Abstract
Aiming at the problem that the existing cigarette machine detection technology has low precision in detecting the soft and hard points of cigarettes in the cigarette production process, an intelligent online soft and hard point detection method of the cigarette machine is designed. The method features segmented detection with adjustable segment numbers and lengths, real-time threshold adjustment, and an adjustable baseline. Application results show accurate calculation, rapid response, and convenient adjustment, achieving a detection accuracy rate of over 95%.
Introduction
In high-speed cigarette manufacturing, maintaining density uniformity is critical for physical properties like suction resistance and hardness, as well as chemical delivery of tar and nicotine. Fluctuations in tobacco quality or equipment pressure often create "hard points" (over-packed) or "soft points" (under-packed). Traditional detection systems typically calculate the average density of the entire cigarette and compare it against a fixed limit. This approach often fails due to the "bowl effect," where tobacco density naturally follows a distribution that is higher at the ends and lower in the middle.
Methodology
The proposed "Intelligent Online Detection Method" treats a cigarette as a series of segments rather than a single unit. The workflow involves high-speed sampling using microwave sensors to collect density data points per cigarette, which are then processed through a dynamic control logic.
The core steps include:
- Segmentation: Dividing collected data points into $n$ adjustable segments.
- Dynamic Baselines: Maintaining a moving average matrix based on the previous $m$ cigarettes to create a "rolling" expectation of density.
- Segmented Thresholds: Calculating Upper Control Limits (UCL) and Lower Control Limits (LCL) for each specific segment.
- Real-Time Comparison: Each new cigarette is compared segment-by-segment against these dynamic limits. If any segment falls outside the bounds, the cigarette is rejected.
- Automatic Recalibration: Periodic re-extraction of the density matrix to adjust for environmental or material changes.
Results and Discussion
The method was tested on 10 groups of 100 cigarettes each, using a TEWS MW4420 cigarette smoke density detector as the ground-truth reference. The intelligent method achieved an average accuracy of 98%, a significant improvement over the 55% accuracy of traditional absolute-limit methods. Furthermore, the consistency was much higher, with a standard deviation of 1.7% compared to 9.4% for the traditional approach.
The primary innovation lies in the system's spatial awareness and contextual sensitivity. By accounting for the natural density distribution of the cigarette rod and using a moving average to handle systemic shifts in raw materials, the system distinguishes between normal production variance and genuine defects. While highly effective, future work could optimize the computational overhead for ultra-high-speed production lines and refine the mathematical framework for segment optimization across different cigarette formats.
Conclusion
The intelligent segmented detection method provides a robust solution for identifying density anomalies in cigarette production. By moving away from static global averages toward dynamic, segment-based thresholds, the system significantly reduces false rejections and improves overall product quality.
Summary updated