Machine learning techniques are increasingly used in monitoring and optimizing industrial production due to recent advances in computing capacities as well as the spread of developer-friendly software libraries. Essential challenges such as high initial manual effort as well as the general data quality remain as obstacles to widespread use.
DualSys strives to replace labor-intensive, human-related activities such as the annotation of process data and the selection of suitable preprocessing methods or to relocate them to later project phases. Furthermore, the special potentials of unsupervised and self-monitored learning processes to solve the problem of the subjective ground truth are examined. The aim is a holistic approach that spans industrial manufacturing processes as well as production systems in the form of application scenarios for process optimization, process monitoring and the prediction of product quality.
With the help of the developed methodology as well as developed procedures and software libraries, a return on investment can be achieved even in previously unprofitable use cases and companies can be enabled to use ML across the board.