The Fraunhofer Institute for Process Engineering and Packaging IVV (Freising, Germany) developed a machine vision-based approach to jet cleaning industrial tanks, such as those used in food and beverage processing.
The institute, which specializes in applied research on industrial processes, plans to test version 1.0 of the system at a customer’s site beginning as early as February 2026. The customer will likely be either a brewery or dairy.
Fraunhofer also is developing version 2.0, which will be fully automated and AI-enabled, supporting continuous improvement in cleaning efficiency as the algorithms learn.
Inefficient Legacy Tank-Cleaning Processes
Typically, manufacturers perform tank cleaning based on the worst-case scenario. A standard program is performed after a tank has been in use for a specified number of hours. The standard cleaning process is the same for all tanks, which vary in size from 3,000-50,000 L and may have different amounts of contamination. The entire program includes a pre-rinse, several cleaning cycles, and final disinfecting cycle.
Because the cleaning isn’t customized, manufacturers often clean already-clean areas repeatedly while missing contamination stuck in hard-to-clean areas such as the fill line, sensor connectors, and manholes.
“Basically, all tank cleanings that are happening in the world right now are wasting extraordinary amounts of resources,” explains Max Hesse, chief engineer for processing and cleaning systems at Fraunhofer.
Related: Optimizing Machine Vision Lighting for Food and Beverage Inspection
Hesse and other team members at Fraunhofer set out to change this dynamic. They wanted to cut the volume of resources—such as water and chemicals—used in the cleaning process as well as the amount of time involved. They focused their work on the pre-rinse and cleaning processes, which occur before the final sanitation cycle.
An Adaptive Approach to Tank Cleaning
The Fraunhofer approach uses the geometry of the tank and level and location of contamination to guide the cleaning process for a specific tank.
Fraunhofer developed an adaptive cleaning nozzle, called AJC. It is in commercial use at food processing companies in Europe and the United States.
AJC has two independent and freely rotatable axes that allow cleaning paths to be adjusted based on the geometry of a tank. The AJC can reach every corner of the tank, including difficult-to-clean areas of a tank. Through control software, an operator can select a specific cleaning path and motion, such as zigzag or spiral.
The pre-programming method reduced cleaning time by about 25% compared to traditional methods.
The Importance of a Machine Vision Approach
Better cleaning coverage of a tank was only a partial solution, Hesse says. What was missing was visual in-line intelligence that is updated as the tank cleaning process occurs. That’s where AJCsens 1.0 comes in. It includes a custom and miniaturized vision system that is integrated into the cleaning head of the AJC. System control is managed by hard-coded software algorithms.
Related: Harnessing UV-C LEDs for Advanced Inspection and Quality Control
The camera system combines visible and UV image data to detect food residue on the surface of the tank. It does not detect individual microorganisms.
