Feature extraction based on scalable hypothesis tests from photodiode data in laser welding processes

Abstract

Inline monitoring technologies gain increasing importance in industrial laser welding applications. Multiple sensing technologies can be implemented for classification of welding defects. Feature extraction from sensing data is a time-consuming process due to various algorithm possibilities for the identification of relevant features. Feature extraction based on scalable hypothesis tests (FRESH) allows for feature extraction with a combination of various time series characterization methods. FRESHs feature selection promises a quick extraction of relevant features from sensing data with an automatically configured hypothesis test. In the example of classification of spatter events, we show that FRESH can be used for the extraction of relevant features with photodiode sensing data in laser welding of copper.