Optical dirt detection for a demand-controlled cleaning of woven filter cloth

To cope with ever rising quality demands while in constant transition to flexible product lines, cleaning should be considered an integral part of any process design and processing operation. For many unit operations feasibility has been proven. However, in the cross-sectional technology of solid-liquid separation, progress in hygienic design of filtration equipment did not keep pace. As an answer we propose a closed-loop control for cleaning intensity and present a method to evaluate the cleanliness of any woven filter cloth. The weave's periodicity is evaluated automatically, based on a reported algorithm. By means of image analysis, particle deposits are detected as irregularities on the near-regular weave pattern. Employing background subtraction, we do not rely on overall contrasts in color or brightness between cloth and soiling. If there is a contrast, we reliably find single particles (of 200 µm in our example), as we demonstrate on different color channels of RGB images. The proposed algorithm was implemented in Matlab and executed on a desktop PC in on-line speed. It will enable economical, safe and reliable cleaning of belt filters and filter presses.