Enabling Layers of Analytics with the PI System from the Edge to the Cloud

Enabling Layers of Analytics with the PI System from the Edge to the Cloud Presenters: Craig Harclerode, O&G Industry Principal, OSIsoft Michael Kanellos, Technology Analyst, OSIsoft October 11, 2017

Today s Presenters Michael Kanellos Technology Analyst, OSIsoft Email: mkanellos@osisoft.com Craig Harclerode O&G Industry Principal, OSIsoft Email: charclerode@osisoft.com

What s Driving Hybrid Analytics? Three Numbers to Know $1 Trillion: The amount of money consumers and businesses will save annually through lower maintenance and consumables through IoT by 2022.

What s Driving Hybrid Analytics? Three Numbers to Know 51%: The population of Things on the Internet by 2021

What s Driving Hybrid Analytics? Three Numbers to Know 40%: IDC predicts that at least 40% of data by 2019 will be stored, analyzed and acted upon at or close to the edge.

Are we in an IOT Bubble? Longer Term Risk vs Near Term Reward.

Outline Context Top 5 Observations Concluding Remarks Q&A

About OSIsoft Founded in 1980 19,000+ Installations, 4,000+ Customers in 123 Countries Top Customer Support Privately Held Company Global Presence, 27 Offices Worldwide Power & Utilities Oil & Gas Chemicals Metals & Mining Pharma Life Sciences Pulp & Paper Datacenters Critical Facilities Over 20% of Revenue Invested in R&D 1,400 Employees 65% of Global 500 Process & Manufacturing Market Leader Enterprise Infrastructure for Streaming Data, Analytics, & Events

PI System Operational Technology (OT) Infrastructure Notifications PI System Access Programmatic methods Line of Business Systems Event Frames Streaming Analytics PI Cloud Services PI Cloud Connect PI Interfaces (Tag Based) Asset Framework PI Integrators Business Analytics PI Connectors & Relays (Asset based, meta data, & tags Including IOT/edge) PI Data Linear Archive (Historian) drag & drop self serve BI Visualization Tool SAP HANA PI Server Bundle Configurable Functionality Perpetual License

Outline Context Top 5 Observations Concluding Remarks Q&A

They evolve an OT Chart of Accounts leveraging configurable smart asset object templates The language of the PI System

OT Chart of Accounts Enables Transformative Business Value Abstraction? Tag Names Asset Names UOM Time Zones

Physical Compressor Stations Centrifugal Compressor Templates Health Index Templates Anomaly Detection Templates Digital Compressor Stations TransCanada Smart OT Infrastructure

Physical Compressor Stations Centrifugal Compressor Templates Health Index Templates Anomaly Detection Templates Digital Compressor Stations TransCanada Smart OT Infrastructure

Exception based KPI Dashboard system Physical Compressor Stations Centrifugal Compressor Templates Contextual Drill Down Health Index Templates Anomaly Detection Templates Digital Compressor Stations TransCanada Smart OT Infrastructure

Shell s Global Monitoring PI System UC2017

Shell central structure Template based calculations per equipment Output Parameters on Facility Element Level calculations visible via standard clients Input Parameters on Facility Element Level

COMPANY and GOAL Company Logo Picture / Image CHALLENGE SOLUTION RESULTS

COMPANY and GOAL Company Logo CHALLENGE SOLUTION RESULTS

They leverage a hybrid data lake strategy

Hybrid Data Lake - Leveraging Fit for Design Technologies Tabular & Unstructured + IT Data Lake/Data Warehouse/Big Data Self Serve Batch & Streaming Integration Predictive ARP Dashboards Statistical Modeling Geospatial & ML/AI multi-dimensiona assessment OT Data Lake - Optimized for real-time data & streaming analytics OT Data Lake/OT Infrastructure Linear/Time, Event, Asset Context Structured Prescriptive, Empirical, & Physics based Streaming Analytics

IT Data Lake/Data Warehouse/Big Data Self Serve Batch & Streaming Integration OT Data Lake - Optimized for real-time data & streaming analytics OT Data Lake/OT Infrastructure

OT Data Lake - Optimized for real-time data & streaming analytics

Tabular & Unstructured + Predictive ARP Statistical Modeling ML/AI Dashboards Geospatial & multidimensional assessment OT Data Lake - Optimized for real-time data & streaming analytics Linear/Time, Event, Asset Context Structured Prescriptive, Empirical, & Physics based Streaming Analytics

Refocusing on Data Use vs Cleansing and Preparation Hybrid Data Lake Approach vs a pure RDB Data Lake Approach

The OT Data Model(PI AF) is Foundational for Higher Level Analytics & The Power of Choice and Self-Serve BI Leading US based E&P Company On Demand/Scheduled

They define analytics and a Layers of analytics framework

Community ML/AI Strategic ML/AI Tactical ML/AI Real-time Streaming Analytics Human Analytics Edge Analytics

Community ML/AI Strategic ML/AI Predictive Statistical Modelling & Machine Learning/AI (Pattern Recognition) Level 2+ Predictive Visual Dashboards & Multidimensional Assessment Tactical ML/AI Real-time Streaming Analytics Descriptive & Prescriptive Level 1 Predictive Time, Event and Asset Context Human Analyti cs Edge Analytics Real-time, contextual, exception based decision support Descriptive & Prescriptive Level 1 Predictive Machine Learning/AI/M2M

Community ML/AI Strategic ML/AI Predictive Statistical Modelling & Machine Learning/AI (Pattern Recognition) Level 2+ Predictive Visual Dashboards & Multidimensional Assessment Tactical ML/AI Real-time Streaming Analytics Descriptive & Prescriptive Level 1 Predictive Time, Event and Asset Context Human Analyti cs Edge Analytics Real-time, contextual, exception based decision support Descriptive & Prescriptive Level 1 Predictive Machine Learning/AI/M2M

Calculating Expected Heat Rate Lookup curve-fit coefficients from SQL Table (Manufacturer Performance Curves) Apply curve-fit to calculate Nominal Heat Rate Calculate Actual Heat Rate

Example of Predictive Analytics in AF Expected vs Actual EA Finding using KPI Strategy Found partially damaged compressor valve. The valve was replaced in a planned & controlled manner.

Smart Asset Object Templates Configuring the Digital Drilling Rig Mud Motor Template Mud Motors Top Drive Draw Works Drilling Phase Analytics Physical Rig #1 Digital Rig #1 Top Drive Template Draw WorksTemplate Mud Motors Top Drive Drilling Phase Events Draw Works Drilling Phase Analytics Physical Rig #2 Digital Rig #2 Drilling Rig OT Infrastructure

Delivering $1B Business Value from Digital Transformation in ~5 years COMPANY and GOAL Deliver $1B in EBITDA by a business transformation enabled by a digital transformation leveraging the PI System as a strategic OT data infrastructure for advanced predictive and proactive analytics Company Logo EMEA UC2016 CHALLENGE Deliver strategic business value to respond to increasing competitive threats; Change a diverse culture to act as one with Operational Excellence & continuous improvement enablement Increasing competitive environment in Eastern Europe Variable cracked spread Diverse culture across 8 countries Low use of data and analytics Poor business performance 4 th Qtle SOLUTION Evolved the use of the PI System as a tag based historian to an asset based infrastructure to support cultural change and data based decision making and support with advanced predictive and proactive analytics. Evolved from Tag to AF based infrastructure across the MOL fuels value chain Normalized tag, asset, UOM, and time using AF as an abstraction layer Used data and information to support business transformation RESULTS Delivering on the MOL Downstream business transformation goal of $1B and more importantly, a sustainable cultural change based on data and information to drive operational excellence going forward into he 21st century. Leading Process Safety Management 1 st Quartile in energy, yields, loss, and utilization OT infrastructure enabling time to value and value momentum with advanced analytics including machine learning

Strategic Machine Learning/Big Data/Advanced Analytics Enabled by the OT Infrastructure Tactical Machine Learning/Big Data/Advanced Analytics Enabled by the OT Infrastructure Real-time Analytics In the OT Infrastructure Analytics and Predictions for : - Dynamic or smart IOW/targets/APM/PSM - How do you smooth operations? - How do you optimize the yields? - How do we optimize the fuels value chain? Analytics & Predictions for : - Coker Hotspot - Hydro treater sulfur in product - Hydro treater cloud point - Bromine Index Benzene - Coke drum filling & removal Analytics and Predictions for: - Corrosion analytics (HTHA, chlorides, etc.) - Natural gas & electrical peak exceedances - CBM exchangers, rotating equipment,etc. - Environmental Limit predictions. UC2016 27+ Tactical ML Apps in production Integrated Control & Safety Systems Excel Files Human Analytics Enabled By and In the OT Infrastructure Enablement of: - Data Based Decisions - Real-time situational awareness - Management by exception 61,000 Event frames across 6 plants

Natural Gas Consumption Prediction BackGround Huge saving possibilities in the decrease of contracted natural gas daily maximum amount Problem High penalty on daily amount exceedance Alerting system was needed Solution Consumption prediction calculations in PI Analysis Detailed information on PI Vision display (about consumption, prediction, contacts of decision makers) E-mail alerting system in Notifications

Integration of the OT Infrastructure & SAP PM 40

Improving DCU Yield and Safety with Azure Machine Learning COMPANY and GOAL Improve Delayed Coking Unit yield and Reduce the risk of coke hot spot steam explosions from feed and operational variability by using Azure machine learning. Company Logo Picture / Image CHALLENGE Opportunity crudes provided incentive to raise DCU yields but resulted in an increase likelihood hot spots resulting in steam eruption events while decoking. Economic incentive to run opportunity crudes $6M for each 1% increase in DCU yield Increase feed variability to the DCU Increased risk of hot spots and steam explosion events during decoking 4X increase in Q1 2016 SOLUTION Leverage existing OT data infrastructure to enable the use of advanced analytics and machine learning to improve yields and reduce the risk of steam eruption. AF infrastructure in place from prior digital transformation from tags to assets Use Microsoft Azure Machine Learning to do massive high fidelity data correlation of DCU feed properties to yields and explosions RESULTS Reduced DCU steam explosion events while improving DCU yields from the processing of more opportunity crudes Enabled the sustainability of increased DCU yields of over 2% by processing opportunity crudes Reduced steam explosions by 75% Calculated savings for 1 DCU unit = $6M/yr for each 1% yield = $12M/yr Rapid rollout to other DCUs

They see IOT as an extension of the PI System Infrastructure

Typical Industrial IoT Architecture 3 rd Party Cloud Services/Apps 3 rd Party Cloud Services/Apps 3 rd Party Cloud Services/Apps Public Private Private Automation System Challenges: Data Ownership Data Security Data Quality Data Context Automation System Edge Gateway Sensor Sensor Sensor Sensor Asset Asset Asset Smart Sensor Smart Sensor Asset Asset Asset Gateway PLC Sensor Sensor

A Need to Mesh Traditional and IIOT Data - Context Traditional Data Sources DCS, SCADA, etc. A few Big Data Pipes IIOT Data Sources Lots of Little Data Pipes.

Brining DCS/SCADA & IOT/Edge Data together in Context 3 rd Party Cloud Services/Apps 3 rd Party Cloud Services/Apps 3 rd Party Cloud Services/Apps Public Private PI System OT Data Infrastructure Automation System Automation System Private Edge Sensor Sensor PI Connector Gateway Sensor Sensor Asset Asset Asset Open Edge Module Smart Sensor OMF App Smart Sensor Asset Asset Asset PLC Edge Data Store Gateway Sensor Sensor

Enterprise PI Server Applications Private Domain Admin DMZ PI Connector Relay PI Connector Admin Server PI Connector Relay Edge PI Connector Windows PI Connector Linux OMF Application Any OS Open Edge Module Linux Edge PI System Windows Subset of Data Edge Analytics Edge PI System Windows Edge Data Store Windows Edge Data Store Linux Assets Automation Systems Edge Devices Sensors Remote Assets Edge Devices Sensors

A Portfolio of IIOT Options and Edge Management Enterprise PI Server Applications Private Domain Admin DMZ PI Connector Relay PI Connector Admin Server PI Connector Relay Edge Subset of Data Edge Analytics PI Connector Windows PI Connector Linux OMF Application Any OS Open Edge Module Linux Edge PI System Windows Edge PI System Windows Edge Data Store Windows Edge Data Store Linux Assets Automation Systems Edge Devices Sensors Remote Assets Edge Devices Sensors

OSIsoft Embedded Technology Examples Service Provider Monico (OMF application) IT Hardware HPE (PI System deployment) IT Hardware Dell (PI System deployment) IT Hardware Cisco (PI Connector on Linux) Automation Hardware Partner (PI Connector on Linux) Service Provider Stratus IoT Solutions (OMF application)

They see new opportunities via the digital value chain the community model

Community Model The Digital Value Chain Coil Tubing Analysis

Outline Context Top 5 Observations Concluding Remarks Q&A

Enabling Layers of Analytics with the PI System from the Edge to the Cloud Evolve Configurable Smart Templates Use a Hybrid Data Lake Strategy Define Layers of Analytics See IIOT as an Extension of the OT Infrastructure Embrace the Community Model The Digital Value Chain

Questions and Answers

Next Stayed tuned for follow-up email o Link to OSIsoft On-Demand webinars o Visit PI Square Have conversation with Michael and Craig