基于递推PLS的置换蒸煮终点软测量建模研究

Research on Cooking End Point Modeling Based on Recursive PLS for Displacement Cooking

中文关键词:  置换蒸煮  软测量建模  蒸煮终点  递推PLS

Key Words:displacement digester  optical fiber on-line measurement  cooking end point  multivariable binomial regression

中文摘要:

      基于离线数据得到的卡伯值软测量模型无法随蒸煮过程变化而更新,针对这一问题,应用基于移动窗口递推的偏最小二乘法(PLS)建立了新的卡伯值软测量模型。通过对制浆过程和蒸煮终点影响因素的分析,确定了建模所需的辅助变量；利用特定大小的移动窗口中的最新数据来更新该测量模型,能够确保模型对蒸煮过程的跟踪能力。仿真结果表明,与传统的PLS模型相比,该模型可更好地预测置换蒸煮的蒸煮终点。

Abstract:

      Kappa number model based on offline data can not be updated with the pulping process changes. By combining rolling window with recursive partial least squares (PLS), a new adaptive Kappa number model was proposed in the paper. In order to determine the supplementary variables for modeling instrumental, the pulping process and the factors influencing the cooking end point were analyzed. The model could be updated by using the latest data in a rolling window with specific dimension, if could ensure the ability of the model to follow the track of cooking process. Simulation results with real industrial data showed that the adaptive Kappa number model had higher precision and better prediction of cooking end point of displacement cooking than the traditional PLS model.