Standardizing an Ontology of Physics

1
Standardizing anOntology of Physics

• Joseph B. Collins
• Naval Research Laboratory

2
An Ontology of Physics (OoP)

• Meta-Physics, not metaphysics!
• The language we must use to describe physics
• Applied physics, not physics research
• Standardized concepts, constructs for
  standardizing physical descriptions
• Applicable to all physics-based models
• Environmental models are physics-based

3
Why a Standard Ontology?

• No black boxes anymore
• Documentation, transparency
• Enables informed use of models
• Automated understanding
• Enables search, retrieval of models
• What http did for the web, OoP can do for
  models
• Composability
• Facilitate meaningful model composition

4
SEDRIS

• Provides Spatial Reference Model (SRM) for
  spatial relationships
• Provides Environmental Data Coding Specification
  (EDCS) for concepts
• Many physical concepts here!
• Provides Data Representation Model (DRM) for
  (static) discretized model transmittals
• No math (derivatives, integrals)
• No dynamics ( like F ma )

5
Semantic RelationshipsOntologies Taxonomies

• Taxonomy implies hierarchy (tree)
• Ontology is a web (graph) of relationships
• Semantic relationships MAY form a taxonomy in
  special cases
• Graphs (webs) can represent any relationship
  topology
• They teach concept webbing in Elementary
  School! (Inspiration)

Trees
Graph / Web
6
A Physics-Based Model Ontology Layercake

• The referent
• Language of physics
• The easier solution
• Lets get an answer!
• The next guys grid
• Each layer to layer transition is informal,
  one-to-many

Physical concepts
Mathematical Expr.
Mathematical approx
Discretized approx
Interpolate
Can we infer the Physical Concept from the last
layer? No!
7
Composing Physical Objects
Electron

• Physical objects combine - composition
• Composition is not simple plugging together
• Division between Internal and External forces and
  motion

Proton
Hydrogen
Photon
8
Why is Measurement Important?

• The Real World
• The ultimate source of input to simulation
• The thing referred to
• Sensor Model
• Observation and detection (analog / digital)
• Measured Parameters / uncertainty, error
• Translation to other model parameters
• Data normalization, inference, transformations

9
Example MODAS

• Oceanographic Model
• Highly appreciated by the ASW community
• A Dynamic Climatology (whats that?)
• Measurement Input, Model Type, Output Variables
• Measures (daily) sea surface temperature,
  elevation, historical temp, salinity profiles
• Empirical (Regression) Model
• Output Ocean temperature, salinity profiles

10
Ocean, Atmosphere

• Navier-Stokes basis (mathematical equations)
• Derived from conservation laws for fluid continua
• Approximations
• Inviscid, compressible / incompressible
• Turbulence model, eqn. of state
• Named discretized models
• Princeton Ocean Model, NCOM, NLOM, ECOM, MODAS,
  NOGAPS, COAMPS
• Data Sources
• measurement data, data drivers (BCs)
• Similarly characterized (high-level to low-level)

11
Marking Up

• Physics Concept an applied force causes a mass
  to accelerate - (marking up a math statement)
• Math Concept Equation - A three dimensional real
  valued vector, F, is equal to the scalar, m,
  multiplied by the three dimensional real valued
  vector, a.
• Numerical (discrete math) Concept

ltforce unitnewtongt F lt/forcegt ltPhysObj.mass
unitkilogmgt m lt/PhysObj.massgt ltPhysObj.accelera
tiongt a lt/PhysObj.accelerationgt
12
The IdealModel
Physics-based description
physics gtltphysics
Mathematical description

SEDRIS description
model.xml gt
Production code version
Validation data
Config. Mngmnt. data
13
Benefits

• Open access to models by non-experts
• Search and retrieval
• Begin automation of documentation
• Decision aids for simulation builders
• Anticipate environmental effects in sim
• Animation engines and dead-reckoning
• Approach to true composability
• Plug and play (not quite)

14
Tools

• Need to use Next Generation Language support
• There are tools that support math formalism
• Mathematica, Maple, MuPad
• Look for support for objects discrete methods
• OpenMath, MathML - Mathematical Markup Language
• PyCES and PhysicsML