1 of 36

Local control in topology optimization: from additive manufacturing to stress constraints

Author:

Jose Antonio Torres Lerma

PhD Research Plan

Supervisors:

Dr. Alex Ferrer Ferre

Dr. Fermin Otero Gruer

2 of 36

Presentation Outline

Introduction

Additive manufacturing constraints

A new topology optimization algorithm with limited parameters

Stress constraints

Summary

03

09

19

25

31

3 of 36

Introduction

4 of 36

Motivation

4

Need to reduce fuel consumption and environmental impact in the transport industry

Aerospace industry

Automotive industry

Topology

Optimization

5 of 36

Topology Optimization

5

Problem formulation

Design variables

The original design variable is discontinuous

Common functionals:

Density

Level set

6 of 36

Additive manufacturing

6

Use of additive manufacturing (3D CAD data printing)

3D printing process failed!

Not enough sensitivity of the machine

50%

75%

85%

01

02

03

Topology optimized design

Minimum length scale constraint

7 of 36

Additive manufacturing

7

Use of additive manufacturing (3D CAD data printing)

3D printing process failed!

Material fell during deposition process

50%

75%

85%

01

02

03

Topology optimized design

Overhang constraint

8 of 36

Objectives overview

8

Objective 1

Local AM constraints

Objective 3

Stress constraints

Objective 2

SQP + TO multiple constraints

Manufacturability and integrity in TO

9 of 36

Additive manufacturing constraints

10 of 36

Additive manufacturing constraints

10

Length scales

constraints

Overhang

constraints

Bound formulation and three density fields representation

Implicit penalization with larger nº inequality/PDE constraints

Isotropic Perimeter as penalty term in the cost function

Efficient fulfillment of a global length

Overhang density filtering through threshold projection

Layer-by-layer computation of the gradient

Mechanical constraints in shape optimization

Complex shape derivatives

?

11 of 36

Objective

To define a simple and efficient local control of additive manufacturing constraints in topology optimization, working for density and level set

12 of 36

Global perimeter

12

Overhang

constraints

The Perimeter is a functional that computes the length of Ω boundaries.

Relative perimeter

INTERNAL BOUNDARIES

Total perimeter

INTERNAL + EXTERNAL BOUNDARIES

Shape derivative (level set)

Gradient methods (density)

1. Domain filtering

2. Perimeter computation

13 of 36

Global perimeter

13

Overhang

constraints

1. Domain filtering

Global isotropic relative perimeter: H1(D)

projection with Neumann boundary conditions.

Similar

Finite gradient

(S. Amstutz, C. Dapogny & A. Ferrer, 2022)

2. Perimeter computation

14 of 36

Global perimeter convergence

14

Overhang

constraints

15 of 36

Global perimeter

15

Other filtering strategies

1. Isotropic perimeter

2. Anisotropic perimeter

3. Non-linear anisotropic perimeter

Implementation

Already seen

Pending to implement

16 of 36

Current findings of global perimeter

16

Cantilever beam

17 of 36

Current findings of global perimeter

17

Gripper compliant mechanism

18 of 36

Extension to local perimeter

18

New motivation: control the length of boundaries and overhang more locally.

01

Global

constraint

02

Multiple

local constraints

One Lagrange multiplier per subdomain.

19 of 36

A new topology optimization algorithm with limited parameters

20 of 36

Optimization schemes

20

Goal: to solve a non-linear optimization problem with multiple constraints (PDE and others) in a versatile way

Augmented

lagrangian

Interior point

methods

MMA

Null space

algorithm

Gradient methods

Shape derivatives

Topological derivatives

Parameter dependent

Gradient methods

Shape derivatives

Topological derivatives

Gradient methods

Shape derivatives

Topological derivatives

Parameter dependent

Parameter dependent

Parameter free

Gradient methods

Shape derivatives

Topological derivatives

21 of 36

Topological derivatives

21

Null space algorithm (multiple constraints and parameter free)

Well suited for gradient methods, but not enough versatile

Spherical linear interpolation scheme

(Slerp)

Topological derivative

How the cost changes with an inclusion

Dr. Ferrer dissertation

22 of 36

Objective

To couple null space algorithm with topological derivatives and multiple local constraints

23 of 36

Null space optimizer

23

Null space algorithm (multiple constraints and parameter free)

Null space flow

Range space flow

Complete gradient flow

24 of 36

Current findings of null space/Slerp

24

Analytic problem with η=1

Topology optimization with density and η=1

Topology optimization using topological derivatives and η=1

25 of 36

Stress constraints

26 of 36

Additive manufacturing

26

Use of additive manufacturing (3D CAD data printing)

3D printing process completed!

50%

75%

100%

01

02

03

Topology optimized design

27 of 36

Three-point flexural test

27

Now, we should test if the printed design fulfills its function, according to the application.

Plasticity/damage generated!

The maximum stress has been exceeded

28 of 36

Objective

To include stress constraints in the entire framework

29 of 36

Stress computation

29

Elastic problem weak form

Stress tensor

But… what do we understand for homogenized stresses?

Intermediate material (micro). Which one?

30 of 36

Stress constraint

30

We choose a metric for the maximum stress and then… we obtain more constraints to consider!

The maximum function is not differentiable, and we can not use bound formulation since the Von Mises stresses are defined at each gauss point.

Thus, we may relax the problem:

31 of 36

Summary

32 of 36

Main goal and its objectives

32

We will ensure the manufacturability and integrity of topology optimized designs by addressing:

The methodology will be based in constraining locally the anisotropic perimeter and the stresses, and then solving the optimization problem with the versatile null space method.

33 of 36

Applications

33

Structures with maximum stiffness

Compliant mechanisms

Microstructures with optimized homogenized properties

Finite elements for linear elasticity. Using yield stress as maximum value

(2D and 3D)

Finite elements for linear periodic cell problem

(2D and 3D)

34 of 36

Next steps

34

35 of 36

Publications and participation in conferences

35

Publications:

  • Torres J., Otero F., Ferrer A. Global length and overhang control for level set and density approaches via perimeter minimization. (Under revision)

  • Torres J., Ferrer A., Otero F. Topological derivative in constrained optimization problems via the null space algorithm. (Ongoing)

Participation in conferences:

  • 19th Pegasus Student Conference (Rome, April 2023).

  • The Fourth International Conference on Simulation for Additive Manufacturing (Munich, July 2023).

  • 9th European Congress on Computational Methods in Applied Sciences and Engineering (Lisbon, June 2024).

  • The Congress on Numerical Methods in Engineering 2024 (Aveiro, September 2024).

  • Math SOMMa Junior Meeting 2024 (Barcelona, October 2024).

36 of 36

Thank you for your attention

36

Jose Antonio Torres Lerma