Improving algorithms of video sequence data recognition for identification of transition processes in a floatation machine of potassium ore

Potash fertilizers are important for the Russian national agriculture and have become an export item. This fact results in increasing potash fertilizer production and improving potassium procession management. The object of research is floatation processes of potassium in the example of “Uralkaliy”, PJSC (Berezniki, Perm region). The aim of the research is improving algorithms of bubble recognizing in the video stream and using them to identify transient processes and situations in a flotation machine. Methods of researches include the system analysis, mathematical modeling, regression analysis, elements of automatic control theory and object identification. Algorithms for recognizing foam in the sylvinic floatation machine have been modified, which significantly increased the speed of recognizing bubbles in images of the low-quality video stream. Experiments were carried out on laboratory and industrial flotation machines, the results showing the possibility of using modified algorithms both in laboratory and industrial conditions. Video sequences of such quality were obtained and processed on the industrial floatation machine and could be used on the industrial flotation machine to identify situations and to control the machine operation. Using modified algorithms in experimental data processing allowed to identify the transient process and to clarify the time of the transient process. It has been shown that the obtained values are comparable with the data of other researchers and are not at variance with the experimental data. The error of bubble recognition has been estimated. The ways of using the data obtained for the decision support systems of the floater or of the automated control systems of the floatation machine have been shown.