Cigarette Factory Equipment Health Assessment System Design And Implementation

The drying process is a critical step in the cigarette manufacturing process that dehydrates,dries and shapes the tobacco leaves to achieve the required moisture content for cigarette manufacture and ensures the quality of the tobacco.The roaster,as the core piece of equipment in the drying process,plays a critical role in the quality of the tobacco.At present,the inspection of the equipment status in the cigarette production workshop still relies on manual inspection,which has disadvantages such as poor timeliness,heavy workload and dependence on the experience of the inspectors,and lacks a scientific and reasonable evaluation system for the operating status of the drying machine equipment.In addition,bearing components are an important part of the drying machine and other mechanical equipment,which have a significant impact on the operating status of the equipment.However,in actual industrial production,the data generated by bearing components has high dimensionality,non-linearity,lack of fault data,noise,and changes in operating loads.Traditional machine learning methods require manual selection of fault features,which cannot guarantee reliability.Therefore,this article focuses on the following research:(1)To address the problem of relying solely on experienced inspectors to manage equipment health in the cigarette production workshop,this paper proposes an equipment health assessment model based on variable weight hierarchical analysis.Experimental results using test cases in five states show that the average evaluation accuracy of the model is 96.6%,which is 28% higher than the fixed weight evaluation.(2)To solve the problem of insufficient fault data,noise and load changes during the operation of bearing components,which make it difficult to identify some faults,this paper proposes a Siamese CNN fault detection model and conducts experiments on the CWRU data set.The results show that the average accuracy of the model under four working conditions reaches 99.6%,which is 0.4% higher than the CNN model.Even with fewer training samples and more noise,the detection accuracy of the Siamese CNN model is higher than that of the CNN model.Under six different loading conditions,the average detection accuracy of the Siamese CNN model reaches 98.87%,which is 3.98% higher than the CNN model,and the detection accuracy under all six loading conditions is over 97%.The experiment fully verifies the effectiveness of the model.(3)Most commonly used fault detection models belong to the supervised learning category and require a large amount of labelled data for model training.However,when the labelled data is insufficient or missing,the performance of these models often falls short.Therefore,this paper proposes an unlabelled defect detection model based on the self-supervised contrastive learning method Sim CLR.Experimental results show that under four different working conditions,the average accuracy of the Sim CLR model reaches 99.2%,which is 0.7%higher than the Res Net-18 model in supervised learning.Even with fewer training samples,the detection accuracy of the Sim CLR model is still higher than that of the CNN model in Chapter 3,with a final average accuracy of 89.5%.Under six different load conditions,the average detection accuracy of the Sim CLR model reaches 92%,which is 2.8% higher than the Res Net-18 model.The experiment fully demonstrates that even in the absence of labelled data,the Sim CLR model can achieve relatively good detection accuracy.(4)Based on user needs and application scenarios,system requirements analysis and system architecture design are performed to determine specific functional modules.Then the system design and development is completed and each functional module is tested to realise a drying machine equipment health assessment system based on B/S architecture.