Project description
Robotic solutions for efficiency gaps in manufacturing
In the heart of modern manufacturing, many production steps hinge on previous results, which makes their automation difficult, causing inefficiencies and resource wastage. Typical examples include inspection and rework stages, where rework depends on the deviations that were found. With this in mind, the EU-funded SeConRob project will address the challenge of non-automatable manufacturing steps that rely on previous outcomes. Pioneering self-configuring robotic processes and using AI-driven data analysis, the project will extract insights from inspection data, generating robot programmes and parameters for downstream tasks. A feedback loop, powered by reinforcement learning, will refine the process in the long term. Test cases will encompass multi-stage processes (inspection, gouging, welding, grinding, and polishing), and demonstrations will target sectors like automotive and aerospace.
Objective
The SeConRob project aims at developing methods for the self-configuration of robotic processes, where each manufacturing step depends on the results of the previous step. In this case a lot of productivity, energy and resources are lost, because the processes currently cannot be automated for technical and economic reasons. Such situations typically occur during inspection and re-work, where the (downstream) re-work process depends on the results of the (upstream) inspection process of each individual part. SeConRob will develop technologies that enable the automation of such processes, by creating robotic processes that can be automatically configured for each individual part. This will build upon AI-based data analysis that extracts information from the inspection data, that used in turn to automatically generate a robot program and process parameters for the downstream re-work process. Physical process models will the basis for the initial planning and a long-term feedback loop based on reinforcement learning will be established to optimize the process and account for properties that are not included in the initial model.
Two use cases with multi-stage manufacturing processes including inspection, gouging, welding, grinding and polishing will provide test cases for the developments. Demonstrations are planned on a real-world production line to raise interest in sectors such as automotive and aerospace, where safety-critical parts are manufactured. The estimated market potential for such multi-stage self-configuring robotic process is about 2000 robotic workcells, corresponding to a market of 600 M€.
Fields of science
- natural sciencescomputer and information sciencesdata science
- engineering and technologymechanical engineeringmanufacturing engineering
- natural sciencescomputer and information sciencesartificial intelligencemachine learningreinforcement learning
- social scienceseconomics and businesseconomicsproduction economicsproductivity
- engineering and technologyelectrical engineering, electronic engineering, information engineeringelectronic engineeringrobotics
Programme(s)
Funding Scheme
HORIZON-IA - HORIZON Innovation ActionsCoordinator
4407 STEYR GLEINK
Austria