为了提高生物发酵过程的控制与优化水平，试对发酵底物甘油和目标产物丁醇的含量进行快速无损检测进行研究。首先对 80份甘油与丁醇的单体系样本进行近红外扫描并进行光谱分析。为 提高模型的预测能力，分别采用了偏最小二乘法、间隔偏最小二乘法、向前间隔偏最小二乘法、向后间隔偏最小二乘法和窗口移动最小二乘法 5种定量校正方法建立甘油和丁醇含量的近红外检测模 型并对模型进行分析与比较。结果表明，向后间隔偏最小二乘法建立的模型效果较好，甘油和丁醇的单组份溶液的检测模型相关系数分别达到 0.99932（甘油）和 0.98843（丁醇）。为测量真实的 甘油发酵液中甘油和丁醇含量，搭建并建立了甘油和丁醇浓度监测平台，验证得相关系数分别达到 0.99074（甘油）和 0.99261（丁醇）。结果表明建立的近红外快速检测模型在检测的准确性和快 速性上均有优异的性能，为发酵行业快速检测提供了新的检测手段。

In this paper, to improve the control and optimization of the bio-fermentation process, rapid and non-destructive detection of glycerol and butanol was studied. To this end, 80 samples of glycerol and butanol were scanned and analyzed by means of near-infrared spectroscopy. Furthermore, in order to improve the predictive ability of the model, five quantitative calibration methods, namely, the partial least squares method, interval partial least squares method, forward interval partial least squares method, backward interval partial least squares method, and minimum forward interval method were used to establish the near-infrared detection model for glycerol and butanol content. The models were then analyzed and compared. The results showed that the model established by means of the backward interval partial least squares method was effective, and the correlation coefficients of the single component solution of glycerol and butanol were 0.99932 (glycerol) and 0.98843 (butanol), respectively. To measure the content of glycerol and butanol in the glycerol fermentation broth, a monitoring platform for glycerol and butanol concentrations was established. The correlation coefficients were found to be 0.99074 (glycerol) and 0.99261 (butanol), respectively. The results show that the NIR rapid detection model demonstrates excellent performance with regards to accuracy and rapidity and provides a new detection method for the rapid detection of glycerol and butanol in the fermentation industry.