OSIsoft plays a crucial role in oil and gas and other process industries. Its PI System acts as a buffer from disparate process control systems and provides operators with a ‘single source’ of current and historical data. Some 1400 attended (60% from oil and gas) the recent 2017 EAME users conference in London and heard from founder and CEO Pat Kennedy on how digital transformation is impacting diverse industries. ‘Whatever you think of the Internet of things (IoT), it’s way low, there will be a trillion sensors… and we love that!’ OSIsoft is working on its systems’ scalability and configurability to adapt to the new ‘big data’ normal.
OSIsoft itself is changing. Earlier venture capitalist investors have been bought out by Mitsui (in 2016) and Softbank earlier this year. Kennedy sees synergies with Mitsui’s energy division and with Softbank portfolio companies including the Vision Fund and ARM. He also highlighted the promise of the OneWeb joint venture that is to launch a constellation of low earth orbiting satellites starting in 2018 that will constitute a ‘global IoT/M2M system.’ Kennedy is moving OSIsoft in the direction of the ‘community system.’ ‘While we need to be more attuned to IP and data ownership these need not be barriers to openness.’
Matt Ziegler and co-presenter Daniele Farris explained how PI fits with the data lake paradigm. The data lake notionally is a single data system ‘that does everything.’ In practice, this goal remains elusive. Raw data needs to be marked-up with context ahead of ingestion by people who understand the process. PI and the PI Integrator act as a bridge into the immutable data repository/data lake. The data lake represents a major shift, ‘from Oracle to open source stuff.’ But then, ‘the risk is transferred to your IT.’ PI can mitigate this risk. PI is ‘climbing the value ladder,’ from monitoring to optimization. Currently this means integration with Tableau, Spotfire and SAP Hana. The future will see multi variate analytics with R, Python and MS Azure.
SVP Martin Otterson announced a new IoT ‘edge’ strategy based around the OSIsoft message format (OMF) and a collaboration with Dianomic. A new Windows/Linux Open Edge module adds a read/write to PI capability to field devices and a PI mini historian. Christian Leroux introduced ‘Pervasive data collection’ (PDC) for Industrie 4.0. Sensors and equipment that were previously too hard or expensive to connect can now be brought into the PI ecosystem. Leroux expects that as control system vendors move to the cloud, we may well be moving towards multiple, competing proprietary clouds. PDC makes a direct connection without going through a vendor’s control system. At the heart of this system-independent connectivity is OMF, the OSIsoft message format. The solution is also applicable to stand-alone kit such as vibration sensors which can be plugged into PI with an ‘OMF app.’ PDC is grandly presented as the ‘OSIsoft IoT architecture for the community.’ Along with OMF connectors, PDC includes an ‘Edge data store’ (Windows/Linux), an Open edge module (Linux/RTOS) and OMF Apps (any OS). OSIsoft is refactoring PI Connectors with OMF. PDC is free, ‘just buy your hardware.’
Ed Knutsen (Siemens) and Morten IIleby (AkerBP) presented on the Norwegian Ivar Aasen platform’s IT that underpins both an ongoing de-manning program and a shift from calendar-based maintenance to onshore maintenance planning, a.k.a ‘predictive’ maintenance, with PI/AF* analytics for condition and performance monitoring. A fiber data link replicates the platform control room to another, onshore in Trondheim. This is currently used in a monitoring/advisory capacity but in the future, with further de-manning, the plan is to run the field from onshore. The joint Siemens/OSIsoft PI-AF solution is productized as MindSphere for Offshore.
David Coe reprised the Microsoft/OSIsoft Red Carpet Incubation Program (RCIP) announced last year, vaunting the merits of Azure’s ‘intelligent cloud,’ where now ‘30% of workloads are Linux-based.’ ‘There almost certainly an Azure presence where you are, an Africa region cloud will be running real soon now.’ RCIP leverages the ISO 14224 oil and gas maintenance standard. Azure is based on NIST security and ‘more standards than anyone else in the industry.’ RCIP now includes artificial intelligence, natural language processing, semantics and cognitive. Coe cited Repsol and DCP Midstream as RCIP users.
Toby Grimshaw described how ARM (a Softbank unit) is ‘seizing the trillion-device opportunity’ of the IoT with its Mbed IoT platform. ARM’s chips currently power ‘95% of smartphones’ and ARM claims that over 300,000 developers use the Mbed IoT operating system. Mbed provides ‘chip to cloud’ security and ARM is to extend OSIsoft Cloud services connectivity to devices from multiple vendors.
Randy Chitwood (Brüel & Kjaer/Setpoint) asked ‘are you serious about vibration monitoring. Setpoint captures high frequency vibration data directly into PI while preserving critical waveform data. There is a perception that ‘PI can’t handle high frequency data.’ But the reality is that ‘if you have good edge processing, you can.’ Setpoint’s specialist hardware collects data from multiple sensors on a machine and runs analytics on a separate stream from safety and operations data. High-speed data is transmitted in standard PI tags and lossless compression concepts from the video streaming domain are leverages to optimize bandwidth use. PI AF connects back into the Setpoint hardware and builds a tag hierarchy. Bidirectional connectivity can see into PI enterprise data to tune analytics.
Bruce Taylor, with help from IT Vizion, has successfully built a master asset model of Sinclair Oil’s refinery in PI AF. Over the last few years, Sinclair has acquired a lot of technology in its digital transformation, Maximo, PI suite, Meridian, LIMS and more. Each was deployed independently within a functional unit. The problem is that ‘in refining, everything is tied to everything else.’ PI AF was not really being used despite its potential role in solving this classic challenge. There is much talk about building an asset model. In fact, building is easy. What’s hard is keeping the model fresh, particularly when it underpins analytics. A counter example of this is using an Excel spreadsheet used as management tool, ‘over time, nobody knows or understands what’s inside.’ The US OSHA regulator mandates up to date plant information, especially P&IDs. These were the foundation of the asset model. Sinclair leveraged IT Vizion’s technology in its own-brand P&ID management system, ‘Smart P&ID iDINO*.’ Systems plug-into the iDINO register. CAD integrates to PI AF with ISO 15926 class templates. Additions and modifications are broadcast to SAP and Maximo. Sinclair now plans to deploy advanced analytics and ‘leapfrog predictive and go straight to prescriptive.’ The PI AF data structure is ‘ready for anything!’
Element Analytics is ‘unlocking’ operational data by capturing PI and other data into an HDFS-based data lake a.k.a. an operational ‘digital twin.’ Andrew Soignier opined that oil and gas and other asset intensive industries are challenged by ‘data readiness.’ ‘Even with PI, data may lack context.’ EA leverages a standard, normalized object data model with quality labels, data relationships and financial impacts. Cleansed data from the lake can be reused in PI AF for cross-asset studies. EA speeds modelling, tagging and data prep across PI, Excel and SAP to ‘figure out where key stuff is located.’ EA is a Microsoft Azure cloud solution provider/partner. Other technologies deployed include PI AF XML, GE JSON and Energistics.
We were intrigued by exhibitor Dianomic which, according to a release timed with the conference, is helping OSIsoft with its ‘edge and open source’ strategy. The collaboration is to bring the ‘treasure trove of open source development tools to the PI System community.’ Dianomic is building Linux-based micro-service modules to connect and manage smart sensors using the OSIsoft message format. Prototype software was up and running on a Rasberry PI-based edge device. OSIsoft has part-funded Dianomic in response to third party DIY/open source developments that bypass PI.
Keary Rogers and Brendon Bell reported on the use of statistical QC (SQC) at TransCanada. Data from its 50k km pipeline network including 800 compressor units streams into its in-house developed analytics system. Early fault detection makes for a cheaper fix and minimizes disruption. Some 1600 data streams are monitored for anomalies (3 out of 4 sensors out of tolerance). Trans-Canada’s ‘enterprise analytics’ system uses physics-based models where they exist, otherwise it’s SQC. Here, simple models catch basic anomalies as trends go out of normal. PI AF templates and the Excel PI-builder plug-in are used to build SQC models. PI SQL Commander queries legacy data and SAP and TransCanada’s compressor book for performance models. The system saved around $10 million in 2017 with 129 anomalies detected. In the Q&A, Rogers addressed the problem of false positives. Although there are thousands of measurements on a compressor, EA only deals with a subset selected with input from compressor experts to minimize the likelihood of a false positive. Seven years of EA operations have been a learning process. ‘We don’t want to bother maintenance teams with trivial interventions.’
Gavin Halls (with Khairil Azwan Khabri) presented Petronas’ rotating equipment analytics (Protean), a PI system-based predictive maintenance program to that identifies incipient failures and opportunities for improvement. Petronas figured that a vendor solution would be expensive and elected to build its own. The current version of Protean leverages an extensive PI suite with data consolidated to a PI System Explorer dashboard. Data points are classified to ISO 14224. Halls asked, ‘why pay the OEM for remote monitoring and diagnostic services? They should be paying us to use our data!’
Anthony Teodorczuk (SBM), with help from Veolia reported on the use of OSIsoft Connected Services to meet water injection targets across SBM’s 14-strong fleet of FPSOs. Under SMB’s ‘lease and operate’ FPSO model, water targets are mandatory. The low oil price means more stringent targets to remove solids and sulfates and a constant attention to optimization. SMB were inspired by a Flowserve presentation at the 2015 Prague UC where a hybrid onshore/offshore support model was presented. The solution has greatly enhanced maintenance of nanofilter membranes. Alexander Dixon from systems integrator Servelec Controls added, ‘Out of the box PI can work but it might not meet all expectations without custom code. We connected SMB’s infrastructure using the cloud connect service. This is on premises at SBM and Veolia is configured as a subscriber. We found cloud connect better than PI-to-PI – as it improves security. Analytics on PI that have saved membranes from damage leverage complex rules that could not be captured in event frames.’ There remain some issues, PI Vision is not multi tenancy so sharing views with clients is hard. SBM is now extending the OSI/Connect concept to other vendors (equipment, consumables) so that all see the same information.
Peter van den Heuvel traced Shell’s journey to advanced analytics. Shell started using PI in 1998 and today has 15,000 PI users, 7.5 million tags and 100,000 calculations per minute. 2016 saw the start of advanced analytics with PI data and in 2018, PI Vision is to be the tool of the future, bringing all Shell’s downstream sites together in a PI ‘super collective’ containing 25 million equipment items in one PI AF structure. Atop of this Shell deploys a constellation of software including Matlab and Alteryx’ for orchestration.
Dan Jeavons took over to announce that Shell’s artificially intelligent solutions have brought ‘quite spectacular business benefits in some cases.’ Jeavons’ team makes data and ‘smart’ analytical applications available in end user workflows. Tools of the trade include Matlab, R, Python and Spark. Shell believes that engineers can learn machine learning and is supporting this with demos to senior managers. Best practices are shared in monthly meetings and a data science workbench is available ‘for a small fee’ across the business. ‘Get with it because many silicon valley companies are ahead of us!’
The Holy Grail is predictive asset maintenance but here, ‘nobody including us is there’ even though there are ‘exciting initial results.’ Shell has been testing a predictive algorithm on PI data from its Shearwater asset. Here highly instrumented equipment is being studied to see which of 200 tags give the best indication of failure. Results are ‘very positive, with huge implications.’ Although, ‘even if it doesn’t work, we are learning a lot about how to combine first principles and data-driven analyses.’ Early trials on Canadian carbon capture and storage include work on massive sparse matrices of laser sensor scans.
The London EAME also saw the launch of a ‘PI + GIS User Group’ a cross industry forum for users of PI and ESRI tools.
** A reference to Sinclair’s Dino brand.
© Oil IT Journal - all rights reserved.