Process-Level Modeling and Simulation for HPs Multi Jet Fusion 3D - - PowerPoint PPT Presentation

▶
process level modeling and simulation for hp s multi jet
SMART_READER_LITE
LIVE PREVIEW

Process-Level Modeling and Simulation for HPs Multi Jet Fusion 3D - - PowerPoint PPT Presentation

Jan 23, 2023 563 likes •932 views

Process-Level Modeling and Simulation for HPs Multi Jet Fusion 3D Printing Technology Hokeun Kim , Yan Zhao and Lihua Zhao Pa3DL, HP Labs CPPS 2016 The 1st International Workshop on Cyber-Physical Production Systems April 12, 2016,

slide-1
SLIDE 1

Process-Level Modeling and Simulation for HP’s Multi Jet Fusion 3D Printing Technology

Hokeun Kim, Yan Zhao and Lihua Zhao

CPPS 2016 – The 1st International Workshop on Cyber-Physical Production Systems April 12, 2016, Vienna, Austria

Pa3DL, HP Labs

slide-2
SLIDE 2

CPPS 2016, Vienna, Austria Hokeun Kim, HP Labs April 12, 2016 2

Table of Contents

  • Introduction
  • Motivation
  • Background
  • Modeling and Simulation Techniques
  • Preliminary Results
  • Conclusion
slide-3
SLIDE 3

CPPS 2016, Vienna, Austria Hokeun Kim, HP Labs April 12, 2016 3

Introduction

  • 3D Printing Technology (Additive Manufacturing)

– Expected to revolutionize the way of production

  • Highly customized and complex parts
  • Small scale manufacturing (<1000 units)

http://www.engineering.com/3DPrinting/ 3DPrintingArticles/ArticleID/8283 http://www.3ders.org/articles/20160105-hp- reveals-more-multi-jet-fusion-3d-printer- expected-in-late-2016.html

slide-4
SLIDE 4

CPPS 2016, Vienna, Austria Hokeun Kim, HP Labs April 12, 2016 4

Introduction

Techniques used for 3D Printing

  • Sintering / Fusion

– Process of compacting and forming a solid mass of material – By heat and/or pressure – Example of material: metals, ceramics, plastics

http://www.substech.com/dokuwiki/ doku.php?id=sintering_of_ceramics

slide-5
SLIDE 5

CPPS 2016, Vienna, Austria Hokeun Kim, HP Labs April 12, 2016 5

Introduction

Techniques used for 3D Printing

  • Fused Deposition Modeling (FDM)

– Laying down fused material with ejecting nozzle

https://en.wikipedia.org/wiki/ Fused_deposition_modeling

slide-6
SLIDE 6

CPPS 2016, Vienna, Austria Hokeun Kim, HP Labs April 12, 2016 6

Introduction

Techniques used for 3D Printing

  • Selective Laser Sintering (SLS)

– Heating powder material by focusing laser to shape the object

https://en.wikipedia.org/wiki/ Selective_laser_sintering

slide-7
SLIDE 7

CPPS 2016, Vienna, Austria Hokeun Kim, HP Labs April 12, 2016 7

Introduction

HP's Multi Jet Fusion (MJF) 3D Printing Technology

– Fast and inexpensive technology – Can provide new levels of quality (different colors, strengths,

flexibility, conductivity, etc.)

slide-8
SLIDE 8

CPPS 2016, Vienna, Austria Hokeun Kim, HP Labs April 12, 2016 8

Introduction

HP's Multi Jet Fusion (MJF) 3D Printing Technology

  • Process Details

– Selectively apply fusing/detailing agent that amplifies/reduces fusion effect – Apply energy on the whole area, layer-by-layer production (significantly

faster than point-by-point production with FDM/SLS)

(a)

Material recoat Ap

(b)

recoat Apply fusing agent Ap

(c)

t Apply detailing agent

(d)

agent

(e)

slide-9
SLIDE 9

CPPS 2016, Vienna, Austria Hokeun Kim, HP Labs April 12, 2016 9

Introduction

HP's Multi Jet Fusion (MJF) 3D Printing Technology

  • Video clip for demonstration of MJF 3D Printer (USA Today, Oct, 2014)
slide-10
SLIDE 10

CPPS 2016, Vienna, Austria Hokeun Kim, HP Labs April 12, 2016 10

Motivation

  • HP's Multi Jet Fusion 3D Printer as a Cyber-Physical

Production System (CPPS)

– Printing process, mechanical parts (cyber part) – Build material layer (physical part)

  • Need for modeling & simulation tool

– To provide modeling and simulation tools for prediction of quality

  • f printed part that is determined during 3D printing process

– To give guidance for future materials/processes development and

  • ptimization

– For fundamentally understanding Multi Jet Fusion technology

slide-11
SLIDE 11

CPPS 2016, Vienna, Austria Hokeun Kim, HP Labs April 12, 2016 11

Motivation

  • Current widely used 3D printing simulation technique

– Finite Element Method (FEM)

  • A numerical method to find approximate solutions for partial differential equations

(PDEs) by dividing large problem into small, simpler parts called finite element

– Pros and cons of using FEM for 3D printing simulation

  • +Accurately represent complex geometry
  • +Capture local physical/chemical effects
  • - Very slow (> 2-3 hours) even when simulating a single layer of material on a

small area (1cm2)

  • - Difficult to simulate cyber part (e.g. control of printing process)

– Not proper for process-level simulation for printing a 3D object with

hundreds or thousands layers

We needed a proper tool for process-level simulation that can simulate cyber part as well, and that is much faster to simulate >100 layers in a reasonable simulation time

slide-12
SLIDE 12

CPPS 2016, Vienna, Austria Hokeun Kim, HP Labs April 12, 2016 12

Background

What is Ptolemy II?

  • An open-source software for research on cyber-physical

systems

– Developed at UC Berkeley since1996 (its predecessor, Ptolemy

Classic started in 1990)

– Supports modeling of both the cyber part (computation,

communication) and physical process (continuous dynamics)

– Quite stable, easy to learn and use (supports GUI, one can build a

model by drawing components)

– Based on actor-oriented design – More information on http://ptolemy.org

slide-13
SLIDE 13

CPPS 2016, Vienna, Austria Hokeun Kim, HP Labs April 12, 2016 13

Background

Actor-Oriented Design in Ptolemy II

  • Actors

– Concurrently executed components – Interact with other actors through

input/output ports connected to each

  • ther

– Can model computation,

communication, physical processes, etc.

slide-14
SLIDE 14

CPPS 2016, Vienna, Austria Hokeun Kim, HP Labs April 12, 2016 14

Background

Actor-Oriented Design in Ptolemy II

  • Directors

– Implement Models of Computation

(MoCs)

– Orchestrate behavior of actors, for

example, when each actor should be executed (=fired)

slide-15
SLIDE 15

CPPS 2016, Vienna, Austria Hokeun Kim, HP Labs April 12, 2016 15

Background

Actor-Oriented Design in Ptolemy II

  • Actor hierarchy

– An actor can have sub-actors

(composite actor)

– Atomic actor = non-composite actor – A composite actor can have its own

director (opaque composite actor)

– Actors in a transparent composite

actor are governed by the upper-level director

slide-16
SLIDE 16

CPPS 2016, Vienna, Austria Hokeun Kim, HP Labs April 12, 2016 16

Background

Models of Computation (MoCs) in Ptolemy II

  • Model of Computation

– A set of rules

  • rchestrating behavior of

actors

  • E.g. When to execute actors,

How actors react to inputs

slide-17
SLIDE 17

CPPS 2016, Vienna, Austria Hokeun Kim, HP Labs April 12, 2016 17

Background

Models of Computation (MoCs) in Ptolemy II

  • Finite State Machines

and Modal Models

– States and state

transitions are used to describe behavior

– Each state can represent

different modes of

  • peration (modal models)
slide-18
SLIDE 18

CPPS 2016, Vienna, Austria Hokeun Kim, HP Labs April 12, 2016 18

Background

Models of Computation (MoCs) in Ptolemy II

  • Discrete Events (DE)

– Time-stamped events

(e.g. timer event, arrival

  • f messages) trigger

execution of actors

– Good for modeling

computation and communication

slide-19
SLIDE 19

CPPS 2016, Vienna, Austria Hokeun Kim, HP Labs April 12, 2016 19

Background

Models of Computation (MoCs) in Ptolemy II

  • Continuous Time

– Continuous behavior of actors is

simulated by sampling and advancing time steps

– Includes ODE solvers for

physical processes modeled in ODEs (similar to Mathworks Simulink)

– Proper for modeling physical

processes (e.g. temperature, thermal transfer)

slide-20
SLIDE 20

CPPS 2016, Vienna, Austria Hokeun Kim, HP Labs April 12, 2016 20

Modeling and Simulation Techniques

Inputs and Outputs of Ptolemy II Model

Configuration parameters for printer control & processes

Config File

Parameters for physical environment & material characteristics

Env File

Image Information

  • f each layer

3D Image File

Simulated Values Time

Surface Layer Physical Characteristic 1

Ptolemy II Model

Physical Characteristic 2 of each layer Layer Processing Time

slide-21
SLIDE 21

CPPS 2016, Vienna, Austria Hokeun Kim, HP Labs April 12, 2016 21

Modeling and Simulation Techniques

CPPS Model Top-Level View

3D Printing System Printer

Printing Controller & Process Modules (Cyber Part)

Layer

Multiple Layers of Build Material (Physical Part) Actions Sensor Readings

Inputs Outputs

Actors in Ptolemy II Message flows between actors

slide-22
SLIDE 22

CPPS 2016, Vienna, Austria Hokeun Kim, HP Labs April 12, 2016 22

Modeling and Simulation Techniques

Cyber Part of CPPS Model

  • Controller

– Sends commands to operate process modules

  • Process modules

– Take actions on build material, and sense physical characteristics of the

surface of build material

Printer Model

Process Modules Printing Controller

Commands Signals

Finite State Machine Preheating Module Fusing/Detailing Agent Jetting Module Fusing Module Material Recoating Module

Actions Sensor Readings

slide-23
SLIDE 23

CPPS 2016, Vienna, Austria Hokeun Kim, HP Labs April 12, 2016 23

Modeling and Simulation Techniques

Example of Printing Process Modeling

  • Fusing Process Model with a Finite State Machine (FSM)

Build Material Fusing Agent Applied

Fusing effect ends Final position Fusing effect begins

Fuse (Outputs action) Moving After Fuse

Fusing Source

Initial position

Idle Moving to Fuse

slide-24
SLIDE 24

CPPS 2016, Vienna, Austria Hokeun Kim, HP Labs April 12, 2016 24

Modeling and Simulation Techniques

Basic Ideas for Modeling Physical part

  • Unlike FEM, We use approximation to simulate physical

characteristics of build material for each layer and each area

  • Each layer/area is modeled as a single actor
  • However, even modeling each layer, if layer grows to 1,000

layers, we will need 1,000 actors, leading to too much

  • verhead for process-level simulation?
  • How can we deal with additive layers efficiently?
slide-25
SLIDE 25

CPPS 2016, Vienna, Austria Hokeun Kim, HP Labs April 12, 2016 25

Modeling and Simulation Techniques

Approximating Physical Part of CPPS Model

  • Dividing build material layers

into three categories

– Surface layer (currently printed) – Internal layers (printed previously) – Bottom layers

Bottom layers Internal layers Surface layer Part area Support area

  • Dividing each layer into two

areas

– Part area (to be fused) – Support area (remains unfused)

slide-26
SLIDE 26

CPPS 2016, Vienna, Austria Hokeun Kim, HP Labs April 12, 2016 26

Modeling and Simulation Techniques

Factors Affecting Surface Layer Temperature

Internal layer Surface layer ④ Heat transfer between layers ③ Heat lost to ambient ① Preheating/fusing source ② Agent jetting device ⑤ Material recoat

slide-27
SLIDE 27

CPPS 2016, Vienna, Austria Hokeun Kim, HP Labs April 12, 2016 27

Modeling and Simulation Techniques

Modeling Additive Layers with Fixed Number of Actors

Before material recoat

Layer info transferred

New Internal Layer 1

Layer info transferred

New Internal Layer 2 New Surface Layer

New layer info

New Bottom layers

Layer info aggregated

After material recoat

Surface layer Internal layer 1 Internal layer 2 Bottom layers

  • Information aggregation example for temperature

!

"#$%&''&( = *%&''&(×! %&''&( + !

  • .'#/.012

*%&''&( + 1

slide-28
SLIDE 28

CPPS 2016, Vienna, Austria Hokeun Kim, HP Labs April 12, 2016 28

Modeling and Simulation Techniques

Physical Part of CPPS Model

Layer Model

Actions Sensor Readings

Internal Layer 1 Internal Layer 1 Physical Characteristics

Part Area Physical Characteristics Support Area Physical Characteristics

Heat Transfer Internal Layer 2 Internal Layer 2 Physical Characteristics

Part Area Physical Characteristics Support Area Physical Characteristics

Bottom Layers Bottom Layers Physical Characteristics

Part Area Physical Characteristics Support Area Physical Characteristics

Surface Layer Surface Layer Physical Characteristics

Part Area Physical Characteristics Support Area Physical Characteristics

slide-29
SLIDE 29

CPPS 2016, Vienna, Austria Hokeun Kim, HP Labs April 12, 2016 29

Modeling and Simulation Techniques

Complete CPPS Model

Printer Model

Process Modules Printing Controller

Commands Signals

Finite State Machine Preheating Module Fusing/Detailing Agent Jetting Module Fusing Module Material Recoating Module

Layer Model

Actions Sensor Readings

Internal Layer 1 Internal Layer 1 Physical Characteristics

Part Area Physical Characteristics Support Area Physical Characteristics

Heat Transfer Internal Layer 2 Internal Layer 2 Physical Characteristics

Part Area Physical Characteristics Support Area Physical Characteristics

Bottom Layers Bottom Layers Physical Characteristics

Part Area Physical Characteristics Support Area Physical Characteristics

Surface Layer Surface Layer Physical Characteristics

Part Area Physical Characteristics Support Area Physical Characteristics

Each box represents a (composite) actor in Ptolemy II, possibly with its own director (MoC, e.g. Discrete-Event, Continuous Time,

  • r FSM)
slide-30
SLIDE 30

CPPS 2016, Vienna, Austria Hokeun Kim, HP Labs April 12, 2016 30

Preliminary Results

Simulated Values vs Experimented Values

  • Reasonable accuracy for each area

Simulated Values Time

(a) Simulation results (b) Experimental results

vDetails are excluded for HP's confidential information

Time Experimental Values

Area 1 Area 2

slide-31
SLIDE 31

CPPS 2016, Vienna, Austria Hokeun Kim, HP Labs April 12, 2016 31

Preliminary Results

Simulation Time – Proposed vs Finite Element Method

Simulation Platform (Workstation) two Intel Xeon E5 @2.60 GHz (6 cores each, total 12 cores) and 64 GB RAM Simulation time for one layer of 1cm × 1cm area (reduced scale) 127 minutes Expected simulation time for one layer of 10cm × 10cm area (normal scale) 100 × 100 × 127 minutes = 7.62 × 107 seconds

  • Material recoat process simulation using finite element method (FEM)
  • Simulation time of FEM is at least quadratic to the

number of particles (≈ area)

  • 100 times area è 100 × 100 times simulation time

v Visualization of FEM simulation for material recoat process

slide-32
SLIDE 32

CPPS 2016, Vienna, Austria Hokeun Kim, HP Labs April 12, 2016 32

Preliminary Results

Simulation Time – Proposed vs Finite Element Method

  • Proposed approach – simulation time for process-level simulation for

all processes including material recoat

Simulation Platform (Laptop, HP Z-book) Intel Core i7 2.8 GHz (4 cores) and 16 GB RAM Simulation time for 100 layers for all processes 592 seconds Simulation time for 1 layer for all processes 5.92 seconds

  • FEM for material recoat process: 7.62 × 107 seconds / layer
  • Proposed approach for all processes: 5.92 seconds / layer

Indication: proposed approach is faster than FEM by approximately 7 orders of magnitude!

slide-33
SLIDE 33

CPPS 2016, Vienna, Austria Hokeun Kim, HP Labs April 12, 2016 33

Conclusion

  • Proposed modeling and simulation techniques of HP’s Multi Jet

Fusion 3D printing technology as a CPPS using Ptolemy II

  • Significantly faster speed than FEM, with reasonable accuracy

– By approximation of layers of build material – Information aggregation for additive layers

  • Flexible design in configuration, can be easily extended and

improved

  • Future Work

– Supporting more complex geometry (Currently we assume printed shapes are

identical for all layers)

– Improving accuracy with equations extracted from experimental data

slide-34
SLIDE 34

CPPS 2016, Vienna, Austria Hokeun Kim, HP Labs April 12, 2016 34

Thank you!

  • Q&A
  • Contact Info

– hokeunkim@eecs.berkeley.edu

  • Acknowledgement

– Prof. Edward A. Lee and Ptolemy Project group at UC Berkeley