As a solution to operating status monitoring of ultrahigh-speed cigarette packer, a monitoring and diagnosis method based on operating status and model matching was proposed via analyzing its running characteristics. At offline modeling stage, steady status and transition status were defined through calculating the stability factor within a sliding time window, the data of steady status were classified into several categories with self-adaptive k-means clustering method, then principal component analysis was employed to establish a statistical monitoring model for each steady status data category. At online monitoring stage, the type of operating status was identified by the real-time stability factor within sliding window. In case of transition status, the monitoring statistics was defined as 0. In case of steady status, in order to create a matching PCA model for real-time monitoring, the Euclidean distance between current valid data and each cluster center was computed; for any monitored statistic exceeding the limit, the variable which caused machine failure was isolated by contribution plots. Off-line validation was conducted on the basis of actual operating data, the results showed that the proposed method was adaptable for the change of operating status of ultrahigh-speed cigarette packer, it could detect machine failures timely and isolate failure causes effectively, and the level of failure monitoring and diagnosis of cigarette packer was promoted.%为解决超高速小盒包装机中采用传统数据报表和人工目测等方式进行设备状态监测，无法满足高效卷烟生产等问题，通过运行特性分析，提出了一种基于工况划分和模型匹配的超高速小盒包装机故障监测与诊断方法。离线建模阶段，计算滑动时间窗口内的稳定度因子以识别稳定工况和过渡工况，采用自适应k-means聚类方法对稳定工况数据进行划分形成若干个稳定工况数据类，再利用主元分析方法对每类稳定工况数据建立统计监测模型。在线监测阶段，根据当前滑动窗口内的稳定度因子判断工况类型，若为过渡工况，则将监测统计量赋值为0；若为稳定工况，计算当前有效数据与各个聚类中心的欧式距离，获得匹配的PCA（Principal Component Analysis）监测模型进行实时监测，任一统计量超限时采用贡献图分离故障原因变量。基于设备实际运行数据进行离线验证，结果表明：该方法能够适应超高速小盒包装机的运行工况变化，及时检测出设备故障并有效分离出故障原因变量，提高了小盒包装机多工况过程的故障监测与诊断水平。
展开▼
