摘要
在磷铵生产过程中,料浆的氟含量预测对生产具有重要意义。本文将径向基函数网络(RBFN)与循环子空间回归(CSR)相结合,设计了RBFN-CSR建模方法。RBFN-CSR方法在确定隐含层结构和参数时,将隐单元数取为训练样本数,径向基函数中心矢量取相应样本值,宽度参数根据样本分布情况采用尝试方法选取,隐含层到输出层的网络权系数运用CSR求解。CSR求解过程包容了最小二乘回归(LSR)、主成分回归(PCR)、偏最小二乘回归(PLSR)以及很多中间的回归方法,它可在非常广泛的范围内根据某一准则选择最优的网络结构参数。运用RBFN-CSR方法建立了酸性磷铵料浆浓缩过程中氟含量的预测模型,交叉验证表明,该模型具有较高的预测精度和良好的稳定性能,有一定的实际应用价值。
In the process of producing ammonium phosphate, predicting the fluorine content of acidic ammonium phosphate slurry is very significant. The radial basis function networks (RBFN) was combined with the cyclic subspace regression (CSR) in this paper, and a modeling approach by RBFN-CSR was designed. When confirming the structure and parameter of hidden layer in RBFN, the RBFN-CSR approach took the hidden layer unit number to the training swatch number, and took the center vector of radial basis function to the corresponding swatch numerical value, and chose the extent parameter by tentative method according the distributing of swatch, and solved the weigh coefficient between hidden layer and output layer by CSR. The whole solving process of CSR subsumes the least square regression (LSR), the principal component regression (PCR), the partial least square regression (PLSR) and many medial regression algorithms. According to a certain rule, the optimal structure and parameter of RBFN was selected from the very broad space by CSR. A prediction model of measuring the fluorine content of acidic ammonium phosphate slurry in the process of concentration was obtained by RBFN-CSR approach. The cross validation indicated the accuracy and stability of predicted value of the model was fine. So the model has certain of actual application value.
出处
《计算机与应用化学》
CAS
CSCD
北大核心
2005年第1期77-80,共4页
Computers and Applied Chemistry
基金
浙江省高校青年教师资助项目
关键词
酸性磷铵料浆
氟含量
预测模型
径向基函数网络
循环子空间回归
acidic ammonium phosphate slurry
fluorine content
prediction model
radial basis function networks
cyclic subspace regression
