I was honored to be able to give a short presentation in Markus Sunela's dissertation party about artificial intelligence in water sector. Such an interesting topic that we yet know so little. How would you utilize an AI in water sector?
3. Artificial Intelligence
TERMINOLOGY
27.11.2017 Kalervo Kylätie / Fluidit LTD
Algorithm
These are the building blocks of an any
A.I and a computer program.
AlgorithmAlgorithm
Neural Networks
For storing data and
processing data relations
MachineLearning
Supervised Learning
“A Cat” “A Cat”
“A Cat”
“A Cat”
Reinforced learning
Deep Learning
0% Cat 10% Cat 25% Cat 75% Cat
100% Cat
Learning in use. A.I. starts to learn why something is
a cat. “Animal, four legs, whiskers….."
“Not a
cat”
4. WHAT IS A.I. USED FOR ALREADY?
27.11.2017 Kalervo Kylätie / Fluidit LTD
Bottom line is, that A.I. is
used to enhance and
transform human labor.
Education Finance Heavy industry Healthcare
Marketing and
Sales
Customer
service
Transportation Infrastructure –
Water sector?
My hypothesis:
1 doctor does one diagnosis in
1 hour.
1 doctor aided with A.I pre-
diagnosis could do 10-30
diagnoses in 1 hour.
5. WHAT COULD BE THE POSSIBLE USABLE
APPLICATIONS IN WATER INDUSTRY?
27.11.2017 Kalervo Kylätie / Fluidit LTD
Difficulty
Advantage
Billing
Customer
service
Pipe connection
application process
Monitoring water
quality
Maintaining and
updating network
information
Monitoring overall
health of a network
and it’s systems
Optimizing water
production
Planning new
pipelines
Auto repair bots
Automated
excavation
Predicting effects of
different events (human or
environmental related)
Leak detection
Communications
Forecastig failures
Optimizing
maintenance routines
Automated investment
optimization
Online simulations
6. LEARNING MATERIALS FOR
WATER SECTOR A.I.
27.11.2017 Kalervo Kylätie / Fluidit LTD
Planning
standards
Best
practices
Scada data
Billing data
Electricity
price
Groundwate
r levels Social media
behavior
Events in the
area
Google
analytics
Network
models
Network
information system
Planning
information
Pipe and machinery
prices
Material
Stock
Financial
information
Human
resources
Soil
information
Weather
information
Emergency
center
Supervised Learning
Reinforced learning
Deep Learning
“Ok sure, no problem.”
Point cloud and
height models “Ah…too much data.”
+Much more+
7. UTILIZATION: NETWORK PLANNING
AI PLANTRON
27.11.2017 Kalervo Kylätie / Fluidit LTD
Algorithms and
Neural networks
Supervised Learning
Reinforced Learning
Learning in action – Deep Learning
Hello
World!
• Daily consumption = 120 l/ person (get
data from SCADA.AI)
• Average velocity in pipes: 0.6 m/s (try to
keep it)
• Depth depends on frost / other infra (ask
infra data from infra AI)
• Cost of pipes per meters is listed here.
(GoogleIT)
• These are all the earlier plans that have
been made
…
Plantron.message
(“Hello World!”)
“Plantron: What is
the best water main
plan for this area?”
“With my current
calculations:”
8. UTILIZATION: NETWORK PLANNING
AI PLANTRON
27.11.2017 Kalervo Kylätie / Fluidit LTD
“With my current
calculations:”
“How about this?” “How about this?” “How about this?”
“No”
“No” “Accepted”“No”
Education
“Ashok Goel created a teaching assistant for his class in artificial Intelligence. After some tinkering by the research team, Jill found her groove and soon was answering questions with 97 percent certainty.”
Finance
Robot trading, no need for people running and screaming in stock exchanges.
Heavy industry
Assembly robots and automatons, Machine vision to spot failures, or welding bots and testing bots.
Automatized harbours – In Tampere Kalmar has pilot Inland in Rusko.
Hospitals and medicine
A.I. Diagnosis based on persons behavior and symptoms compared to all data in all over the world yields better results than single persons opinion.
Predict best cancer treatment for a patient based on his/hers medical data (Microsoft)
Human Resources & Recruiting
A.I. that generates specific question patterns based on candidates CV for preliminary job interview
Marketing and sales
Google, Facebook countless A.I. implementations
Vainu.io – Business lead A.I.
Online and telephone customer service
Customer Helpdesk robots that regocnize specific keywords and forward people to right subsites
Siri etc.
Transportation
Autonomous cars that can operate without a driver (24 000 driverless cars to Uber from Volvo)