KHS will present a new, self-learning filling valve using artificial intelligence at BrauBeviale.
The feasibility of this flexible valve has been verified by KHS in the DnSPro research project, which also involved Infineon, WIBU Systems, EPOS, the Ruhr University in Bochum and Ostwestfalen-Lippe University of Applied Sciences.
With six other partners, the project aims to develop a self-learning filling valve with which beverage producers can fill all liquids into all existing types of container.
This would do away with the need for manual conversions and the effort required for operation and maintenance would be greatly reduced, says Jochen Ohrem, expert of R&D
Management at KHS in Bad Kreuznach, Germany.
“We developed cyber-physical systems for this purpose, with the help of which the valve can determine how to best fill a certain beverage into a certain container as quickly as possible,” Ohrem continues.
The filling process is analyzed with the assistance of a camera.
This continuously monitors the inclusion of bubbles and foaming to prevent excessive foaming and thus product loss.
Using microcontrollers and the camera’s evaluation electronics, the filling valve is opened to varying degrees by a stepper motor depending on the fill level.
“The focus was on ‘learning’ a number of skills: self-configuration, analysis, self-diagnosis and, ultimately, self-optimization,” explains Ohrem.
The future objective behind all this is to increase flexibility and energy and resource efficiency in production through the application of an autodidactic system of artificial intelligence.
KHS will be presenting the key data on this intelligent filling valve which fully satisfies all of the previously specified project requirements at BrauBeviale.
“The development is now entering the next phase where we’ll be gathering further experience with this prototype,” Ohrem states.
Instead of a filling computer centrally positioned on the machine that regulates the process of all valves, in the future this task is to be managed locally by miniaturized computers installed on each valve group.
This would allow a simple sensor such as a pressure sensor to be inserted into each filling valve which documents and analyzes the pressure curve, resulting in a process which optimizes itself.