Study On Soft Sensing Method Of Permeability Of Tuna Paper Based On Least Squares Regression

At present, the hazards of tobacco products on human health has become one of the hottest global issues, and the tobacco enterprises have paid more efforts to develop various techniques to reduce the intakes of tar in cigarettes for smoking group. Among them, the laser perforating technology for tipping paper (i.e., a special kind of paper wrapped around the cigarette filter) has aroused the widespread concerns of tobacco enterprises, and which is gradually becoming a major approach to reduce the intakes of tar especially for middle-and-high grade products. In order to satisfy the requirements on porosity for different kinds of products, it must ensure that the laser powers and optical systems can be adjusted at any time when perforating for the tipping paper. It is important to note that the porosity can be defined as the volume of air through the tipping papers in unit time、unit area and unit differential pressure (international units for CU). The literature surveys show that the holes’areas in the observing region is a key factor for calculating the porosity, but how to make use of the limited information deriving from the area of holes is still a research difficulty to realize both the high accuracy measurements for porosity and its model solution.This dissertation researches on the factors that affect the porosity of tipping paper, and then least squares regression (LSR) is utilized from qualitative analysis according to the characters extracted from the holes’areas and sample data. Moreover, an effective measuring model based on the differential evolution and least squares regression is proposed for porosity of tipping paper. The main works of this paper can be listed as follows:(1) This dissertation reviews the research and development of measuring porosity technology for tipping paper in the domestic and other countries and compares the difference between online measuring and offline measuring of porosity.(2) On the basis of qualitative analysis and adopts the data which can be obtained, selects holes’areas as a main factor which affect porosity of tipping paper. For the reason that holes’areas of tipping paper cannot be measured directly, this paper introduces the basic knowledge of image processing technology and extracts the key feature data to establish the soft measurement model by processing and analyzing the capture image of tipping paper.(3) Selects the LSR as the detection model according to the theory of regression analysis. The model contains certain nonlinearity between holes’areas and porosity according to actual measured data which obtained by designed detecting equipment. Then, combines the characteristics of linear and nonlinear modeling of LSR, and the nonlinear LSR is selected as the soft measurement model. For the uncertainty problem of LSR initial values, this paper used the weighted orthogonal least squares regression (WOLSR) to develop a detection model of porosity and then, can use this model for calculating porosity. Simulation experiments based on actual production datum show that WOLSR model is better than the classical nonlinear LSR.(4) For the problem of the WOLSR see the squared error and minimum as the objective function may cause low measuring accuracy. This paper by using improved differential evolution (IDE) algorithm to optimize the WOLSR model parameters, then the combination model of IDE-WOLSR is derived, which can greatly enhance accuracy of measuring and has important practical significance. The test results of the actual production datum show that the proposed method is real effective.The IDE-WOLSR model of new detection device has been applied in a domestic factory of tipping paper, and it woks well.