基于统计空间映射的模式识别及在线质量推断

A Pattern Recognition Method based on Statistical Mapping Space

采用回归分析的方法，建立特征变量与产品质量之间的统计对应关系，把产品质量表达成特征变量的回归函数，进而得到特征空间与产品质量空间在统计意义上的映射关系．在产品质量空间进行聚类，在特征空间进行分类，而后提出了一种基于统计空间映射的在线模式识别方法．利用唐钢烧结厂的实测数据进行了仿真，验证了本方法的正确性．从算法分析和仿真结果看，这一算法可以有效地克服模式交叉现象的影响，并可对复杂生产过程进行在线质量推断．

The statistical relationship between feature vector and quality index is built by regress analysis and then the quality index is expressed to the regress function of feature vector. By this means, a statistical mapping relation between in.feature space and quality index space is built and a online pattern recognition method based on the statistical mapping space is provided through the clustering in quality space and classification in feature space. The validity of the method is verified by the simulation results of the data from Sinter factory of Tangshan Steel Corporation. From the algorithm analysis and simulation results, this method can effectively overcome the pattern intercross and can be used for complex production process quality online prediction.