基于拉曼光谱的电线绝缘材料老化状态评估

电线绝缘材料老化状态的准确评估有助于减少因电线绝缘老化引起的火灾，该实验基于拉曼光谱检测平台及自行搭建的老化设备，对13种电线绝缘材料(聚偏氟乙烯、聚丙烯、聚四氯乙烯、尼龙、亚大尼龙、聚氨酯、乳胶、聚全氟乙丙烯树脂、橡胶、聚乙烯、聚氯乙烯、硅胶、进口硅胶)进行加速温度老化以及加速紫外老化试验并定期检测，温度老化10个时间段，时间间隔为32 h,每个老化时间15个样本数据，获得温度老化的每种材料共150个样本光谱数据；紫外老化13个时间段，时间间隔16 h,每个老化时间15个样本数据，获得紫外老化的每种材料共195个样本光谱数据。依据老化时间段，温度老化分为10类，紫外老化分为13类，采用线性回归分类和支持向量机对原始光谱数据进行分类，两种分类算法准确率均在80%以上的材料有尼龙、聚氨酯、特氟龙、橡胶等，但部分材料的分类准确率却低于70%,在对原始光谱数据进行支持向量机分类时，由于样本数量多以及光谱维度高，支持向量机分类所需时间较长，为进一步提升分类准确率以及分类速度，对原始光谱数据进行迭代自适应加权惩罚最小二乘法、五点三次平滑等预处理方法，采用PCA压缩，样本光谱维数从2 048维降至...

Evaluation of Aging State of Wire Insulation Materials Based on Raman Spectroscopy

An accurate evaluation of the aging state of wire insulation materials can be used to reduce fire incidences caused by wire insulation aging. In this study, Raman spectrum detection platform, self built aging equipment, accelerated temperature aging and accelerated UV aging tests were applied to evaluate the aging state of 13 kinds of wire insulation materials(polyvinylidene-fluoride,polypropylene,polytetrachloroethylene,nylon,Yada-nylon,polyurethane,latex,perfluoroethylene-propylene-resin,rubber,polyethylene,polyvinyl-chloride). The samples were tested regularly based on temperature aging for 10 time periods. Using 32 hours interval and 15 sample data per aging time, the spectral data of 150 samples of each material (aged) were obtained. Similarly, 13 time periods of UV aging, at a time interval of 16 hours and 15 samples data per aging time, spectral data of 195 UV aging samples were recorded. According to aging period, temperature aging is divided into 10 categories, and UV aging was divided into 13 categories. Linear regression classification and a support vector machine was used to classify the original spectral data. It was found that nylon, polyurethane, Teflon, rubber, etc., have more than 80% accuracy of the two classification algorithms. However, the classification accuracy of some materials was less than 70%. The support vector machine classification of original spectral data consumed a longer time due to alarge number of samples and high spectral dimension. In order to further improve the classification accuracy and speed, the original spectral data were preprocessed by iterative adaptive weighted penalty least square method and five-point cubic smoothing. PCA compression was used to reduce the sample spectral dimension from 2048 to 3.Because the spectral dimension of the reduced sample is less than the number of samples, it can not meet the requirements of linear regression classification.So support vector machine was used for classification. After preprocessing and feature extraction, the classification effect of data was greatly improved, and the classification accuracy of temperature aging and UV aging of all the materials was more than 90%. Furthermore, the classification speed of the support vector machines has also been greatly improved. These results provide a theoretical basis for the effective evaluation of the aging state of wire insulation materials and provide technical support for preventing accidents caused by insulation aging.