Navigant Research estimates that utility companies will spend almost $50 billion on asset management and grid monitoring technology by 2023. Today many organizations are facing budgetary challenges in order to increase reliability, uptime and safety within their facilities.
The industry is adapting to new technologies including utilization of advanced sensors and sensor fusion, edge devices, artificial intelligence, and machine learning to create the maintenance center of the future.
Bernie Cook, former Director of Maintenance and Diagnostics at Duke Energy and now VP of Woyshner Service consulting, will join us to provide practical guidance and examples of how utilities can begin adapting these next generation technologies within their facilities to drive significant reduction in maintenance costs.
Following Bernie, Stuart Gillen, Director of Business Development at SparkCognition, will give examples of how machine learning technologies are augmenting current practices that make maintenance engineers more efficient at predicting critical asset failure.
Join this webinar to learn about:
- Real examples of ways utilities are moving to more advanced monitoring and diagnostic capabilities and the technologies involved.
- How machine learning can improve equipment reliability and performance, and reduce operational and maintenance costs.
- How machine learning can augment or even supplement human subject matter experts by providing significant advance notice of asset performance issues.
3. Large Power Utility - Smart Technology Application
Bernie Cook
Retired Duke Energy Director – Maintenance & Diagnostics
Currently Executive Consultant – WSCInc.
• Low-cost Sensor Technologies and New
Measurement Capabilities
• Data Analytics, Integration and Visualization
• Advanced Controls and Automation
• Monitoring and Diagnostics
• Digital Worker Technologies
Real Time
Information
Distributed
and
Adaptive
Intelligence
Action and
Response
4. Large Power Utility – investment in Smart Technologies
2010 -Catastrophic Equipment Failure
Executive Directive – Leverage new
technologies to improve reliability
GAPS
Lack of online Sensors on
critical equipment
CBM Programs – manual 80%
collection & 20% Analytics
M&D Centers limited to available
instrumentation
Need more Auto Diagnostics
5. Smart Technology Design
RemoteMonitoring
• M&D
Network
• Cyber
Security
• New Sensors
for Reliability
• Operations
Zero Events
• Enhance
M&D Center
• Wireless
Sensors
Diagnostics&Prognostics
• Asset Fault
Signature
• Big Data
Analytics
• Rule Based
Diagnostics
• Dynamic
RUL
DataVisualization
• Data
Integration
• Data Hubs
• Dashboards
• Equipment
Health
Visuals
• Digital
Worker
Data --------------- Information-------------Insight ---------Actionable Intelligence
6. 6
Smart, Connected Plant Assets
Assets Sensors Data Acquisition-
M&D Network
Monitoring & Diagnostics Integration &
Visualization
10,000+ 33,000+ 2,400+ Nodes
Turbine Critical Equipment
Steam Turbine
Combustion Turbine
Generator
Boiler
Balance of Plant
Motors, Pumps, Gearboxes, Fans
Transformers
Iso-Phase Bus Ducts
Electrical Buses
Phase I = Base Installation
Temperature
Accelerometers / Vibration
Turbine Vibration Monitoring (VDMS –
Vibration Diagnostics Monitoring System)
Proximity
Oil Analysis
Phase II = Advanced Sensors
Cameras
Thermal Cameras
Infrared Sensors (IR)
Electro Magnetic Signature Analysis
(EMSA)
Motor Current Signature Analysis
Sensors (MCSA)
CT (foreign object & leak detection)
Phase III =
Advanced Sensors II
New Sensors for Major Component
Zero Event Operations
Focuses on reducing operational risk;
event free index
NI CompactRIO
(reconfigurable embedded control and
acquisition system)
NI cRIO-9068 / NI cRIO-9074
NI cRIO-9024 Turbine
NI InsightCM™ Enterprise
NI InsightCM™ Data Explorer
NI InsightCM™ Serve
APR Models
Efficiency Monitoring
& Thermal Modeling
ADVANCED ANALYTICS
Industrial Internet of Things – Interconnectivity Direction
Business Intelligence
7. SmartGen Sensor Technologies
H2 and NH3
Leak Detection
Oil Levels
Oil Dielectrics
Particle Counts
NI cRIO Monitoring Node SENSORS
Turbine Monitoring
Systems
Balance of Plant Systems
Embedded Turbine
Monitoring System
Vibration Sensors
3rd Party Systems
Sensors
Vibration
Temperature
Pressure
Flow
Position/Displacement
Oil Quality
Ultrasound
Infrared Thermography
Power/Current
Leak Detection
Dissolved Gas Analysis
Electromagnetic
Interference
Partial Discharge
Optics
Acoustic
Optics
Accelerometers,
Proximity Probes
8. Utility Example of ROI – Technology Application
FINDS
An investigated notification
identified an equipment issue
that requires corrective action
384 COST AVOIDANCE
Based on the difference
between probability and
impact of failure with and
without M&D center
interaction
$31.50M
198 $18.26
134 $4.39M
52 $8.85M
Equipment Finds
2013
2014
2015
One Utility Example:
By adding new
technologies…..
From 2013 to 2015
4x increase in
Equipment Problem
Detection
$10M+ annual increase
in avoided cost.
9. Increasing Challenges drives more Innovative Solutions
Power Generation Challenges
Energy Efficiency lower load demands
Renewables affecting dispatching
Run plants with less staff & $$ Fourth Industrial Revolution
Age of Innovation
Battery Storage
Wind
Solar
Power
Generation has to
embrace innovative
new process and
technology solutions to
survive
People
Enhanced Processes
New Technologies
10. Smart Plant Connected Assets combines sensors, microprocessors, data acquisition, data storage and software
with critical hardware across the fleet such as steam turbines, combustion turbines, generators, transformers, and
large balance-of-plant equipment. The smart critical assets are also connected to each other via wired and wireless
technology. The resulting “smart, connected plant assets” have intelligence and connectivity that enable an entirely
new set of functions and capabilities:
PLANT ASSETS
Physical components e.g. combustion turbines,
steam turbines, generators, turbines, pumps,
motors etc.
SMART PLANT ASSETS
Sensors, microprocessors, data acquisition,
data storage, controls, software, embedded
operating system, enhanced user interface etc.
SMART, CONNECTED PLANT ASSETS
Ports, antennae, protocols enabling wired /
wireless connections with plant asset: one-to-
one, one-to-many, many-to-many.
Source: What is “Internet of Things” (IoT)? www.educatingplanet.com
Integration of People + Processes + Assets + Data = Better Decision Making
Interact Compute Connect
Smart Plant Connected Assets
11. Data → Information → Insight → Actionable Intelligence
11
Next Innovation Wave:
Advanced Analytics
Descriptive
Analytics
Diagnostic
Analytics
Predictive
Analytics
Prescriptive
Analytics
What happened?
Why did it happen?
Difficulty
Value
Source: Gartner
What is likely
to happen?
What should I do
about it?
We are Here
12. Data in Context: Analytics for Equipment & Component Condition
Monitoring
& Trending
Pattern
Recognition &
Identification
Early
Identification &
Fault/Source
Determination
Big Data Analytics
Fault Signature
Recognition
Dynamic RUL
Current State
Data --- Information ---- Insight ------ Actionable Intelligence
13. 13
Apply new Prescriptive Analytics -
On Line Vibration Analytics
Predictive Auto Diagnosis
Failure Signature Recognition - Prognostics
PI
Historian
DCS
Existing & New Sensor Data
M&D Center
Models
Asset Health
Information
M&D
IT
Data Integration
Big Data Analytics
Digital Worker
Dynamic field interaction
Equipment Health, Dwgs,
Work orders, etc…
14. En g i n e e rin
g A n a l y s i s
En g i n e e rin
g A n a l y s i s
Monitoring
Tools
SMEsSMEs
DCSDCS
Walkdown
ALARMSALARMS
Walkdown
Troubleshooting an Issue: How This Could Change
TODAY TOMORROW
Monitoring
Tools
Prescriptive
Analytics
15. In Summary – How to get started??
Power
Generation has to
embrace innovative
new process and
technology solutions
to survive
People
Enhanced Processes
New Technologies
• Power Generation challenges are real
• Utilities need to embrace new technologies
and processes to survive
• Starts with Leadership Sponsorship & Buy-in
• WSC provides a Workshop & Assessment
• Educate Leadership
• Identify & Prioritize Issues
• Develop an Implementation Plan
• New Processes
• New Technology Application
• Fossil // Hydro // Nuclear
17. How would you write code to tell the difference
between a banana, an apple, and a grape?
Bananas are Yellow Apples are Red Grapes are Green
18. How does Machine Learning tell the difference
between a banana, an apple, and a grape?
• Feed in measurable characteristics to an algorithm
• Height
• Width
• Height/Width
• Color
• Color variation
• Shape
• Let the algorithm define the relationships between
the measurable characteristics and the fruit they
embody
19. SparkCognition Cognitive Analytics – Beyond Machine Learning
Natural language processing
• Enables recall of answers, in context
• Analysis of human readable text for clues,
insights and evidence
Deep Learning and Reasoning
algorithms
• Improves accuracy
• Learns complex patterns
• Scales efficiently: High speed, large data
implementations
• Make decisions in the absence of training
data
Automated Model Building and Infinite
Learning
• Watches data and derives rules
• Incorporates human feedback to
strengthen or dismiss conclusions
• Automatically learns from feedback and
greater volumes of data
• More data = more accuracy, capability &
insight.
Powerful Visualization with Evidential
Insights
• Provides transparency and evidence about
what the cognitive system is learning and
proposing
• Presents data elegantly – Analyst friendly
interface, easy feedback
• Elevates evidence / reasoning for machine
decisions
Powerful advancements in state of the art
20. Machine Learning can make sense of data from your
enterprise
Years
ExponentialDataGrowth
• Data explosion across all
departments (i.e. Lots of data!)
• Complex relationships exist
between data
• Silo’ed data sets limit visibility
into enterprise risks
• Limited ability to add people to
analyze
• Ability to codify knowledge
retention
Ideal Environment for Machine
Learning
21. Asset Health Architecture
Detailed Evidence
• Provide evidence behind the
insights
• Provide tools for expert analysis
System Optimization
• Optimize not at local but at a
global level
• Plug insights into platforms such
as BI, Inventory mgmt., PLM etc.
Actionable Insights
• Extend asset life
• Avoid downtime
• In-field, real-time
recommendations
• Cyber Security Threats
Data Collection
Output
Analytics
Platform
Assets
Data Lake
24. Challenge of analyzing transient events
A Day in the life of a Turbine
• Transients are when systems are stressed (think aircraft landings and takeoffs)
• Graph shows one variable (Speed)…and what is state of the art today, actually
thousands of variables are being collected.
• Transients can last for different time periods (startups can be 20 min to 40 min long)
• Steady state can be analyzed today, with some difficulty, no tools to analyze transient
events
25. Machine Learning compared to traditional analytics approach
Traditional Machine Learning
Time As a Variable
Single point event prediction
Good at steady state, non time varying
events
Can’t handle transient events, where
time is a variable rather than an index
Complete time series treated as single event
Very good at comparing both steady state &
transient events (Startup, coast-down,
thermal vector, cooldown etc.)
User Skill Required
Knowing the relevant data is key to
building a model
Model effectiveness depends on the skill
of the user
Relevant data and key features are identified
automatically by the tool
The information is inherent to the process
Information can be captured on day one
Diagnostic Data
Identification
Fault diagnostic library must be defined
explicitly
Key features calculated and added to the
model for diagnosis
Fault diagnostic data is inherent to the event
1st, 2nd, & 3rd order features automatically
created
User input integrated into system to “learn”
from SME experience
Data Cleansing &
Model Training
Model training is done by Manually
excluding bad points from a data set
Model identifies normal (median) behavior
and 3 standard deviations automatically
27. The picture changes as more features applied
As we add more variables the picture changes
• Now add thousands of
features/turbine
• Identify new events coming in
real-time
• No modeling required from SMEs
to gain insights
• Model keeps learning SMEs
can “nudge” model over time.
28. Initial Screen
• Classify Gas Turbine transient events via
supervised learning (i.e. we know the “labels”)
• Automatically classifies multiple operating states
• Ability to utilize and capture SME expertise,
then LEARNS
• Top down vs bottoms up approach
• Utilizes National Instruments
cRIO and OSI PI historian data
for vibration and speed
Plant 1 Plant 1
Plant 1 Plant 1
Plant 1 Plant 1
Thermal
Event
Bad Bearing
31. Empowering the end-user to improve business operations
SparkCognition and the client developed an IBM Watson
powered “Advisory” application for maintenance
Application enables Directors of maintenance and
technicians to:
• Conduct machine to human dialog to troubleshoot with high
accuracy
• Speedy identification to map the right fault codes and
troubleshooting tips using Natural Language Processing (NLP)
queries
• Optimize work flow and deliver relevant documentation for a
faster turnaround of planes
Lowered the cost of maintenance and improved asset availability for operators by up to 10%
32. SparkCognition Differentiators
• At our core we are a Machine Learning company. This isn’t the case for many organizations just
hiring data scientists.
• We have developed cutting edge, Patented, Cognitive Analytics IP. Patented algorithms are rare.
• Our products rely on data driven, automation of model creation. Takes the burden off the
customer.
• We provide operators solutions on day ONE (SparkPredict, SparkSecure), not just analysis tools.
• Reasoning utilizing NLP drives toward prescriptive maintenance
• Top Down approach vs. Bottom Up
• Outcomes Learn, and Adapt based on SME input (i.e. training)
• We have addressed multiple use cases with multiple techniques to provide value to nearly every
customer in the last two years.