Testing technology is an elementary component of industrial production. The testing phase ensures that specified limits are not exceeded and that legal standards are met. For the production of electric vehicles, many innovative production steps have to be introduced outside of already existing structures. Consequently, equally innovative test procedures adapted to the new tasks are indispensable.

The high-voltage battery unit (HVB) of an electrically powered vehicle is subjected to a test for possible leaks during the leak test and, in the event of a find, is subsequently forwarded to the leak detection and seal step. To ensure a high degree of safety, the volume flow of the leakage must not exceed a prescribed limit. For this purpose, the goal is to develop a method in which gas leaks from a pressurized HVB can be detected and quantified with the aid of a highly specialized camera guided by a robot. A neural network is to be used that can locate a leakage based on the image or video recordings and quantify the volume flow. For this purpose a convolutional neural network as well as a long-short term memory neural network are trained for a regression task while also improving the experiment setup to create a superior dataset.