EaSy ML - Assistance Evaluation System for Machine Learning
problem
Cost-optimized production combined with high process and product quality is a central promise of Industry 4.0 and the associated digitalization.
There are numerous examples of how machine learning (ML) techniques help to analyze data, gain new insights into production, and optimize production.
However, a superficial understanding of the algorithms and procedures is often not enough to draw meaningful conclusions from the production data.
There are numerous examples of how machine learning (ML) techniques help to analyze data, gain new insights into production, and optimize production.
However, a superficial understanding of the algorithms and procedures is often not enough to draw meaningful conclusions from the production data.
objective
The aim of the EaSyML project is now to significantly reduce the high costs of production optimization using machine learning for SMEs by empowering the workers themselves to apply the methods and algorithms of machine learning to the collected data.
approach
Thanks to artificial intelligence, an analysis system that complements ODION GmbH's machine data acquisition provides an intelligent tutor who supports the worker in the selection and application of machine learning methods and thus in the analysis of production data.
As an expert in production, the worker can easily identify interesting questions and scenarios and check the plausibility of the analysis results provided. Insights can then be incorporated into a further, more differentiated or completely different question or evaluation by means of the AI tutor.
The system should thus offer every SME the opportunity to use the know-how of the workers about their own production and machines in such a way that this knowledge is used to uncover even difficult and, due to the complex data burden for people, incomprehensible relations and facts within the production.
As an expert in production, the worker can easily identify interesting questions and scenarios and check the plausibility of the analysis results provided. Insights can then be incorporated into a further, more differentiated or completely different question or evaluation by means of the AI tutor.
The system should thus offer every SME the opportunity to use the know-how of the workers about their own production and machines in such a way that this knowledge is used to uncover even difficult and, due to the complex data burden for people, incomprehensible relations and facts within the production.
Recovery concept:
The ML methodologies implemented within the project based on the MoSes-Pro algorithms developed by ZeMA, supported by the assistance of an intelligent tutorial system, can be applied to various scenarios within production. Here, SMEs participating in the project can actively influence the decision as to which scenarios or which issues that are important or urgent for them should be considered.
So far, the following application scenarios have been identified:
• Identify complex relationships in production
• Anomaly detection
• Sensor self-monitoring
• Predictive maintenance
• Prediction of product quality
Contact person: Titian Schneider
Project management: Prof. Dr. Andreas Schütze
Duration: 01.03.2019 – 28.02.2021
Category: Multimodal smart sensing
EN 
DE 
FR 
