基于动态数据的上浆率在线增量软测量

Incremental online soft sensor for sizing percentage based on dynamic data

针对浆纱过程产生的实时动态数据,提出一种基于增量学习的在线软测量建模方法,实现对上浆率的预测.将增量学习的思想引入软测量算法,去除冗余数据,提高算法效率;使用改进式山峰算法确定数据中心,通过自适应方法确定去噪半径,完成对噪声数据的筛选;选取软测量算法进行建模.实验所用数据采集自真实浆纱过程.仿真结果表明:该算法预测精度较高,具有一定的抗噪性能,均方根误差最小可达0.263 3,最大绝对误差最小为0.633 1,适用于多种智能算法的在线更新.

b Aiming at the dynamic data come from sizing process, a new incremental online algorithm was proposed to establish soft sensor model for the prediction of sizing percentage. In order to improve the efficiency of algorithm, incremental learning method was introduced to reduce the redundant. The improved mountain method was used to determine the center point of data, and with the help of adaptive radius, the noisy data had been deleted. Finally, the soft sensor method was chosen to build the model. The experimental data came from the real sizing process, the simulation result demonstrated that the soft sensor model based on incremental online algorithm has the best accuracy and anti-noisy performance. The minimum of RMSE is 0.263 3 and the minimum of MAE is 0.633 1. At the same time, it is suitable for the online update of a variety of intelligent algorithms.