The document discusses using machine learning for industrial sensors to enable on-device intelligence by recognizing patterns in sensor data. It provides an example of a typical industrial sensor that collects large amounts of high-frequency data but discards 99% of it due to constraints. On-device machine learning can analyze the full sensor data to detect abnormal vibrations, temperatures, or patterns that could indicate faults. The document also outlines the steps to develop machine learning models from raw sensor data collected on edge devices.
3. 3
Typical industrial sensor in 2020
Vibration sensor (up to 1,000 times per second)
Temperature sensor
Water & explosion proof
Can send data >10km using 25 mW power
Processor capable of running >20 million
instructions per second
4. 4
But... what does it actually do?
Once an hour:
• Average motion (RMS)
• Peak motion
• Current temperature
5. 5
99% of sensor data is discarded due to
cost, bandwidth or power constraints.
https://www.mckinsey.com/~/media/McKinsey/Business%20Functions/McKinsey%20Digital/Our%20Insights/
The%20Internet%20of%20Things%20The%20value%20of%20digitizing%20the%20physical%20world/The-
Internet-of-things-Mapping-the-value-beyond-the-hype.ashx
8. 8
On-device intelligence is the only solution
Vibra&on pa+ern
heard that lead to fault
state in a weekTemperature
varies in a way that
I've never seen
before
Machine
oscillates different
than all other
machines in the
factory
12. 12
TinyML
Inspired by "OK Google"
Focus on inferencing, not training
Machine learning model is just a mathematical
function with lots of parameters
Accuracy vs. speed, reducing parameters, hardware-
optimized paths
Targeting battery-powered microcontrollers
Pete Warden
Neil Tan
13. https://www.flickr.com/photos/oceanyamaha/7091324605
13
What is it good for?
Recognizing sounds Detecting abnormal vibration
https://pixabay.com/photos/washing-machine-wash-cat-4120449/
Biosignal analysis
https://www.flickr.com/photos/sheishine/16696564563
Anything with messy, high-resolu3on sensor data
14. 14
Enabling new use cases
Sensor fusion
http://www.gierad.com/projects/supersensor/
15. Arm Cortex M3
Real Time Clock
Watchdog Timer
16b Timer
2 Channel ADC
FLASH
512KB
SRAM
256KB
8KB ROM
UART (2)
GPIOs (32)
PDM (4), I2S
Dual MAC DSP
96KB SRAM
64b RTOS Timer
Clock Generation
32kHz, 16.384 MHz XTAL
8MHz HFO
Power Control
PMU, POR, Buck
AHBLITEBusPeripheralBus
BRIDGE
DSP DMA Controller
SPI (3), I2C (3)
SERIAL INTERFACES
ANALOG
Temp Sensor
SYSTEM
Cortex-M plus DSP
(to make ML fast)
+
Continuous Voltage
and Frequency Scaling
=
Very efficient machine learning
algorithms on device
Eta Compute ECM3532
16. 16
ECM3532 runs inferences in mJ
Area Task Model Dataset Inferences
per second
Energy per
inference (mJ)
Vision Image
classification
Eta Compute – CNN
32x32
CIFAR10 20 0.15
Vision Person detection MobileNet V1
96x96
COCO 2 1.5
Vison Object counting MobileNet V1
256x256
COCO 1 3
Audio Command
recognition
Eta Compute - GRU Google 2 to 5 0.5
Motion Activity detection Eta Compute - CNN MotionSense
50 0.02
17. 17
ECM3532 AI Sensor Board
Available on
▪ 1.4 x1.4-inch board with sensors
⎯ECM3532
⎯2 x Microphones
⎯1 x Pressure/temp sensor
⎯1 x Accel/Gyro
⎯Serial Flash for data
⎯Battery socket
▪ Bluetooth on board from ABOV
▪ Extension for other types of RF
▪ Easy programming through UART
connector
▪ Debug port for advanced
development
23. Classification
What's happening right now?
Anomaly detection
Is this behavior out of the ordinary?
Forecasting
What will happen in the future?
23
3. Letting the computers figure it out
24. 24
Picking the right algorithm
Classification
Neural network
Anomaly detection
K-means clustering
Forecasting
Regression
27. 27
Get some hardware
Eta Compute AI Sensor Board
https://www.digikey.com/product-detail/en/eta-compute/ECM3532-ASBK/2742-ECM3532-ASBK-ND/12359709
Any smartphone
28. 28
Edge Impulse - TinyML as a service
Embedded or edge
compute deployment
options
Test
Edge Device Impulse
Dataset
Acquire valuable
training data securely
Test impulse with
real-time device
data flows
Enrich data and
generate ML process
Real sensors in real time
Open source SDK
Free for developers: edgeimpulse.com