Motion visualisation and analysis through vision-based digital twins

Deadline

Motion visualisation and analysis through vision-based digital twins

Challenge

Motion applications show complex behavior due to flexibility, backlash, friction, parameter & load variations, non-linearities, etc … Often, solely classical sensors are present in a motion system, which are insufficient to fully visualize the complex motion and to provide richness towards the operator, designer or control engineer.

In order to improve accuracy and speed in multi-coordinated motion systems, kinematic coordination is an important prerequisite. Although low-cost vision techniques are emerging, an adequate architecture is required to integrate these to answer the above needs. Existing 3D CAD Motion simulators are not capable of modelling the aforementioned complex behaviors. If advanced visualisation of the complex motions and operation of the machine would be added, in addition to the classical discrete sensors, valuable insights would become available. Operators would be able to better judge (and increase) the quality of the manufactured parts.

In addition, the system would be able to operate closer to its physical limits, increasing the production quantity. It also would allow easier intervention in the manufacturing process, as it reduces machine downtime and increases robustness.

Finally, the machine would become a more flexible asset, as the control designer can modify the machine’s behavior given the increased insights in its complex motion and operation. The term advanced visualization is used deliberately, as it concerns not solely camera vision itself, but also the deduction of valuable motion information from this vision data.

In this project we will achieve this by fusing vision technology with kinematic & dynamic motion simulation and geometry data which is crucial knowledge embedded within the 3D CAD Motion simulator, to form a vision-based digital twin.

Project goals

The project goal is to develop an architecture for vision based monitoring of single and multiple coordinated motion systems:

  • To achieve motion and parameter comparison/overlay based on digital twin with virtual 3D CAD motion.
  • To improve kinematic coordination in multiple coordinated motion systems.

A low cost vision sensor is added to a machine, which classically only contains discrete sensors. This enables full capture of complex motions and direct observation on the tool and workpiece. This vision is fused with the classical sensor data and motion and control simulation based on 3D CAD geometry to achieve a vision-based digital twin. We will leverage on the 3D CAD + Motion & control simulation methodology developed in the AMOCAD project, Visual AI techniques for feature extraction to obtain motion from vision, and recent developments in hybrid modelling. The aforementioned techniques can require a high cpu power. Moreover, large data-streams are involved, and real-time operation is the aim for certain tasks. Therefore, a proper IT architecture is developed where sufficient cpu power and the right communication requirements are allocated at the right level, cfr. I3oT. The vision-based digital twin is a.o. used within this project to improve the kinematic coordination of multiple coordinated motion systems. Experimental validation is foreseen on at least one single motion and one multiple motion application.

Interested?

CADAIVISION_SBO is a Strategic Basic Research (SBO) project. We are looking for companies to join the User Group and work with us on the valorisation of the project.

Interested? Complete the form below and we will contact you as soon as possible.

Deadline

Strategic Basic Research (SBO)

Deadline

Strategic Basic Research (SBO)

Deadline

Strategic Basic Research (SBO)