at the edge or in the cloud? - schneider electric

16
At the Edge or in the Cloud? Where to get the best data insights for improved efficiency #DigitalEvolution #InnovationDay #EcoStruxure

Upload: others

Post on 05-Apr-2022

0 views

Category:

Documents


0 download

TRANSCRIPT

At the Edge or in the Cloud?Where to get the best data insights for improved efficiency

#DigitalEvolution#InnovationDay#EcoStruxure

Page 2Schneider Electric |

Presenter

Fahd SaghirDigital Solutions Manager Industry BusinessSchneider Electric

Fahd Saghir

Edge Analytics – Definition

“Edge computing pushes applications, data and computing power (services) away fromcentralized points to the logical extremes of a network.” (Wikipedia)

Page 3Confidential Property of Schneider Electric |

Edge Analytics – Accelerated Decision MakingV

alu

e

Time

Data Captured

Data Historized

Data available at Enterprise Level

Action Taken

Data Latency

Analytics Latency

Decision Latency

Data loses impact

Dat

a St

ora

ge C

ost

Time

EDGE SCADA

HISTORIAN

MILLISECONDS SECONDS MINUTES HOURS

CLOUD

Page 4Confidential Property of Schneider Electric |

Typical Real Time Data Aggregation Architecture

SCADA

Data Aggregation occurring in MILLISECONDS

Data Aggregation dependent on communication backbone

Limitations due to Industrial Protocols (MODBUS vs. DNP3)

Data Aggregation dependent on HISTORIAN capabilities

Challenge sharing data from HISTORIAN with applications ENTERPRISE

EDGEPage 5Confidential Property of Schneider Electric |

Hardware Configuration, Logic Development

Ethernet, Serial, CanBUS

ARM Microprocessor

VxWorks, Quadros RTXC

4-20mA, 1-5V, Counters, Relays

Firmware Upgrade, Unified Abstraction Layer

Ethernet, GSM/GPRS, Wi-Fi, Bluetooth

IoT Framework – MQTT, AMQP

Intel Microprocessor

Ubuntu, Windows IoT

Data Management, Machine Learning Models

Edge Analytics – RTU vs. Gateway

Page 6Confidential Property of Schneider Electric |

Firmware Upgrade, Unified Abstraction Layer

Ethernet, GSM/GPRS, Wi-Fi, Bluetooth

IoT Framework – MQTT, AMQP

Intel Microprocessor

Ubuntu, Windows IoT

Data Management, Machine Learning Models

Edge Analytics – RTU vs. Gateway

Page 7Confidential Property of Schneider Electric |

Edge Analytics – Edge Enabled Architecture

EDGE

SCADA

ENTERPRISE

IoT Capable Device

DOCKER Container capable OS

Open Protocols

Security Standards

Page 8Confidential Property of Schneider Electric |

Edge Analytics – Building and Deploying a ML Model

HISTORICAL DATA DATA PROCESSING MODEL TRAINING

TRAINING DATA SET

MODEL VERIFIED MODEL

CLOUD OR ON-PREMISE

EDGE

VERIFIED MODEL

REAL TIME DATA RULE ENGINE ANALYTICS

Page 9Confidential Property of Schneider Electric |

Edge Analytics – RAW vs. FILTERED DATA

Page 10Confidential Property of Schneider Electric |

Page 11Confidential Property of Schneider Electric |

• Critical assets such as pumps can be managed more efficiently through an edge analytics solution

What is an example of an edge analytics application?

Centrifugal Pump SCADAPack

Reactive to changes in thepump’s behaviour

Automation tightly coupled toa specific type of pump

Limited availablecommunication protocols

RTU OS is only capable ofexecuting simple programs

To be able to manage the pumps moreeffectively, the analytics application shouldaim to:

Connect and integrate with the existing cloudand field infrastructure

Learn the different behaviours of the pump fromthe data generated

Identify and highlight patterns of irregular oranomalous behaviour

Communicate the analysis effectively tocontrollers and operators

Provide ahead of time feedback on the pump’sperformance

Some of the key requirements to achievethe goals:

Machine learning capable operating system torun complex programs

Support for open protocols to connect with agreater number of services

Data from the field, for the different stages ofthe pump’s performance.

Be an Industrial IoT capable device

Application

Page 12Confidential Property of Schneider Electric |

How does the unsupervised model work on sample data?

After pre-processing, data from the three variables (speed,

load, pressure) is converted into a two dimensional space. The

machine learning algorithm then clusters these points based

on their spatial distance. Colours represent the different

clusters the model has identified at each stage.

The 6 clusters identified by the model in relation to the data in its original form.

Clustering Progression Clustering of Original Data

700 --

600 --

500 --

400 --

300 --

200 --

100 –

0 --

-100 --

Application

Clusters Learnt by Model

How to interpret the results?

Clusters Applied to Original Data

700 --

600 --

500 --

400 --

300 --

200 --

100 –

0 --

-100 --

Page 13Confidential Property of Schneider Electric |

Application

Page 14Confidential Property of Schneider Electric |

How does the model work on real data?

Pump Data Analysis

Application

#DigitalEvolution #EcoStruxure #InnovationDay

Join the conversation

Visit our Innovation Hub.

©2018 Schneider Electric. All Rights Reserved. Schneider Electric | Life Is On and EcoStruxure are trademarks and the property of Schneider Electric SE,its subsidiaries, and affiliated companies. All other trademarks are the property of their respective owners.

Page 16Confidential Property of Schneider Electric |

Which application in your business would be most suited for Edge Analytics?

Round Table Questions

Do you have high quality Raw Data to improve your Machine Learning models?

How would you improve your Machine Learning models over time?