undergraduate research - final presentation

16
SLS Simulation Research Summary – Summer 2015 Aevyrie Roessler 1 under the direction of: Prof. Tim A. Osswald 1 Prof. William Aquite 1 1 Polymer Engineering Center University of Wisconsin-Madison

Upload: aevyrie-roessler

Post on 10-Feb-2017

166 views

Category:

Engineering


2 download

TRANSCRIPT

PowerPoint Presentation

SLS Simulation Research Summary Summer 2015Aevyrie Roessler1

under the direction of:Prof. Tim A. Osswald 1Prof. William Aquite1

1 Polymer Engineering CenterUniversity of Wisconsin-Madison

1

Continuing work on the MATLAB simulation created in Jochen Kettemann and Galip Yilmazs Masters Thesis work.

Predicts the deformation in a part produced using Selective Laser Sintering (SLS) 3D printing

Non-uniform cooling (crystallization and shrinkage) causes internal stresses in the part, leading to deformation/warping.Research OverviewIntroduction

2

Research OverviewIntroduction

(Kettemann, 2014)

3

Powder bed held below the melting temperature.Laser selectively heats powder in the shape of the desired parts cross section to melting point.Heated powder is now sintered to its neighbors.A new layer of powder is rolled on top of the previous layer.The part is built up layer by layer in this process.Research OverviewSelective Laser Sintering (SLS)

4

The MATLAB simulation emulates this process with a few key sections:Heat Transfer simulates the laser heating and subsequent heat conduction and cooling.Crystallization uses temperature data from heat transfer to calculate the crystallization in the material.Solid Mechanics uses the crystallization/shrinkage to calculate the internal stresses and deformation.Research OverviewSimulation Overview

5

Documenting and formatting the code:Inconsistent indentationComments in TurkishUncommented variables and functionsResearch OverviewContributions - Documentation

6

Updating the crystallization model:Started by making a video renderer to visualize changes in the simulation.Renders a frame for each time step, showing the percent crystallinity as the part cools.Makes visualization of data easy.Great heuristic for quickly seeing if changes in the code have catastrophic consequencesResearch OverviewContributions - Crystallization

7

Research OverviewContributions - Crystallization

Percent CrystallinityExample from a simple rectangular part with 5 layers:

8

Avrami model did not agree with experimental data taken in Anna Mllers crystallization research for.Used linear interpolation to directly use data to find %crystallinity given any time in the simulationMade so any data (text file) could be imported into the simulation to use for this calculationProblems: Need an equation if work is to be published.Lots of computation needed.

Research OverviewContributions - Crystallization

9

Solution: Fit the data with an equationUsed a Piecewise function so I could have two simple equations instead of one huge polynomial with an x^6.Used video renderer to check that formula had the same results as linear interp. Profiled simulation performance in MATLAB to check that it was also computationally faster. (22%!) Success!

Research OverviewContributions - Crystallization

Integral of % crystallizationTime (s)Formula 1Formula 2

10

Updating simulation to work with complex shapes:Started by making another video renderer to visualize temperature in a cross section of the part.Used to spot errors in heat transfer modifications.Modified laser heating code to heat an L-shape into the powder bedResearch OverviewContributions Complex Geometry

11

Research OverviewContributions Complex Geometry

Results of L-Shaped heating profile:Temperature (C)

12

Research OverviewContributions Complex Geometry

Expected heatingActual heatingProblem: Heating a small section of the layer leads to instant conduction through the width of the layer.

What went wrong?Remaining issues with boundary conditions not matching the new L-shaped heating profile.

13

Research OverviewContributions OverviewAccomplishmentsCode documentation/commentsCrystallization Avrami -> Experimental data -> New formulaVideo renderer (crystallinity)Complex geometryComplex heating profileVideo renderer (temperature)

14

Research OverviewContributions OverviewWhere to go from here?Convert heat transfer to Finite Element instead of Finite Difference more flexible with complex geometryGet it working in 3D right now the simulation is 2D/3D.Optimizations for faster computing move to a compiled language e.g.: Fortran.Need to write an .stl importer for 3D model import. Make everything run in reverse! End goal is to compute the required geometry such that after deformation, it matches the input geometry.

15

William Aquite, PhDProject Advisor

Jochen Kettemann, MSSLS Simulation Author(Mechanics, general code)

Anna Mller, MSCrystallization research dataTim A. Osswald, PhDPEC Faculty

Galip Yilmaz, MSSLS Simulation Author(Heat transfer)Acknowledgements Thank you!

16