Presentation on theme: "Ontology-based Knowledge Management in the Steel Industry"— Presentation transcript:
1 Ontology-based Knowledge Management in the Steel Industry Chapter 11B. Ramamurthy
2 IntroductionAn important aspect for businesses is knowledge and intelligence generation and management.Right knowledge and intelligence is important for right and timely decisions.We will discuss the approach used by steel industry to address knowledge and intelligence management.
3 Steel Industry Context Arcelor Mittal: world’s number one steel company330,000 employees60 countriesGeographical diversity: Industrial activities in 27 countries across Europe, Americas, Asia and Africa.Arcelor Research Knowledge Innovation (KiN) Center aims to classify, model and put into service the knowledge of this group.Knowledge-intensive tasks steer business processes (how?)Business processes are realized using services (WS) in the implementation (how?)
4 Critical Business areas Business optimizations: supply chain, sales, purchasing, marketingCustomer solutions based on knowledge (ex: American relationship with Cuba has been improving… steer business to pay attention to customer needs in this region).Industrial process support: Factory-wide, line piloting, process modelsCross-cutting service assistance (transversal service assistance) (ex: services spanning multiple domains)
5 Solution basis Data mining Knowledge-based systems Simulations of optimization techniquesSemantic webArcelorMittal collaborates with CTIC Foundation (Center for the Development of Information and Communication Technologies) for semantic web related activities.Together they provide steel industry standard for W3C semantic web activity
6 Motivation and Use Cases Knowledge capitalization toolsUnified data description layerSupply chain management: raw materials to finished productsOntologies are not new: used for knowledge representationOntologies will be used here to integrate:
7 Ontologies for integration Structural clarity : hierarchical structure vs. RDBMSHuman understandingMaintainabilityReasonability: infer new knowledgeFlexibilityInteroperability (OWL suite)In summary, ontology is a powerful tool for knowledge management, information retrieval and extraction, and information exchange in agent-based as well as in interactive systems.
8 Knowledge Capitalization Group of applications devoted to manage content, documents, and information, structured so that users can access knowledge, add and modify them.Content management systems, document management systems, wikis, dynamic web portals, search engines, etc.What is required?Ontologies and tools to exploit themtools: semantic search, human resources networking and management
9 Knowledge capitalization: human resources and networking Human resources in multinational companyDepartments need to exchange professional information: contacts, employee profiles, etc.Typically reside in department’s hard driveHRMS: Human Resource Management System: to describe people, job requirements & qualifications.Extensive Ontologies and taxonomies are available:HierarchyE-recruitmentExperts Assignment
10 Unified data description layer Huge company built from many smaller companies incrementallyAll kinds of software + widely varying levels of usagesXML has emerged as a syntactical solution for inter-application data communication
11 XML can do’s and not Promotes reuse (XML parsers) XML instances can be checked for syntactical correctness against grammar (XML Schema)Can be queried (XQuery, XPath)Can be transformed (XSL)Can be wrapped using commodity protocols (web services)However they convey only structure… they are meaningless (no semantics)Ontologies have the potential to fix this situation by providing precise machine-readable semantic descriptions of the data.
12 Adding Semantics to content How to do it?Managing legacy DB:Choice 1: transform into relational db to ontology collections (R2O) √Choice 2: Wrap relational databases with semantic interfacesSteel producers use models and simulation tools to predict or control impact of various events: semantics can help in re-use of many existing models across departments, countries and organizations.Distributed searches: can index multiple repositories, esp. in multilingual environments
13 Supply Chain Management (SCM) Supply chain is a coordinated system of organizations, people, processes, and resources involved in moving a product or service from suppliers to customers.In AM (ArcelorMittal) is indeed quite complexIndependent business unitsMitigate delays in production processVariances in production times and product qualityManaging orders and sub-ordersHeterogeneous processesSupply chain modeling and simulationHighly dynamicMost data reside in heterogeneous systemsIslands of automationNeed to form a global model
14 SCM Solution at AMOntology engineering to support supply chain modelingIdentify data and knowledge required for specific modelDevelop mechanisms to extract the above informationPopulate Ontologies with required knowledgeBuild simulation models and implant a generic procedure to fill the necessary input values
15 A Business process Abstraction AM will use Supply Chain Operation Reference (SCOR) model developed by supply chain council.Ontology will be developed based on SCOR.SCOR is structured around five processes: Plan, Source, Make, Deliver and ReturnAll these can be semantic (composite) web services in the modelProcesses are decomposable
16 Ontology for Business processes Ontology will address categories of the supply knowledge:Process: process cost, process qualityResource: capacity of resourceInventory: control policyOrder: demand or order quantity, due datesPlanning: forecast methods, order scheduleDevelop supply chain ontology: help simulations and future system designs.
17 Modeled Factory and Metallurgical Routes Application of ontology design and semantic web.A metallurgical route involves set of processes (realized using web services) from order to production.How can it help? What was the situation before introduction of semantics?Lack of modularityLack of standardsLack of integration between business models and production rulesSolution: formal description of the concepts that occur in metallurgical routes.All concepts are formalized as ontology classes.These concepts or blueprints have to be agreed upon by different plants.This framework represents a common understanding of the products and production lines.
18 Semantic Metallurgical route: HotRollingMill Maximum/minimum entrance widthMaximum/minimum exit widthProductivityThickness reduction capacityInput material is of type SlabOutput material is of type HotRollAdding semantic enabled each facility to add values to a semantic instance of the concept.Web services could query the facilities before processing orders (p.255): that is HotRollingMill will be available via a web service to the applications that need its information details.Ontology is centrally developed, and instances are kept at decentralized locations and served by WS.More intelligence is embedded in WS through addition of semantic to data… results in less number of rules.Here is an example of services-enabled enterprise (AM).
19 AM, The Ultimate Service-enabled Enterprise Semantic search: Ontologies, metadata, thesauri and taxonomies (ARIADNE project)H.R. and networking: Ontologies, international classifications and rulesUnified data description layer: Ontologies and data mediationExpert knowledge and industry process modeling: Ontologies and rulesSupply chain management: Ontologies, SCOR model, semantic web services, rulesModeled factory: Ontologies and rules (metallurgical routes, Visonto)
20 Practical Experiences Ontologies are powerful mechanisms to capture knowledge.Knowledge is key factor in productivity.Sharing knowledge among employees perform similar tasksOverall productivity can be improved by transfer of knowledge from experienced employees to inexperienced ones.This is needed for spanning the gap in multilingual world, to improve understanding and productivity and to avoid industrial accidents and to provide best practices.
21 Expert Knowledge and Industrial Process Modeling Metal working and factory modeling: how to manage bottlenecks, solve inventory, and work in progress problems like line stoppages, and material defects, optimize production rates, determine plant capacity etc.Solution: build a shared ontological abstraction of metallurgical concepts and to use it as an interoperable framework in production lines and product life cycle management.An ontology that focuses on process, equipments, problematic and best practices of continuous annealing line has been built.Different models are developed at different production lines which share many concepts; there is need for reuse and interoperability.Solution: ontology based services-enabled framework
22 Generic Production Line (p.2527-258) ProcessPerforms/Performed byIs composed of/is component ofLineToolEquipmentSupplies/Supplied byProducts
23 Enhancing Ontology Reuse and Interoperability Ontology language: (OWL-Full, OWL-DL, OWL-Lite)OWL-DL (Description Language) was chosen for its expressiveness and for its support of computational completeness and decidability.Common semantics: need to share same vocabulary and points of view.Meta-modeling: multi-layering of concepts: highest level described more general concepts and the lowest specific for each line; intermediate layers describe common processes and equipment and tools.
25 Usage of OntologiesUsed for streamlining industrial equipment to perform steel fabricationAlso help staff to maintain devices, control of processes, test product quality and other operations involving human intervention.RDF model allows information (from experts) as web resources.OWL has a annotation feature to add metadata information to any resource of an ontology.Ex: rdfs: comment, rdfs: seeAlsoAlso applying a social network enhances the utility of the factory ontology.Experts share the same model of the whole process and they can interchange information and documents by means of the ontology.
26 Visonto: A tool for ontology visualization Ontology authoring: protégé?No, they developed their own in collaboration with CTIC foundation.Can be customized within the ontology.View: tree view heavily linked to web pages for knowledge disseminationMultilinguism is a key feature: language-agnostic for domain knowledge with annotation in multiple languages, other subtle details such as units of measurement, monitory units and dates/time etc.Simple string-search based search; query-based search based on SPRQL.Query by example interface: a good choiceFilter of information through points of view and other filters.
27 Visonto ArchitectureVisonto is a web application, without any substantial software installed by the client.Knowledge sharing and collaborative environment. A common pool of Ontologies and comments.Long term plan involves adding reasoners, semantic web services.
29 ARIADNE: Enrichment of syntactic search Another internal projectVerity/autonomy K2 productIndexing spider gathers and builds repositories of all internal documentsJ2EE web user interface was built on top of the search engine API.Result is a powerful capitalization of company information.Web interface in Java and Jena framework.Search comparison in multiple languages.
30 Open Issues Development of large ontologies Semantic web services Combining ontologies and rulesDevelopment of more tools for leveraging knowledge base