基于核独立成分分析的发酵过程在线监测方法

提出了一种基于核独立成分分析（KICA）的发酵过程在线监测方法，该方法结合了发酵过程数据的特点，采用了一种新的过程监测指标U_s²，对发酵过程数据各时刻独立分量与该时刻所有批次独立分量均值的偏差信息进行特征提取，具有较强的抗干扰能力。青霉素发酵检测的实验结果表明，采用新监测指标的发酵过程监测方法能更好的识别较小的故障，降低漏报率，提高发酵过程在线监测的准确性。

A fermentation process monitoring method based on kernel independent component analysis

A fermentation process monitoring method based on kernel independent component analysis (KICA) is proposed, and the method is combined with the characteristics of batch process data with a new indicator being used. The indicator effectively extracts information about the deviation between independent components at each moment and their mean in the intermittent process, which overcomes the effects of disturbance more effectively than traditional indicators. The results of penicillin fermentation detection experiments show that fermentation process monitoring based on KICA with the new index is very effective. This method has a greater ability to detect small faults than traditional methods and a low false alarm rate, as well as giving improved accuracy during the monitoring process.