| Abstract:In this study, a soft-sensing model of paper quality based on gradient boosting decision tree (GBDT) was proposed. This method could soft-measure the key physical indicators of paper such as tensile strength, softness and bulk online. The results showed that the average relative errors of tensile strength, softness and bulk when using GBDT for soft measurement of paper quality were 7.21%,7.38%, and 3.5%, respectively. Comparing the new data collected for verification, the average relative errors of tensile strength, softness, and bulk were 6.87%, 6.88%, and 3.12%, respectively, indicating that the model had high accuracy in predicting the new verification data, which could provide a reference for stabilizing product quality, optimizing the production process and reducing production costs. |
