Ontological Foundations For Sys ML Henson Graves September

Ontological Foundations For Sys. ML Henson Graves September 2010 1

INCOSE MBSE Roadmap MBSE Capability Reduced cycle times System of systems interoperability Design optimization across broad trade space Cross domain effects based analysis June 15, 2008 Well Defined MBSE Maturity Institutionalized MBSE across Academia/Industry This talk fits here Distributed & secure model repositories crossing multiple domains Defined MBSE theory, ontology, and formalisms Architecture model integrated with Simulation, Analysis, and Visualization Matured MBSE methods and metrics, Integrated System/HW/SW models Ad Hoc MBSE Document Centric Emerging MBSE standards 2010 Refer to activities in the following areas: • Planning & Support • Research • Standards Development • Processes, Practices, & Methods • Tools & Technology Enhancements • Outreach, Training & Education 2020 2025 INCOSE IW 09 MBSE Workshop 2

Outline § Why is an MBSE reasoning formalism so important § Lessons from applying OWL to engineering applications § Steps toward integrating OWL reasoning with Sys. ML 3

Reasoning Is Required Multiple Places In The Systems Engineering Process § Are requirements consistent § Are implementations feasible § Is design sufficiently detailed for implementation § Can an implementation satisfy design requirements § Do proposed modifications stay within design constraints Deployment Requirements Test & Verification Design Develop requirements specifications Implementation Develop design specifications Check specification consistency Check integration design consistency Perform integration tests Verify that implementatio n realizes specifications Perform verification tests Verify product satisfies requirements Produced by Engineering Tools Produced by Reasoning Tools … Long history of attempting to use formal methods for engineering, with mixed success, often too hard to use, doesn’t scale 4

Size And Complexity Put Bounds On Manual Analysis Design WBS Partition 150 Build Architecture Decomposition Design for Implementation Support As Built Design As Maintained Design 15000 1500 2 Rush Time 3 Last Skier 4 Check Line 5 Impatient Skier 6 Line 7 Lift 1 8 Lift 2 9 Ski 10 Go Home Many enterprises use modeling extensively but the result is an enormous collection of non-integrated models - Situation is worse than in document centric development 5

Where Does One Look Formal Logical Foundation For Modeling Systems: OWL § Designed for conceptual modeling § represents more than 20 years of research § Extensive experience model complex physically structured systems in the life sciences and medicine § Logic based modeling language § Optimized reasoning algorithms § designed so as to be decidable (arbitrary queries can be answered) § W 3 C language standard with tool support § Designed so as to allow for extensibility … its clear that OWL and UML/Sys. ML have significant overlap 6

Ian Horrocks Helped Me Develop An OWL Air System Ontology in Protégé To Answer: n Can OWL provide semantic foundation and integration for MBSE l Could OWL work where other approaches have failed? l Can ontologies capture meaning of concepts independent of interpretation by subject matter experts? l Can automated reasoning be used to check design properties such as consistency and conformance with specification? Paper in OWL Experiences and Directions 2008 7

The OWL Experience § One can build OWL ontologies to represent static structure of systems § Used reasoning to verify design consistency with requirements and other questions § We made extensive use of an Upper Ontology § Some requirements not (easily) expressible OWL § Weight of product is sum of weights of components § Behavioral requirements § Representing a “detailed design” in OWL is difficult § (where all valid implementations have the same parts structure and connection relationships) The experience led me to focus on how to represent detailed designs, first in OWL and its extensions, and then in Sys. ML 8

Which Models (Ontologies) Have The Property That All Valid Implementations Have The Same Structure? Model Rule out implementations with § two fingers sharing distal phalange § Hands with 500 fingers § Non-connected fingers as part of hand Then can argue from a specific implementation to any implementation 9

How Can a Model Be Characterized So All Implementations Have The Same Structure? A Water Model Used To § Generate 3 D Visualization of implementation § Answer questions about mass, size, geometrical shape, … Develop examples and generalize 10

Yvonne Bijan and I Have a Sys. ML Model And a Proof That All Implementations Of Are The Same § This is a prototypical design analysis/verification problem § If all implementations are the same you can calculate or measure weight of individual molecule § Another version of problem is when is a design sufficiently complete that it can be implemented The two diagrams which are part of the water model show both the parts structure and the bond connection … we are building the models directly in Sys. ML but have to go outside Sys. ML for reasoning 11

We Had To Add Some Additional Axioms For Water That May Be Implicit In Sys. ML Issues arising in the proof § Does any valid implementation of Water have exactly three atoms Language concepts and constructs match modeling domain § Maybe it is implied by the Sys. ML spec, but we had to add it § The covalent bond is a Sys. ML connection between parts § We added the equation has. Hydrogen. Atom. covalent. Bond = has. Oxygen. Atom § We also had to make assumptions about restrictions of properties … we have (we think) a general concept of structural template that can be validated by Sys. ML tools 12

Sys. ML Has Language Constructions Not In OWL § § Variables & operations Constraints Behavior Role properties, e. g. , part properties and other component properties OWL reasoning can be made to work where languages overlay, but the reasoning requires extension for full Sys. ML Constraints used to generate 3 D visualization 13

Conclusions: Providing Sys. ML With A Logical Foundation is feasible and it … § Enables engineers to work in a good user friendly language integrated with valid reasoning tools § Engineers are able to employ benefits of logic without having to learn special logical language syntax § Provides better integration with simulation § Provides a check on expressiveness and coherence § Provides potential language candidate extensions for Sys. ML and OWL 2 There is a strong case that OWL and Sys. ML can be unified with benefits to both 14

Next Steps § Develop Sys. ML use cases for inference § Develop rules to translate Sys. ML to an extended OWL 2 § Export Sys. ML to reasoner and reimport results § Develop template validation code for Sys. ML tools § Verify Sys. ML logic retrofit is computationally tractable 15