In-situ process monitoring can generate a significant amount of data, which makes it difficult to identify relevant parts of the signal correlating to process features. To be able to identify and extract these parts, we employ traditional statistics and machine learning, developed both in-house and using external implementations. Using these tools, we can focus on not only extracting data, but also understanding data, and thereby the process itself.

When working with statistical and ML-based models, we always focus on both the scientific and industrial relevance. While a research-project might benefit from a more complex model, an industrial project might benefit from a more simple model to achieve faster turn-around times. In our work we therefore can use our broad and fundamental knowledge of ML-based and statistical models to select the model which fits the best on the current requirements, to achieve the best possible results independent on varying requirements and conditions.

​The laser process is highly complex in the sense that multi-physic interactions take place on a large scale in time and space. In order to better understand the process dynamics, numerical models have been developed and are continuously improved. The gained knowledge is used for better controlling the process by applying the appropriate monitoring strategy and process parameter selction.