3d physics-based numerical simulations: advantages...

1

3D PHYSICS-BASED NUMERICAL SIMULATIONS: ADVANTAGES AND LIMITATIONS OF A NEW FRONTIER

TO EARTHQUAKE GROUND MOTION PREDICTION

Roberto Paolucci

Department of Civil and Environmental Engineering, Politecnico di Milano

6th National Conference on Earthquake Engineering&

2nd National Conference on Earthquake Engineering and Seismology

June 14-16, 2017

Roberto Paolucci

2Contents

Motivation for 3D physics-based earthquake ground motion simulations

the spectral element code SPEED

3D physics-based earthquake ground motion simulations in Istanbul

producing broadband ground motions

2

Roberto Paolucci

3Approaches to earthquake ground motion prediction

Physics-based

Roberto Paolucci

4

NGA West2 database (Ancheta et al., 2013)

range of potential

damage to structures

Limitations in the use of GMPEs

3

Roberto Paolucci

5Limitations in the use of GMPEs

Roberto Paolucci


4

Roberto Paolucci


NGA2 recordsM>6 R<20km

Roberto Paolucci


INGV wordlwide near source dataset6 < M < 7 & PGA > 100 cm/s2

5

Roberto Paolucci

9Physics-based 3D earthquake ground motion simulations

To create a numerical laboratory to simulate “earthquakeground shaking scenarios” as realistic as possible in terms of:

the complexity of the seismic source

the complexity of the geological and morphologicalenvironment

the frequency range of the seismic excitation and ofresulting ground motion

Objective

Roberto Paolucci

10

… but you need

3D numerical models

high frequency

verifications on real earthquakes

GMPEs3D physics based GM simulations

Physics-based 3D earthquake ground motion simulations

6

Roberto Paolucci

11Contents





Roberto Paolucci

12SPEED (speed.mox.polimi.it)

7

Roberto Paolucci

13

Faccioli E, Maggio F, Paolucci R, Quarteroni A. 2D and 3D elastic wave propagation by a pseudo-spectral domain decompositionmethodJournal of Seismology, 1997

Stupazzini M., Paolucci R., Igel H. Near-fault earthquake ground motion simulation in the Grenoble Valley by a high-performance spectral element code Bulletin of the Seismological Society of America, 2009

Mazzieri I., Stupazzini M., Guidotti R., Smerzini C. SPEED-Spectral Elements in Elastodynamics with Discontinous Galerkin. A non-conforming approach for 3D multi-scale problems. International Journal for Numerical Methods in Engineering, 2013

SPEED: some references

Roberto Paolucci

14

Features 3D unstructured conforming and non-conforming hexahedral meshes ( e.g., between sub-domains Ω1,2, Ω3 and Ω4) Non uniform polynomial approximation orders(e.g., between sub-domains Ω1 and Ω2) leap-frog FD time advancing scheme visco-elastic and non-linear elastic soil behaviour

Kernel

hybrid parallel programming based on MPI and Open-MP METIS software library to handle partitioning and load balancing designed for multi-core machines or large clusters -optimized for HPC clusters (e.g., FERMI Blugene/Q)

SPEED: main features

8

Roberto Paolucci

15SPEED: spatial discretization

Legendre-Gauss-Lobatto points

Lagrange polynomials

Roberto Paolucci

ikjj

kik

kk

ij nnV

tMtm

),x(),x( 0

);(),x(0kR

kkkkk ttsAutM

shear modulus

Slip source function

co-seismic slip

volume of the kth subfaultslip and normal fault

vectors f(f,f,f)area How to introduce

high-frequencycomponents ?

How to introduce spatial

incoherency?

SPEED: Treatment of seismic input

Kinematic modeling of an extended seismic source

16

9

Roberto Paolucci

SPEED: soil modelling17

clay

sand

Viscoelastic models

Q = cost

Q = Q0f

Rayleigh damping

Non-linear elasticity

Roberto Paolucci

18Previous applications of SPEED (1)

Stupazzini M, Paolucci R, Igel H (2009) Near-fault earthquake ground-motion simulation in the Grenoble valley by a high-performance Spectral Element code. BSSA , 99: 286–301.

Smerzini C, Villani M (2012)Broadband numerical simulations in complex near field geological configurations: the case of the MW 6.3 2009 L'Aquila earthquake, BSSA, 102: 2436–2451

10

Roberto Paolucci

19Previous applications of SPEED (2)

Gubbio (Central Italy)

Sulmona (Central Italy)

Santiago de Chile

Christchurch earthquake February 22, 2011 (New Zealand)

Wellington (New Zealand)

Po Plain earthquake May 29, 2012 (Northern Italy)

Marsica earthquake 1915 (Central Italy)

Thessaloniki (Greece)

Istanbul (Turkey)

Beijing (China)

Roberto Paolucci

20Contents





11

Roberto Paolucci

21Past century earthquakes along the North Anatolian Fault

after Bohnhoff et al., 2013

Roberto Paolucci

22Princes Islands segment of the NAF

Yellow stars mark estimated epicentres of major earthquakes along the Princes Islandssegment during the last 2,000 years. After Bohnhoff et al., 2013

12

Roberto Paolucci

23Setup of the 3D physics-based numerical simulations

Roberto Paolucci

24Construction of the numerical model (1)

Combined digital elevation/bathimetry model

13

Roberto Paolucci


Vs model

Roberto Paolucci


Fault geometry model

14

Roberto Paolucci


Spectral element numerical model (resolution: fmax= 1.5 Hz)

Roberto Paolucci

28Slip distribution along the fault

HB94: Herrero and Bernard (1994)CA15: Crempien and Archuleta (2015)

Kinematic slip distribution models

15

Roberto Paolucci

29Construction of scenario earthquakes

45 scenario earthquakes of M ranging from 7 to 7.4 where constructed by considering randomly generated hypocenter and slip distribution models, with both HB94 and CA15 approaches

Roberto Paolucci

30Post-processing 45 shaking scenarios (7.0 Mw 7.4)

Forward directivity PGV (cm/s) Scen Antidirective PGV (cm/s)Neutral directivity PGV (cm/s) Backward directivity PGV (cm/s)

Comparison of Mw7 scenarios: effect of directivity

16

Roberto Paolucci


Ground motions for a selected scenario

Roberto Paolucci


Ground motions for a selected site

17

Roberto Paolucci


Neutral Forward Backward

Animation of ground motion: effect of directivity

Roberto Paolucci

34

when 3 segments are ruptured, the FW directivity scenario wrtIstanbul is the most likely to occur

Dependendance of results on ruptured segments

18

Roberto Paolucci

35Comparison with GMPEs

Bray & Rodriguez-Marek (2004)only FW directivity

Chiou and Youngs (2008)

Roberto Paolucci

36

Bray & Rodriguez-Marek (2004)only FW directivity

Chiou and Youngs (2008)

Comparison with GMPEs

19

Roberto Paolucci

37Components of single-station standard deviation

between-event s.d.within-event s.d.

total s.d.

records of Christchurch earthquakes(Chen and Faccioli, 2013)

Roberto Paolucci

38Contents





20

Roberto Paolucci

39

(1) A hybrid broadband is first produced by combining the low-frequencysimulation with the high-frequency ground motion generated from stochastic approaches

Producing broadbands from SPEED

lack of correlation of the LF and HF parts100

10-1

100

101

102

f (Hz)

FA

S (

cm/s

)

LF(SPEED)HF(SP96)BB

0 1 2 3 40

0.5

1.0

f (Hz)

fl f

h

LF filter HF filter

0 5 10 15 20 25 30

-200

0

200

LF

(cm

/s2 )

0 5 10 15 20 25 30

-200

0

200

HF

(cm

/s2 )

0 5 10 15 20 25 30

-200

0

200

BB

(cm

/s2 )

t (s)

LF: SPEED (f < 1.5 Hz)

HF: SP96 (f > 1.5 Hz)

BB ( 0 < f < 25 Hz)

Roberto Paolucci

40Producing broadbands from SPEED

(2) An ANN is trained (una tantum) based on a strong motion dataset(SIMBAD, Smerzini et al., 2014)

21

Roberto Paolucci

41

(3) The "hybrid" broadband is scaled in order to match the ANN trained response spectrum

T

Sa

T*

region of validity of numerical simulations

hybrid response spectrum

ANN responsespectrum

3D physics-basedsimulated response

spectrum

hybrid

ANN-matched


Roberto Paolucci

42

(4) Verification on real case studies: simulation of Po Plain eqk, May 29 2012


response spectra Fourier spectra

rec velocity

rec acceleration sim acceleration

sim velocity

22

Roberto Paolucci

43


Spatial correlation of PGAs


reco

rded

sim

ulat

ed

Roberto Paolucci

44


Spatial correlation of Sa=1s


reco

rded

sim

ulat

ed

23

Roberto Paolucci

453D numerical simulations of the May 29 2012 Po Plain earthquake

Roberto Paolucci

463D numerical simulations of the May 29 2012 Po Plain earthquake

24

Roberto Paolucci

47Producing ground motion movies: 2012 Po Plain eqk

Roberto Paolucci

48

Pros

accounting for specific geologic and tectonic conditions

capturing physics of earthquake ground motion, especially near-source

proper spatial correlation of motion between adjacent sites

Cons

large number of scenarios needed: how to cluster them?

limited reliability in the high-frequency range, although in progress

need of vast amount of input data: suitable for large urban areas

Conclusions

25

Roberto Paolucci

49Acknowledgements

Reluis (Italy)

RS2 Project : "3D numerical simulations and near-source effects"

Munich Re (Germany)

"3D physics-based earthquakeground shaking scenarios in large urban areas"

3d physics-based numerical simulations: advantages...

Documents