SPDM* Online 2021 heard from Sun Maria Lehmann and Jørn Ølmheim on Equinor’s data mesh journey that started in 2016. The data mesh philosophy (see also our report on BP’s data mesh) revolves around the concepts of design thinking and providing the best possible data products for different users that focus on use cases. Lehmann compared this to making pancakes to order. A federated domain-driven model targets different user personae with brainstorming and scrum/agile rapid iterations to establish data product principles. Data products involve more than data, including also code for ingestion, transformation and update. The mesh consists of several layers. The approach allows for different technologies for different parts of the business. A minimal governance layer assures interoperability with connected data products constituting the mesh, ‘there is a lot of value in existing models’. In the Q&A the mesh approach was contrasted with OSDU’s ‘all data in one place’ approach. For Equinor, OSDU’s data footprint is currently limited and in any case, ‘you will never have everything in OSDU, Industry 4 (process), financials and so on’. We want to see how data in OSDU fits with data products not in OSDU. Asked on the state of deployment, Lehmann admitted that the mesh today is not widely deployed, Equinor is working on a data platform for production data and is ‘very much at the beginning’.
Michael van der Haven (CGI) sees the data mesh as way of easing pain points in data migration. Van der Haven compared the OSDU promise with reality. OSDU opponents cite vendor support and ‘expensive’ cloud deployment. Van der Haven envisages OSDU as a spider in the web of the data mesh. This differs from the usual perception of OSDU. Not all data in is necessarily in the data lake. It may be on prem or in the cloud, and not necessarily in an open format. The market is slowly but surely moving to provide the connectors that enable a data mesh. Pulling metadata into OSDU is a good idea for data browsing.
Gunnar Staff described how Cognite is ‘rethinking data and digital in oil and gas’, asking ‘why are we still struggling?’. Industry is drowning in data but ‘consumer tech is leading the way forward’. Staff cited Google map as a typical ‘disrupter’. In the field, oils are ‘still struggling with paper-based solutions and manual input!’ Enter the ‘unified semantic data model’ that combines IT and OT data with a focus on the digital twin. However, few digital twin proofs of concept have scaled successfully. The answer here is (commercial plug) a Cognite Data Fusion-enabled pipeline into data science. In the Q&A Staff was asked if CDF was a competitor to OSDU. He claimed not, ‘OSDU is not an implementation, you don’t buy OSDU. CDF has promised to be OSDU compliant – it is not a competitor.’
Robert Swiergon from the UK Oil & Gas Authority (now rebranded as the North Sea Transition Authority) presented the new Cessation of production (COP) initiative. COP provides operators with a standard reporting template user guide that allows them to check for data inconsistencies and reporting lacunae. Even today, data captured is inadequate. The OGA is now playing catch-up, requesting data from operators to fix multiple issues in reported production. The result is (will be?) a cleaned-up and accurate dashboard showing daily production of UK fields and wellbores. OGA is also working to embrace carbon capture data and on a PowerBI-based front end.
Jill Lewis (Troika) retraced the evolution of the SEG’s seismic data recording formats from the ‘too successful’ SEG-Y that is still with us. But SEG-Y does not do 3D, microseismic or OBC and is (strictly speaking) a spec for the now defunct 9 track tape. Rev 1 addressed many of the shortcomings but had, ‘zero take-up’. The current Rev 2 is more promising but would benefit from a more flexible metadata capability. Troika has proposed* adding an XML file to capture extra information and allow for auto-read and safe data transfer. XML allows for a more robust data description than legacy formats that require knowledge of bit/byte locations, these are ‘things of the past’. Lewis expressed hope that the OSDU folks would be listening in and opt for Rev 2 rather than the older versions. A JSON representation of Troika’s XML add-on is under discussion.
* Back in 1999 Oil IT Journal opined, ‘The venerable SEG-Y format for seismic data has suffered over the years from inconsistent use, and a desire to stuff more and more information into the format [ … ] SEG-Y is therefore a prima facie candidate for what the French would call a ‘re-looking’ à la XML.
Jamie Cruise (Schlumberger) asked ‘Can AI help OSDU enterprise adoption?’ For Cruise, we are now ‘a couple of years into the OSDU era’. This means that we are now starting to think about how to solve the data management challenges and achieve the nirvana of a single repository. We need to move toward AI supported by physics (or physics supported by AI). This implies managing regular data along with information produced by automation and AI to provide ‘democratized access through de-siloization’. Also, we are leaving the database era. The database did a good job but now we need big data and the cloud. Operators are working with help from Schlumberger on OSDU. It is not always simple, we have been going a few times around the block but have produced ‘glorious open source code’. Cruise sees OSDU in the context of a data mesh, contributing content to single, extended version of the truth. While there is a place for conventional, interactive QC, Cruise is most interested in AI-enhanced rapid ingestion. This involves updating master data records, mining reports with feature extraction and creating a virtual source for master/gold records. AI is cheaper than a person and can be applied to very large numbers of documents. NLP in also useful for data ingestion. The aim is for a corporate knowledge base in OSDU that ‘puts every piece of data in context’. In the Q&A Cruise was asked if anyone has already retired their legacy data stores (OpenWorks, Delfi…) in favor of OSDU. Cruise was not sure that Delfi is ‘legacy’. There is a lot of Delfi in OSDU! But this is early days for OSDU adoption. It is currently running as a supplement to other systems. ‘There will be customers turning-off their legacy systems real soon now for cost reasons, although there is still work to do’. Another question covered data clean-up strategies. Is it better to perform a mass clean-up before migration to OSDU or can this be more easily realized from inside OSDU once all data is has been liberated to it. Cruise opined that you don’t need to do all remediation up front. Just migrate raw source data using OSDU’s flexible data migration services and then build or use clean-up and dedupe services inside the platform. On the issue of data residency legislation, OSDU partner IBM provides in-country deployment. The solution can be transparent across different data regimes.
Harvard Gustad presented Slimm, Equinor’s spatial location information model and media files. Slimm is designed to spatially-enable various data types. Slimm is built on Omnia, Equinor’s cloud-based data platform and allows for interaction with remote devices such as handhelds and tablets. One use case is corrosion monitoring using a photograph taken by a plant operator. Slimm locates the photo on a 3D plant model for e.g. corrosion evaluation. One third of Equinor data has a spatial component. The 3D location problem is a meaningful problem to address as there are ‘no standard off-the-shelf solutions available’. Equinor is adopting an open source strategy to develop minimum viable data products for operations and maintenance. Equinor’s Technology, Digital & Innovation (TDI) is working on autonomous inspection robots. These are tested at the Kaarstoe K-Lab in a 3D virtual plant model. Slimm data feeds into Equinor’s Echo asset digital twin.
Henri Blondelle described how his company Agile DD has pivoted from oil and gas (Blondelle was previously a geophysicist with CGG) into the mining arena. Agile is now capturing assay table in mining documents or ‘(data) mining for mining’. He thanked clients, including BRGM, Orano and Barrick Gold, for their welcome into these new fields. Mining leverages real-time data from Lidar drones to monitor activity. One Nevada asset has some 40 million documents for capture. These are similar to oil and gas reports, tables, composite logs. Blondelle contrasted manual data entry from documents to a database with smarter approaches. Documents with a similar layout can be collated and rules defined to capture key elements. This is ‘quite a good approach but limited’. Agile’s approach iQC/IDP (intelligent document process) leverages AI to capture and populate a structured data set. Domain-specific training is the key, generic AI models are no good. The model must handle all aspects of a document, layout and content. Training by users is also important. OCR can be useful – even on handwritten documents. For automatic capture of tables in documents, Agile has developed TABIO*. Why not just use existing tools like PDFTables, Tabula or MIT’s Camelot? Because these are quite deterministic and inflexible. They are not trainable so don’t improve with use. Blurry text and other defects will affect the results but even a partial success on a large data set can be very useful in providing extra data points. In the Q&A Blondelle was asked if, after document capture, operators could throw their paper records away. He replied that, ‘No, you will never extract everything. Only a human eye/brain can do this’. If you do have to lose the paper, at least, keep the scans.
* Tabio development was co-funded by Saipem, Schlumberger, Subsea7 and Total. The open source GIT repository can be found here.
David MacKinnon (seconded from Total) presented the UK Technology Leadership Board which describes itself as ‘one of seven industry task forces [that] works with government and other stakeholders to [adapt] oil and gas technologies that support the energy transition’. For MacKinnon, this means robots and autonomous systems, his passion in particular, corrosion inspection robots roaming offshore platforms controlled remotely from Aberdeen. The Energy Systems Catapult, OLTER (the offshore low touch energy center where industry, supply chain, academia, developers and other sectors collaborate) and the ‘National Robotarium’ currently under construction at Herriot Watt university.
The Society for Professional Data Managers (SPDM) was established in 2017 to support the professional development of the worldwide community of data and information managers working in the energy sector. SPDM was founded by CDA* and Norway’s ECIM. The next ECIM Annual International E&P DM meeting, the 25th, will take place in Haugesund, Norway
* Common Data Access has now been rolled-into Oil & Gas UK, which recently rebranded as Offshore Energies UK.
© Oil IT Journal - all rights reserved.