Predictive Analytics for Cigarette Manufacturing Machine Operational Status Using Machine Learning

Cigarette manufacturing machines operate continuously at high production speeds, making them vulnerable to operational disturbances, increased product rejects, and unplanned downtime. This study develops a machine learning-based predictive analytics model to classify machine operational status into three categories: Normal, Warning, and Failure using industrial sensor data. The dataset consists of 500 observations containing motor CPM, motor torque, motor current, gearbox vibration, gearbox temperature, panel temperature, sensor status, and reject data. Several machine learning algorithms, including Decision Tree, Random Forest, Gradient Boosting, and Bagged Trees, were evaluated. The experimental results show that the Bagged Trees model achieved the best performance, with an accuracy of 96%, precision of 0.92, recall of 0.80, and F1-score of 0.86 for the Failure class. In addition to status classification, the proposed framework identifies potential failure modes, including Tool Strain Failure (TSF), Heat Dissipation Failure (HDF), Power Failure (PWF), Overstrain Failure (OSF), and Sensor Malfunction Failure (SMF). To improve model transparency, SHAP (SHapley Additive Explanations) was applied to explain the contribution of sensor features to prediction outcomes. The results demonstrate that the integration of Bagged Trees and SHAP provides an accurate and interpretable predictive maintenance framework for early failure detection, root cause analysis, reduced downtime, and improved machine reliability.