基于便携式激光诱导击穿光谱技术的耐热钢老化等级评估

长期承受高温高压耐热钢的运行状态严重影响着设备系统的安全生产,传统检验手段在时间和操作上存在一定的局限性。将开发的便携式激光诱导击穿光谱(LIBS)设备应用于耐热钢失效分析,结合化学计量学方法实现T91钢老化等级的预测评估。首先对不同老化等级T91样品的光谱特性进行了分析,研究了不同表面状态样品的元素特征谱线随脉冲数变化趋势,获取各样品的典型光谱;然后采用K折叠支持向量机递归特征消除算法(K-SVM-REF)对光谱变量进行特征选择,以提高模型预测性能。结果表明基于特征选择的光谱变量较全谱变量建立的支持向量机(SVM)预测模型测试集的分类准确率从84.38%提升到90.63%。此外还研究了样品表面状态对模型性能的影响,为开发的便携式LIBS设备用于实现金属受热面失效诊断的实际测量提供了有效的依据和方法。

Estimation of the aging grade of heat-resistant steel based on portable laser-induced breakdown spectroscopy

The state of heat-resistant steel that has been subjected to high temperature and high pressure for a long time seriously, would affect the safe production of equipment systems. Traditional inspection methods have certain limitations in time and operation. The developed portable laser-induced breakdown spectroscopy (LIBS) equipment was applied to the failure diagnosis of heat-resistant steel, and combined with chemometric methods to predict and evaluate the aging grade of T91 steel. Firstly, the spectral characteristics of T91 samples with different aging grades were analyzed. The variation trend of the element characteristic spectral lines of samples of different surface states with the number of pulses was studied, so that the representative spectra of each sample were obtained. Then the spectral variables were selected by the K-fold-support vector machine-recursive feature elimination (K-SVM-REF) to improve the prediction of the model. The results showed that the classification accuracy of the support vector machine (SVM) prediction model based on spectral variable of feature selection compared with full spectrum variables was improved from 84.38% to 90.63%. In addition, the influence of the sample surface state on the performance of the model was also studied, which provided an effective basis and method for the practical measurement of the failure diagnosis of the metal heating surface with portable LIBS device developed.