针对氧化铝生产过程中无法在线测量苛性比值和溶出率的问题,建立了一种基于PCA-WLSSVM的预测模型.利用主元分析(PCA)消除样本共线性,降低样本维数.根据样本映射点到最小包含超球球心的距离确定样本的权值,以优化最小二乘支持向量机(LS-SVM)的参数,并提高加权LS-SVM的松散性和鲁棒性.仿真结果表明,此模型能有效地在线预测苛性比值及溶出率.
A prediction model based on PCA-WLSSVM is proposed to online measure the ratio of soda to aluminate (RSA) and the leaching rate (LR) in alumina manufacturing. Principal component analysis (PCA) is used to eliminate redundancy and reduce dimension of the samples. According to the distance from the sample's innuendo point to the core of the least hypersphere containing all the innuendo points, the weights are determined to optimize parameters of the least squares support vector machines (WLS-SVM) and to increase looseness and robustness of the weighted LS-SVM. The simulation result shows that the presented PCA-WLSSVM model can online measure the ratio of soda to aluminate and the leaching rate effectively.