ACRA4DT

ACRA4DT

Automated Configuration of Robots and Analytics in I4.0 with Digital Twins

ACRA4DT

Within the ACRA4DT project (Automated Configuration of Robots and Analytics in I4.0 with Digital project Automated Configuration of Robots and Analytics in I4.0 with Digital Twins (ACRA4DT), supplementing robot-based manufacturing processes with semantic knowledge provides machine-based learning anomaly recognition technology with additional context. This enables the automated integration of data analyses for small-batch production.

Project description

Setting up high-grade digitalized industrial processes requires a multitude of configuration and programming steps. This starts with the programming of the control logic for individual robots and continues with the configuration of the analytics that monitor the production process.

These manual steps are extremely time-consuming and require a high degree of specialized knowledge. As a result, they are economically feasible only when carried out with static production plans in which the operating time significantly exceeds the programming time, or when high-grade systems are involved. On the other hand, the demand for flexible production lines and single-batch manufacturing using middle-grade systems is constantly on the rise. Although robots and 3D printers promise the required tooling flexibility, the reality is that the effort required to adapt programs and analyses is still too high for deploying them on a broader basis.

Research contribution

Traditional analytic approaches in industrial automation and robot systems rely on raw data from multiple sensors and require a high degree of manual effort for every new area or application scenario. When using a knowledge-based approach for the development of production systems, the product, process and resource know-how is formally represented in semantic representation languages.

By employing such a semantic Digital Twin model, raw sensor data can be automatically enhanced with semantic labels and context information such as the current task and its respective part or parameters.

In this project fortiss is examining how machine learning approaches for anomaly recognition can be automated and optimized by integrating this type of information, particularly for small-batch production systems.

Project duration

01.11.2019 - 31.12.2020

Dr. Markus Rickert

Your contact

Dr. Markus Rickert

+49 89 3603522 43
rickert@fortiss.org

Project partner

IBM