Neural Network Model for Quality Indicators Assessment: Case of Paper Manufacturing Industry

Analysis of roll products at the final stage of production is the main stage for identifying defects, disrupted integrity or homogeneity, etc. Paper production is a typical example of such an operation in an industrial setting. It is proposed to use the results of the assessment for opacity, described by several standardized and statistical estimates, as the main characteristic for the quality of paper products. Studies of the dependence of the quality of the paper web on the production conditions and the properties of raw materials produce considerable variance, so that it is impossible to make accurate predictions. For this reason, we used a neural network modeling technology to develop an intelligent system for monitoring the quality of the paper web. Online quality control allows to assess the efficiency of the paper machine and rapidly adjust the manufacturing execution system. Special technologies such as computer vision systems can be introduced for this purpose, making it possible to make a transition from subjective assessment of the structure and defects of the paper web to obtaining objective quantitative estimates of these indicators. We considered a procedure for determining the estimates of structural heterogeneity of the paper web at the final stage of its production. We suggest to expand the classification of finished product samples by using neural fuzzy interpolation of linguistic values of such indicators. The approach introduced is aimed at improving the efficiency of the production process.