State of IoT review. beyond predictive maintenance and asset management. Value based IoT solutions. Data driven and digital transformation. IoT platform
Cloud based simulation
High end Edge computing
Simulation via digital twin
Massive digital twin simulation
6. McLaren digital twin / master
• Constant product development and improvement
• Difference between good and bad race is 4% product performance
• Top 5 0,15% product performance difference
• Regulations …. V12 -> V6 still same performance.
Narrative :
- Cost of sensors are going to 0
- Cost of storage and connectivity is going to 0
- Focus on mass simulation
- Close the loop to design and performance
7. • IoT platform
• Cloud based simulation
• High end Edge computing
• Simulation via digital twin
• 4mb per second
• 17000 parameters per car (even eye gaze of driver)
• 10gb per car per day
• All shared in the car digital twin design model
8. McLaren realm
• Continuous delivery of a car via digital design and simulation
• Every car is changed 5% per race.
• Models to optimize performance via data
• Artificial intelligence models to optimize model on goal (speed /
performance / efficiency )
• Massive digital twin simulation
• Now the digital twin becomes the digital master for simulation and
design
9. Manufacturing solutions
• Not just condition monitoring… focus on outcome based on your digital strategy
• Take 2 – 3 years to reap the products of your data strategy
• Trust you are doing something different…. Different architecture and different support
• Group them based on solution category
• Digital transformation = value chain transformation
Old: Sell machine, service contract, visits, spare parts
Now you are responsible for output (pay per product):
• Making the machine more usable (engineers)
• Make the machine more productive (head of prod)
• Make the machine easier maintainable (service)
• New business models (CFO)
Machine helps you….
11. What if
What happens when an algorithm cuts your
healthcare?
The Verge, Mar 21, 2018
Don't Grade Teachers With a Bad Algorithm
Bloomberg, May 15, 2017
Who will you decide to kill with your
self-driving car?
Verge, Aug 9, 2016
How to prevent artificial-intelligence
Directory of ethics and guidelines can be found at :
https://algorithmwatch.org/en/project/ai-ethics-guidelines-global-inventory/
13. Rights to know about machine learning models
ComputationDesired output
Input data
ML Model
Machine learning
AI
application
Label’
Input data’
Right to know
the bias in
input data or
bias in
selection
Right to know the
explainability of the
model
Impact on use:
- Precision
- Exception
14. How to use a responsible machine learning mandate
Stewardship: As AI extensively relies on data and analytical models, they need to be
protected to ensure data privacy, security and compliance
Fairness : AI bears the risk of amplifying human biases, potentiallyresulting in unfair
and Unintended consequences
Transparency : Success of AI adoption is directly linked to trust. It is imperative to be
transparent about the use and decision of AI and the associated data supply chain.
Accountability : Given its novelty and risk associated with AI, organizations need to
properly govern and self-regulate AI programs
Agency: Embracing human + machine interaction through amplifying human’s talents
and ability, empowerment, and education.
15. Can we trust unbiassed AI
“there is no such thing as an unbiassed data set” (bias in data and bias in
selection)
16. The problem with AI is that some vendors are using data that was
bought from someone else, who bought it from someone else.
They cannot trace the data. That brings a new ethical issue.
But what we should do is retain responsibility on the human side.
We should never allow a machine to be held responsible for
anything, because then we would be surrendering responsibility.
That would be a dangerous thing.
113 “Thank you for everything”
18. What is HD map
• HD maps (high definition maps) are essential for self-driving cars. They
have a high accuracy of object locations, up to 10 cm. Self-driving cars use
HD maps for multiple purposes. It helps to solve the localization problem:
figuring out where exactly the car is in the world.
• Automatic map creation via computer vision
• Add roads
• Add object
• Add temporary changes
• Correct errors
19. What is an HD map
https://www.youtube.com/watch?v=--qtSy8k5_s&feature=emb_title
24. Shifting trends
• IoT becomes “digital” or “data-enabled”
• Beyond “predictive maintenance”
• New Services, new sources of value
• Added data services
• Outcome based
• Pay-per-use
• Integrate trough the whole value chain
• Movement of value across value chain
• Transition to SAAS model is 2-5 years
25.
26. General Aircraft Engines
• Pay per hour flight
• From condition monitoring to pay-per-use
• Optimize maintenance
• Capex to Opex
• Return guarantee
• Long term relationship
27. Example : Trash collection (BigBelly)
• Traditional sell waste bins and empty
them on route
• Smart waste bin saves money on
emptying and labor
• And transfers the investment in waste
bins to OPEX costs for cities
• From product to subscription sales
Pay-per-use bin
• Automatic checks and maintenance
28. Hospital beds
• Sensors in beds
• Utilization
• Localization
New services
• And also usage based cleaning
• Replacement on usage
• Monitor sleep patterns / anti-decubitus
• Track wandering of Alzheimer patients
• Pay per use bed with these services
31. Embracing digital twin
• Vehicle for providing information interoperability of data
• Standardization
• Standardized electronic name plate of your asset
• Think system (and same experience as physical asset)
• Architecture similar to service oriented architecture
• Round trip simulation
32. Next step / Digital Twin API (standardization)
• Make digital twin interchangeable between
departments and companies
• Administrate API
• Information model semantics API
• Data exchange API
• Publish / Subscribe bus API
• Disconnected operations and replicate API
• Discovery API
36. Problems
• Broken tools cause heavy losses
• Repairs are taking a long time
• Tools are stolen or alternatively used
• Tool costs need to be spread across
multiple projects
• Tools are not always used evenly frequent
• Risks causes constructers to divert to
higher stock or cheaper tools
38. Added value
• Subscription model
• Pay-per-use
• High reliability, near-defected models are quickly replaced
• Capacity management (scale up and down) / peek capacity
• Tools are geo fenced
• Automatic repair service based on usage
• Long term subscription contracts
• No CAPEX / OPEX costs are assigned to the project
• Firm customer connection
39. Was this easy … no
• Took 2-3 years
• Longer sales cycle
• Family business / strategic decision
• Facing massive disruption
• Focus on digital strategy
• Optimize with data and algorithms
40. Titel van de presentatie 40
Bedankt voor uw aandacht
Hinweis der Redaktion
(Azure, OSI, PTC, ABB)
Now:
IoT stack to differentiate in product design trough sensor
Also for : connected rail, connected healthcare, connected sports