砂磨分散的计算机预报

通过研究砂磨工艺参数对纳米二氧化钛分散效果的影响，建立了有关砂磨分散各项主要参数的数学模型．并对砂磨效果进行预报。研究发现使用基于支持向量机算法的模式识别方法较好，其对砂磨分散效果的预报误差比传统实验数据处理方法如多元回归、最小二乘法以及人工神经网络低一个数量级．并发现在较高介质填充率下．介质填充率对砂磨效果的影响将大于介质粒径．此方法能够更好地指导生产实践。

Computer prediction of sand milling

Effect of sand milling parameters on dispersing of nanosized titanium oxide in water was studied in this paper. A set of mathematic models concerning sand milling parameters were set up and sand milling results were predicted. Research results disclosed that the pattern recognition on the basis of support vector machine was better; and its prediction errors were lower than that of the traditional data analysis methods up to one order of magnitude, such as partial least-squares,artificial neural network and multivariable linearity regression.The effect of the filling ratios of sand medium on sand milling results was higher than that of the sizes and the new mathematic model on the basis of support vector machine could direct actually production process better.