Meeting Minutes
===
###### tags: `AI Proficient` `Meeting' 'INEOS' 'Continental' WP1'
Nov 16th, 2020, online

* Alexandre Voisin (UL)
* Karën Fort (UL)
* Aitor Arnaiz, Santiago Fernandez, Iker Esnnaola, Kerman Lopez de Calle, Egoitz Konde (Tekniker)
* Uroš Milošević (TF)
* Lazar Berbakov, Dea Pujic (IMP)
* Sirpa Kallio, Peyman Davvalo Khongar, Lotta Sorsamäki, Marjut Suomalainen, Karhan Özdenkci, Juho Peltola (VTT)

# Welcome and quick roundtable

# Continental use cases presentation

Conti presents the combiline, with 4 areas: Extrusion, Cooling, Sidewall Packing and Tread packing.

Conti presents 10 cases grouped into three of the for areas (as sidewall is not considered that much relevant for the introduction of AI technologies)

Conti wants to add new sensors to be included in AI-PROFICIENT solution. New sensors could be also created from the partners of the project with Conti support. Special interest in vision related sensors.

access to [CMMS](https://en.wikipedia.org/wiki/Computerized_maintenance_management_system), yes, information will be available

[Digital Twin](https://en.wikipedia.org/wiki/Digital_twin), Conti would like to have one but don't know if some use case requires it or if it is feasible

Attention has to be paid on the terms Human...
From previous meeting, definitions have been provided:
* Human-in-command - AI is providing just a decision support, while operator executes the control manually;
* Human-on-the-loop - operator with a supervisory role, while AI is suggesting the control strategies and executes after approval; and
* Human-in-the loop - implicit knowledge by the operators is fed into the AI optimization cycle

Fully automated, semi-automated and manual control to be considered as part of the use cases; real-world limitation will be encountered (what is feasible or not to be demonstrated)

One way should be to give different priorities to the use case scenario in order to order them in consistence with project objectives, impacts, constraints.

## Virtual visit of the combiline

TCU: thermo control unit

# INEOS use case presentation
## Geel use case
Polyproperlyne reactor

The reactor, idealy, stops only once a year.

In geel – simulator – very basic – Not really fit to the detaill level – Not possible to simulate well
There is pilot plant – 10-15 meters longs – for basic tests. But not really reflecting the real challenges.

Geel use case "image recognition":
* recognise additive name and lot numbre. Name not always at the same place in the sticker, sticker can be a little flod, dirty and plastic film wrap it.
* 99% of time the label is quite visible, but some time it is not

only 5 operators to drive the asset:
* 1 for extrusion section (where the errors occurs)
* 4 others ?

The error can be feedback to the operator.

## Cologne use case

Polyethyline reactor (Christophe - can you include a sentence about the case??)

Digital twin – some previous experiences have not worked so far.

# Task 1.1 work session

presentation of the template.

Linked to the impacts and STOs to be done in T1.3 ==> Paul - to try to identify use case with STOs in separate table (not template)

Use case template will have several pages dedicated to differents tasks (1.1,1.2, 1.3...)

Tags to be added:
* augmented reality: support the user with the proper instruction

the tags address two different aspects: goal and technology.

Template validated with 5 tags.

# Task 1.2

## Karen presents Ethic in AI-PROFICIENT.

Existing ethical recommandation from every partner.

Ethics in/by/for design(ers): needs a presentation to clarify these concepts. Mainly concrete examples (in more conventional domains) could be relevant to investigate how these examples could be (or not) translated in manufacturing area and more precisely AI-PROFICIENT uses cases.

## Marc Anderson's presentation

EC Coordinated plan stresses implementation from beginning of design process
Procedural principles: proactive, embedded, end to end
Example: Biased hand dispensers (no proactive questioning)
EU Commission Guidelines on AI
Including: transparency, explainability, allowing human oversight, aligning with human values
Direct vs Indirect Impacts. Direct human impacts are the practical place to begin.
Academic vs Business view of Ethics (should we do it? vs we are doing it, how can we do it ethically?)
Focus on physical and mental human-machine interactions
Value for the manufacturer in ethical integration.
Machine centered vs Human centered outlook (food for thought)

# Task 1.3 - Presentation of Aitor Arnaiz

Task 1.3 details the inputs, outputs and activity route during next months (see slides)

Of particular importance, for next weeks, is the identification of the potential scenarios that can be drwan from the use cases and the potentiaal technologies available at each partner.

Maintenance action points for next month:
* Tech. providers (Now to 30 Nov.)::
* Send name of representative per partner that will act as lead in T1.3 TO ([e-mail])
* Review slides – ask for doubts. Send comments/ corrections/ additions…

* Tech. providers (Now to 15 Dec.)
* Review info on pilots & use cases
* Develop questions/requests for pilots & use cases
* Propose technologies … & complete solutions

# Wrap of the meeting