Application of big data algorithms in fault analysis of air supply cut tobacco feeding in cigarette production lines

For real-time analysis of pneumatic feeding on cut tobacco running status, and to realize fast fault warning, data was collected over nearly three years from a tobacco factory cigarette machine, focusing on 12 key parameters of various production running states including dust removal airflow, tobacco feed negative pressure, tobacco feed air speed, tobacco feed frequency, air leakage coefficient, dust removal power, and valve opening degree. Utilizing the random forest algorithm for parameter selection, a fault prediction model for the air supply cut tobacco system was established through the SVM algorithm. Experimental results show the three parameters, namely the dust removal airflow, tobacco feed negative pressure, and tobacco feed airflow are the most significant factors contributing to air supply cut tobacco abnormalities. The built prediction model exhibited an accuracy of 86.87%. Application of the model in the cigarette production line led to a 72% reduction in cut tobacco delivery failure rate and reduced problem analysis time from 20 minutes to 3 minutes. This intelligent discrimination method of air supply cut tobacco feeding state holds significant application value for enhancing the intelligent manufacturing level of enterprises [1].