With the growing accessibility of high-performance computing (HPC), 3D physics-based simulations (3D PBS) have become a standard tool for simulating earthquake ground motions. While the spatial extent of regions modelled in 3D PBS is expanding, small-scale features such as highly heterogeneous and/or non-linear soft sediments as well as the built environment in urban areas are often simplified or omitted altogether. This is largely due to the exponential computational overhead required for their accurate representation, but also to the complexity in getting a detailed level of information in large areas. However, the localised effects of these features on ground motions are well-documented in smaller-scale simulations. For instance, detailed site response analyses routinely incorporate realistic nonlinear constitutive models in 1D and 2D, while soil-structure interaction (SSI) studies examine the influence of building response on ground motions in small domains. However, these effects remain underexplored at urban scales (Site-City Interaction, SCI), where they are likely to be more pronounced. In this context, the primary objective of this work is to advance the state-of-the-art in 3D PBS by enabling the simulation of realistic ground motions in large urban areas while maintaining computational efficiency. The key contributions of this research mostly include (1) the progress on validated case studies of simulated ground motions in urban areas, (2) the development and implementation of advanced numerical tools within the high-performance spectral element code SPEED \url{https://speed.mox.polimi.it/} developed at Politecnico di Milano, (3) the incorporation of realistic nonlinear soil behaviour and (4) the implementation of the module SPEED-SCI for coupling of simple nonlinear building models in 3D PBS to analyse SCI effects. Nonlinear soil behaviour is modelled by extending Iwan's multi-surface plasticity approach from 1D to 3D. This method discretises the nonlinear stress-strain curve of soil into linear segments, effectively eliminating the need for computationally intensive Newton-Raphson iterations. In SPEED-SCI, buildings are simplified as nonlinear multiple degree-of-freedom (MDOF) shear systems. Base reaction forces from buildings are applied as point forces on the ground surface, while the buildings are subjected to ground accelerations at each time iteration. This approach avoids the need for mesh refinement and eliminates additional computational overhead. A pre-processing tool is also developed to calibrate nonlinear structural parameters from simple building inventory data, streamlining the workflow and enhancing usability. Validation of 3D PBS was performed within a Blind Prediction exercise, organized within the Effects of Surface Geology (ESG6) conference, simulating broadband ground motions from weak and strong events of the 2016 Kumamoto earthquake sequence. The region's highly heterogeneous subsurface model and complex finite-fault seismic source presented significant challenges. The SPEED workflow is updated to account for these complexities. The 3D PBS (visco-elastic) were first performed by omitting the thin heterogeneous sediments to generate the low-frequency ground motions, which were then enriched with high frequencies using a routine based on artificial neural networks (ANN2BB). A subsequent 1D site response analysis was carried out to include the seismic response of the shallow sediments down to about \qty{50}{\meter} depth. While these simulations are not fully coupled, they satisfactorily reproduced recorded ground motions, demonstrating the efficiency of the 3D PBS and ANN2BB approaches. The Kumamoto region model was further refined to include highly heterogeneous soft sediments, enabling fully coupled nonlinear simulations employing the nonlinear model developed in this work. These simulations were stable and required only twice the computational time of viscoelastic simulations. Results indicated that while shear modulus degradation is concentrated at the basin edge near the seismic source, long-period ground motions are deamplified on the other side of basin, as they traversed the basin. The accuracy and stability of the SPEED-SCI implementation was validated against CAMUS-III shake table experimental data. Further, its practicality was demonstrated through a 3D plane wave simulation of the Kumamoto region incorporating a cluster of 298 tall buildings in the city centre. Observed amplification and deamplification of ground motions near buildings were consistent with previous studies, though the spatial variation is case-dependent for this configuration of buildings. While fully coupled fault-to-structure simulations were not yet performed in this work, all necessary components are developed, and tested, which can be exploited in immediate future.
Con l’accessibilità crescente del calcolo ad alte prestazioni (HPC), le simulazioni fisiche tridimensionali (3D PBS) sono diventate uno strumento standard per la simulazione dei movimenti sismici del suolo. Sebbene l’estensione spaziale delle regioni modellate nelle 3D PBS sia in continua espansione, caratteristiche su piccola scala, come sedimenti soffici altamente eterogenei e/o non lineari, nonché l’ambiente costruito nelle aree urbane, vengono spesso semplificate o del tutto omesse. Ciò è dovuto principalmente all’elevato costo computazionale necessario per una rappresentazione accurata, ma anche alla complessità nell’ottenere informazioni dettagliate su vaste aree. Tuttavia, gli effetti localizzati di queste caratteristiche sui movimenti del suolo sono ben documentati in simulazioni su scala più piccola. Ad esempio, analisi dettagliate della risposta del sito incorporano di routine modelli costitutivi non lineari realistici in 1D e 2D, mentre gli studi sull’interazione suolo-struttura (SSI) esaminano l’influenza della risposta degli edifici sui movimenti del suolo in domini limitati. Tuttavia, questi effetti rimangono poco esplorati su scala urbana (Interazione Sito-Città, SCI), dove è probabile che siano più pronunciati. In questo contesto, l'obiettivo principale di questo lavoro è avanzare lo stato dell’arte delle 3D PBS, consentendo la simulazione di movimenti sismici realistici su vaste aree urbane, mantenendo al contempo l’efficienza computazionale. I principali contributi di questa ricerca includono (1) il progresso negli studi di caso validati di simulazioni di movimenti del suolo in aree urbane, (2) lo sviluppo e l’implementazione di strumenti numerici avanzati all’interno del codice agli elementi spettrali ad alte prestazioni SPEED \url{https://speed.mox.polimi.it/}, sviluppato presso il Politecnico di Milano, (3) l’incorporazione di un comportamento realistico non lineare del suolo e (4) l’implementazione del modulo SPEED-SCI per l’accoppiamento di modelli semplificati non lineari di edifici nelle 3D PBS per analizzare gli effetti SCI. Il comportamento non lineare del suolo è modellato estendendo l’approccio della plasticità multi-superficie di Iwan da 1D a 3D. Questo metodo discretizza la curva tensione-deformazione non lineare del suolo in segmenti lineari, eliminando efficacemente la necessità di costose iterazioni di Newton-Raphson. Nel modulo SPEED-SCI, gli edifici sono modellati come sistemi di taglio non lineari a più gradi di libertà (MDOF). Le forze di reazione alla base degli edifici sono applicate come forze puntuali sulla superficie del suolo, mentre gli edifici sono sottoposti ad accelerazioni al suolo ad ogni iterazione temporale. Questo approccio evita la necessità di un raffinamento della mesh ed elimina il sovraccarico computazionale aggiuntivo. Inoltre, è stato sviluppato uno strumento di pre-processing per calibrare i parametri strutturali non lineari a partire da semplici dati di inventario edilizio, ottimizzando il flusso di lavoro e migliorando la fruibilità. La validazione delle 3D PBS è stata effettuata nell’ambito di un esercizio di previsione alla cieca, organizzato durante la conferenza Effects of Surface Geology (ESG6), simulando movimenti sismici a larga banda per eventi deboli e forti della sequenza sismica di Kumamoto del 2016. Il modello geologico altamente eterogeneo della regione e la complessa sorgente sismica finita hanno presentato notevoli sfide. Il workflow di SPEED è stato aggiornato per tener conto di tali complessità. Le 3D PBS (visco-elastiche) sono state inizialmente eseguite omettendo i sottili sedimenti eterogenei per generare movimenti del suolo a bassa frequenza, successivamente arricchiti con alte frequenze utilizzando una routine basata su reti neurali artificiali (ANN2BB). Un'analisi della risposta del sito in 1D è stata poi effettuata per includere la risposta sismica dei sedimenti superficiali fino a circa \qty{50}{\meter} di profondità. Sebbene queste simulazioni non siano completamente accoppiate, hanno riprodotto in modo soddisfacente i movimenti del suolo registrati, dimostrando l’efficienza dell’approccio 3D PBS e ANN2BB. Il modello della regione di Kumamoto è stato successivamente affinato per includere sedimenti soffici altamente eterogenei, consentendo simulazioni completamente accoppiate non lineari con il modello sviluppato in questo lavoro. Tali simulazioni sono risultate stabili e hanno richiesto solo il doppio del tempo computazionale rispetto alle simulazioni viscoelastiche. I risultati hanno indicato che, mentre la degradazione del modulo di taglio si concentra sul bordo del bacino vicino alla sorgente sismica, i movimenti del suolo a lungo periodo risultano deamplificati sul lato opposto del bacino, dopo averlo attraversato. L'accuratezza e la stabilità dell'implementazione di SPEED-SCI sono state validate confrontandole con i dati sperimentali della tavola vibrante CAMUS-III. Inoltre, la sua applicabilità è stata dimostrata attraverso una simulazione tridimensionale di un’onda piana nella regione di Kumamoto, incorporando un cluster di 298 edifici alti nel centro città. L’amplificazione e la deamplificazione osservate dei movimenti del suolo in prossimità degli edifici sono risultate coerenti con studi precedenti, sebbene la loro variazione spaziale dipenda dalla configurazione specifica degli edifici. Sebbene in questo lavoro non siano state ancora eseguite simulazioni completamente accoppiate dalla sorgente sismica alla struttura, tutti i componenti necessari sono stati sviluppati e testati, rendendoli sfruttabili nell’immediato futuro.
From the fault rupture to city seismic response: 3D multi-scale physics-based scenarios of earthquake effects
Sangaraju, Srihari
2024/2025
Abstract
With the growing accessibility of high-performance computing (HPC), 3D physics-based simulations (3D PBS) have become a standard tool for simulating earthquake ground motions. While the spatial extent of regions modelled in 3D PBS is expanding, small-scale features such as highly heterogeneous and/or non-linear soft sediments as well as the built environment in urban areas are often simplified or omitted altogether. This is largely due to the exponential computational overhead required for their accurate representation, but also to the complexity in getting a detailed level of information in large areas. However, the localised effects of these features on ground motions are well-documented in smaller-scale simulations. For instance, detailed site response analyses routinely incorporate realistic nonlinear constitutive models in 1D and 2D, while soil-structure interaction (SSI) studies examine the influence of building response on ground motions in small domains. However, these effects remain underexplored at urban scales (Site-City Interaction, SCI), where they are likely to be more pronounced. In this context, the primary objective of this work is to advance the state-of-the-art in 3D PBS by enabling the simulation of realistic ground motions in large urban areas while maintaining computational efficiency. The key contributions of this research mostly include (1) the progress on validated case studies of simulated ground motions in urban areas, (2) the development and implementation of advanced numerical tools within the high-performance spectral element code SPEED \url{https://speed.mox.polimi.it/} developed at Politecnico di Milano, (3) the incorporation of realistic nonlinear soil behaviour and (4) the implementation of the module SPEED-SCI for coupling of simple nonlinear building models in 3D PBS to analyse SCI effects. Nonlinear soil behaviour is modelled by extending Iwan's multi-surface plasticity approach from 1D to 3D. This method discretises the nonlinear stress-strain curve of soil into linear segments, effectively eliminating the need for computationally intensive Newton-Raphson iterations. In SPEED-SCI, buildings are simplified as nonlinear multiple degree-of-freedom (MDOF) shear systems. Base reaction forces from buildings are applied as point forces on the ground surface, while the buildings are subjected to ground accelerations at each time iteration. This approach avoids the need for mesh refinement and eliminates additional computational overhead. A pre-processing tool is also developed to calibrate nonlinear structural parameters from simple building inventory data, streamlining the workflow and enhancing usability. Validation of 3D PBS was performed within a Blind Prediction exercise, organized within the Effects of Surface Geology (ESG6) conference, simulating broadband ground motions from weak and strong events of the 2016 Kumamoto earthquake sequence. The region's highly heterogeneous subsurface model and complex finite-fault seismic source presented significant challenges. The SPEED workflow is updated to account for these complexities. The 3D PBS (visco-elastic) were first performed by omitting the thin heterogeneous sediments to generate the low-frequency ground motions, which were then enriched with high frequencies using a routine based on artificial neural networks (ANN2BB). A subsequent 1D site response analysis was carried out to include the seismic response of the shallow sediments down to about \qty{50}{\meter} depth. While these simulations are not fully coupled, they satisfactorily reproduced recorded ground motions, demonstrating the efficiency of the 3D PBS and ANN2BB approaches. The Kumamoto region model was further refined to include highly heterogeneous soft sediments, enabling fully coupled nonlinear simulations employing the nonlinear model developed in this work. These simulations were stable and required only twice the computational time of viscoelastic simulations. Results indicated that while shear modulus degradation is concentrated at the basin edge near the seismic source, long-period ground motions are deamplified on the other side of basin, as they traversed the basin. The accuracy and stability of the SPEED-SCI implementation was validated against CAMUS-III shake table experimental data. Further, its practicality was demonstrated through a 3D plane wave simulation of the Kumamoto region incorporating a cluster of 298 tall buildings in the city centre. Observed amplification and deamplification of ground motions near buildings were consistent with previous studies, though the spatial variation is case-dependent for this configuration of buildings. While fully coupled fault-to-structure simulations were not yet performed in this work, all necessary components are developed, and tested, which can be exploited in immediate future.File | Dimensione | Formato | |
---|---|---|---|
Sangaraju_thesis_v2.pdf
non accessibile
Descrizione: PhD Thesis - Srihari Sangaraju
Dimensione
33.86 MB
Formato
Adobe PDF
|
33.86 MB | Adobe PDF | Visualizza/Apri |
I documenti in POLITesi sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/10589/234619