). On the other hand, the slightly higher value of performance factor for 2500 years in C0 (1.8) compared to C1-C2 (1.6) is reflected in the increased percentages of sites with max S_e ratios higher than 1.0 for this return period. For spectral acceleration at 1.0 s, the percentages of sites exhibiting an increase of seismic demand with C0 are lower, which is attributed mainly to the narrower plateaus acquired when applying C0. The percentages are only slightly affected when considering local site effects. We should highlight, however, that these remarks reflect only the selected industrial sites at European scale and thus cannot be considered as conclusive. Increasing the number of studied sites and/or a more detailed evaluation of the specific soil conditions might differentiate the herein obtained results.

Table 6 Percentages of number of sites with C0/C1 max S_e and S_e (T = 1.0 s) ratios higher or less than 1.0 for the return periods of 475, 2500 and 5000 years, for rock conditions and considering local site effects

6 Case studies

6.1 Introduction

In the following two structures with different resonance frequencies of vibration are analysed to study the impact on the seismic action effects due to the use of the ESHM20 hazard maps. In the first example a large storage tank is analysed, in which hydrodynamic pressure components are characterized by eigenfrequencies in different ranges of the response spectra. In the second example, the seismic response of a representative steel frame structure with various technical components is investigated. The purpose of these comparisons is not to derive general rules and certainly the results cannot be transposed to other sites or countries. Nevertheless, the evaluation is meant to sensitize the designers and engineers to certain effects and not to draw general conclusions and apply them to other problems. Furthermore, the significant dependence of the seismic demand and thus structural response on the site amplification effects needs to be underlined, as this is provided in the national annexes or codes respectively and is not part of the basic document prEN 1998-1-1 (2022).

6.2 Tank for the storage of liquified natural gas (LNG)

The first case study is conducted for a single containment LNG tank that stores the liquid with an operational temperature of about -100 °C in a cylindrical steel tank. The product vapour is contained by an outer steel tank with dome roof and integrated suspended insulated ceiling. The outer tank is designed to contain the product vapours and to protect the thermal insulation placed between the inner and outer shell. The tank is surrounded by a bund wall to contain product leakage. The structure and graded wall thickness over the height of the inner and outer tank are shown in Fig. 9. The outer tank with dome roof is made of steel 16MnDR (EN 10028-4 2017) and stabilized by T-shaped ring stiffeners and a strong ring top ring stiffener with a shell thickness of 40 mm. The inner tank is made of steel X8Ni9 640 (EN 10028-4 2017), has a radius of ${\text{R}} = 23.0{ }m$, a height of ${\text{L}} = 32.0{ }m$ and is filled up to an operating level of $H = 31.5 m$ with a product density of $\rho_{{\text{L}}} = 5.7{ }t/m^{3}$. The ratio of the filling height to the tank radius is $\gamma = 1,37.$ The unanchored inner tank is stabilized by an additional T-shaped top ring stiffener.

Fig. 9

Structure and graded wall thicknesses of the a inner and b outer tank

The design for the inner and outer tank is carried out according to EN 14620–1 and EN 14620–2 (2006) for the plant site Leverkusen in the western part of Germany (DE-1) near to the Rhine River with low seismicity. Both tanks are designed for the Operating Basis Earthquake (OBE) and the Safe Shutdown Earthquake (SSE) with the primary container filled to the maximum normal operating level. In the following investigation only the decisive SSE event is considered. According to EN 14620-1 (2006) the SSE event shall be represented by 5% damped design response spectra according to EN 1998-1 (2004) with a 1% probability of being exceeded within a 50-year period (ergo 4975 years return period²). DIN EN 14620-1 (2006) specifies further, that the SSE spectrum needs not be greater than two times the OBE spectrum according to EN 1998-1 (2004) with a 10% probability of being exceeded within a 50-year period (ergo 475 years return period). In the following, the seismic actions according to DIN EN 14620-1 (2006) are compared with the actions to be applied according to the standards C0, C1 and C2. The comparison is conducted for PGA values determined with importance factors and probabilistic hazard maps (DIN EN 1998-1/NA 2021; ESHM20 2020) with a return period of 4975 years. The objective is to show the impact on the seismic action effects due to the use of the ESHM20 hazard maps. The LNG tank is well suited for a comprehensive comparison because the seismically induced hydrodynamic pressures in the inner tank are assigned to different period ranges. Table 7 summarizes the resulting PGA values for the two design concepts. Figure 10 shows the corresponding response spectra for soil class C (C0, EN 14620–1) and subsoil condition C-T (C1, C2).

Table 7 PGA values according to specific codes and design concepts

Fig. 10

Response spectra for codes C0, C1, C2 and PGA values given in Table 7

6.2.1 Comparison of the design results for the inner tank

The inner tank is analysed with the force-based approach according to prEN 1998-4 (2022) considering the sloshing of the liquid at the free surface (convective component), the horizontal and vertical movement of the liquid together with the rigid tank (impulsive component) and the fluid–structure vibrations of the tank and the liquid (impulsive flexible component). The dynamic effects of the convective and impulsive modes of vibrations are described by equivalent static pressure distributions applied on the tank wall and bottom. The fundamental natural period of the joint bending vibration of the tank and the liquid due to horizontal seismic excitation is calculated using the approximate equation according to Rammerstorfer und Fischer (2004) with the tank wall thickness at one third of the filling height $s_{w} \;=\; {\text{ 13 mm}}$ and the correction factor ${\text{F}}\left( {\upgamma } \right) = 3.15:$

$${\text{T}}_{{{\text{if}}}} = 2 \cdot {\text{R}} \cdot {\text{F}}\left( {\upgamma } \right)\sqrt {\frac{{{\text{H}} \cdot {\uprho }_{{\text{L}}} }}{{{\text{E }} \cdot s_{w} }}} = 2 \cdot {23}{\text{.0}} \cdot {3}{\text{.15}} \cdot \sqrt {\frac{{{31}{\text{.5}} \cdot {0}{\text{.568}}}}{{{\text{200e6 }} \cdot {0}{\text{.013}}}}} = 0.38 {\text{s}}$$

(1)

The fundamental natural period for sloshing is estimated according to prEN 1998-4 (2022):

$$T_{{{\text{con}}}} = \frac{{2\pi \sqrt{\frac{R}{g}} }}{{\sqrt {1,841\tanh \left( {\frac{{1,841{ }H}}{R}} \right)} }} = 7.14 s$$

(2)

The impulsive rigid pressure component is calculated using the spectral acceleration of the design response spectrum at period $T_{{{\text{ir}}}} = 0 s$. The spectral accelerations according to the different codes and design concepts are presented in Table 8. Due to limited energy dissipation of thin-walled tanks, a reduction of the spectra with behaviour factors was not considered. The convective pressure component is calculated using the elastic design spectrum with dam** corrections factors for 0.5% dam**. As the convective period is higher than 4 s, the corresponding spectral accelerations are calculated according to prEN 1998-1- 1:2022, 5.2.2.2(13) for C0 and EN 1998-1, Annex A (2004) for all other applied codes.

Table 8 Spectral accelerations $S_{{\text{e}}}$(g) and $S_{{\text{r}}}$(g) for the three pressure components

The total base shear and overturning moment above the base plate due to hydrodynamic pressure effects are compared to clarify the consequences of applying the different response spectra, since they are decisive for the buckling verification at the tank base and the critical occurrence of the uplift. The loads according to EN 14620-1 (2006) are used as reference values since this code is the basis for the tank design. The total base shears $F_{{{\text{b}},{\text{if}}}}$(impulse flexible), $F_{{{\text{b}},{\text{c}}}}$(convective), $F_{{{\text{b}},{\text{ir}}}}$(impulse rigid) and the overturning moments $M_{{{\text{b}},{\text{if}}}}$(impulse flexible), $M_{{{\text{b}},{\text{c}}}}$(convective), $M_{{{\text{b}},{\text{ir}}}}$(impulse rigid) above the base plate are calculated according to prEN 1998-1-1 (2022) as integrals from the impulsive rigid, flexible and pressure functions:

$$F_{{{\text{b}},{\text{if}},{\text{h}}}} = C_{{{\text{F}},{\text{if}}}} { }\Gamma_{{{\text{if}}}} { }m_{{\text{l}}} { }\left( {S_{{\text{r}}} \left( {T_{{{\text{if}},{\text{h}}}} } \right) - S_{{\text{r}}} \left( {T_{{{\text{ir}}}} } \right)} \right)$$

(3)

$$M_{{{\text{W}},{\text{if}},{\text{h}}}} = C_{{{\text{MW}},{\text{if}}}} { }\Gamma_{{{\text{if}}}} { }m_{{\text{l}}} { }H{ }\left( {S_{{\text{r}}} \left( {T_{{{\text{if}}}} } \right) - S_{{\text{r}}} \left( {T_{{{\text{ir}}}} } \right)} \right)$$

(4)

$$F_{{{\text{b}},{\text{c}}}} = C_{{{\text{F}},{\text{c}}}} { }\Gamma_{{\text{c}}} { }m_{{\text{l}}} { }S_{{\text{e}}} \left( {T_{{{\text{con}}}} } \right)$$

(5)

$$M_{{{\text{W}},{\text{c}}}} = C_{{{\text{MW}},{\text{c}}}} { }\Gamma_{c} { }m_{{\text{l}}} { }H{ }S_{{\text{e}}} \left( {T_{{{\text{con}}}} } \right)$$

(6)

$$F_{{{\text{b}},{\text{ir}},{\text{h}}}} = C_{{{\text{F}},{\text{ir}}}} { }\Gamma_{{{\text{ir}}}} { }m_{{\text{l}}} { }S_{{\text{r}}} \left( {T_{{{\text{ir}}}} } \right)$$

(7)

$$M_{{{\text{W}},{\text{ir}},{\text{h}}}} = C_{{{\text{MW}},{\text{ir}}}} { }\Gamma_{{{\text{ir}}}} { }m_{{\text{l}}} { }H{ }S_{{\text{r}}} \left( {T_{{{\text{ir}}}} } \right)$$

(8)

Participation factors ${ }\Gamma_{{{\text{if}}}} , \Gamma_{{\text{c}}} , { }\Gamma_{{{\text{ir}}}}$, coefficients $C_{{{\text{F}},{\text{if}}}} , C_{{{\text{M}},{\text{if}}}} ,C_{{{\text{F}},{\text{c}}}} , C_{{{\text{M}},{\text{c}}}} ,C_{{{\text{MW}},{\text{ir}}}}$ and total liquid mass $m_{{\text{l}}}$ are given in Table 9.

Table 9 Participation factors, coefficients, and total liquid mass

The total base shear and overturning moment are calculated using the SRSS-rule as given in (9) and (10):

$${\text{F}}_{{\text{bh,tot }}} { = }\sqrt {\left( {{\text{F}}_{{\text{b,c}}} } \right)^{{2}} { + }\left( {{\text{F}}_{{\text{b,ir}}} + {\text{ F}}_{{\text{b,if}}} } \right)^{{2}} }$$

(9)

$${\text{M}}_{{\text{W,tot}}} { = }\sqrt {\left( {{\text{M}}_{{\text{W,c}}} } \right)^{{2 }} { + }\left( {{\text{M}}_{{\text{W,ir}}} + {\text{M}}_{{\text{W,if}}} } \right)^{{2}} }$$

(10)

The resulting base shears and overturning moments of the pressure components and their superposition are presented in Table 10. The percentage deviation of the superimposed reaction forces and moments with respect to the reference calculation according to DIN 14620-1 (2006) shows that the application of national hazard maps (DIN EN 1998-1/NA, 2021) and ESHM20 lead to 1.6–2.0 higher support forces for a return period of 4975-years.

Table 10 Inner tank: Base shear and overturning moments according to specific codes and design concepts

If the concept of importance factors is applied, minor maximum deviations between the considered standards C0, C1, C2 and DIN 14620-1 are obtained. However, major differences for the convective and impulsive rigid pressure components are calculated due to the differing normative definitions of the spectral accelerations in the short and long period range of the response spectra. The vibration periods of the pressure components in different period ranges lead to significant differences for the individual pressure components that are important for the design. Therefore, it is obvious that the alternative concepts lead to different safety levels. It should be noted that the differences in the safety levels of the two concepts are site dependent.

6.2.2 Comparison of the design results for the outer tank

The comparison of the reaction forces and moments for the outer tank is performed for the fundamental vibration period $T = 0.2 s$ with a seismic active mass of 1750 t. The overturning moment is calculated with a corresponding lever arm of 21 m. The results in Table 11 show again differences between the alternative concepts using importance factors and probabilistic hazard maps with higher return periods. The differences between the codes in case of importance factors result from differing soil factors. Again, the alternative concepts lead to different safety levels.

Table 11 Outer tank: Base shear and overturning moments according to specific codes and design concepts

6.3 Five-storey moment resisting frame with components

In the second case study, a typical five-storey moment resisting frame with technical installations is analysed. In this example, the variation of seismic response for two representative regions in Europe is explored, given the seismic demand according to DIN 4149 (2005) and prEN 1998-1-1 (2022). The facilities the DE-1 in Germany and GR-3 in Greece are selected as examples for low and medium/high seismicity regions. Within industrial facilities, frame structures with multiple stories are very common as supporting structures in industrial plants, as they allow a flexible arrangement of process components.

6.3.1 Deterministic approach

The frame structure with installations and the simplified calculation model is shown in Fig. 11. A 5 m wide strip of the example unit is considered, and all components are modelled as lumped masses at girder height. Permanent and live loads are simply concentrated at the floor levels. The self-weight of the frame is 197.5 kN, the floor self-weight (dead load) is assumed as 8.6 kN/m² and the variable live load with 15 kN/m² (of which 30% are to be applied). The storey loads to be considered in the seismic design situation are idealized as lumped masses at each storey level with M_DL+LL = ½ · 5 m · 9 m (8.6 kN/m² + 0.30 · 15 kN/m²)/9.81 m/s² = 30 t. Both girders and columns are executed in S235 steel with the standardized profiles as illustrated in Fig. 11. The modal analysis of the frame structure results in the eigenfrequencies, and mode shapes as illustrated in Fig. 12. These first three vibration modes activate more than 90% of the effective total mass.

Fig. 11

a 5-storey frame structure with components and b equivalent structural model

Fig. 12

First three vibration modes of the frame structure

The design force is calculated for the vessel on the top floor with a mass of 15 t, a lever arm of 1.0 m above the floor level and a fundamental frequency of 5 HZ is used as a comparative basis. For the design the acceleration of the vessel A_v is calculated, given the seismic input at the selected locations in Germany and Greece according to Sect. 5.4 for the codes C0 ( prEN 1998-1-1 2022), C2 (EAK 2003; DIN 4149 2005) and the return periods 475, 2500 and 5000-years. The resulting equivalent static loads are determined using the SRSS rule for the first three modal contributions. The resulting component accelerations A_v are summarized in Tables 12 and 13 for the selected locations in Germany and Greece. For the code C0 both options according to Table 4 are considered to investigate the significance of site amplification factors and to see the dispersion of the results. For further information the ratio of the spectral acceleration at T = 0 s (PGA on soil) of the input spectra is given in the last two columns of Tables 12 and 13.

Table 12 Component acceleration A_v and S_a(T = 0 s) for industrial site DE-1, codes C0, C2 and different return periods

Table 13 Component acceleration A_v and S_a(T = 0 s) for industrial site GR-3, codes C0, C2 and different return periods

Based on the comparison it is apparent, that the change of the response of the structure may vary depending on the country specific design spectra implementation. In Germany, an increase of S_a(T = 0 s) from 20 to 45% leads to a lower or almost equivalent structural response acceleration for the regarded vessel. In contrast, in Greece the increase of S_a(T = 0 s) results in a significant increase in the structural response up to 77% for the 475-years return period. Also, the comparison of soil acceleration values for the 5000-years return period shows opposite results. In Germany, the ratio C0/C2 is almost constant, whereas in Greece a significant increase between 16 and 36% can be observed. The comparison illustrates the strong dependence of the changes in seismic design depending on the country-specific regulations and implementations.

6.3.2 Probabilistic seismic risk assessment of industrial structure

The above assessment provides a deterministic comparison for the different seismic inputs. In order to also appreciate the effect of the hazard over various return periods in a probabilistic sense, the site-specific hazard curves were utilized to convolve them with a representative fragility curve of the storage vessel investigated in the previous section. A simplified reliability-based probabilistic verification according to prEN-1998-1-1 (2021), Annex F was performed to assess the performance of the vessel in fully probabilistic terms. It should be noted that the applicability depends on the national choices and the target values are a national decision.

In this case study, fragility is defined in terms of seismic demands only. Resulting fragilities represent the probability of reaching a specific limit state at a specific loading intensity (defined by an appropriate intensity measure) without considering uncertainties in the capacity of the system. A correlation between the PGA and the demand parameter is assumed.

Generic fragility curves for a liquid filled storage vessel in two directions are used with failure modes corresponding to the failure of the anchor bolts fixing the vessel to the supporting structure. To be able to make a fair comparison of the changes in results due to the new hazard input, the same generic fragility curve for both sites was used. As the assessment is performed on a component this simplification is tolerable, as pieces of equipment are usually designed based on envelo** seismic input motions which are not site specific. The fragility functions are expressed using a lognormal cumulative distribution function: F(x) = P[D ≥ d^LS |IM = x] = P [ln(x/A_m)/β], where P is the probability for a damage (D) to be larger than a damage limit state (d^LS) given a certain intensity measure (IM) of x in terms of PGA, and A_m and β are the median and dispersion of the damage fragility. The dispersion β is usually expressed by the aleatory term β_U and the epistemic term β_R. For the sake of simplicity, only the aleatory term β_R is considered in the following example. The median capacities conditioned on PGA are given in Table 14 and the resulting fragilities are plotted in Fig. 13. The first fragility corresponds to a seismic loading perpendicular to the longitudinal axis of the vessel (0°) and the second is the result of a loading parallel to the longitudinal axis of the vessel (90°). For longitudinal loading, the initial stiffness of the component is almost 50% higher than in transversal direction. Further details on the derivation of the fragility function are not of importance for this case study, but they represent realistic component response and cover the range of outcomes.

Table 14 Fragility parameters for the vessel on top floor level conditioned on PGA

Fig. 13

a Fragility functions of the storage vessel (left) and b hazard curves for sites DE-1 and GR-3

Splitting the site-specific hazard into nine acceleration bins associated to initiating events (see Fig. 13b) and calculating the random variables for the component for each seismic event can be summarized as given in Table 15. As ESHM20 is providing the hazard in terms of probability of exceedance in 50 years (POE₅₀), the annual probability of exceedance (APE) is calculated for the risk evaluation (modified from current Eurocode 8, Chapter. 2.1):

$$APE = 1 - EXP\left( {\frac{{LN\left( {1 - POE50} \right)}}{50}} \right)$$

(12)

Table 15 Bins and parameters for the risk calculation of the vessel at the industrial sites DE-1 and GR-3

Performing the convolution of the fragility (row #3 and #4) and the hazard curves (row #5 and #6) by using the values of the nine bins in Table 15 results in computed total vessel failure probability of 8.15E−4 and 1.18E−2 (sum of row #7 and #8, respectively) perpendicular to the longitudinal direction and 2.21E−8 and 4.16E−6 (sum of row #9 and #10) in longitudinal direction, respectively for DE-1 and GR-3. It should be noted that prEN1998-1-1 (2022) proposes for structures of consequence class CC2 a default target annual probability of 2E−4 for the case of near collapse (NC). For consequence class (CC3 or higher) an annual target annual failure probability of down to 10^–7 for fundamental and fatigue design situations at the ultimate limit state is proposed in prEN 1990 (2020). Therefore, for this example the probabilistic verifications of the vessel on the top floor are not satisfied in both loading directions.

6.4 Discussion on consequences for structural assessment

On the example of the LNG tank and verification for 5000 years return period it has been shown, that even in low seismicity regions the application of national hazard maps (DIN EN 1998-1/NA, 2021) and new ESHM20 can lead to 1.6–2.0 higher support forces in comparison to a scaling of the PGA values with importance factors ranging from 2.0 to 2.2. Simply scaling the seismic input based on 475 years return period to higher spectral values, in the attempt to reproduce the spectral values at larger return periods, has several limitations, but is of course a very pragmatic practical approach. Nevertheless, as with changing return period also the spectral shape of the UHS is expected to change within some boundaries, making use of response spectra derived from a probabilistic seismic hazard assessment should be preferred. But more importantly, it needs to be recognized that scaling by a factor does not bring the response spectra to the same hazard level (usually lower) as using a higher return period and they cannot be converted into each other. In the case of critical facilities, the practice is to assume that e.g. a UHS for an APE of 1E−4 (10.000 years) is representative in its shape also for lower and higher APEs. Furthermore, those structures are usually designed in a very conservative manner where ground motions at low return period really do not matter. This is to a certain degree acceptable and enables e.g. in the safety margin analysis to make shortcuts by simply using scaled values. In this paper the starting point is the 475-year return period UHS with scaling towards much lower APEs. In this case, it is not really defendable to justify that the spectral shape would remain identical also for 10.000 years and that the amplitude is a result of a linear extrapolation, as the hazard curve is nonlinear in that transition.

Thus, in terms of safety level both outcomes are still useful but not directly comparable, and their use depends on the needs of the engineer. The definition of the required safety levels is a national task, but a consistent procedure all over Europe would be desirable and a benefit for engineers in terms of comparability and achievement of the same safety targets.

With the second example of the storage vessel on the top floor of a typical steel frame it has been observed that even though the PGA increases with the new ESHM20 model, the structural response does not necessarily increase in comparison to the national codes, given the considerations of the response spectral analysis of the contributing eigenmodes and differing spectral shapes. This fact can be seen for the industrial site in Germany due to different site amplification factors, but not for Greece. Therefore, it can be generally summarized that the differences between the application of ESHM20 and national codes are highly country specific.

The assessment for the LNG tank and the storage vessel was performed with the conventional, deterministic verification approach based on target safety levels associated with performance factors or return periods to comply with. Alternatively, a risk assessment was conducted using the site-specific hazard curves now available in ESHM20 to also appreciate a risk assessment in terms of complying with certain performance goals. A simplified reliability-based probabilistic verification according to the prEN-1998-1-1 (2022), Annex F was carried out for the storage vessel. The verification only considered the vessel fragility and not a combined failure assessment of all components together with the steel frame itself, nor a fault tree analysis to keep it simple. First of all, it can be seen that the result is of course highly dependent on the utilized fragility functions parallel and perpendicular to the longitudinal axis direction of the vessel. But comparing the difference in input hazard to the structure with the resulting risk leads to orders of magnitude difference, which shows the sensitivity of the probabilistic approach in comparison to the partial safety concept. However, the resulting risk number raises also the question about the practical comparison with proposed target values given in prEN 1998-1-1 (2022) and EN 1990 (2020), since very high requirements are imposed with these target values that are roughly comparable with power plants for high-risk structures. Of course, the observations for only one vessel cannot be turned into general conclusions, and further guidance seems certainly necessary for the designers, as the derivation of adequate fragility curves is difficult and time consuming and there is yet not a large experience when comparing the results to target risk values for conventional civil structures in the context of societal acceptance (Calgaro and Gulvanessian 2001; Proske 2019). This discussion should take place at the European level in order to get uniform levels for risk analyses, which will become increasingly important in future.

7 Conclusions

The newly released European Seismic Hazard Model-ESHM20-has been adopted as a reference for the definition of seismic action in the upcoming revision of Eurocode 8. The design of industrial facilities is based on the choice of specific limit states and consequence classes depending on the consequences of damage, on their importance for public safety and civil protection in the immediate post-earthquake period, and on the economic, social and environmental consequences. It is preferable to define the corresponding seismic actions in terms of specific return periods, rather than by a simplified linear scaling of reference spectra, which is still the standard in most member states due to the lack of hazard maps with different return periods. However, the presented study for 79 industrial sites in nine countries shows significant country- specific differences between the scaled spectra used for the original design according to the national codes and the spectra according to ESHM20 for higher return periods. The use of the ESHM20 maps avoids the bias associated with simply scaling of the code spectra and provides a consistent basis for the engineering practice in Europe in terms of comparability and achievement of the same safety targets. This is of great importance for industrial plants, as they usually have to be designed for higher return periods than 475 years due to the higher risk potential compared to conventional building structures. The impact through the use of ESHM20 hazard maps on the design of industrial facilities was clearly demonstrated for a single containment LNG tank and a five-storey moment resisting frame with technical components. Furthermore, the implementation of a reliability-based verification of industrial facilities and components with a fully probabilistic approach has made one step forward with ESHM20 through the provision of hazard curves at pan-European scale. The practical use of ESHM20 hazard curves was demonstrated with a simplified reliability-based probabilistic verification according to prEN-1998-1-1 (2021) to assess the performance of a vessel in a five-storey moment resisting frame in fully probabilistic terms. However, the example also showed that the interpretation of the target probability of exceedance given a consequence class and limit state deserves some further discussion among experts to put it into the right practical context. Although these definitions concerning safety are made at national level, a harmonization at European level is needed and hence recommended.

The present pilot study for selected industrial sites has shown that the new hazard maps based on ESHM20 provide an excellent basis for the seismic risk assessment of existing facilities in Europe, which are often designed with outdated seismic codes or even without any consideration of seismic actions. Therefore, it is planned to continue the study to assess the safety level of industrial plants in Europe on a more comprehensive data basis.