Design for Circularity (and 2nd Life of Components) and Artificial Intelligence

Subject

To sum up, this research project aims to answer the question: How can we help find solutions for reusing components recovered from the disassembly of end-of-life products, other than in their initial context?
It is therefore important to formalise the business expertise of the designers and products involved (ontologies, heuristics, knowledge bases, etc.), and to develop or find AI approaches that help to navigate through the alternatives (identify them, evaluate them and help to propose the most relevant solutions) while integrating information on the criticality of materials and components (market view, material view, demand projection, etc.), as well as modelling the loss of performance of these components at the end of their first life and disassembly operations. 
The scientific challenges are to use feedback to analyse products and contribute to rating processes (such as the demountability index proposed by the AGEC law, but also the future sustainability index planned for 2025) and to provide factual and quantified feedback to those who place products on the market. By translating the constraints and opportunities of disassembly/recycling solutions, as well as systematically analysing construction solutions, it is also possible to propose design orientations and strategies, product architectures and also to evaluate assembly solutions by developing methods for finding the best compromise between technical/economic/circularity performance. This work will help to enrich the material and product circularity indicators initiated by the E.McArthur Foundation. The I2M is taking advantage of its strong expertise in eco-design for end-of-life, built up through work with eco-organisations (Ecologic and Ecosystem) and their members (designers and recyclers). This subject is beginning to be addressed in the field of power converters in the Vivae ANR (bringing together academics and industry), with a PhD student starting in January 2022 on design methods in this sector, and with a post-doctoral student to be recruited in 2023 who will focus on the development of dedicated converter disassembly solutions.
In addition, it seems necessary to develop an environment to assist in the search for second-life components (from the designer's point of view) and to search for fields of application for secondlife materials or components and to assist in the search for cases of application (from the recycler's point of view). An approach along the lines of Ashby's performance maps, applied to recycled materials on the one hand and components on the other, could be envisaged.
One of the aims of this work is to use AI technologies, data mining and learning to analyse a set of existing and known cases (multi-sectoral), and to classify the material/component associations and use cases into families of cases, looking in particular at the assembly or connection options. These approaches should be based on and inspired by the material performance indices and design choice aids developed by M. Ashby.

The first step will be to transpose the materials vision of Ashby's approach with a component/module vision and define the nature of the databases and indicators required. 
In this way, a second life component or recycled material is analysed as a set of performances. It is therefore necessary, in the same way as for material data sheets, to propose structures for component performance data sheets by crossing multi-sectoral visions.

Finally, we will need to evaluate the various AI technologies, as well as knowledge formalisation technologies (in design, among other areas), to assess their suitability for use in the type of preliminary design assistance context. It seems necessary to combine guided learning approaches with ontology-based approaches for formalising knowledge.
The ultimate aim is to provide an environment to help choose usage domains and products. This is currently a key issue for the development of component and often material remanufacturing processes and solutions. This will make it possible to respond to the criticality of components and materials.
Design objectives and existing solutions must therefore be defined on the basis of functional definitions, enriched by the functions/solutions database, and processed using the AI solution. The method proposed by Prof. Ashby for the selection of materials based on material performance indicators will be extended and adapted to the search for component/module applications based on the performance characteristics of second-life components/products.
The various phases that should be carried out could be as follows:

  • a formalism must be selected or defined for the functional representation and transform a design solution into knowledge for the AI (ontologies, heuristics, knowledge book, ..)
  • a guide to help with the functional breakdown of the product to translate the overall design objectives and the specific local solutions (based on existing approaches such as Fast, SysML, SADT, etc.) to translate the design process,
  • structure and populate a specific database linking the functionality and the solution, feeding design knowledge into the AI models
  • propose an inference algorithm to ensure navigation and bijection between functional definitions and existing solutions, paving the way for automatic processing,
  • propose automatic evaluation indices for proposed solutions to justify proposed choices. 

This work will be applied to stimulate the circular economy with an assistant for a use case based on second life components (and their residual functionalities) or secondary primary materials (and their performances). This topic extends the work on design assistance, steering the way towards AI for end-of-life design.

Mise à jour le 13/07/2023

Informations

Please forward your application to the supervisors with Curriculum Vitae, a short message or covering letter on the proposed subject and optional: one or more names of teachers to contact for recommendations. If your application is accepted, it will be followed by a face-to-face or video interview.

Supervisors

Nicolas Perry


Contacter par courriel

Stéphane Pompidou

Contacter par courriel

Emmanuelle Abisset-Chavanne


Contacter par courriel