Soft Sensing for pH Value of Raffinate Solution Based on Nonlinear Partial Robust M-regression
-
摘要: 提出了一种径向基函数网络(Radial basis function networks, RBFNs)与偏鲁棒M-回归(Partial robust M-regression, PRM)相结合的非线性PRM (Nonlinear PRM, NLPRM)建模方法, 用以解决鲁棒非线性系统建模问题. 该方法首先通过RBF变换获得扩展的输入数据矩阵; 接下来PRM算法通过反复迭代计算, 自适应地为变换后的数据分配不同的连续权值, 用以克服离群点对模型的影响. 本文通过仿真实验, 验证了方法的有效性; 并将其应用于湿法冶金萃取过程萃余液pH值软测量建模问题, 获得了相比于偏最小二乘法(Partial least squares, PLS)、PRM以及RBF-PLS方法更高的预测精度.Abstract: A nonlinear partial robust M-regression (NLPRM) modeling method combing radial basis function networks (RBFNs) and partial robust M-regression (PRM) is presented to solve the robust nonlinear system modeling problem. First, an extended input data matrix is formed by RBF transforming. Then, PRM algorithm is used, so that through iterative computation, consecutive weights are adaptively distributed to transformed data to diminish the effects of outliers on modeling. Simulation experiment is performed to show the effectiveness of this method. And it is utilized to develop a soft sensor model for pH value of raffinate solution in hydrometallurgy extraction process, and highly precise prediction results are obtained compared with partial least squares (PLS), PRM, and RBF-PLS methods.
计量
- 文章访问数: 2746
- HTML全文浏览量: 61
- PDF下载量: 1367
- 被引次数: 0