4. Microsoft Confidential, For Internal Use Only
• Wave 1 - 1990s' fixed Internet: 1 billion users
• Wave 2 - 2000s' mobile wave: another 2 billion connected endpoints
• Wave 3 - Internet of Everything: connect 10X - 20X
as many "things" to the Internet by 2020
EXPLOSION OF CONNECTED POSSIBILITY
5. Microsoft Confidential, For Internal Use Only
Mobile
Gateway
Car
Gateway
Desktop
Gateway
In-room
Gateway
Wearable
Gateway
Portable
Devices
Desktop
Devices
Floortop
Devices
Ambient
Sensors
Wearable
Sensors
Wearable
Devices
• Context - Each “thing” or connected device is part of the digital
shadow of a person, essential to enable deeply contextualized services.
• Assistance & Task Completion - This informs business
operations and services as much as it informs an individual’s
engagement and behavior.
• Understanding Intent - Every new dimension of
data increases predictive power, enabling an agent, bot,
or application to contextually answer the question:
“What does the human want?”
• Knowledge & Insight - From the data streams that
implement the “digital shadow”, we can use predictive
analytics to understand people’s needs and behaviors
better than ever before.
‘THINGS’ - THE DIGITAL SHADOW OF A PERSON
The most interesting aspect of the Internet of Things is the world of humans that use it
7. Real World IoT Use Cases
Electric
charging
stations
Street
sweepers
Postboxes
Aircrafts
Auto
Elevators
Factory floor
Oil equipment
Cows
Engines
Vending
machines
Buildings
Fryers
Medical devices
Vaccine
dispensers
Trucks
BusesDogs
Oil distribution
Smart meters
Internet
of Things
Power plant
Surveillance
Power tools
Racing
Mining
equipment
Smart grids
10. AZURE REGIONS
42Announced Azure
regions world wide
Hyper-Scale Capacity
3.5 Trillion Messages / Week
Hyper-Scale Azure Footprint
AZURE IOT REGIONS
12
Azure IoT regions world wide
11. Azure IoT Suite Remote Monitoring - BasicDevices
Back end
systems and
processes
Cosmos DB
Web App
Logic AppsIoT Hub
C# simulator
Microservices
Active
Directory
24. Azure IoT Starter Kits
Raspberry Pi 3 Kit
Windows 10 and Raspbian
Samples in C and C#
Feather M0 Wi-Fi Kit
RTOS
Samples in Arduino IDE and C
Feather Huzzah ESP8266 Kit
RTOS
Samples in Arduino IDE and C
ThingDev Kit
RTOS
Samples in Arduino and C
Intel Edison Kit
Linux Yocto
Samples in JavaScript (Node.js)
http://azure.com/iotstarterkits
Notes de l'éditeur
http://www.goldmansachs.com/our-thinking/outlook/iot-infographic.html
Last year, 4.4 ZB of data was created…
Human Genome has 3B Base Pairs = 750 MB, Humanity’s Genome = 7.2B x 750 MB = 4.7 EB…equals data created every 9 hours
The Library of Congress has 24M books = 189 TB of information…equals data created every 1.26 seconds
In 2016, we will create 13 ZB of data
In 2020, we will create 44 ZB of data
Source: Digital Universe Forecast, IDC, April 2014
So why choose Azure? The answer is our unmatched hyper-scale footprint.
As you can see Azure provides truly hyper scale with the largest public cloud infrastructure in the world. There are 38 announced datacenters worldwide with more planned each year and Azure IoT is present in a large number of these with more added each quarter depending on business demand. Azure is the only cloud provider recognized in the industry to have leading solutions in IaaS, PaaS, and SaaS. Azure PaaS platform services can help you be more productive and increase your ROI. In addition, Azure has the most comprehensive compliance coverage of any cloud provider with 50 compliance offerings. To protect your organization, we embed security, privacy and compliance in our development methodology.
Describe Rod Pumps:I am sure you have seen those pumps around, extracting oil from wells that sometimes are 1 mile deep. These pumps are usually in very remote locations with little to no connectivity and no human supervision most of the time.
Traditionally the pumps have been running continuously at the same pace, without taking into account the changing conditions such as temperature, oil pressure downhole, making them all but really efficient.
Finally, maintenance teams have to visit pumping sites often for no reason most of the time and too late in some cases to prevent damage or loss.
Schneider solution:
Schneider Electric created a SCADA solution called Realift to optimize efficiency, extend well life and lower operating expenses.
Realift consists in monitoring well sites and oilfields, collect real time data from sensors attached to pumps and analyze this data to detect anomalies and optimize efficiency.
Today, when a problem occurs, the SCADA system will notify over SMS or email an expert who will interpret data from sensors to diagnose the issue.
Schneider ML models
Schneider wants to be proactive and started working with Azure Machine Learning to do predictive maintenance.
Considering the remote location of the pumps and their limited connectivity, Schneider needs to run the ML models as close as possible to the pumps.
Data analyzed by ML models:
Sensors measuring pump position and fluid load at the surface of the well allow creating an X-Y plot ‘dynacard’ that provides visibility of downhole conditions that can impact pump fill.
Machine Learning models trained with millions of dynacards can instantly and automatically detect anomalies and downhole conditions. No need for a human to look at hundreds, thousands of dynacards to diagnose an issue or determine when next maintenance is needed.
Here is the challenge:
Rod pumps being in remote locations with little to no connectivity it is often difficult and costly to have them send all sensors data to the cloud for analysis.
Furthermore, if a defect or anomaly is detected that requires immediate action to stop the pump, then the round trip to the cloud and bad connection can be in the way of taking action in a timeline manner.
Last but not least, in the oil & gaz industry (like in many other ones) this data is very sensitive and customers prefer not seeing it leave the production site.
Schneider Realift SCADA solution allows real time monitoring of well sites.
[Animate]
A SCADA Remote Terminal Unit installed onsite collects data from pump sensors through a field bus connection like CAN.
[Animate]
A local SCADA application, monitors sensors through the RTU over a Modbus connection
[Animate]
When an issue occurs, the local SCADA application will notify the supervision site sending an SMS or email.
[Animate]
As most often sites are not connected or transmitting data is very expensive, an expert has to be dispatched onsite to diagnose the issue looking at dynacards data.
[Animate]
If an expert cannot be dispatched, data needs to be brought back to the supervision site over an expensive connection or manually.
[Animate]
Data collected when issues occur or during regular onsite maintenance is also used to optimize pump efficiency.
But as you can imagine, having an expert looking at every single dynacard to determine issues and trends is subject to optimization.
For this, Schneider experts developed and trained Machine Learning models in the Cloud with the dynacards data from sites all over the world.
These ML models help determine trends to optimize pumps usage efficiency and also plan for onsite maintenance.
While this solution has already demonstrated it can save a lot of energy, and minimize costs, it is still reactive. The intelligence is in the Cloud and data is hard to bring up there.
What Schneider wants to do is to bring the intelligence at the Edge to make their SCADA solution proactive, implementing predictive maintenance.
And we have it working here for real!
Here is what you will see in the demo in a minute
[Animate]
On the Well site, we have added a gateway device.
[Animate]
This device runs Azure IoT Edge
[Animate]
An edge module has been developed to collect dynacards data from the RTU over ModBus. If the IoT Edge device is connected, this module can send real time data to Azure for remote monitoring as well as further training ML models in the cloud
[Animate]
The data is then converted by another module in a format that will be readable by the ML model, itself running on IoT Edge as a module.
[Animate]
The ML model scores the dynacards and if it detects an anomaly it can notify the SCADA system for it to adjust the pump operation or stop it. In case of alert the SCADA system will send an alert over SMS or email (if connectivity permits)
[Animate]
Once again, if the IoT Edge device is connected, it can send the alert up to the cloud itself to trigger maintenance processes automatically in LOB applications.
[Animate]
For the demo we are replaying readings from a real pump and I will be playing the local operator directly connected to the RTU to show you what’s happening