TOWARDS COGNITIVE COMPUTER AIDED...
Transcript of TOWARDS COGNITIVE COMPUTER AIDED...
June 7th-9th | Seattle, WA2016 NAFEMS Americas Conferencenafems.org/americas
TOWARDS COGNITIVE
COMPUTER AIDED ENGINEERING
Stefan Suwelack1, Markus Stoll2, Annika Meyer1, Steffen Slavetinsky1, Manuel Serf1, Nikola Bursac1, Albert Albers1, Rolf Bendl2, Rüdiger Dillmann1 and Stefanie Speidel1
1Karlsruhe Institute of Technology2German Cancer Research Center
1
June 7th-9th | Seattle, WA2016 NAFEMS Americas Conferencenafems.org/americas
2
June 7th-9th | Seattle, WA2016 NAFEMS Americas Conferencenafems.org/americas
3
Surgeon
Cognitive
Surgical
Assistant
June 7th-9th | Seattle, WA2016 NAFEMS Americas Conferencenafems.org/americas
Knowledge Base
Factual Knowledge
Experience
Semantics
Data
describes
June 7th-9th | Seattle, WA2016 NAFEMS Americas Conferencenafems.org/americas
Cognitive Computing
• Deep Blue 1997
• DARPA Grand Challenge
2007
• Watson 2011
• AlphaGo 2016
5
June 7th-9th | Seattle, WA2016 NAFEMS Americas Conferencenafems.org/americas
Cognitive Computing: Deep Learning
6
“Two pizzas sitting on top of a stove top oven”
June 7th-9th | Seattle, WA2016 NAFEMS Americas Conferencenafems.org/americas
Vision: Cognitive Simulation Assistant
7
June 7th-9th | Seattle, WA2016 NAFEMS Americas Conferencenafems.org/americas
Challenges in CAE
8
License Cost
SaaS
Hardware Cost CAE-Experience
CognitiveComputing
June 7th-9th | Seattle, WA2016 NAFEMS Americas Conferencenafems.org/americas
CAD und CAE Today
9
Design-
engineer
Numerical
analyst
June 7th-9th | Seattle, WA2016 NAFEMS Americas Conferencenafems.org/americas
Cognitive CAE
10
Design-
engineer
Cognitive
Simulation
Assistant
Numerical
analyst
June 7th-9th | Seattle, WA2016 NAFEMS Americas Conferencenafems.org/americas
Simulation Apps I
11
June 7th-9th | Seattle, WA2016 NAFEMS Americas Conferencenafems.org/americas
Simulation Apps II
12
June 7th-9th | Seattle, WA2016 NAFEMS Americas Conferencenafems.org/americas
Generalized Cognitive Simulation App
13
June 7th-9th | Seattle, WA2016 NAFEMS Americas Conferencenafems.org/americas
Model Factual CAE Knowledge
• Philosophie: Study basic categories of being
• Computer science: Knowledge representation
• Compatibility through shared
Upper Ontologies
• Well suited for
data integration
14
June 7th-9th | Seattle, WA2016 NAFEMS Americas Conferencenafems.org/americas
Data ontology
What is represented by the
data?
• Information content entity
How is the data represented?
• Information structure entity
15
June 7th-9th | Seattle, WA2016 NAFEMS Americas Conferencenafems.org/americas
Algorithm ontology
• Draws on existing standards,
e.g. OpenAPI model
• Based on data ontology
• Integrates different software
modules and services
• Alternative representation
by YAML/JSON
16
June 7th-9th | Seattle, WA2016 NAFEMS Americas Conferencenafems.org/americas
Simulation ontology
• Representation of simulationknowledge
• Models, boundaryconditions, solvers…
• Based on existing work, e.g. HuPSON ontology, Step AP 209
• Workflow representationand simulationautomatization
17
June 7th-9th | Seattle, WA2016 NAFEMS Americas Conferencenafems.org/americas
Current work: Learn from experience
• Semantic database ofsimulation data
• Use machine learning toextract knowledge
• Deep learning for shapeunderstanding
• Enables: BC transfer, automate defeaturing, meshing etc.
18
Danglade et al.
June 7th-9th | Seattle, WA2016 NAFEMS Americas Conferencenafems.org/americas
19
Life is cheap.
It's the accessories that kill you!
June 7th-9th | Seattle, WA2016 NAFEMS Americas Conferencenafems.org/americas
Computer Aided Engineering modeling
language
• Semantics-based CAE middleware
• Interfaces, converters and runtime for CAE
modules
• Simulation data management
• Open source license (autumn 2016)
• Vision: Help to create OSS stack for CAE
20
CAEML
June 7th-9th | Seattle, WA2016 NAFEMS Americas Conferencenafems.org/americas
Computer Aided Engineering Modeling
Language
21
Simulation Ontology
YAML
CAEML Runtime Python API
Python Data API
CAEML
June 7th-9th | Seattle, WA2016 NAFEMS Americas Conferencenafems.org/americas
CAEML core features
• Framework for data conversion
– Data is represented as acyclic directed graph
– Format conversion is graph matching
– Easy creation of new convertes
• Definition of simulation workflows
• Runtime for simulation workflows
• Simulation data management
22
June 7th-9th | Seattle, WA2016 NAFEMS Americas Conferencenafems.org/americas
Applications
• Intuitive visual programming of simulation
workflows
• Cloud-based simulation services
• Cognitive Computer Aided Engineering
23
June 7th-9th | Seattle, WA2016 NAFEMS Americas Conferencenafems.org/americas
Towards Cognitive CAE
24
Cognitive
Simulation Apps
CAEML
middleware
Lifelong
learning,
data analytics