Srikanth Muralidhara, Flutura Decision Sciences and Analytics
October 2015
Algorithms &
Industrial IOT
Forces at Play - The New World Order
• Increased focus on Super Optimization and Automation in mature markets
• Innovation will be led from emerging markets and then brought to mature markets
Source : Oxford Economics
2021 G8
Compounded by Scale of Inefficiencies
Electric power transmission and distribution losses (% of output) in India was 27%
Global retail industry inventory preventable loss at $300 billion
And many more…..
Wide range of unsolved problems – Need for Innovative SolutionsA core part of the solution would be Scalable Algorithms
Industrial IOT Areas and Applicability of Algorithms
4Source : Industrial Internet Consortium
Think “Narrow + Deep”
5Formulate Problems - Narrow and Deep
What should be the tilt in the runner to
maximize RPM?
When can the first service be deferred for a
given seismic region ?
How soon would the pump be affected due to
cavitation?
Think “Subsytems”
6Define Sub System Boundaries and Interaction Points Clearly
Think “Breadth+ Depth”
7Richness in the breadth and optimal depth in data. This is where it all starts !
8
Grey-Box Model
Gracefully marry mechanistic and statistical models
𝑃 = 𝑃𝑜𝑤𝑒𝑟𝜂 = 𝑡𝑢𝑟𝑏𝑖𝑛𝑒 𝑒𝑓𝑓𝑖𝑐𝑖𝑒𝑛𝑐𝑦𝜌 = 𝑑𝑒𝑛𝑠𝑖𝑡𝑦 𝑜𝑓 𝑤𝑎𝑡𝑒𝑟
𝑔 = 𝑔𝑟𝑎𝑣𝑖𝑡𝑎𝑡𝑖𝑜𝑛𝑎𝑙 𝑐𝑜𝑛𝑠𝑡𝑎𝑛𝑡ℎ = ℎ𝑒𝑖𝑔ℎ𝑡 𝑞 = 𝑓𝑙𝑜𝑤 𝑟𝑎𝑡𝑒
Think “Greybox”
Think “Edge + Central ”
9
Location of the algorithm - Edge versus Centrally Controlled Intelligence
Think “Autonomous”
10Extent of Automation and Control
Unmanned systemsRemotely Assisted Remotely Controlled
Tying it all together – Predicting Turbine Failure
Data Preprocessing
•Mahalonobis Outlier Detection
•Pearson Correlation
•Missing Value Imputation
Feature Engineering
•Dominant Frequency Analysis
•Principal Component Analysis/Factor Analysis
•Mean Time Between Failure
Model Development
• Survival Analysis
• Logistic Regression
•Boosted Trees
•Random Forest
• What is the probability the coupler will survive for the next 6 months ?
• Which components are in the last stage of degradation ?
• Is the turbine X axis vibration in the normal operating range?
• Which are the main lead indicators affecting the plant load factor?
• What is the operating range of turbine X axis vibration?
• How does the lubricant oil temperature affect the bearing temperature?
Closing thoughts – 3 core points
• Age of Super Automation taking shape• Intersect of IOT + Analytics can pave the way for responding to this market condition
• Algorithms for unsolved problems
• Grey Box algorithms for better outcomes• Mechanistic + Statistical Models
• Breadth of data key to realizing successful outcomes• Uncovering and fixing data blind spots
13
“The price of light is less than the cost of darkness”Arthur C. Nielsen
IOT
Analytics