Detecting a boundary layer using a machine learning algorithm

A method for detecting a boundary layer of a product in a container, comprising:

• providing a measuring system for detecting the boundary layer of the product in the container, including:
• a radar-based measuring device, including:
• a transmission unit designed as a measuring probe extending into the container via which high-frequency signals can be transmitted in a direction of the product and, after reflection at the product surface, can be received as received signals;
• a signal generation unit that is designed to generate according to a time-domain reflectometry (TDR) method the high-frequency signals to be transmitted;
• and
• a receiving unit that is designed to record the received signals;
• and
• an evaluation unit in which a machine learning algorithm is formed in order to detect the boundary layer on the basis of the received signals;
• wherein, in a learning phase, the machine learning algorithm is exposed to teaching situations with known measuring system parameters, including:
• a fill level of the product, a thickness of the boundary layer, a thickness of an additive product layer, a dielectric value of the product, and a dielectric value of the additive product layer,
• emitting the high-frequency signals according to the TDR method via the transmission unit into the container;
• recording the received signals after reflection of the high-frequency signals;
• and
• evaluating the received signals via the machine learning algorithm such that the boundary layer is detected, wherein the evaluating includes:
• determining along the measuring probe a distribution of an attenuation coefficient and/or of a dielectric constant in the container;
• and
• determining from the distribution of the attenuation coefficient and/or of the dielectric constant in the container a mass or a volume fraction of the boundary layer.