The World Wide Web Consortium’s (W3C) Workshop on the Semantic Web in Oil and Gas* was held in Chevron’s Houston offices with attendance from a good cross-section of industry decision makers and semantic practitioners. Attendees hailed from BP, Oxy, Shell, Total, Halliburton, Energistics and Schlumberger inter alia. On the oil company side, Chevron is the most enthusiastic proponent of the semantic web although the technology still hovers between academic proof of concepts and enterprise deployment.
W3C CEO Steve Bratt traced the evolution of the web from hyperlinked documents (Web 1.0) to ‘one web’ of creators and consumers (Web 2.0) and now, linked data (Web 3.0) a.k.a. the Semantic Web. RDF is the core semantic web standard – using a simple ‘subject, property, value’ triple to describe anything. This turns the web into a ‘big global relational database.’ According to Bratt, the Gartner Group is ‘very positive’ but sees full take up of semantic data by 2017 and by 2027 for a ‘semantic environment.’
The healthcare/life sciences (HCLS) industry is the W3C’s poster child for semantic web take-up. HCLS, like oil and gas, has been plagued by barriers to interoperability including commercial applications, external resources, a lack of APIs and data mismatches. A possible outcome of the workshop would be the inception of an Energy/Oil and Gas interest group along the lines of the HCLS above. In the Q&A, Bratt acknowledged that growth in HCLS has been slow.
Schlumberger’s Bertrand du Castel offered a short history of upstream data initiatives noting the failed effort in the 1990s to agree on a common data model. Du Castel spoke with first hand knowledge, he was head of POSC when it was an $8 million per year operation. In terms of data management standards, ‘nothing came out of it, an ROI of zero!’ More recent attempts to find ways to speak to each other through common exchange formats like WITSML and PRODML have taken root. Today, we have a good model of what exploration is and an understanding of what we are all talking about. Now we need go to the next level and reap the value. Mistakes were made in the 1990s, but now ‘we are getting back on track.’
Chevron’s main semantic effort is the three year Integrated Asset Management (IAM) R&D project that was carried out at the University of Southern California’s CiSoft department. CiSoft’s Ram Soma described IAM as a ‘comprehensive transformational approach’ to integrated oilfield operations that sets out to increase integration, enable ‘what if’ scenarios, create a knowledge base and reduce risk. The ‘non disruptive’ technology’ works across previously non interoperable data silos. IAM is now seeing real world deployment through a technology transfer project, with UK-based Microsoft developer Avenade. IAM’s metacatalog is an OWL triple store. Semantic web technology provides an expressive and rich data model suited for inference and rule based reasoning, it is also vendor-independent. The three level design includes a domain-independent upper ontology. Beneath this are domain-level models of asset elements and finally application and workflow specific ontologies.
Frank Chum presented Chevron’s position paper. Chevron has published ontologies for information integration in oil and gas on the W3C site**. Chevron’s problem areas include the difficulty of ‘semantic reconciliation’ of enterprise metadata, the standardization of information and integration across WITSML, PRODML, ISO15926, PPDM etc. The ‘N° 1 role of the semantic web is data integration across applications.’ Chevron’s Exploratory Pilot seeks to achieve the ‘holy grail’ of enterprise search by linking technical data and documentation. Chevron developed a semantic metadata store for technical data – ‘in a way that brought value and did things that were not possible before.’ The Exploratory Pilot mustered metadata from Unix-based SeisWorks and GoCad projects, building an ontology and RDF data store. RDF proved ‘extremely useful to us in bringing together factoids of unrelated information.’ Chevron found that the semantic web was like a ‘souped-up business intelligence system.’ It does require a lot of effort to ‘corral’ metadata.
In a wide ranging Q&A, several oil companies expressed their hopes for semantic technologies. For Total, the expectation is that the semantic web will help with knowledge transfer to the next generation of oil industry workers and that the semantic web can help locate information that you didn’t know existed. BP stated that it had started down the path of a service oriented architecture, but didn’t seem to be getting huge benefits. BP was attending the workshop to check out the ‘next IT wave.’ For Schlumberger, the move to deepwater and harsh environments requires new technology that goes beyond human capabilities. All of which points to a long term move to automation and technologies to support this. Shell noted the semantic web’s potential in exploration for capturing ‘story telling’ and subtle inference from geological models. Speaking from an engineering perspective, Fluor noted that everyone is now losing money due to the lack of interoperability. ISO 15926 is changing this, enabling interoperability between engineering partners. Semantic techniques will become mainstream if they are used.
Fluor Corp.’s Onno Paap presented a paper on ISO 15926 data modeling with RDF/OWL. Today’s engineers are data mappers, they spend all their time ‘yellow lining’ documents, checking PDF documents against the database. With better data management, ‘one engineer could do the work of three!’ Commenting on the W3C’s preferred semantic modeling building block, Paap said, ‘triples are too limited in data description, you need more than just ‘subject,’ ‘relation,’ ‘object.’’ Hence the ISO 15926 use of ‘templates,’ ‘a pattern for facts.’ Facades are grouped in a ‘confederation of participating facades.’ A laptop with a facade browser sends SPARQL queries to the ‘confederation.’ Fluor uses 10,000 equipment vendors on 800 current projects. Paap’s ‘façade’ concept did not go down well with the W3C’s purists!
Jean François Rainaud, a researcher at the French Petroleum Institute (IFP) showed how the semantic web has been used to enable ‘intelligent’ document search in the context of a C02 storage project. The Energy-Web Ontology Knowledge Hub (E-WOK) process begins with the annotation of the document collection down to the document ‘fragment’ level. The semi-automated tagging process uses natural language processing to extract significant words which are ‘conceptualized’ in a domain ontology. Existing semantic resources have be re-used including Dublin Core, the Geon, the NADM, and GeoSciML. Interaction with a subject matter expert is required to define concepts. OWL is used to describe data modeling of geological formations including formation boundaries, faults and fractures.
Speaking on behalf of ENI, Brooke Aker (Expert Systems) stated that the Italian major has adopted its Cogito semantic platform widely. Cogito offers natural language processing includes morphological and grammatical analysis, used to ‘disambiguate’ words with more than one meaning. ENI’s semantic network contains 350,000 words and 2.8 million relationships. A search for ‘China’s nuclear energy strategy in 2020’ winnowed 25 relevant documents from millions in a collection. The technique is good at capturing ‘weak signals.’
David Norheim described Computas’ Active Knowledge System for Integrated Operations (AKSIO). This was designed to avoid Norwegian operators repeating errors and also to help with the ‘big crew change.’ AKSIO is an ‘active socio-technical system for experience transfer in drilling,’ sponsored by StatoilHydro. AKSIO regards a drill rig crew as a ‘friend of a friend’ (FOAF) network. AKSIO generates ‘experience reports’ with semantics to screen and annotate knowledge and to build a searchable, knowledge base. This can be filtered by discipline, operation, equipment state, etc. The AKSIO drilling ontology (in OWL-DL) was created by subject matter experts (SMEs) and knowledge engineers using ‘question driven query scripting.’ Incident reports are routed to SMEs for screening and annotating with the domain ontology. The result is an increased rate of knowledge reuse, good take up of best practices, avoiding repeating one’s mistakes and process improvement.
This report is an extract from The Data Room’s Technology Watch report from the W3C Semantic Web in Oil and Gas event—more from tw@oilit.com.
© Oil IT Journal - all rights reserved.