Scholarlink

In order to improve the efficiency of insulation condition assessment of polypropylene (PP) cable insulation, a prediction model of AC breakdown strength of PP insulation based on near-infrared (NIR) spectroscopy was established to realize non-destructive and efficient detection of PP breakdown strength. In this paper, the NIR spectra of 26 PP samples with different aging conditions were obtained. 20 samples were used as calibration set and the 6 other samples were chosen as prediction set. The interval combination optimization (ICO) algorithm was used to select the key wavelengths with high correlation with breakdown strength from 151 variables. The partial least squares regression (PLSR) model with and without wavelength selection were established, respectively. The PP samples in the prediction set was further used to test the accuracy of the models. Only a 2.580% error is presented in ICO-PLSR model while a 3.362% error is shown in PLSR model without wavelength selection. Therefore, it is suggested that the model with wavelength selection is qualified to rapidly and precisely obtain the breakdown performance of unknown PP samples, which can be further used in sampling of new PP cables and aging assessment during their service.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A Non-Destructive Method for Breakdown Performance Prediction of Polypropylene Cable Insulation Based on Near-Infrared Spectroscopy by Interval Combination Optimization-PLSR Model

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Dangguo Xu,
  • Yamei Li,
  • Yanfeng Gao,
  • Zhaowei Peng,
  • Yi Lu,
  • Kaiying Chang,
  • Kangning Wu

摘要

In order to improve the efficiency of insulation condition assessment of polypropylene (PP) cable insulation, a prediction model of AC breakdown strength of PP insulation based on near-infrared (NIR) spectroscopy was established to realize non-destructive and efficient detection of PP breakdown strength. In this paper, the NIR spectra of 26 PP samples with different aging conditions were obtained. 20 samples were used as calibration set and the 6 other samples were chosen as prediction set. The interval combination optimization (ICO) algorithm was used to select the key wavelengths with high correlation with breakdown strength from 151 variables. The partial least squares regression (PLSR) model with and without wavelength selection were established, respectively. The PP samples in the prediction set was further used to test the accuracy of the models. Only a 2.580% error is presented in ICO-PLSR model while a 3.362% error is shown in PLSR model without wavelength selection. Therefore, it is suggested that the model with wavelength selection is qualified to rapidly and precisely obtain the breakdown performance of unknown PP samples, which can be further used in sampling of new PP cables and aging assessment during their service.