To solve the problem that the prediction accuracy of coal ash slagging behavior is generally low by single index, a model has been built up for the prediction purpose based on partial least squares (PLS) method and cross-validation theory, which includes four input variables, such as the softening temperature, base-acid ratio, percentage of silicon content and silica-alumina ratio, and one output variable, i.e. the slagging rate. A new concept of isotropic and anisotropic index is proposed, according to which the in- fluence of each index on the slagging behavior is qualitatively analyzed, and subsequently an evaluation cri- terion is obtained combined with relevant fitting equations. Measurement results show that the proposed PLS model is much more accurate in prediction than that of single index, proving the model to be reasonable and feasible.%针对煤灰结渣特性单一评判指标的预测精度普遍偏低的问题,基于偏最小二乘算法(PLS)和交叉验证理论建立了煤灰结渣特性预测模型,该模型有4个输入变量,即煤灰的软化温度、碱酸比、硅比和硅铝比,1个输出变量,即结渣程度.提出了指标的同向性和异向性的概念,依据所提出的概念对各单一指标对煤灰结渣特性的作用进行了定性分析与讨论,结合拟合方程,给出了煤灰结渣特性的评判依据.通过对测试样本进行检验,结果表明:所提出的PLS预测模型的预测精度远高于单一评判指标的预测精度,所建模型是合理可行的.
展开▼