中文摘要：

利用FTIR/ATR光谱技术建立糖厂废水化学需氧量（COD）的快速测量方法.基于光谱Savitzky-Golay（SG）平滑和偏最小二乘法（PLS）建立光谱优化模型.建立计算机算法平台,把540种平滑模式和1-40的PLS因子数任意组合分别建立SG-PLS模型,预测效果最优的模型平滑模式为2阶导数平滑,6次多项式,51平滑点数,其对应的PLS因子数（F）、预测均方根偏差（RMSEP）和预测相关系数（RP）分别为11,24.4 mg/L和0.968,预测相关很高,并且明显优于未做SG平滑的直接PLS模型.

英文摘要：

The rapid determination method of wastewater chemical oxygen demand（COD） for sugar refinery was established by FTIR/ATR spectroscopy.The optimal spectral analysis model was established based on partial least squares（PLS） and Savitzky-Golay（SG） smoothing method.By establishing the computer algorithm platform,for any combination of 540 SG smoothing modes and 1-40 PLS factors,the SG-PLS models were constructed respectively.The SG smoothing mode of the optimal degree po al model is the 2nd order derivative smoothing,6lynomial,51 smoothing points,and the corresponding PLS factor（F）,the root mean square error of prediction（RMSEP） and prediction correlation coefficient（RP） are 11,24.4 mg/L and 0.968 respectively.The prediction relation was very high and this model was obviously superior to the direct PLS model without SG smoothing.