应用模式识别法区分镁水泥耐水性能的研究

Computerized Pattern Recognition Applied to Study of Distinction of Water-Resistant Property of Magnesium Oxychloride Cement

以镁水泥组分及13种添加剂为特征参数,试样在空气中养护28天与水中浸泡一定周期的强度相对变化值表征镁水泥的耐水性能,应用模式识别法对掺入各种添加剂的镁水泥就其耐水性能的优劣在高维空间进行分类,并将高维空间中的点集压缩到二维空间,结果表明,不同耐水性能的训练点分别落入二维空间的不同区域.利用这个分布规律对耐水性能未知的八组镁水泥试样进行了耐水性能的预报.实验数据与预报结果一致.

The computerized pattern recognition for distinction of resistance to water of magnesiumoxychloride cement doped with various additives has been done by using compositions and 13additives as the characteristic parameters and by using resistance to water of the cement ascriterion,which is determined by the changes in compressive strength of magnesium oxychloridecement measured on the same samples hardening for 28 days in air and immersing in waterfor 3 months.The distinction has been realized in both high-dimensional and two-dimensionalspace.The result shows that the training-points with different water-resistant properties ofthe cement are clearly located in two regions in two-dimensional space.The water-resistantproperties of 8 samples are predicted by using the distribution rule,which consists with theexperimental results.