Über die Implementierung eines User Elementes

(UEL) in Abaqus für dynamische hydro-mechanisch

gekoppelte Probleme

Seminar ZG 28.04.2020
Antal Csuka
Structure Presentation
• General Aspects and why do we need a UEL
implementation

• Formulation of the implemented UEL

• Verification of UEL

• General Aspects in working with the UEL

• Linear vs. Nonlinear Geometry

• Possible Extensions
 Hydro-mechanical coupling

• Firstly, what means „coupling“ ?

• Example of 2 system of diff. equations:

𝑥ሶ = 𝑎𝑥 + 𝑏 𝑥ሶ = 𝑎𝑥 + 𝑏𝑦
a, b, c, d constants
𝑦ሶ = 𝑐𝑦 + 𝑑 𝑦ሶ = 𝑐𝑦 + 𝑑𝑥

uncoupled coupled

• Hydro-mechanical coupling in a soil means that the pore pressure and fluid
flow affect the deformation and strength of the soil skeleton
3 of 32
Hydro-mechanical coupling 1D

• Terzagi – Consolidation Theory

• „Simple“ Analytical Solution for 1D

4 of 32
Hydro-mechanical coupling 2D/3D

• 2D&3D cases for stationary Problems -> Flownet (Strömungsnezt) also for
anisotropical permeability but homogenous

• For instationary (time dependent) Problems Numerical Solution are required

(e.g. FEM, FDM, FVM,..)

• E.g. instationary Problem: Calculation of the phreatic surface in an earth dam

Vermeer and
Verruijt (1981):
𝛾
∆𝑡 ≥ 𝑤 (∆ℎ)2
6𝐸𝑘

Ende Aufstauung 10h nach Aufstauung Steady-State

5 of 32
 Why the Title: „About the implementation…“

Vibro Compaction, Earthquake …

Stationary case dynamic processes
(uncoupled) Assumption a = 0 still correct?
Solutions?

FEM-Softwares
Just Slow-Procesess (Abaqus, Plaxis, Ansys…) High coupling (undrained)
Quasi-Statical e.g. Abaqus with UMAT
Intertia Forces = 0

Instationary case Plaxis VIP 2018

(coupled) Dynamic Consolidation

For the general case of moderate coupling

very limited solutions given by the Standard Packages

6 of 32
Why the Title: „About the implementation…“

For the general case of moderate coupling

very limited solutions given by the Standard Packages

• Technically is possible … old stuff (Zienkiewicz 1984; Biot 1941, 1956, 1962)

• Main Problem: Time-Stepping

𝛾
• 𝑤
Vermeer and Verruijt (1981): ∆𝑡 ≥ 6𝐸𝑘 (∆ℎ)2

• Consider the Problem:

• 𝛾𝑤 = 10 kN/m³
• E = 60000 kPa
∆𝑡 ≥ 0.011 𝑠 >> time step for incremental non linear
• k = 1e-4 m/s² constitutive eq., e.g. 1e-5 s
• ∆ℎ = 0.2 m

7 of 32
 For FEM Discretization
Porous Media
Displacement & Porepressure

Water disp. U
Pore pres. p
Total disp. is better
Soil disp. u than relative disp.

Zienkiewicz, 1984

homogenization u-p-U Formulation

Saturation (no air)

Simplifications
WATER u-U Formulation
Compressible Fluid,Volumetric Locking

SOIL u-p Formulation

𝒖ሷ = 𝑼ሷ

8 of 32
Does 𝒖ሷ = 𝑼ሷ ??

Staubach&Machacek, 2019

Most of the cases yes,

For very big hydraulic conductivity, PW

gradients and frequency no
Does Darcy hold? Is FEM still capable?

9 of 32
Implementation of a User Element (UEL) for the
u-p Formulation (Chrisopoulos, 2018)

Ingredients for the strong form:

Strong form -> Differential

Weak form -> Integral -> Discretization -> Algebraic Equations (Matrix form)

10 of 32
 Implementation of a User Element (UEL) for the
u-p Formulation (Chrisopoulos, 2018)

Ingredients for the strong form:

• Conservation of Momentum for the whole continuum

• Conservation of Momentum for the fluid

• Constitutive equation for the fluid phase

• Constitutive eq. for the soil

11 of 32
Finite element discretisation
Weak form => Integral (over a domain)

Strong Form <=> diff. equation = 0

=> ∫diff equation = 0 (whole domain)

 Σ ∫ diff equation = 0 (smaller domain)

 discretization & interpolation (Ansatz Functions)

=> Algebraic Equations (Matrix form)

12 of 32
Finite element discretisation
Weak form => Integral (over a domain)

• Approximation for u&p:

• Possibilities: Requests:

Spatial derivatives
• Bilinear u and Constant p
Ladyzhenskaya–Babuška–Brezzi
• Bilinear u and p
condition
• Biquad u and Bilinear p
Physicality
(Stress & Porepressure same
• Biquad u and p
Interpolation order)
• …….

13 of 32
Finite element discretisation
Weak form => Integral (over a domain)

• Approximation for u&p:

• Possibilities: Requests:

Spatial derivatives
• Bilinear u and Constant p
Ladyzhenskaya–Babuška–Brezzi
• Bilinear u and p
condition
• Biquad u and Bilinear p
Physicality
(Stress & Porepressure same
• Biquad u and p
Interpolation order)
• …….

14 of 32
Finite element discretisation
Weak form => Integral (over a domain)

• Approximation for u&p:

15 of 32
 Finite element discretisation
Weak form => Integral (over a domain)

Residuum for displacement

Residuum for pore pressure

Missing in the Standard FEM-

Packeges
16 of 32
 Time discretisation and what Abaqus wants

Hilber-Hughes-Taylor
HHT - Algorithmus

Iteration during Taylor series

one increment

Coupling

17 of 32
Time discretisation and what Abaqus wants

Numerical integration

18 of 32
 dU

Steps Inside UEL U

• RHS – Residuum vector

• AMATRX – Stiffness Matrix
• SVARS – state variables
• PROPS(NPROPS) – property values
• COORDS – Node coordinates at the beginning
• U, DU, V, A – Displ., Displ. Inc., Velocity and Acceleration vectors
• PARAMS – alfa, beta, gamma – HHT
• LFLAGS – Solution Procedure
• TIME; DTIME
• NDOFEL, KSTEP, KINC, JELEM, MCRD…
19 of 32
 Steps Inside UEL
• Subroutine UEL is called

• Rearrangement of U, dU, V, A

• Start Gauss Loop over Gauss Points

Check Inc number => read SVARS (stress, strain, void ratio…)
Determine BMATRIX and strain increments
Call UMAT to get stress update
Calculate components of AMTARX& RHS; e.g. Kuu = Kuu + Kuu(Gausspoint)
Save SVARS in STA(jelem,actualG,1:n) for visualisation
• End Gauss Loop

• Calculate AMTRX&RHS coresponding to procedure

• END SUBROUTINE UEL

20 of 32
Verification: Terzaghi 1D Consolidation

σ = 100 kPa, E = 1300 kPa, ν = 0.4, k = 0.0001 m/s

Schümann, 2015

21 of 32
Verification: Cavity Expension (Chrisopoulos, 2018)

σ' = -50 kPa, p = 50 kPa

kf = 100 MPa

22 of 32
Verification: Cavity Expension (Chrisopoulos, 2018)

Good agreement More detailed Verification through

Difference due to UMAT?? Master Thesis (Sha,2020)

23 of 32
 UEL – General Working Aspects

• UNSYMM=YES, very important for the numerical procedure for matrix inverting

• Visualisation

No Visualisation Elements Visualisation Elements I Visualisation Elements II

Machacek, 2016

24 of 32
 UEL – General Working Aspects

• UNSYMM=YES, very important for the numerical procedure for matrix inverting

• Visualisation

• Use of Visualisation Elements for applying loads

• Contact – just Node to Surface (possible extension also using vis. elements??)

25 of 32
From linear geometry to nonlinear

Schümann (2015) about large deformations:

• Lin. Geom: Determination of all Elementmatrixs w.r.t. reference config
• Nlin. Geom: Determination of all Elementmatrixs w.r.t. current config.
Nodes coordinates x(t) = x0(t)+u(t)
• In *STEP activate NLGEOM=yes

Is it enough ???

26 of 32
1D Compresion Linear- vs Nonlinear-geom

u = 0.5

H = 2.0
E = 10000
v = 0.3

27 of 32
Simple Shear Test

28 of 32
„Pure“ Rotation
σ = -50 kPa

Linear: differences ca. 0.7%

Nonlinear: σ𝐸𝑛𝑑𝑒,𝐴𝑏𝑎𝑞𝑢𝑠 = -50 kPa
σ𝐸𝑛𝑑𝑒,𝑈𝐸𝐿 = 5.5 kPa

Open Question: Same behaviour also for pure rotation?

29 of 32
 Objective stress rate

• Cauchy stress tensor is objective

• Stress rate is not objective

• Objective are (most cited):

Truesdell rate of the Cauchy stress

Green-Naghdi rate of the Cauchy stress

Zaremba-Jaumann rate of the Cauchy stress

30 of 32
 Implemented in
ABAQUS
Hughes Winget Algorithm, 1980

Molenkamp,1986 ->difference between

Spin and Material Rotation Rate
significant for eps,dev > 10%

31 of 32
Some possible extensions

• Unsaturated case

• Variable conductivity (void ratio, effective stress…)

• Nonlinear geometry

• 3D - case

• Infinite Elements

32 of 32