CSIRO Process Science and Engineering, Lucas Heights Science and Technology Centre, Locked Bag 2005, Kirrawee NSW 2232 Australia.
Appl Spectrosc. 2010 Dec;64(12):1335-41. doi: 10.1366/000370210793561600.
Laser-induced breakdown spectroscopy (LIBS) and partial least squares regression (PLSR) have been applied to perform quantitative measurements of a multiple-species parameter known as loss on ignition (LOI), in a combined set of run-of-mine (ROM) iron ore samples originating from five different iron ore deposits. Global calibration models based on 65 samples and their duplicates from all the deposits with LOI ranging from 0.5 to 10 wt% are shown to be successful for prediction of LOI content in pressed pellets as well as bulk ore samples. A global independent dataset comprising a further 60 samples was used to validate the model resulting in the best validation R(2) of 0.87 and root mean square error of prediction (RMSEP) of 1.1 wt% for bulk samples. A validation R(2) of 0.90 and an RMSEP of 1.0 wt% were demonstrated for pressed pellets. Data preprocessing is shown to improve the quality of the analysis. Spectra normalization options, automatic outlier removal and automatic continuum background correction, which were used to improve the performance of the PLSR method, are discussed in detail.
激光诱导击穿光谱（LIBS）和偏最小二乘法回归（PLSR）已被应用于对来自五个不同铁矿床的原矿（ROM）铁矿石进行多种物种参数的定量测量，该参数称为烧失量（LOI）。基于来自所有矿床的 65 个样本及其重复样本的全球校准模型，LOI 范围在 0.5 至 10wt%之间，对于预测压制球团和散装矿石样本中的 LOI 含量是成功的。使用一个包含另外 60 个样本的全球独立数据集对模型进行验证，结果得到了最佳的验证 R²为 0.87，预测均方根误差（RMSEP）为散装样品的 1.1wt%。对于压制球团，验证 R²为 0.90，RMSEP 为 1.0wt%。数据预处理被证明可以提高分析质量。详细讨论了用于提高 PLSR 方法性能的光谱归一化选项、自动异常值去除和自动连续背景校正。
