(中南大學(xué) 信息科學(xué)與工程學(xué)院，長沙 410083)
摘 要: 由于浮選性能受多種因素的制約，適宜的礦漿pH值是高效泡沫浮選的關(guān)鍵。針對pH值在線檢測儀易受干擾、維護(hù)保養(yǎng)成本高等不足，結(jié)合泡沫浮選過程機(jī)理分析，以泡沫視頻圖像特征為輔助變量，將局部核函數(shù)和全局核函數(shù)加權(quán)組合，提高模型的學(xué)習(xí)和泛化能力，利用Schmidt正交化理論約簡多核矩陣，減小計算量，建立基于稀疏多核最小二乘支持向量機(jī)的浮選礦漿pH值軟測量模型。工業(yè)運(yùn)行數(shù)據(jù)測試結(jié)果表明：所建模型具有預(yù)測精度高、反應(yīng)迅速、穩(wěn)定性好等優(yōu)點(diǎn)，適于工業(yè)應(yīng)用。
關(guān)鍵字: pH值；軟測量；多核最小二乘支持向量機(jī)；稀疏性；泡沫浮選
sparse multiple kernels least squares support vector machines
（School of Information Science and Engineering, Central South University, Changsha 410083, China）
Abstract:The pH value of pulp can directly influence the mineral froth flotation efficiency. Considering the poor stability of detectors and serious manual detection time-delay, a novel soft-sensor of pH is proposed combined with the analysis of flotation mechanism and convex combination of Gaussian and linear kernel function based on the sparse multiple kernels least squares support vector machines using image features as instrumental variable. Furthermore, the kernel matrices were reduced by Schmidt orthogonalization theory to lower the computational complexity. The experiment has verified the presented model performs high prediction accuracy, high efficiency and good stability.
Key words: pH value; soft sensor; multiple kernels least squares support vector machines; sparsity; froth flotation
