Condition monitoring and diagnosis of ultrahigh-speed cigarette carton packaging machine based on operational mode recognition

Due to the multi-mode characteristic of cigarette carton packaging machine, the traditional monitoring method such as data report and manual observation can not meet the need of high-efficient cigarette production. Based on operational mode recognition, a condition monitoring and diagnosis approach of ultrahigh-speed cigarette carton packaging machine was proposed. The stability factor within a sliding time window and adaptive k-means clustering method was used to partition the operational space and form several steady mode sub-spaces, fully considering the multi-mode characteristic of operation process. The statistical monitoring model was established by utilizing principal component analysis (PCA) for each steady mode sub-space. The mode recognition of online data was realized by using stability factor within the current sliding time window. In steady mode, the matching monitoring model was obtained based on the similarity between the current valid data and each cluster center, then the monitoring statistics of online data was calculated based on the matching monitoring model. When any monitoring statistic exceeds the control limit, the fault cause variable was separated by using contribution plot. At last, the efficacy of the proposed method was illustrated by applying it to ultrahigh-speed cigarette carton packaging machine. Testing result showed that, the proposed method can more timely and effectively detect fault and separate fault cause variable.