back to list

Project: Decision Support tool for Product Release

Description

Vlisco releases 250 - 300 new dessins every year, and every new dessin is manually judged on feasibility and producibility. A risk profile is generated based on likelihood of typical errors related to the product drawing.

This process is called Product Release and relies heavily on the availability, interpretation and level of expertise of the product quality team.

Vlisco would like the product release decisions to be supported by data science based on available risk profile information of similar designs:

  • Support decision making of risks for manufacturing of new designs
  • Speed up the product release process
  • More consistent and reliable risk profile, less dependent on experts
  • Learn more about the relationship between product attributes for design and
  • manufacturability

Internship Assignment proposal

  • Investigate strongest combination of variables and machine/deep learning model with regard to risk profiling our images.
  • Develop a decision support interface for the product release process to predict more consistent and structured the feasibility of new dessins in manufacturing and implement a learning- and feedback loop between design, technical support and manufacturing.

Assignment objectives:

  1. Determine optimal way of measuring product similarity with regard to manufacturability (risk profiling), using image recognition and product attributes.
  2. Use deep learning to find hidden/new features in images that relate to risk profiling
  3. Develop the data science model based on your findings, having a risk profile (e.g. score 1-10) as output for each new design.
  4. Design an interface to be used as decision support by product quality experts during the product release process.
Details
Student
JM
Jon Maessen
Supervisor
Robert van Liere
Secondary supervisor
Anna Vilanova
External location
Vlisco
Link
Thesis