Online Near-infrared Nicotine Detection Technology Study For Flue-cured Tobacco

Improving the monitoring and control capabilities of the production process,ensuring the stability of the flue-cured tobacco index and uniform chemical composition are modern threshing and redrying enterprises'pursuit of product quality.As the State Tobacco Monopoly Administration continues to increase the requirements for homogenization of threshing and redrying products,the chemical composition testing of threshing and redrying products has changed from sampling to full inspection,and laboratory testing to online inspection.Therefore,the near-infrared spectroscopy analysis technology,which has the advantages of fast,accurate,and low detection cost,has gradually become an important means for threshing and redrying enterprises to detect the composition of tablet tobacco products.However,tobacco leaves are a natural and complex product.The diffuse reflectance spectrum not only contains the physical and chemical information of the sample,but also carries a variety of complex background information such as detection conditions.Especially the poor environment of online detection will cause deviations in the detection results,In some production emergencies,detection data will be invalid.In order to solve this problem,this article takes the modeling of nicotine index as an example.Starting from the basic principles of near-infrared spectrum analysis and actual production conditions,a near-infrared spectrometer for online detection of flue-cured tobacco is designed and installed,and configured according to the production line situation The probe and related auxiliary facilities were used;then this set of equipment was used for sample collection of online flue-cured tobacco,and the experimental comparison of selected samples,singular sample rejection,spectral pretreatment,spectral optimization and modeling methods was used to determine the use of 1^stDerivative&MSC,the best method of using GA algorithm to optimize the wavelength and PLS modeling,established an online nicotine detection model with R² of 98.22,RMSECV of0.0821,and RPD of 7.33.At the same time,for the abnormal conditions such as head,tail,and material breakage in the actual production process,this paper uses a qualitative analysis method to discriminate the collected spectra.Through the comparison and use of modeling parameters,a correct calibration set is established.Qualitative analysis model with 99%accuracy and 97%accuracy of the validation set.Finally,in order to achieve the purpose of automatically removing the abnormal spectrum,the spectrometer scanning process control is used.In each scanning cycle,the average spectrum abnormality judgment is performed every 64 scans of the spectrum,and the abnormal spectrum is eliminated in time to ensure that the detection data can be accurate and real-time.To the label printing system.