Your SlideShare is downloading. ×
Infovision ramsu sundararajan & srihari narasimhan _ analytics on the industrial internet
Upcoming SlideShare
Loading in...5
×

Thanks for flagging this SlideShare!

Oops! An error has occurred.

×
Saving this for later? Get the SlideShare app to save on your phone or tablet. Read anywhere, anytime – even offline.
Text the download link to your phone
Standard text messaging rates apply

Infovision ramsu sundararajan & srihari narasimhan _ analytics on the industrial internet

304

Published on

Analytics on the Industrial Internet …

Analytics on the Industrial Internet
Ramsu Sundararajan
Srihari Narasimhan
Software Sciences & Analytics
GE Global Research
Bangalore, India

0 Comments
1 Like
Statistics
Notes
  • Be the first to comment

No Downloads
Views
Total Views
304
On Slideshare
0
From Embeds
0
Number of Embeds
0
Actions
Shares
0
Downloads
35
Comments
0
Likes
1
Embeds 0
No embeds

Report content
Flagged as inappropriate Flag as inappropriate
Flag as inappropriate

Select your reason for flagging this presentation as inappropriate.

Cancel
No notes for slide

Transcript

  • 1. Analytics on the Industrial InternetRamsu SundararajanSrihari NarasimhanSoftware Sciences & AnalyticsGE Global ResearchBangalore, India
  • 2. The Industrial InternetSource: www.gereports.com 2
  • 3. The analytics challenge Energy Transportation Financial Services Healthcare … 3
  • 4. Analytics in energy services Capture Create Discover, Create, Validation of Raw Modeling Actionable Decision support Services results and Grow Value Data Information Analytics Data Acquisition methods Pervasive Analytics and Modeling to Equipment Advanced Decisioning Engines Robust, Secure, Information Convert Data into Through Artificial & Long- Long-Lasting Acquisition Information Computational Intelligence Decision Systems - Sensors - Computing & storage resources - Performance optimization - Real-time adaptive - Cameras differ along the line (e.g. Fault - Improved CFD models, feedback for behaviour - Field reports (text) detection on on-site monitors vs. design of newer equipment - Systems that evolve - Data from smart RM&D centers) - Diagnostics & prognostics – along with the data grid? - Visualization tools for multi- performance monitoring, fault detection, dimensional time series data root cause analysis - Working with compressed data - Plant-level analytics and decisioning - Text mining - Flow up to contractual services 4
  • 5. Performance monitoring → Diagnostics → Prognostics → Lifecycle management Performance monitoringSource: Rajagopalan C., Debasis Bal, Roopesh Ranjan. The Power Of Analytics In Equipment Diagnosis. JFWTC Journal. Vol. 7 (3-4). 2011. 5
  • 6. Scalable, resource-aware analytics• Big data (X) + Small data (Y)• Data quality• Scalable, reusable analytics• Complexity challenge: How to use domain knowledge to improve the bias-variance trade-off• Computational challenge: “Big” is in the eye of the computer! 6
  • 7. Computing Challenge Processors are getting wider and not faster? • Power wall • Memory Bottleneck • ILP (Instruction Level Parallelism)• More cores: need change in programming paradigms/architecture• Power consumption • Supercomputing challenge – exascale computing <= 20MW • Low end devices need more features, more performance, better battery life – low power many-core computing• Memory – bandwidth, cost, power (low memory footprint computing)• Higher performance at lower cost 7
  • 8. Computing Challenge Processors are getting wider and not faster? • Power wall Many analytics and applications in • Memory Bottleneck GE may be required to run on a • ILP (Instruction Level Parallelism) heterogeneous computing platform• More cores: need change in programming paradigms/architecture• Power consumption • Supercomputing challenge – exascale computing <= 20MW • Low end devices need more features, more performance, better battery life – low power many-core computing• Memory – bandwidth, cost, power (low memory footprint computing)• Higher performance at lower cost 8
  • 9. Analytics and Computing• Low-cost sensors combined with low-cost low- power processors to pre-process the data and send back only the bits that need further analysis• Low-cost computing infrastructure to enable analytics capabilities• Self-provisioning systems to connect analytical algorithms with the cloud• Examples include remote monitoring, engineering applications, image analytics, etc. 9
  • 10. Computing @ GE Global Research Low pressure turbine simulation @ Oak Ridge National Lab (Jaguar) Turbojet engine simulation at Aragonne National Lab (BlueGene/P) 10
  • 11. Fast Recon on GPU Dynamic Volume GPU-Based Recon as a Imaging Solution Vibrant -Flex CPU Cores GPU Cores Lava-Flex fMRI + Fewer powerful CPU cores vs. 100’s of massively parallel GPU ASL cores + Low incremental cost Larger bubble => more phases Data Size - Parallel Algorithm design andTowards Higher Resolution, Higher Acceleration, developmentMultiple Phases ! More complex iterative Algorithms!!Low Recon Lag at Low Cost !!! 11

×