乳酸是谷氨酸发酵过程中的主要副产物,及时准确地检测其浓度变化对谷氨酸发酵过程控制和优化具有重要意义。该文采用近红外光谱技术结合偏最小二乘的方法建立并优化谷氨酸温度敏感突变株发酵过程中乳酸浓度预测模型。在光谱预处理为一阶导数+矢量归一化(1st+SNV)、波数为5 500~7 500 cm-1+4 200~4 900 cm-1条件下获得最优乳酸浓度预测模型。该模型交叉验证误差均方根(RMSECV)、决定系数(R2)以及剩余预测偏差(RPD)分别为0.482 g/L、0.912和5.98。分别以谷氨酸正常发酵和低溶氧发酵条件下获得的发酵液验证该模型的准确性和可靠性,与实际浓度相比,该模型R2分别为0.914和0.923,平均相对误差分别为7.86%和6.58%。上述结果证明,该预测模型能够很好地检测发酵过程中乳酸浓度,可为谷氨酸温度敏感突变株强制发酵过程中乳酸浓度实时监控及发酵过程优化提供理论和实践依据。

Lactate is a main by-product during glutamate fermentation and is important for control and optimization of the process. In this study,the calibration models for monitoring concentration of lactate in the temperature-triggered glutamate fermentation process were developed by the near infrared spectroscopy and partial least-squares regression. The models were developed and optimized by selecting different spectral pretreatment methods and wave number regions. The spectral pretreatment method was first derivative + vector normalization and the wave number region was5 500 ~ 7 500 cm- 1+ 4 200 ~ 4 900 cm- 1. The root-mean square error of cross-validation( RMSECV) of lactate,determination coefficient( R2) and predicted residual deviation( RPD) were 0.482 g/L,0.912 and 5.98,respectively. These results showed that the lactate model had good predictive ability. Fermentation liquor samples from normal fermentation and low dissolved oxygen fermentation were used as external validations to check the model. Compared with the predict values,the determination coefficient were 0. 914 and 0. 923,respectively. And the average relative errors were 7. 86% and 6. 58%,respectively. These results showed that the model could predict and monitor lactate concentration during glutamate fermentation process accurately and quickly,and it will provide theoretical basis for the real-time control of lactate concentration during the temperature-triggered glutamate fermentation process.