AWS Manufacturing Day Philadelphia-Boston-April 2019

© 2019, Amazon Web Services, Inc. or its Affiliates. All rights reserved.© 2019, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
AWS in Manufacturing
Douglas Bellin, WW BD Lead Manufacturing
bellin@amazon.com

© 2019, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
Improving
manufacturing operations
is everything!

Manufacturing
Industry Drivers
Emerging Markets
Complex, Dynamic Value Chains
Security, Cyber & Physical
Truth in Data
Demanding Customers
Converging Technologies
Ubiquitous Connectivity
Traceability & Transparency,
Brand & Reputation
Workforce
Shortage of expertise, Knowledge
Transfer (Attrition), Tech Savvy

Product-as-a-
Service
Digitally
“Executed”
Manufacturing
Data is the
New Oil
Connected
Products
Manufacturing Industry Trends
Sustainability

Responding to
Business Demands
24 × 7 × 365
Operations
Asset
Lifecycle
Enabling the
Workforce
Protecting and
Security IP
Unleashing Data and
Bringing Insights
Global & Regional
Collaboration
Cost
Reduction
Manufacturing Industry Challenges

If you knew the state of every thing in the world,
and could reason on top of the data:
What problems would you solve?

DIGITAL MANUFACTURING IS A IMPORTANT
ELEMENTOF THE DIGITAL SUPPLYCHAIN
A connected supply chain generates demand signals that are used to manage the supply (manufacturing process) creating a tightly-knit
‘sense and respond’ supply chain
SUPPLIER MANUFACTURING DISTRIBUTION
The signal is used to manage
downstream activities.
In the factory, this is used to plan the
production run.
Customer and consumer
provides demand signal
Supplier
Network
Raw
Materials
Inventory
Production Phases Shipped Out Finished
Good Stocks
Transit
Stocks
Last Mile
Fulfilment
SALES

What is preventing the industry from moving ahead?
AI/ML & BIG
DATA
expertise is
rare
Building and
scaling AI/ML &
BIG DATA
technology is hard
Deploying and operating
models/solutions in
production is time-
consuming and
expensive
A lack of cost-effective,
easy-to-use, and
scalable AI/ML & Big
Data services

…realizing value from Big Data is challenging
What’s
holding
you back
from
using Big
Data?1:
Unable to link data
together
96%
of Manufacturers
state data is not used
Data collected too
infrequently
Data difficult
to access
39%
of Manufacturers do not
regularly collect data
66%
of Manufacturers find
data is difficult to access
Without the right platform, data insights remain elusive

REVENUE
GROWTH
OPERATIONAL
OVERHEAD
Empowered
Sales Teams
Increased
Efficiency
Intelligent Decision
Making
Products that Get
Better with Time
Better Relationship
with Customers
Data Driven
Discipline

AWS in Manufacturing

About Georgia-Pacific
Georgia-Pacific, owned by Koch
Industries, is an American wood
products, pulp, and paper company
based in Atlanta, Georgia. The
organization is one of the world’s
largest manufacturers and distributors
of pulp, towel and tissue paper and
dispensers, packaging, and wood and
gypsum building products.
Industry: Manufacturing
Headquarters: Atlanta GA
Employees: 35K
Website: www.gp.com
We are using AWS data analysis technologies to predict ... precisely
how fast converting lines should run to avoid tearing. By reducing
paper tears, we have increased profits by millions of dollars for one
production line.
Steve Bakalar, VP of IT & Digital Transformation
“
”
Challenges Solution Benefits
Georgia-Pacific wanted to gain new
insights from manufacturing data
collected at paper production plants,
but it relied on disparate sources to
analyze data on material quality,
moisture, temperature, and other
features.
Georgia-Pacific uses an AWS
advanced analytics solution,
featuring Amazon Kinesis and
Amazon SageMaker, to
collect and analyze data from
equipment at manufacturing
facilities across North America.
• Boosts profits by millions of dollars
• Predicts equipment failure 60-90 days
in advance
• Runs more production lines in a
predictable manner
• Ensures highest quality products
Case Study: Paper & Building
Product Manufacturing

About SKF
Founded in 1907, SKF is the world’s
largest bearing manufacturer. The
company also manufacturers seals,
lubrication and smart lubrication systems,
maintenance products, mechatronics
products, power transmission products
and condition monitoring systems.
SKF has a large distributor network with
17K distributor locations spanning 130
countries.
Industry: Manufacturing
Headquarters: Sweden
Employees: 46K
Website: www.skf.com
I see a lot of speed of innovation coming from AWS, and we are
confident that this the platform we are going forward with.
Johan Tommervik, CIO
“
”
Challenge Solution Benefits
• Move beyond selling only products
to a “Rotating Equipment
Performance” model
• Ensuring automatic lubrication of
bearings to maximize performance
• Gather data from customers to
improve product design
• Add new placement part revenue
Connected System 24 single point
lubricator feeding Data Lake with
Amazon S3 to ingest and analyze
data; AWS ML to analyze products
in the field; AWS Database to
manage large amounts of complex
vibration and equipment data;
AWS IoT and Lambda to speed
time to market and lower costs.
• Revenue expansion beyond ship-and-
forget to a services enhanced model
• Grow sales even if raw product
shipment numbers do not increase
• Innovate faster with lower costs
• Focus on value for customers instead
of managing IT resources
Case Study: Smart Bearing and
Smart Manufacturing

New Format
Manufacturing Functional Areas
Marketing and Sales
Market Analysis
Portfolio Planning
Demand Creation
Selling Products /
Services / Outcomes
Marketing & Sales
Analytics
Brand Management
Product &
Production Design
Refine Plan and Define
Product/Services
Design and Release
Product/Services
Validate Product/Service
Design to Requirements
Prepare and Validate
Production Environment
Prepare and Validate
In-Service Environment
Manufacturing
Operations
Plan and Schedule
Production
Manage Supply and
Inbound Logistics
Make or Assemble
Product
Fulfill Production Order
Supply Chain
Develop Supply Base
Manage Production
Support and Materials
Manage Warehousing
and Distribution
Conduct Supplier
Aftermarket Support
Service Chain
Manage Technical
Maintenance and
Engineering Information
Plan Maintenance
Perform
Maintenance/Repair
Manage Services Supply
Manage Service
Contracts and Warranty
Analyze and Report
Service and Quality Data
Mitigate Risk in SLAs
Business
Operations
Provide Administrative
Services
Manage Business
Provide Human
Resource Support
Provide Financial
Support
Manage Public Affairs
Provide Legal Counsel
Provide Information
Technology and
Communications

Marketing and Sales
Product &
Production Design
Manufacturing
Operations
Supply Chain Service Chain
Business
Operations
Manufacturing Applications Map
Enterprise Resource Planning (ERP)
Enterprise
Resource
Planning (ERP)
Customer
Relationship
Management
(CRM)
Product Lifecycle
Management (PLM)
Customer Relationship
Management
(CRM)
Enterprise Asset
Management
(EAM)
Supply Chain Management (SCM)
Manufacturing Operations
Management (MOM)

Marketing and
Sales
SMART PRODUCT
SMART FACTORY
Solution Area
DATA LAKE ON AWS
Use Cases AWS Services
Amazon Forecast
Amazon Sagemaker
• ML for Demand Forecasting
• ML for Campaign Generation and
Execution
• Opportunity Scoring
• Pricing/Trade-spend modeling
• Suggested Next Steps
Amazon EC2
Amazon EBS
Amazon IAM

Product &
Production Design
PRODUCT DESIGN
Solution Area
DATA LAKE ON AWS
Amazon EC2
Amazon EFS
• HPC workloads for EDA, CFD, FEA, Crash
Simulation.
• Simulation model optimization with AIML with
live performance model as baseline
• Comparison of virtual sensor data with
physical sensor data
• Comparison of test performance data with
product performance data from field
• Condition monitoring of physical assets tested
• Failure prediction for physical assets tested
Amazon S3
Amazon Glacier
Amazon EBS
Amazon Appstream

Manufacturing
Operations
SMART FACTORY
Solution Area
DATA LAKE ON AWS
• Production/Process optimization
• Preventive / predictive maintenance for machines
• Additive manufacturing with AI / ML support
• Condition monitoring of machine tools augmented by ML
• Worker safety
• Knowledge Transfer
• Digital Twin
• Plant as a service
• Computer vision for Quality
• Early machine tool replacement reduction
• Scrap reduction & quality optimization
• Power optimization
• Fleet management
• Integration with SAP & MRP & MES systems (logistics /
inventory / JIT use cases)
• Streamlining logistics
Amazon Sagemaker
AWS IoT AWS Outposts
AWS SiteWise AWS IoT Analytics
AWS IoT Events AWS Greengrass
Amazon Timestream

Supply Chain
SUPPLY CHAIN VISIBILITY
SUPPLY CHAIN FORECAST
Solution Area
DATA LAKE ON AWS
Amazon Forecast
Amazon Sagemaker
• ML for Demand Forecasting
• ML for Supply Side Availability
• Improved Inventory Management
Amazon API Gateway

SMART MANUFACTURING
This leads to better efficiencies and quality,
and provides competitive advantage in
manufacturing. Smart Manufacturing solutions integrate different layers on the ISA 95 standard architecture.
8
To achieve the benefits of Digitization in its Manufacturing operations, companies should implement a comprehensive solution, that covers
Level 1-4 of the ISA 95 architecture.
Smart Manufacturing, also referred to as
Industry 4.0 or Digital Factory, capability
includes data generation (via sensors,
actuators), data collection, aggregation,
visualization and analytics.
Factory operations can be improved via the
availability of real time information, that
improves the machine operations and
enables line operators and factory leaders to
make immediate, data driven decisions.

Problem with current model
Functional Layer Data
ERP 10%
MES 25%
SCADA 35%
PLC 50%
Sensors 100%

AWS Manufacturing Reference Architecture
Greengrass
Edge/GW
S3
Data Lake
Kinesis
MES
Factory Machines
ML
Inference
IoT Core
Sage Maker
ML
QuickSight
Business
Intelligence
Athena
Historian
Storage GW
EMR
EBS EC2 Batch AppStreamEBS EC2
E&D Workloads
(PLM/HPC/CAE)
Enterprise Workloads
(SAP ERP/CRM)DMS RDS
Local Servers
Redshift
Data Warehouse
DataIngestion
API
SiteWise
Snowball Edge
Smart Products
DynamoDB Lambda
IoT Core
Amazon Forecast
Plant Maint. Planning
Sample Business Functions
Greengrass
Connectors
IoT Analytics
Timestream
Outposts
IoT Events
EC2
Lambda
Business Logic

• Factory visit / Amazon FC visit
• DI workshops
• Working Backwards workshops
• PR/FAQ
Next Steps

Thank you.

Bill Durham
Principal IoT Sales Specialist
Securely Connecting and
Managing Industrial IoT
Devices at Scale

Industrial IoT Market
Focused on next-generation manufacturing that generates a convergence
between industry, business, and internal functions and processes
Industrie 4.0
in Germany
Society 5.0
in Japan
Made in China
2025
Trends
↓ ↓ ↓ ↓

The Industrial Revolution
1st 2nd 3rd 4th

Industrie 4.0
What’s changed?
• Increasing need to optimize
and predict system performance
• Need for geographically scattered assets
that function together as a system
• Scalable systems that support a growing volume
of instrumentation and data accessibility
• Improve security of devices and systems
• Integrate multiple protocols and standards
• Solutions require a mix of legacy and newer equipment
including intelligent sensors and actuators

Challenges
Security Downtime Legacy Equipment

Operations (OT)Enterprise (IT)
Challenge: Brownfield Environments
IT Systems
CRM
Asset Management
ERP
Supply Chain
Finance
Maintenance
Compliance
SCADA,
DCS, etc.
Various Protocols

Opportunities
IoT Drives Manufacturing Innovation
Event-based digital monitoring for
optimized operations, stock
handling, improve OEE, and reduce
MTBR
Automated alerting connected to ERP,
Asset and operational services to
create fully automated, data driven
operations
Data logging and analytics
platform. Integrated data
types reduce MTBF and
optimize productivity

AWS Industrial IoT Technology Stack
TLS
CONNECTEDINDUSTRIALAPI
INDUSTRIAL PLANT
AMAZON
FREE RTOS
CONNECTIVITY
PROTOCOLS MQTT MQTT + WebSockets
INGESTION [AWS IoT] DEVICE GATEWAY REGISTRY RULES ENGINE DEVICE MANAGEMENT
DATA SERVICES AMAZON S3 AMAZON DYNAMODB AMAZON RDS AMAZON REDSHIFT
PRESENTATION
AMAZONAPIGATEWAYAMAZONCOGNITO
CONNECTEDINDUSTRIAL
PLATFORM
WEBMOBILE
SECURECOMMUNICATION
APPLICATION SERVICES AWS LAMBDA AWS SNS/SQS AMAZON QUICKSIGHT AMAZON COGNITO
AWSIDENTITY
ANDACCESS
MANAGEMENT
AWS
Device SDK’s
ANALYTIC SERVICES AWS IoT ANALYTICS AMAZON SAGEMAKER AMAZON KINESIS AMAZON ATHENA
AWS
GREENGRASS
ML INFERENCE
AWSIoT
DEVICE
DEFENDER
CELLULAR/FIXED
AWS
GREENGRASS
HTTP

Predictive
Maintenance
Predictive
Quality
Asset Condition
Monitoring
Popular Industrial IoT Use Cases

L2 AB CIP Protocol/Modbus/OPC/Other Industrial Protocols
ISA 95 & ISA 99 Industrial Edge Architecture
L5 Cloud
L4 ERP/SAP
L3 MES
L1 PLC
L0 Industrial Equipment
Greengrass on
Industrial Gateway
AWS IoT
MQTT
Telemetry channel
(MQTT)
File channel
(HTTPS)

Use Case
Predictive Maintenance
Understand current health of equipment and predict machine
failure before business operations is impacted
• Ingest sensor data from PLC’s, MES and Vision Systems
• Improve performance in the factory by monitoring OEE
• Monitor condition of factory machines
• Detect equipment anomalies Kinesis with and trigger notifications with SNS
• Predict equipment failure using advanced analytics with SageMaker
• Run ML models at the edge with Greengrass ML inference
• Filter data at the edge so that all the data does not need to be sent to the cloud
• Visualize and report on Equipment Time to Failure and Predictive Maintenance using dashboards

Predictive Maintenance Architecture
Pinpoint
Greengrass
IoT Rule (all data)
S3 Data Lake
Amazon Kinesis
Firehose
MES/SCADA
Protocol
conversion
Email
SMS
Factory Machines
Vision
Snowball
Kinesis
Analytics
Protocol
conversion
ML
Inference
AWS IoT/Greengrass/
Device Management/
Device Defender
Sage MakerML Models
Amazon QuickSight
Amazon Kinesis Streams
Kinesis
Firehose
IoT Anomaly
Data Repository
Amazon
Athena
Amazon
Athena
IoTRule(alerts)
Realtimeand
HistoricalVisualization
CloudWatch
Cognito
CloudTrail
config
IoT Cert
IAM

Predictive Maintenance Architecture
with AWS IoT Analytics
Pinpoint
Greengrass
IoT Rule (all data)
MES/SCADA
Protocol
conversion
Email
SMS
Factory Machines
Vision
Protocol
conversion
ML
Inference
AWS IoT/Greengrass/
Device Management/
Device Defender
Sage MakerML Models
Amazon QuickSight
AWS IoT AnalyticsIoTRule(alerts)
CloudWatch
Cognito
CloudTrail
config
IoT Cert
IAM
Jupyter Notebook
Anomalies

Problem
An Oil and Gas company had the inability
access their IoT data. Other business units
within the enterprise owned and controlled the
assets in the field and while many had IoT
data, they were not in a position to have that
data leave their on-premise environment.
Solution
By using AWS IoT, this customer is able to
preprocess the IoT data coming from their field
assets, enrich that data with various internal
and external data sources, and provide a time-
series optimized data store. This empowers
their in-house data science team to build and
train machine learning models on top of data
sets derived from the data store.
Impact
The customer’s goals were to validate their
hypothesis that IoT data, with proper analysis,
provides meaningful value to the enterprise. In
the near future, the customer expects to take
the anomaly detection models they authored
and test them for deployment at the edge.

Use Case
Predictive Quality
Quickly pinpoint product quality issues related
factory output, rather than equipment performance
• Ingest industrial sensor data from PLC’s, MES, and Vision Systems
• Ingest quality data (Inspection Images) into S3
• Improve product quality and uptime in the factory by monitoring OEE
• Monitor quality of finished products using Vision Systems
• Use streaming analytics to detect quality anomalies and trigger notifications
• Use advanced analytics to analyze product quality images to detect and predict quality issues
• Analyze product quality at the edge using Greengrass ML Inference
• Visualize and report on product quality using dashboards

Predictive Quality Architecture
Pinpoint
Greengrass
IoT Rule (all data)
S3 Data Lake
Amazon Kinesis
Firehose
MES/SCADA
Protocol
conversion
Email
SMS
Factory Machines
Vision
Snowball
Kinesis
Analytics
Protocol
conversion
AWS IoT/Greengrass/
Device Management/
Device Defender
Sage MakerML Models
Amazon QuickSight
Amazon
Kinesis Firehose
IoT Anomaly
Data Repository
Amazon
Athena
Amazon
Athena
IoTRule(alerts)
Realtimeand
CloudWatch
Cognito
CloudTrail
config
IoT Cert
IAM
S3
Analytics short term
data repository

Predictive Quality Architecture
Pinpoint
Greengrass
IoT Rule (all data)
MES/SCADA
Protocol
conversion
Email
SMS
Factory Machines
Vision
Protocol
conversion
ML
Inference
AWS IoT/Greengrass/
Device Management/
Device Defender
ML Models
Amazon QuickSight
CloudWatch
Cognito
CloudTrail
config
IoT Cert
IAM
Jupyter Notebook
Vision system
images
Sage Master/
Jupyter Notebook
Anomalies

Problem
Valmet delivers technology and automation
with multiple dependent processes running
in parallel. Data analytics is needed to
optimize Valmet’s customers’ processes.
Solution
Valmet is building a new digital twin
capability to allow paper mill operators view
equipment and process data during
production runs. AWS IoT Analytics is at the
core of this solution training ML models for
paper quality forecasting and scheduling
metrics generation for digital twin view-
generation.
Impact
AWS IoT Analytics allows Valmet to
combine historical models of equipment
performance with live data from current
operations to glean insights that help them to
further provide solutions that enable their
customers to produce paper with lower costs
and optimum quality.

Use Case
Asset Condition Monitoring
Monitor and scale industrial equipment and understand asset
condition for one or more monitored parameters of assets
• Ingest sensor data from PLC’s, MES, and Vision Systems
• Improve performance in the factory by monitoring OEE
• Monitor condition of factory equipment through sensor data—temperature, vibration, error codes, etc.
• Filter data at the edge so that all the data does not need to be sent to the cloud
• Use streaming analytics to detect condition anomalies and trigger notifications
• Build ML Models in SageMaker to detect and predict equipment condition deterioration and failure
• Analyze Vibration and other sensor data at the edge with Greengrass ML
• Visualize and report on equipment condition using dashboards

Condition Monitoring Architecture
Pinpoint
Greengrass
IoT Rule (all data)
S3 Data Lake
Amazon Kinesis
Firehose
MES/SCADA
Protocol
conversion
Email
SMS
Factory Machines
Vision
Snowball
Kinesis
Analytics
Protocol
conversion
AWS IoT/Greengrass/
Device Management/
Device Defender
Amazon QuickSight
Kinesis
Firehose
IoT Anomaly
Data Repository
Amazon
Athena
Amazon
Athena
IoTRule(alerts)
Realtimeand
CloudWatch
Cognito
CloudTrail
config
IoT Cert
IAM

Condition Monitoring Architecture
Pinpoint
Greengrass
IoT Rule (all data)
MES/SCADA
Protocol
conversion
Email
SMS
Factory Machines
Vision
Protocol
conversion
AWS IoT/Greengrass/
Device Management/
Device Defender
Amazon QuickSight
CloudWatch
Cognito
CloudTrail
config
IoT Cert
IAM
Jupyter Notebook
Anomalies

Problem
A Global Mining Company was looking to
measure rough roads on mines as potholes can
cause damage to mining equipment that is
extremely expensive. The Mining Company was
looking to understand the degradation of mining
equipment, such as Excavators.
Solution
The Global Mining Company turned to AWS to
place gateways and vibration sensors on trucks.
The customer collects data from equipment,
which allows them to identify potholes and other
problems on mining routes that can contribute to
equipment degradation.
Impact
AWS IoT allows the Global Mining Company
to continuously monitor equipment status, health,
and performance to detect issues in real-time. It
also helps the company detect road issues and
identify equipment degradation over time to
minimize unexpected downtime.

Industrial IoT Use Cases and Solutions
Adds real time contextualization to the
sensor payload from external sources
Provides tools to identify correlation
factors and to predict device failure
Visualizes the anomaly with your
devices for you to proactively remediate
issues

Industrie 4.0 Tenets and Why AWS?
Interoperability Local AWS Lambda with AWS Greengrass to integrate protocol other than MQTT & HTTP
Virtualization AWS IoT Shadows work in both local AWS Greengrass and the AWS Cloud with thing types and custom attributes
Decentralization Leverage 11 AWS Regions to subscribe to AWS IoT topics using selective rules
Real-Time Capability
AWS Greengrass achieves lower latency with local devices to support
critical automated decision making for mission critical industrial use cases
Service-Orientation
Multiple layers of AWS Lambda functions addressing increasingly deeper layers that can
be orchestrated with AWS Step Functions invoked by AWS IoT or Amazon API Gateway
Modularity
AWS Greengrass for a hybrid end-to-end process with local real-time
processing and cloud agility for stream processing, analysis and archival
Security
AWS IoT Device Defender secures your fleet of industrial devices by continuously auditing
the security policies associated with your devices to ensure they are secure at all times
Analytics and Insight
AWS IoT Analytics cleans, filters, transforms, and enriches IoT data before
storing it in a time-series data store for analysis and advanced analytics
Lifecycle Device Management
AWS IoT Device Management makes it easy to securely onboard, organize,
monitor, and remotely manage industrial devices at any scale

What Sets AWS Industrial IoT apart?
Industrial IoT Vision Reference architectures built for popular industrial use cases so you can quickly get started
Service Breadth
and Depth
AWS IoT services allow you to gather data from, run sophisticated analytics on,
and take actions in real-time on your diverse fleet of IoT devices from edge to the cloud
Security
Built-in device authentication and authorization to keep your IoT solutions secure. Continuously audit policies associated with your
devices, monitor your device fleet for abnormal behavior, and receive alerts if something doesn’t look right. You can even take corrective
actions
Scalability Reliably scale to billions of devices and trillions of messages
IoT Analytics
and Machine Learning
Sophisticated analytics including pre-built machine learning models for common IoT use cases, and machine learning inference at the
edge capabilities
Partner Network
and Community
Rich ecosystem of technology and consulting partners such as Intel, TI, Microchip, Bsquare, C3 IoT, Splunk, and Accenture
Trusted and Proven Customers such as Pentair and Kempii have achieved business outcomes such as increased revenue and faster time to market

Thank You!

O N I C A . C O M PREMIER CONSULTING PARTNER
Next Gen Solutions with IoT, Serverless
Applications & Machine Learning
Tolga Tarhan
CTO
Onica

If you knew the state of everything and
could reason on top of that data…
What problems would you solve?

© 2 0 1 9 O N I C AO N I C A . C O M
Our customers problem solve with IoT
Predictive
maintenance
Transform
personal healthcare
Smart buildings
& city systems
Fleet management Energy efficiency New service
creation

© 2 0 1 9 O N I C AO N I C A . C O M
Impacting the bottom line with IoT
Revenue growth
IoT data drives business
growth
Operational efficiency
IoT data decreases OpEx
New services &
business models
Products that get better
with time
Better relationship
with customers
Increased
efficiency
Intelligent decision
making
Data driven
discipline

Case Study:
Resource Extraction Co.
Asset Monitoring with IoT

© 2 0 1 9 O N I C AO N I C A . C O M
The Business Problem
Resource Extraction Co. needed a way
to collect data on the conditions at the
end of the oil well

© 2 0 1 9 O N I C AO N I C A . C O M
The challenge
- Deep underground, not easily accessible
- Harsh conditions for any type of electrical hardware
- Device at the bottom would not be-able to talk to
anything above ground

© 2 0 1 9 O N I C AO N I C A . C O M
The solution
- Designed special sensors that go
all the way down to the oil
- Built a gateway device w/
Bluetooth 5 that sits next to the
well and communicates to the
sensor at the bottom of the bridle
- Architected the cloud side
infrastructure so the gateway
device could communicate with
AWS IoT

© 2 0 1 9 O N I C AO N I C A . C O M
The Architecture
AWS Cloud
sens
or
MQTT over
TLS
AWS Lambda
Function
Web UI
AWS Lambda Function
IoT Rules
Business Logic
Gateway
Amazon Simple
Storage Service
(AmazonS3)
AWS IoT
Greengrass

© 2 0 1 9 O N I C AO N I C A . C O M
The result
Safer work environment
More efficient drilling, more $$$ saved and earned
Solution is globally scalable
Real-time insights
Less guess work if a problem or anomaly arises

Case Study:
Chemical Dispensing Co.
Predictive Maintenance with IoT, Serverless App & Machine Learning

Chemical Dispensing Co. was filling their
tanks too often and making unnecessary
many trips to tanks that were still 75% full,
resulting in costly excess labor

© 2 0 1 9 O N I C AO N I C A . C O M
The challenge
They wanted to optimize operations through
• Reducing frequency of deliveries
• Mapping the shortest possible delivery routes
All without ever having the risk of an empty tank

© 2 0 1 9 O N I C AO N I C A . C O M
The solution
• Designed fill sensors to fit inside the
tank to monitor the liquid level at
1-hour intervals
• Optimized delivery schedules and
routes using algorithms that ingest
data from IoT devices.
‒ Screen IoT data for anomalies
‒ Predict supply depletion at each
site
‒ Algorithmically prioritize delivery
schedule
Legend
Red: 20% Full
Orange: 30% Full
Yellow: 40% Full
Green: 50% Full
1
2
4
6
7
9
103
5
8

© 2 0 1 9 O N I C AO N I C A . C O M
The Architecture
Amazon SageMaker
Endpoint
Amazon S3
AWS Lambda
Amazon
DynamoDB
Amazon API
Gateway
API´s for
interacting
with model and
providing
feedback
Automated Deployment Pipeline
Serverless
Framework
Git Jenkins
Amazon Lex
Existing AWS
Infrastructure
IoT Devices
Chat
Mobile Apps

© 2 0 1 9 O N I C AO N I C A . C O M
Workstream in Amazon SageMaker

© 2 0 1 9 O N I C AO N I C A . C O M
The result
More efficient workforce through route
optimization
Cut their operations costs in half
Reduced staffing needs
100% customer satisfaction

Case Study:
Medical Device Manufacturer
(MDM)
New Revenue Streams: Smart Products

©2019 ONICAONICA.COM
MDM: a global Life sciences company that
seeks to manufacture a new line of cloud
connected endoscopic surgical cameras,
with an AI/ML workflow to correlate surgery
length with positive outcomes.

The challenge
- Store patient data in a HIPAA-compliant manner.
- Synchronize data in real-time between camera
systems, in-hospital server products, EMR systems, the
cloud, and mobile applications.
- Liberate patient information and endoscopic
images/video from the hospital and into the cloud.
- Offer a hybrid approach to connecting physical-world
surgical centers with cloud-based services.

The Solution
• Designed and deployed systemwide cloud side
serverless architecture for devices & iOS application
• Software design for iOS
‒ User experience design
‒ Software engineering leveraging various DevOps practices
‒ Built and deployed in a fully-automated manner using
AWS CloudFormation and Chef

The Solution
Elastic
Load
Balancin
g (ELB)

The result
Delivered IoT-enabled endoscopic cameras
Supports thousands of physicians across the globe
Provides with real-time, remote access to critical surgical
information
Implemented industry-first AI/ML workflow to correlate surgery
length with positive outcomes
Cloud also enables modern collaboration and reinforces
MDM’s leadership role as a technology innovator in its industry.

Nobody buys “IoT”

© 2 0 1 9 O N I C AO N I C A . C O M
IoT solutions are complex & multidimensional
Connecting,
communicating,
securing
Devices &
sensors
Infrastructure
providers,
building blocks
Connectivity &
infrastructure
Incisive,
actionable,
predictive
Analytics &
insights
Engage,
empower,
delight
Applications &
services
Business
transformation,
cultural change
Change
management

© 2 0 1 9 O N I C AO N I C A . C O M
Devices & Sensors
Most of the time, the device you need
doesn’t exist. There are too many
permutations.
• Sensor/actuator
• Power source, battery management
• Connectivity
• Environmental considerations
• Firmware development tools/environment
• Remote management, OTA updates
Devices &
sensors

© 2 0 1 9 O N I C AO N I C A . C O M
It always starts with hot glue and tape
Week 3 – prototyping Week 6 - PoC is ready for
testing

© 2 0 1 9 O N I C AO N I C A . C O M
Connectivity Options
Connectivity Technology Price
Wi-Fi 802.11b/g/n Free
Cellular 4G LTE on VZW and AT&T $
Satellite BGAN on Inmarsat $$$
LoRa LoRa Free (*)
5/1/2019

“
GET THERE FASTER

“IoTanium Rapid Prototype Board:
• Get started immediately with in-stock, fully
assembled hardware
• Begin collecting sensor data on Day 1
• Multiple pre-integrated connectivity options,
including Wi-Fi, BLE & LTE
• Exposed contacts for easy prototyping

© 2 0 1 9 O N I C AO N I C A . C O M
Connectivity & Infrastructure
The first step is to get data off the
device, and into the cloud.
• Secure connectivity
• Gateways vs WiFi-enabled sensors
• Reliable ingestion
• OTA updates
• Provisioning & management
Connectivity &
infrastructure

© 2 0 1 9 O N I C AO N I C A . C O M
Data & Analytics
IoT applications generate a lot of
data. Data is currency.
• Store as much data as you can
• Use inexpensive, object-based,
storage
• Automate your ETL & analysis
pipelines
• Avoid schema lock-in
Analytics &
insights

© 2 0 1 9 O N I C AO N I C A . C O M
Applications!
IoT isn’t always about analyzing data. There
are normally bespoke applications
involved.
• Adds an element of application
development to IoT projects
• These applications are usually event-
driven
• Use serverless technologies
• Automatic scaling & self healing
• Pay as you go consumption model
Applications &
services

© 2 0 1 9 O N I C AO N I C A . C O M
Jump start with IoTanium Platform

© 2 0 1 9 O N I C AO N I C A . C O M
People & process
The best IoT solutions result in new
business models
• Move from selling products to services
• Capture subscription revenue
• New processes, training, and tools are
required
Change
management

© 2 0 1 9 O N I C AO N I C A . C O M
How can an IoT Partner help?
IoTanium Hardware &
Hardware Engineering
Connectivity & Infrastructure
with IoTanium Platform
IoTanium Analytics,
Data Engineering &
Data Science
Onica Professional Services

© 2 0 1 9 O N I C AO N I C A . C O M
Fast track your IoT initiatives
On-Site IoT
Workshop
• AWS IoT: Ingest & store
data in real-time
• Building Serverless
applications w/Lambda
• Strategize design specs
to meet your business
needs
Enterprise IoT
Solutions
• Product & business model
Dev/Design consulting
• Prototype IoT device
hardware, platform &
analytics
• Onica Data Science
Workbench
Six-Week IoT
Accelerator
o
• IoT Architecture & UX Design
On-Site Workshop
• Deliverables: functional IoT
device prototype hardware
& software
• Support of prototype for 90
days

IoTanium
Workshops

Six-Week
IoT Accelerator

© 2 0 1 9 O N I C AO N I C A . C O M
Onica is a large and fast-growing Amazon Web Services (AWS) Partner Network (APN) Premier
Consulting Partner and Managed Service provider, helping companies enable, operate, and
innovate on the cloud. From migration strategy to operational excellence, cloud native
development, and immersive transformation, Onica is a full spectrum AWS integrator.
Santa Monica | Irvine | Chicago | Dallas | Houston | New York | Calgary | Montreal | Toronto | Vancouver

Predictive Maintenance for
Industry 4.0
© 2017 DataRPM –

• Predictive Maintenance in the Life of an Engineer
• Successful Use Cases
• Challenges in Solving Predictive Maintenance Problems
• How Does DataRPM Solve These Issues
• Defining the Path to Success
Agenda

Day in the Life of an Engineer
Pump Shows Erratic
Behavior – Predictive
Model indicated failure
Engineer is Notified on his Mobile
Device through MADP
Engagement Service
Uses AR Technology to assess
the problem
Goes through detailed analysis
onsite even with no network
through offline sync capability
Engineer logs in through his
secure authentication service
Raises a ticket to resolve issue
Conducts Root Cause
Analysis on his analytics dashboard

Welcome To The Asset Health Portal

Issue Chances of
Occurring
Damaged
Impeller
90%
Eroded Casing 35%
Worn Sealing
Rings
45%
Eccentric Impeller 60%
Bearing Damage 25%
Predicted Pump Health Status Assess Predictors and Confidence Problem Identified
Possible Impact Raise a Maintenance Tickets Manage Operations
Case temp
Case Pressure
Pump suction
A/B port temp
Click here for details

Observed a Big Change in Sensor Behavior at 11:54 am on
March 4th
© 2017 DataRPM – Proprietary and Confidential96
Case Temp and
Pressure
showed a huge
anomaly right
at this point in
time

Observed a rise in the number of anomalies from March 1st to
March 4th as compared to the other days
Date Anomaly
Count
26th Feb 4
27th Feb 111
28th Feb 118
1st Mar 242
2nd Mar 128
3rd Mar 145
4th Mar 369
4th Aug
Feb March

Case temperature had been showing signs 3 days prior to
the event of failure
Out of 10 days data the anomalies for Case
temperature clearly has an increasing pattern 3
days prior to the failure event.
Date # of Anomalies Anomaly
Score
1st Mar 3 O.55
2nd Mar 11 0.65
3rd Mar 18 0.69
4th Mar 30 0.79
Degrading Pattern
4th Mar

Objective: Reduce Plant Downtime – Engineers View
UI Elements
Analytics
Data Integration
Rules Engine
Mobile and Web Interfaces

Objective: Reduce HVAC Downtime – Data Science View
Prediction Model
Metrics
Model Validation Metrics

Determine Actionable Insights:
Pipe Blockage, Operator
Issues, Station Failure
Difference
over Time
Pressure
Head,
Power
Analyze Overall Pipeline Efficiency
Identify underperforming Stations
Identify influencing Factors for each
underperforming Station
Create a Prediction Model
Objective: Improve Efficiency of Gas Pipelines

Objective: Improve Quality of Production in Healthcare Devices
Determine Actionable Insights:
Operator Issues, Maker
Issues, Machine Issues,
Process Issues
Man
Machine
Maker
Characteristics ofTransducer– Error vs Non-Error
Identify key influencers
Identify combination of
characteristics for certain factors
Incorporate Sensor Data for More
Machine Level analysis and prediction
Error vs No-Error

Amazing Results From Customers
Leading Commercial
HVAC Manufacturer
Leading Global Luxury
Auto Manufacturer
Leading European
Telco Provider
Fortune 50 Global
Manufacturing Giant
Leading Healthcare
Equip Manufacturer
{ 66X More Field Errors Detected & Predicted}
{ Identified & Reduced Breakdowns by 75% }
{ Predicted Failures for 85% = Millions of Boxes}
{ $27M in Annual Savings Opportunities Identified }
{ Identified the Causes for 75% Scrap Rate }

SelectUseCases|AllDataDriven Meet SLAs For Commercial Washers
Predict Failures of Industrial Washing
Machines
Minimize Downtime For Gas Turbines
Predict Failures of Gas Turbines
When To Send Replacement Set-top
Boxes
Predict Failures of Set-top Boxes
Optimize Locomotives for
Lowering Fuel, Emissions & Noise
Predict Triggers / Mis-Triggers / Non-
Triggers
Prevent Breakdowns In Connected
Cars
Predict Part Failures & Breakdowns of
Cars
Minimize Bad Quality & Recalls in
Automotive Assembly Line & Paint
Shop
Predict Quality Issues On Assembly-
Line
Increase Yields & Minimize Scraps in
Healthcare Equipment
Predict Quality Issues & Identify Data
Gaps
Efficiency Of Heavy Machineries
Detect Transmission Anomalies, Learn
Driving Pattern Impacts & Oil Quality on
Warranty
Reduce Rework on Assembly Line for
Automotive
Detect Components & Configurations
Issues
Automated Inventory Stocking for Parts
of Semiconductor Manufacturing
Equipment
Predict Part Inventory Requirements by
Geography
Root Cause Analysis for Identifying
Quality Issues with Semiconductors
Detect Anomalies that are
Predictors of Quality
Prevent Failures & Improve Efficiency
of Pumps & Other Rotating Equipment
Predict Failures & Detect
Efficiency Predictors

Challenges In Solving Predictive
Maintenance Problems

Source: Capgemini
60%
of organizations are relying on traditional
Analytics capabilities to take advantage of
the data generated from IoT sources

Hope Reality

DATA
80% of the problems
are new, random and
unknown. No prior failure
labels in the data to train
predictive models the
traditional way
SKILL
Annual demand for data
scientists and data
engineers will reach
nearly 700,000 openings
by 2020
SCALE
20+ billion connected
devices world wide. All
unique with constantly
changing environmental
and operating conditions
PRODUCTION
Only 15% of big data
projects make it to
production from PoCs.
Production data science
workflows are complex
Key Problems
Source: IBM Source: Statista Source: GartnerSource: ARC Advisory Group

How Does DataRPM
Solve These Issues

Solutions To The Key Problems
DATA
Using a cognitive
approach to self learn
the normal patterns and
generate labels to detect
known and unknown
issues
SKILL
Using Meta-Learning
based automation of
time-consuming data
science tasks increase
productivity of data
scientists by 10x
SCALE
Using a distributed and
parallel computing on
with a transfer-learning
based approach for
automated digital twin
creation
PRODUCTION
Enterprise grade Data
Science Process flow
Framework enables
seamless transition from
R&D to production at
scale

DataRPM’s Unique Value Proposition
We provide a Machine-First Human-Guided
Approach for teams exploring ways to leverage
their IIoT data to improve quality, yield, and
maintenance of their industrial assets

Pre-built and robust cognitive-process flow for Analyzing Asset Behavior
and Predicting for Failures built after years of research
Pre-Built: 50 different steps, 15 different algorithms tied together to analyze time-series
sensor data
Robust: It has been applied to over 50 different types of assets – from nuclear power
plants to chip-manufacturing machines to set-top boxes
Includes algorithms selection and parameter tuning to enable faster, better and more
accurate results.
Production-Ready platform which removes the complexity of development,
scale, integration and hence, save several months of production effort
Lets data scientist create their own flow and deploy it in a production environment without
worrying about the underlying architecture
DataRPM Product Differentiation

Multiple Sensor Data
Identify sequence-based anomalies of stages within each
time series sensor and generate a score
5
Identify magnitude-based anomalies within each stage for
each sensor and generate a score
4
Detect stages in the time series in each sensor2
Analyze each sensor of every asset individually1… … … …
Establish “normal” baseline of each stage by comparing the
same stage in the sensor within the same asset and other
assets
3
Generate machine state and anomalies by combining sensor
level ones and generate a score
6
Group similar assets by comparing machine states and
transfer learning
7
A bottom-up approach that
extracts Meta-data from the
atomic unit of sensor stages
all the way up to the asset
groups and self learns the
operational characteristics
Meta-Learning Based Approach Delivers Automation @ Scale

Moving Up The Maintenance Value Chain
Working
EquipmentCondition
Broken
High
Cost/Risk
Low
Point where Failure Process starts
Early Indicators appear in Sensors / KPIs
More Indicators in Sensors / KPIs
Major Physical Manifestation
(Noise, Vibrations, Heating etc.)
More Physical
Manifestations
BREAKDOWN
Costs of Maintenance
& the associated Risks increase with Time & closer to the point of Breakdown
Time
Predictive Maintenance
detects possible
Failures early so as to:
take Corrective Action,
Avoid Unplanned
Downtime &
Unscheduled
Maintenance, thus
reducing Costs & Risks
“The Goldilocks Zone”
$
REACTIVE MAINTENANCE
CONDITION-BASED MAINTENANCE
PREDICTIVE MAINTENANCE
COGNITIVE PREDICTIVE MAINTENANCE
PREVENTATIVE MAINTENANCE

The 4 Pillars of Industrial IOT
Sensors &
Gateways
Communication
Cloud
Infrastructure
Apps &
Analytics
Customer
Customer Progress
AWS

Are You Ready To Benefit from IIOT?
e You Deployed Sensors? Do You Have Historical Data for these assets?
Is your asset a high value asset or causing a hig
value failure?
Your Asset Fail Often? Do You Collect Failure Data?
Do You Have IOT Data On the Cloud?
You Have Asset Downtime?

How Many People Got -
0 Yes
3 Yes
5 Yes
All Yes
Unless You Get All Yes – You May Not Be Ready!

The Industrial IOT Step Journey
Define
Pilot
Implement
Infrastructure - H/W and S/W
Data – Sensor or Process
Analytics – ML or BI
Outcome – Data Science or
Business
Business Objectives – Why?
Benefits - Who?
Which Assets – What?
Monetize – Who Pays?
ROI – Internal or External
Scale – multi-Production
Business Model – SaaS or
Private Cloud
Feedback – Data Scientists or
Engr

What Does It Take To Succeed
Data Scientists
Devops Engg
Operations
Teamsneers
Sensors
Gateways
Cloud Security
ROI
usiness
enefit
Problem Statement
Feedback
Analytics
Apps
Data
Integrations

© 2019, Amazon Web Services, Inc. or its Affiliates. All rights reserved.© 2018, Amazon Web Services, Inc. or its Affiliates. All rights reserved. Amazon Confidential and Trademark
AWS IoT Business Development
IoT on AWS
IoT Data Services
AWS IoT Business Development
IoT on AWS
IoT Data Services

AWS IoT Architecture
How can I control, manage, and secure my devices?

AWS IoT Architecture
AWS IoT Analytics
AWS IoT Core
AWS IoT Device Management AWS IoT Device Defender
AWS IoT Greengrass
Amazon FreeRTOS AWS IoT Device SDK
AWS IoT SiteWise AWS IoT Events

AWS IoT Analytics is a fully managed service that collects, pre-processes,
enriches, stores, analyzes and visualizes IoT device data at scale.
AWS IoT Analytics
From raw sensor data
to sophisticated
IoT analytics

Collect AnalyzeProcess Store Visualize
AWS IoT Analytics
AWS IoT Analytics is a service that processes, enriches, stores, analyzes,
and visualizes IoT data for business insights.
Data
services
Collect only the
data you want to
store & analyze
Convert raw
data to
meaningful
information
Store device
data in time-
series data store
for analysis
Get deeper insights
with built-in ML
support
Quickly visualize
your IoT
data sets

How can I unlock
and liberate my
equipment data?

Structure your data and specify
performance metrics for your assets
and processes
Easily browse equipment and process
data, build data views to identify
inefficiencies, diagnose issues, and
improve cross-facility processes
AWS IoT SiteWise collects data from the shop floor with a local gateway, structures & labels that
data, and generates real time KPIs & metrics to make better data-driven decisions.
See your data flowing in minutes
without writing code, just connect
and configure your gateway
Data
services
AWS IoT SiteWise

Customizable
Views
Data Management and
Modeling Tools
Remotely Manageable
Edge Gateway
Access data from local
databases on the
factory floor
Collect data consistently
from different sources
Visually identify
equipment or process
issues
AWS IoT SiteWise

Browse raw data coming from equipment,
production lines, and processes and view
their respective performance metrics
Avoid the hassle of querying individual data
streams for each asset, and then writing
aggregating logic
Easily develop operational
dashboards, mobile applications
for factory staff, or deploy machine
learning models across your organization
Customizable Views
Visually identify equipment or process
issues

IoT SiteWise Console View (Preview Release)
Assets and equipment hierarchies can calculate
metrics and provide drill-down information
Equipment details
Sensor readings
Near real-time calculations

IoT SiteWise Gateway Installed
PLC and Gateway plugged into
network switch
Program Logic Controller (PLC)
Gateway hosting OPC-UA server &
SiteWise software package
Maintain factory compliance with minimal
installation and industrial grade Gateway for
different environments

AWS IoT SiteWise – Manageable Gateway
Stream time-series data from on-premises to AWS Cloud
Industrial
Gateway
Client
VPN AWS IoT
Analytics
AWS IoT
SiteWise
On-premises historian or server Visualize and analyze in the cloud

How can I
monitor changes
across multiple
data streams?
!

Build simple logic to evaluate
incoming telemetry data to detect
events in equipment or a process
Detect events from data
across thousands of sensors
and other sources
Trigger responses to
optimize operations
AWS IoT Events allow you to continuously monitor data from your equipment and fleets of
devices for changes in operation and to trigger the appropriate response when events occur
Data
services
!
AWS IoT Events

Event Detector Models Scalability
Integration with
analytics tools & other
AWS services
Data
services
Reduce the cost of
device maintenance
Uncover new insights Easily automate
operations
AWS IoT Events

Event Detector
Models
Evaluate multiple inputs to derive the
state of processes, equipment, or
products
Schedule maintenance or send alarms or
alerts prior to failure
Improve the efficiency of processes,
products, equipment, and staff
Reduce the cost of device maintenance

AWS IoT SiteWise & IoT Events Workflow
Collect data
Operational
Metrics
System Administrators Operators & Analysts
End-to-end Value Proposition: Reduce time and cost of managing complex infrastructure for equipment data
Condition
Monitoring
Power
Workflows
Customize views
Create models
Explore assets
across all sites
Subject Matter ExpertsStakeholder Value:

Reference Architecture for Industrial IoT Data
Services
On-premises Servers
IoT SiteWise
Enabled Gateway
PLC
Industrial
Time-Series Database
IoT Analytics
Data Services
Enrich, Process, Integrate
Amazon SageMaker
Build, Train, Deploy
Monitor, Detect, Trigger
IoT Events
Condition Monitoring
AWS
Applications
Operational
Dashboards
Machine
Learning
Power
Workflows
On-premises factory systems Cloud intelligence and business insights
AWS Cloud
IoT SiteWise
Interactive Viewer
Metrics, Mapping
IoT Greengrass
Direct Connector
MES / ERP
Accessible

AWS IoT 2019 New Services
Things / Edge
Sense & Act
Cloud
Compute & Analyze
IoT Edge Services
Enhancements for Greengrass ML Inference enable OEMs to
significantly lower the BOM cost of edge devices and still gain1.5x
to 2x gain in ML inference response time.
Greengrass Connectors provide support for pre-built connectors
(from AWS and from our partners) for AWS native and 3rd party
web services at the edge. This enables developers to both boost
the power of industrial applications running on Greengrass by
bringing in data from other services, as well as reduce the time
they spend on writing these connectors themselves.
Greengrass Secrets Manager enables secrets, such as user
credentials, to now be secured locally on a Greengrass device via
Greengrass Hardware Security Integrations, thus securing
Greengrass Connectors.
Greengrass Hardware Security Integrations enables Greengrass
devices to use Hardware Security Modules and Trusted Platform
Modules for private key storage, via the industry standard
PKCS#11 security interface.
Things Graph allows building intelligent workflows visually
between physical devices / sensors and web services from
different vendors that speak different protocols and do not work
with each other out of the box. Things Graph simplifies
application development and enables customers to bring them to
market faster.
Direct Ingest allows AWS IoT Core customers to securely send
large amount of data to over 10 AWS services such as Kinesis and
S3 via AWS IoT rule actions, without incurring additional
messaging charges. Direct Ingest optimizes data flow for high
volume data ingestion workloads by removing the pub/sub
Message Broker from the ingestion path. With Direct Ingest,
customers can now save up to 75% on data ingestion costs while
continuing to benefit from all security and data processing
features of AWS IoT Core.
IoT SiteWise is a managed service that enables industrial
enterprises to collect, structure, search, and analyze thousands of
sensor data streams across multiple industrial facilities by
connecting into factory historians. Customers can now skip months
of developing complex infrastructure integration and data
management tools. Instead, they can focus on using their data to
detect equipment issues, improve process optimization across
factories, view analytics on factory equipment, and more.
IoT Events is a managed service that continuously monitors
equipment, devices, and sensors to detect changes in operations to
trigger alerts. IoT Events reduces development time and integration
requirements of building complex logic systems, manageable event
detectors, and lightweight trigger actions. This will help customers
drive accurate operational metrics, improved up time and higher
product quality across fleets of sensors and factory operations.
IoT Data ServicesIoT Cloud Services

AWS Manufacturing Day Philadelphia-Boston-April 2019

Recommended

Recommended

More Related Content

What's hot

What's hot (20)

Similar to AWS Manufacturing Day Philadelphia-Boston-April 2019

Similar to AWS Manufacturing Day Philadelphia-Boston-April 2019 (20)

More from Amazon Web Services

More from Amazon Web Services (20)

AWS Manufacturing Day Philadelphia-Boston-April 2019