基于高斯混合模型的流程工厂设备集相似度计算

Similarity calculation between process plant equipment collections based on Gaussian mixture models

为了对流程工厂设备集之间的相似性进行度量,提出一种基于高斯混合模型的相似度计算方法。该方法从设备数量和设备构成两方面进行计算,再将二者结合起来得到统一的相似度指标。对于设备数量,利用Jaccard系数思想进行计算;对于设备构成,先通过分类树将设备集映射成二维坐标空间中的点集,再用高斯混合模型描述点集分布规律,最后基于高斯混合模型间的JS(Jensen-Shannon)散度计算相似度。通过流程工厂专家的评级打分及对设备工艺作用的分析,验证了计算结果的合理性和有效性。

To measure the similarity between process plant equipment collections,a similarity calculation method based on Gaussian mixture models was proposed.This method calculated from two aspects which were the amount of equipment in collection and the constitution of equipment collection,then the two factors were integrated to achieve an integrate similar indicator.For the amount of equipment,it was calculated by Jaccard coefficient thought;For the constitution of collection,firstly the collection was mapped to be points set in two dimensional coordinate space with classification tree,then Gaussian mixture models were used to describe the regularities of points distribution,finally the similarity was calculated with the JS divergence of Gaussian mixture models.According to the grading scores and the analysis of equipment process function from process plant professionals,the rationality and effectivity of the calculations were verified.