基于经验模式分解与主元分析的耐火材料声发射损伤分类

Acoustic emission damage classification of refractories based on empirical mode decomposition and principal component analysis

耐火材料微观结构十分复杂,其破坏形式呈现多样化、复杂化,如何对其损伤进行高效准确地判别十分重要。为此,借助声发射技术对镁碳质耐火材料受压过程进行了损伤声发射信号采集,针对海量声发射数据,首先采用了经验模式分解结合主元分析,对声发射信号进行了特征提取与参数化降维:在大量的声发射时域与频域参数中,选取前7个主元来表征损伤,其损伤贡献率可以达到99%;之后将主元作为支持向量机输入进行损伤分类,将损伤分为两个主要类型,基质相与界面相损伤;最后将分析结果与扫描电镜结果进行对比分析,发现两者结论可以较好地吻合,可为运用声发射手段检测耐火材料损伤提供一种高效准确的解决方案。

The microstructure of refractories is complex,and the failure modes of refractories are diversified and complicated.It is very important to distinguish the damage of refractories efficiently and accurately.In this paper,the damage acoustic emission signals of magnesia-carbon refractories under compression were collected by means of acoustic emission technology.Firstly,the feature extraction and parametric dimensionality reduction of tremendous acoustic emission signals were carried out by empirical mode decomposition combined with principal component analysis.A large number of parameters in time and frequency domains of acoustic emission were selected.The first seven principal components represent the damage,and the damage contribution rate can reach 99%.Then the principal components were used as the input of support vector machine to classify the damage.The damage was divided into two main types,for matrix phase and for interface phase.Finally,the analysis results were compared with the results of scanning electron microscopy.The two conclusions are in good agreement.This study can provide an efficient and accurate solution for the detection of refractories damage by acoustic emission.