Image-based condition monitoring for modelling the temporal development of the wear condition of ball screws by means of machine learning

• Applicant: Professor Dr.-Ing. Jürgen Fleischer
• Subject Area: Production Automation and Assembly Technology
• Term: from 2020 to 2024
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 442210122

The aim of the research project is the image-based investigation and modelling of the wear condition of ball screws using methods of machine learning. The surface of the ball screw spindle shall be recorded by camera technology and the recorded images shall be evaluated by intelligent machine learning algorithms. The evaluation consists of the precise local detection, classification and measurement of surface defects on the spindle. To qualify the algorithms, training data must be generated and suitable models must be created. The evaluation takes place at regular intervals which ensures image-based, temporal observation of the development of the damage. The results are evaluated images of the spindle surface over the entire operating time of the ball screw including the development of the damage over time. The correlation of these evaluations with the operating time of the ball screw allows the modelling of the wear progress over time. The aspect of the often small amount of data in technical applications is to be taken into account by investigating the use of modern data augmentation techniques. The model will then be used for new ball screws to derive the remaining service life of the component from its current damage pattern. The validation is carried out with new ball screws and the comparison of the predictions of the model with the real operating time of the component.

• DFG Programme: Research Grants