Introduction to computational plasticity using FORTRAN

37
Introduction to computational plasticity using FORTRAN Youngung Jeong Changwon National Univ.

Transcript of Introduction to computational plasticity using FORTRAN

Introduction to computational

plasticity using

FORTRANYoungung Jeong

Changwon National Univ.

One dimensional elastic rod

โ€ข ๐‘‘๐œ€!" = #""

with #"#$= ๐‘

โ€ข Elastic constitutive law:โ€ข ๐”ผ๐œ€!" = ๐œŽ (elastic stiffness ๐”ผ = 200 [GPa] )โ€ข Assuming ๐”ผ is constant, the above constitutive

law can be expressed as:

โ€ข ๐‘‘๐œ€!" = %๐”ผ๐‘‘๐œŽ

L vs ๐œŽ?

๐‘ฃ = ๐‘ ๐‘‘๐‘ฃ๐‘‘๐‘ก = 0

๐‘™ ๐‘ก = 0 = ๐‘™!

One dimensional elastic rod

โ€ข ๐‘‘๐œ€!" = #""

with #"#$= ๐‘

โ€ข ๐‘‘๐œ€!" = %๐”ผ๐‘‘๐œŽ

โ€ข ๐‘™ = ๐‘™' + ๐‘๐‘ก

โ€ข %๐”ผ๐‘‘๐œŽ = #"

"โ†’ โˆซ'

( %๐”ผ๐‘‘๐œŽ = โˆซ"!

" #""โ†’ %

๐”ผ๐œŽ = ln "

"!

๐œŽ = ๐”ผ ln๐‘™๐‘™!

๐œŽ vs ๐‘ก ?

๐‘ฃ = ๐‘ ๐‘‘๐‘ฃ๐‘‘๐‘ก = 0

๐‘™ ๐‘ก = 0 = ๐‘™!

One dimensional elastic rod๐‘‘๐œ€!" = #"

"with #"

#$= ๐‘

๐‘‘๐œ€!" = ๐”ผ ๐‘‘๐œŽโ€ข ๐‘™ = ๐‘™' + ๐‘๐‘ก

1๐”ผ๐‘‘๐œŽ =

๐‘๐‘‘๐‘ก๐‘™' + ๐‘๐‘ก

โ†’ 6'

( 1๐”ผ๐‘‘๐œŽ = ๐‘6

'

$ ๐‘‘๐‘ก๐‘™' + ๐‘๐‘ก

โ†’1๐”ผ๐œŽ = ln ๐‘™' + ๐‘๐‘ก โˆ’ ln ๐‘™'

๐œŽ = ๐”ผ ln๐‘™! + ๐‘๐‘ก๐‘™!

Confirm following results:โ€ข ๐‘™' = 20 ๐‘š๐‘š

โ€ข #"#$= 0.001 ๐‘š๐‘š

โ€ข ๐”ผ = 200 ๐บ๐‘ƒ๐‘Ž (equivalently, 200000 ๐‘€๐‘ƒ๐‘Ž)โ€ข Loading for 30 seconds

Now that you have stress, you could estimate the force as well.

โ€ข Letโ€™s say, the thickness is 1 [mm] and the width is 12.5 [mm]; The cross-sectional area amounts to 12.5 [mm^2]

๐น = ๐ด โ‹… ๐œŽ = ๐ด ๐”ผ ln๐‘™! + ๐‘๐‘ก๐‘™!

๐ด โ‹… ๐œŽ = ๐‘š๐‘š(๐‘€๐‘ƒ๐‘Ž = 10)*๐‘š ( 10+๐‘๐‘š( = ๐‘

One dimensional Newtonian rod

โ€ข ๐‘‘๐œ€ = #""

with #"#$= ๐‘

โ€ข Newtonian fluidโ€™s constitutive law:

โ€ข ๐œ‚ #)#$= ๐œŽ

โ€ข Assuming ๐œ‚ is constant (Newtonian fluid), the above constitutive law can be expressed as:

โ€ข ๐œ‚ #)#$= ๐œŽ โ†’ ๐œ‚

"###$= ๐œŽ โ†’ *

"#"#$= ๐œŽ

L vs ๐œŽ?

๐‘ฃ = cmms

๐‘‘๐‘ฃ๐‘‘๐‘ก = 0

๐‘™ ๐‘ก = 0 = ๐‘™!

The cross-sectional area amounts to 12.5 [mm^2]

One dimensional Newtonian rod

๐œ‚๐‘™๐‘‘๐‘™๐‘‘๐‘ก= ๐œŽ

๐œ‚๐‘™' + ๐‘๐‘ก

๐‘ = ๐œŽ

Waterโ€™sviscosityisknownasฮท = 1.3ร—10"# [Pa โ‹… s] at 10ยฐ๐ถ๐œ‚๐‘™๐‘‘๐‘™๐‘‘๐‘ก =

๐‘ƒ๐‘Ž โ‹… ๐‘ ๐‘š๐‘š

๐‘š๐‘š๐‘  = ๐‘ƒ๐‘Ž

Newtonian fluid (water) under tension๐‘™! = 5. ๐‘š๐‘š๐‘‘๐‘™๐‘‘๐‘ก = 0.1

๐‘š๐‘š๐‘ ๐‘’๐‘

๐œ‚ = 1.3ร—10"# ๐‘ƒ๐‘Ž โ‹… ๐‘ 

The cross-sectional area amounts to 12.5 [mm^2]

Newtonian fluid (water) under compression

๐‘™! = 5. ๐‘š๐‘š๐‘‘๐‘™๐‘‘๐‘ก = โˆ’0.1

๐‘š๐‘š๐‘ ๐‘’๐‘

๐œ‚ = 1.3ร—10"# ๐‘ƒ๐‘Ž โ‹… ๐‘ 

Letโ€™s compare two different speed of compression.

๐‘™! = 50 ๐‘š๐‘š๐œ‚ = 1.3ร—10"# ๐‘ƒ๐‘Ž โ‹… ๐‘ 

๐‘‘๐‘™๐‘‘๐‘ก = โˆ’0.1

๐‘š๐‘š๐‘ ๐‘’๐‘

๐‘‘๐‘™๐‘‘๐‘ก = โˆ’1.0

๐‘š๐‘š๐‘ ๐‘’๐‘

My Python cheat sheet

Corn starch

https://youtu.be/Vx2DjGwnd44

A reasonable consideration (*VERY* phenomenological constitutive model)

Newtonian fluidโ€™s constitutive law:๐œ‚ "#"$= ๐œŽ

non-Newtonian fluid

๐œ‚(๐œŽ)๐‘‘๐œ€๐‘‘๐‘ก= ๐œŽ

with๐œ‚ ๐ˆ = ๐œ‚! + ๐›ผ๐ˆ

Continuedโ€ข ๐‘‘๐œ€ = #"

"with #"

#$= ๐‘

๐œ‚๐‘‘๐œ€๐‘‘๐‘ก= ๐œŽ โ†’ ๐œ‚' + ๐›ผ๐œŽ

๐‘‘๐‘™๐‘™๐‘‘๐‘ก

= ๐œŽ

โ†’๐œ‚!๐‘™๐‘‘๐‘™๐‘‘๐‘ก+๐›ผ๐œŽ๐‘™๐‘‘๐‘™๐‘‘๐‘ก= ๐œŽ โ†’

๐œ‚!๐‘๐‘™+๐›ผ๐œŽ๐‘๐‘™= ๐œŽ

โ†’๐œ‚!๐‘๐‘™+๐›ผ๐‘๐œŽ๐‘™= ๐œŽ โ†’

๐œ‚!๐‘๐‘™= ๐œŽ 1 โˆ’

๐›ผ๐‘๐‘™

โ†’๐œ‚!๐‘๐‘™ โˆ’ ๐›ผ๐‘

= ๐œŽOkay, analytical solution was easily found.

A reasonable consideration (*quite* demanding phenomenological

constitutive model)

Newtonian fluidโ€™s constitutive law:

๐œ‚๐‘‘๐œ€๐‘‘๐‘ก= ๐œŽ

non-Newtonian fluid

๐œ‚(๐œŽ)๐‘‘๐œ€๐‘‘๐‘ก= ๐œŽ

with a viscosity that is exponentially related with stress?

๐œ‚ ๐ˆ = ๐œ‚! exp๐ˆ๐›ผ

Continuedโ€ข ๐‘‘๐œ€ = #"

"with #"

#$= ๐‘

๐œ‚ ๐œŽ๐‘‘๐œ€๐‘‘๐‘ก= ๐œŽ โ†’ ๐œ‚' exp

๐ˆ๐›ผ

๐‘‘๐‘™๐‘™๐‘‘๐‘ก

= ๐œŽ

I gave up looking for the analytical solution of ๐ˆ(๐‘™)โ€ฆ

Numerical approach?

๐œ‚! exp๐ˆ#

$%$&= ๐œŽ

๐‘‘๐œ€ = $''

$'$&= ๐‘

Newton-Raphson method

Howtosolve?

๐œ‚P expQR

STSU = ๐œŽ

Ourgoalistofindacertain๐œŽ thatgives๐‘“ ๐œŽ = 0

Example NRโ€ข ๐‘“ ๐‘ฅ = ๐‘ฆ = cos ๐‘ฅ + 3 sin ๐‘ฅ + 3 exp(โˆ’2๐‘ฅ) = 0์˜ ํ•ด(์ฆ‰, y=0 ์ผ๋•Œ์˜ x๊ฐ’)๋ฅผ ์ฐพ์•„๋ณด์ž.

ํ•ด: y=0 ๋งŒ์กฑํ•˜๋Š”์ ์—์„œ์˜ x ์ขŒํ‘œ

Visual illustration of NR (ex 1)

Visual illustration of NR (ex 2)

Newton Raphson Method - Algorithmโ€ข 1. Guess x value and letโ€™s name it as ๐‘ฅ! where the subscript 0 means โ€˜initialโ€™.

โ€ข 2. Obtain new guess ๐‘ฅ$ by following the below tasks.

โ€ข Estimate ๐‘“ ๐‘ฅ! and "#"$

. In case "#"$

is a function of ๐‘ฅ. For the first attempt, use ๐‘ฅ!.

โ€ข Obtain the next guess ๐‘ฅ% by drawing a tangent line at the point of (๐‘ฅ!, ๐‘“ ๐‘ฅ! ) and obtain its intercept with x-axis. You can do it by defining the line function derived from the tangent line, i.e.,

๐’š =๐œ•๐‘“๐œ•๐‘ฅ ๐‘ฅ! ร— ๐’™ โˆ’ ๐‘ฅ! + ๐‘“(๐‘ฅ!)

Find the intercept of the line with x-axis, i.e., ๐‘ฆ = 0, which gives ๐‘ฅ$:

0 =๐œ•๐‘“๐œ•๐‘ฅ ๐‘ฅ! ร— ๐‘ฅ$ โˆ’ ๐‘ฅ! + ๐‘“ ๐‘ฅ! โ†’ ๐‘ฅ$ โˆ’ ๐‘ฅ! = โˆ’

๐‘“ ๐‘ฅ!๐œ•๐‘“๐œ•๐‘ฅ ๐‘ฅ!

โ†’ ๐‘ฅ$ = ๐‘ฅ! โˆ’๐‘“ ๐‘ฅ!๐œ•๐‘“๐œ•๐‘ฅ ๐‘ฅ!

โ€ข 3. We are using this intercept as the new ๐‘ฅ.

And repeat 2-1/2-2 steps until ๐‘“ ๐‘ฅ% โ‰ˆ 0.

์ดํŽ˜์ด์ง€๋ฅผํ™•์ธํ•˜์„ธ์š”:https://youngung.github.io/nr_example/

NR summary

โ€ข ๐‘ฅ%&' = ๐‘ฅ% โˆ’( )!"#"$ )!

โ€ข Repeat the above until ๐‘“ ๐‘ฅ% < tolerance

โ€ขOf course, you can do it manually, step-by-step. Usually, people make computer do the repetitive and tedious tasks.

Howtosolve?

๐œ‚P expQR

STSU = ๐œŽ

๐‘“ ๐œŽ = ๐œŽ โˆ’ ๐œ‚! ฬ‡๐œ€ exp๐ˆ๐›ผ

๐œ•๐‘“ ๐œŽ๐œ•๐œŽ = 1 โˆ’

๐œ‚! ฬ‡๐œ€๐›ผ exp

๐ˆ๐›ผ

Letโ€™sconsiderSTSU is given as ฬ‡๐œ€

Now, if you have a reasonable guess on ๐œŽ (say, ๐œŽ!), letโ€™s estimate next guess ๐œŽ$ and so on.

๐œŽ%&$ = ๐œŽ% โˆ’๐‘“ ๐œŽ%๐œ•๐‘“๐œ•๐‘ฅ ๐œŽ%

Cheat sheet

Continuedโ€ข ๐‘‘๐œ€ = #"

"with #"

#$= ๐‘

๐œ‚ ๐œŽ๐‘‘๐œ€๐‘‘๐‘ก= ๐œŽ โ†’ ๐œ‚' exp

๐ˆ๐›ผ

๐‘‘๐‘™๐‘™๐‘‘๐‘ก

= ๐œŽ

I gave up looking for the analytical solution of ๐ˆ(๐‘™)โ€ฆ

We might be able to use NR method to solve the above in combination with Euler method!

Euler + Newton-Raphson

๐œ‚P exp๐ˆ๐›ผ

๐‘‘๐‘™๐‘™๐‘‘๐‘ก

= ๐œŽ โ†’ ๐œ‚P exp๐ˆ๐›ผ

ฮ”๐‘™๐‘™ฮ”๐‘ก

= ๐œŽ โ†’ 0 = ๐œŽ โˆ’ ๐œ‚P exp๐ˆ๐›ผ

ฮ”๐‘™๐‘™ฮ”๐‘ก

๐‘™(ijk) = ๐‘™(i) + ฮ”๐‘™

๐‘ก(ijk) = ๐‘ก(i) + ฮ”๐‘ก

๐‘กP = 0๐‘™P = 0

ฮ”๐‘ก is, as usual, fixed as constant

Note that &'&( = ๐‘. If we apply Euler approximation,

ฮ”๐‘™ฮ”๐‘ก = ๐‘ โ†’ ฮ”๐‘™ = ๐‘ฮ”๐‘ก

โ†’ 0 = ๐œŽ โˆ’ ๐œ‚P exp๐ˆ๐›ผ

cฮ”๐‘ก๐‘™ฮ”๐‘ก

โ†’ 0 = ๐œŽ โˆ’ ๐œ‚P exp๐œŽ๐›ผ๐‘๐‘™

We apply the Newton-Raphson method to as below function:

โ†’ ๐‘“ ๐œŽ(%) = ๐œŽ(%) โˆ’ ๐œ‚! exp๐œŽ(%)๐›ผ

๐‘๐‘™(%)

โ†’๐œ•๐‘“ ๐œŽ๐œ•๐œŽ = 1 โˆ’

๐œ‚!๐›ผ exp

๐œŽ(%)๐›ผ

๐‘๐‘™(%)

โ€ข Outer loop over timeโ€ข Inner loop over NR search

Euler + Newton-Raphson (Cheat sheet)

Euler + Newton-Raphson

๐‘ฃ =๐‘‘๐‘™๐‘‘๐‘ก = cos(๐‘ก) โˆ’ 0.5

๐‘‘๐‘ฃ๐‘‘๐‘ก

= sin ๐‘ก

๐‘™ ๐‘ก = 0 = ๐‘™!

The cross-sectional area amounts to 12.5 [mm^2]

Euler + Newton-Raphson

๐œ‚P exp๐ˆ๐›ผ

๐‘‘๐‘™๐‘™๐‘‘๐‘ก

= ๐œŽ โ†’ ๐œ‚P exp๐ˆ๐›ผ

ฮ”๐‘™๐‘™ฮ”๐‘ก

= ๐œŽ โ†’ 0 = ๐œŽ โˆ’ ๐œ‚P exp๐ˆ๐›ผ

ฮ”๐‘™๐‘™ฮ”๐‘ก

๐‘™(ijk) = ๐‘™(i) + ฮ”๐‘™

๐‘ก(ijk) = ๐‘ก(i) + ฮ”๐‘ก

๐‘กP = 0๐‘™P = 0

ฮ”๐‘ก is, as usual, fixed as constant

Note that &'&( = cos(๐‘ก) โˆ’ 0.5. If we apply Euler approximation,ฮ”๐‘™ฮ”๐‘ก = cos(๐‘ก) โˆ’ 0.5 โ†’ ฮ”๐‘™ = cos(๐‘ก) โˆ’ 0.5 ฮ”๐‘ก

โ†’ 0 = ๐œŽ โˆ’ ๐œ‚! exp๐ˆ๐›ผ

cos(๐‘ก) โˆ’ 0.5 ฮ”๐‘ก๐‘™ฮ”๐‘ก

โ†’ 0 = ๐œŽ โˆ’ ๐œ‚! exp๐œŽ๐›ผcos(๐‘ก) โˆ’ 0.5

๐‘™

We apply the Newton-Raphson method to as below function:

โ†’ ๐‘“ ๐œŽ(%) = ๐œŽ(%) โˆ’ ๐œ‚! exp๐œŽ(%)๐›ผ

cos ๐‘ก(%) โˆ’ 0.5๐‘™(%)

โ†’๐œ•๐‘“ ๐œŽ๐œ•๐œŽ = 1 โˆ’

๐œ‚!๐›ผ exp

๐œŽ(%)๐›ผ

cos ๐‘ก(%) โˆ’ 0.5๐‘™(%)

Results

Euler + Newton-Raphson (Cheat sheet)