基于Bayes判别模型的火场中铜导线短路熔痕 定量金相鉴定方法研究

借助计算机图像处理技术,改进了定量金相分析方法,测量了20个一次短路熔痕和20个二次短路熔痕的金相组织,对数据进行了主成分分析,提炼出了晶粒特征因子、气孔特征因子和复合因子三个主成分,基于Bayes判别法建立了火场中短路熔痕定量金相分析判别模型,经自身检验和交互检验,准确率均高于80%,充分满足鉴定要求,为更加有效地应用定量金相法鉴别短路熔痕提供了理论依据。

Quantitative metallography identification method of copper conductor short-circuit molten mark based on the Bayes discriminate analysis model

Based on image processing and quantitative metallography principle, a quantitative metallography method was developed and applied to identify short-circuit molten marks in fire. Using this method, we measured 20 primary short-circuit molten marks and 20 secondary short-circuit molten marks for testing samples. By principal component analysis, three principal components, including the grain characterization factor, the pores characterization factor, and the complex factor, were extracted. By the Bayes discriminate analysis, a model for identifying copper conductor short-circuit molten marks was established. By self-consistency validation and cross validation, it was shown that the recognition accuracy of this model was above 80%.