(End-to-End) AI for quality assurance in manufacturing
The project will create user-friendly End-to-End AI for automated quality control (QC) of manufactured components that can adapt to different environmental conditions and product designs.
The project will use state-of-the-art AI techniques, such as multimodal large language models (LLMs), zero-shot defect detection, synthetic data generation, and robot motion planning, with the purpose of accelerating the utilization of AI for quality control in Swedish industry.
The project aims to increase the performance and competitiveness of the manufacturing industry and reduce its environmental impact and resource consumption. It will contribute to scientific and societal progress of AI research and applications through collaboration and knowledge exchange among academic and industrial partners, and by sharing its findings and results. The project will support the digital transformation and sustainable development of the Swedish industry and society.
The project will apply multimodal LLM, zero-shot defect detection methods, and public datasets to create a new system for quality control. The potential of synthetic data will be explored, as well as the possibility of using sound as a complement to image data. The project will address the need for smart robot motion planning, which is very important in cases where use of robots for QC is highly desirable, but where it is not feasible to manually program the motion path.
Summary
Project name
AI4QAM
Status
Active
RISE role in project
Coordinator
Project start
Duration
3 years
Total budget
16 943 502 SEK
Partner
Enodo Robotics AB, Husqvarna AB, PVI Hydroforming AB, Scania CV AB, Tekniska Högskolan i Jönköping AB, Thule Group AB
Funders
Vinnova, Avancerad och innovativ digitalisering 2024 - första utlysning
Project members
Larisa Rizvanovic Jonas Lindqvist Tomas Olsson Andreas Thore Lennart Elmquist Rakesh Shrestha