一种消除在线多通道近红外分析仪各通道光谱差异的方法

针对在线多通道近红外分析仪因光纤耦合器件加工精度和装配过程存在细微差异而引起通道间光谱不一致的问题，在对光谱差异进行解析的基础上，提出了一种运算简捷、且在实际应用中易于实现的平均光谱差值校正(MSSC)方法，并与常用的模型传递算法如斜率／偏差(S／B)算法、分段直接校正(PDS)算法，以及通过偏最小二乘-人工神经网络(PLS-ANN)建立多通道混合校正模型进行了对比。结果表明，该方法可有效消除各通道所测光谱之间存在的差异，实现了多通道分析模型的通用性。

A Method to Eliminate Spectral Differences among Multi-channels of an On-line Near Infrared Spectroscopy Analyzer

A spectral pretreatment method, mean spectra subtraction correction method (MSSC), was proposed exclusively to remove spectral difference ~among multi-channels caused by small coupling differences ~among fibers and multiplexer in an on-line near infrared (NIR) analyzer. The goal of the MSSC pretreatment is to perform a calibration built in a channel to predict spectra of other channels directly without any systematic error. The prediction statistics show perfect consistent among the eight channels without any systematic errors by use of the MSSC method. Some typical calibration transfer algorithms such as piecewise direct standardization(PDS) and slope/bias (S/B), and developing a multi-channel global model by partial least squares-artificial neural network (PLS-ANN) method were also attempted. Compared with those methods, the proposed MSSC method gives better results for all the eight channels. Moreover, the MSSC method is relatively simple and convenient to be operated in practical applications.