connected everything 2018 brewnet-v2 · 2018. 7. 3. · intelligent cloud connected sensors for...

22
BREWNET Intelligent cloud connected sensors for economic small scale process optimisation Dr Nicholas Watson University of Nottingham Email: [email protected] Dr Syed Ali Raza Zaidi University of Leeds Email: [email protected] Network Plus: Industrial Systems in the Digital Age Conference 2018 – Looking Beyond Industry 4.0

Upload: others

Post on 07-Sep-2020

1 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Connected Everything 2018 BREWNET-v2 · 2018. 7. 3. · Intelligent cloud connected sensors for economic small scale process optimisation Dr Nicholas Watson University of Nottingham

BREWNETIntelligent cloud connected sensors for economic small scale 

process optimisation

Dr Nicholas WatsonUniversity of Nottingham

Email: [email protected]

Dr Syed Ali Raza ZaidiUniversity of Leeds

Email: [email protected]

Network Plus: Industrial Systems in  the Digital Age Conference 2018– Looking Beyond Industry 4.0

Page 2: Connected Everything 2018 BREWNET-v2 · 2018. 7. 3. · Intelligent cloud connected sensors for economic small scale process optimisation Dr Nicholas Watson University of Nottingham

BREWNET Introduction• Project motivation• Consortium• Project methodology

Ultrasonic sensor• Sensor design• Signal and data processing• Results from brewery• Impact acceleration account follow on work

Internet‐of‐Things

Summary• Lessons Learnt• Future Roadmap

Agen

da

• LPWAN• Sensor Node Design• Overall Solution• Open Issues

Page 3: Connected Everything 2018 BREWNET-v2 · 2018. 7. 3. · Intelligent cloud connected sensors for economic small scale process optimisation Dr Nicholas Watson University of Nottingham

Traditional processes still dominate

1700 Breweries within the UK.

Fermentation effects beer quality and downstream 

processes

Opportunity for fermentation PAT

New technologies must be:1) Cost effective  2) Simple to use 

Motivation

Page 4: Connected Everything 2018 BREWNET-v2 · 2018. 7. 3. · Intelligent cloud connected sensors for economic small scale process optimisation Dr Nicholas Watson University of Nottingham

Consortium

RobNigel Rix (KTN)Quality control

Nik Watson Josep Escrig Escrig Nengjia Ren

Consortium

Syed Ali Raza Maryam Hafeez

Page 5: Connected Everything 2018 BREWNET-v2 · 2018. 7. 3. · Intelligent cloud connected sensors for economic small scale process optimisation Dr Nicholas Watson University of Nottingham

1. Process sensors 2. Cloud empowered gateway3. User‐centeredpredictive analytics

Methodology

Page 6: Connected Everything 2018 BREWNET-v2 · 2018. 7. 3. · Intelligent cloud connected sensors for economic small scale process optimisation Dr Nicholas Watson University of Nottingham

Sensor Prototype

Probe with US and temperature sensor  Tri‐Clamp port for sensor housing 

2000 Litre Fermenter

Probe design

Initial sensor attached to laptop streaming data to the cloud

Page 7: Connected Everything 2018 BREWNET-v2 · 2018. 7. 3. · Intelligent cloud connected sensors for economic small scale process optimisation Dr Nicholas Watson University of Nottingham

Data analysis

1st Reflection 2nd Reflection

Delay between reflections

Amplitude 1streflection

Data analysis methods:• Ultrasonic velocity• Reflection amplitude • STD of reflection amplitude 

• Frequency content analysis• Machine learning  • Temperature calibrations

Page 8: Connected Everything 2018 BREWNET-v2 · 2018. 7. 3. · Intelligent cloud connected sensors for economic small scale process optimisation Dr Nicholas Watson University of Nottingham

Brewery results

Page 9: Connected Everything 2018 BREWNET-v2 · 2018. 7. 3. · Intelligent cloud connected sensors for economic small scale process optimisation Dr Nicholas Watson University of Nottingham

Challenges: 1) Acquiring sufficient labelled data for machine learning 2) Variability 

Brewery results

Page 10: Connected Everything 2018 BREWNET-v2 · 2018. 7. 3. · Intelligent cloud connected sensors for economic small scale process optimisation Dr Nicholas Watson University of Nottingham

Further DevelopmentEPSRC Impact Accelerating funding provided to further develop sensor technology• What is the most suitable data processing method• What is the cheapest cost a sensor can be built for

0.0

1.0

2.0

3.0

4.0

5.0

18/06/2018 19/06/2018 20/06/2018 21/06/2018 22/06/2018 23/06/2018

ABV (%

)

Date

ABV% Batch‐1 VS Batch‐2

batch‐1 ABV% (Tilt density)batch‐2 ABV% (Tilt density)batch‐1 ABV% (Manual density)batch‐2 ABV% (Manual density)

Page 11: Connected Everything 2018 BREWNET-v2 · 2018. 7. 3. · Intelligent cloud connected sensors for economic small scale process optimisation Dr Nicholas Watson University of Nottingham

IoT for Food & Drink Manufacturing

Page 12: Connected Everything 2018 BREWNET-v2 · 2018. 7. 3. · Intelligent cloud connected sensors for economic small scale process optimisation Dr Nicholas Watson University of Nottingham

IoT for Food and Beverage ManufacturingOperational EfficiencyPerformance analytics, Fault Diagnostics and Predictive Maintenance

Process EfficiencyProduction process optimization.Cost and waste reduction.Plant automation.

Food SafetyQuality control.Evidence for regulatory compliance.Traceability.

Supply Chain OptimizationTransparency.Logistic optimization.

Page 13: Connected Everything 2018 BREWNET-v2 · 2018. 7. 3. · Intelligent cloud connected sensors for economic small scale process optimisation Dr Nicholas Watson University of Nottingham

BREWNET Network Architecture

Time

Time

Time

Ultrasonic Transducer

Fermenter

Temperature & pH Sensor

Probe

Housing for Transducer

Microcontroller Unit

Communication Unit

Gateway

Cloud Server

Prediction Engine

Brewery Manager Smartphone Dashboard

Page 14: Connected Everything 2018 BREWNET-v2 · 2018. 7. 3. · Intelligent cloud connected sensors for economic small scale process optimisation Dr Nicholas Watson University of Nottingham

Connectivity: Sensor to Gateway

LoRA

Key Features

• Spectrum Occupancy: Unlicensed• Frequency Band: 868 MHz, 433 MHz and 900 MHz• LoRA Link Budget:

Receiver Sensitivity of -136 dBmTX Power of 14 dBm150 dB Link Budget, i.e., 22 km LoS and 2 km NLoS

Low costLow cost

Extended CoverageExtended Coverage UE Complexity ReductionUE Complexity Reduction

Battery LifetimeBattery Lifetime

Spread SpectrumSpread Spectrum

Low Duty CycleLow Duty Cycle

Claim:  It can operate successfully at ranges exceeding 15 km in suburban settings and more than 2 km in dense urban environments. It can achieve over 10‐year battery life and it can 

operate in networks comprising up to 1 million nodes.https://www.digikey.co.uk/en/articles/techzone/2016/nov/lorawan‐part‐1‐15‐km‐wireless‐10‐year‐battery‐life‐iot

Page 15: Connected Everything 2018 BREWNET-v2 · 2018. 7. 3. · Intelligent cloud connected sensors for economic small scale process optimisation Dr Nicholas Watson University of Nottingham

LoRA Measurements

100% 100% 99%

71%

46%

94%87%

79%

30%

10%

90%

61%

45%

3% 0%0%10%20%30%40%50%60%70%80%90%100%

200 400 600 800 1000

Packets R

eveive Rate 

Distance (m)

Multi‐Spreading Factor against Packets Receive Rate  

SF12SF9

‐10%0%10%20%30%40%50%60%70%80%90%100%110%

200 400 600 800 1000

Packets Re

veive Ra

te

Distance (m)

Multi‐Spreading Factor against Packets Receive Rate  

SF12SF9

Page 16: Connected Everything 2018 BREWNET-v2 · 2018. 7. 3. · Intelligent cloud connected sensors for economic small scale process optimisation Dr Nicholas Watson University of Nottingham

LoRA Coverage

• Theoretical Computation of Coverage• Height factor for Gateway

• Tall Building/Mast• Drone based channel sounding

• Antenna selection• Directional antenna may help

• Protocol Level Optimization• Frequency Hopping• Packet Data reduction and Long SF• Output power configuration• Chain Architecture

Page 17: Connected Everything 2018 BREWNET-v2 · 2018. 7. 3. · Intelligent cloud connected sensors for economic small scale process optimisation Dr Nicholas Watson University of Nottingham

LoRA Coverage

4km 20% PRR

Page 18: Connected Everything 2018 BREWNET-v2 · 2018. 7. 3. · Intelligent cloud connected sensors for economic small scale process optimisation Dr Nicholas Watson University of Nottingham

Sensing Node Design

Ultrasonic Signal Processing

MCU & CommunicationSPI

SPITemperature RTD

Page 19: Connected Everything 2018 BREWNET-v2 · 2018. 7. 3. · Intelligent cloud connected sensors for economic small scale process optimisation Dr Nicholas Watson University of Nottingham

Overall Design

Power Management

RTD

LoPY

Amplifier ASIC

US PROBE

Application Server

MULTITECH CONDUITGateway

MQTT Broker

MQTT Client

Subscribe

LoRA/WiFi

MQTT ClientDB Endpoint

Publish

ML Engine

Dashboard

Node 1…n

MQTT Client API

Visualization API

Page 20: Connected Everything 2018 BREWNET-v2 · 2018. 7. 3. · Intelligent cloud connected sensors for economic small scale process optimisation Dr Nicholas Watson University of Nottingham

Research Directions

Chicken & Egg Problem• Data vs. ConnectivityIntelligent Networking• Self-organization of LPWANContainerization• Cloud and Fog ComputingFuture Proof Design• Hardware Abstraction• Protocol Abstraction

Bor, Martin and Roedig, Utz (2017) LoRa Transmission Parameter Selection. In: 2017 13th International Conference on Distributed Computing in Sensor Systems (DCOSS) :. IEEE, pp. 27‐34. ISBN 9781538639917

Page 21: Connected Everything 2018 BREWNET-v2 · 2018. 7. 3. · Intelligent cloud connected sensors for economic small scale process optimisation Dr Nicholas Watson University of Nottingham

Dissemination & Future Plans

Invited talks:• Appetite for Engineering (October 2017)• UK‐China 2nd Food Detection Forum (November 2018)• Mondelez Global Analytical Forum (December 2017)• KTN 47th Intelligent Sensing Program – Sensing in Manufacturing (March 2018)

• The Future of Food Safety Conference (May 2018)• Two publications in preparation (Sensor design and analytics & IoT)

Future funding plans:• Further BREWNET work (Multiuse sensors, integrating weather data)

• Sensors in harsh environments• Low cost sensors and IoT for agri‐tech/precision farming

Page 22: Connected Everything 2018 BREWNET-v2 · 2018. 7. 3. · Intelligent cloud connected sensors for economic small scale process optimisation Dr Nicholas Watson University of Nottingham

Summary

Lessons Learnt:• Combining technologies is challenging (compatibility, robustness)• Machine learning maybe not the most suitable approach for products with 

acceptable large variability (e.g. craft beer)• Need to publish standardized framework for IoT projects• Need for Self‐organization & hardware abstraction

Summary:• Project showed that sensors can be used to determine the alcohol content 

in a fermenter and be used to predict optimal end point• There are various different data processing methods, ultrasonic velocity is 

the most common but also the most expensive• Commercial prototype sensor under development• IoT is key enabler for manufacturing, however future proofing the IoT

platforms is non‐trivial and design framework itself is evolving.