Fraunhofer Center for Silicon Photovoltaics CSP
The time courses of the individual sensors are sorted and classified according to their similarity.
Sensors record the course of the processes.
We develop analysis methods with the help of training data, consisting of process data of successful and less successful production processes. The time courses of the individual sensors are sorted and classified according to their similarity. Based on the resulting similarity maps, machine-learning algorithms can learn the relationship between the course of events and failure probabilities. For new processes, the current failure risk can thus be predicted at any time. If an increase in risk is detected, countermeasures can be initiated immediately.