基于电子鼻和高斯过程的秸秆固态发酵过程监测技术

基于电子鼻监控数据,建立基于高斯过程的状态监控分类器,实现对秸秆饲料固态发酵过程的有效监测。秸秆饲料固态发酵过程实验周期为7 d,每隔24 h利用电子鼻系统对发酵气体监测数据进行采集。该发酵实验共分20批次,其中10批次实验数据用来训练高斯过程分类器,其余10批次实验数据用来测试所训练分类器的性能。实验结果表明所采用电子鼻系统可以对秸秆饲料固态发酵过程状态进行有效监控。将所训练高斯过程分类器与支持向量机、神经网络分类器进行比较表明,基于高斯过程分类器的正确率为100%,高于基于支持向量机、神经网络分类器的正确率85.71%、94.29%,能够更好地实现对秸秆饲料固态发酵过程的监测。

Monitoring of Straw Solid-state Fermentation Based on E-nose and Gaussian Process

The e-nose system and Gaussian process (GP) classifier were used to accurately monitor physical and chemical changes in solid-state fermentation (SSF) of crop straws to replace off-line chemical analysis in laboratory. The SSF experiment cycle is seven days and the gas monitoring data sets were collected by e-nose every 24 hours. In this experiment 20 data sets corresponding to 20 batches of fermentation processes were collected, and ten of which were used for training GP classifier, while the rest for testing the performance of it. Test results show that the e-nose system could effectively monitor SSF process of crop straws and the classification accuracy of GP classifier was higher than that of support vector machine classifier or neural networks classifier. So the e-nose system combined the GP classifier method could be an effective strategy to monitor SSF process of crop straws.