| Abstrakt: |
As one of the sources of renewable energies, ethanol is produced from lignocellulosic compounds and food waste during the fermentation process. In the present study, the capability of electronic nose (e-nose) system and machine learning approaches was evaluated of classification/prediction of bioethanol production and sensor responses as real-time monitoring. For this purpose, rice and wheat residual were conducted by acid-ultrasound, thermal-acid hydrolysis, and plasma technology, respectively. Then, the prepared substance was poured into both a pneumatic anaerobic digester (PAD) and a mechanical anaerobic digester (MAD). The fermentation process lasted for 20 days and data were acquired every five days. The output of each reactor was examined after the distillation stage. For this purpose, the output bioethanol was injected into an e-nose with an arrangement of 13 sensors, and the feature of the produced bioethanol was investigated by gas sensors. The response of the sensors was predicted and identified by chemometric methods. Moreover, the optimal responses with time and type of digester were investigated by RSM (Response Surface Methodology). The obtained results indicated that given the significance coefficient, the gas sensors MQ135, MQ2, MQ7, MQ9, TGS2620, and TGS822 had the highest rate of detection. Moreover, the obtained results showed that the PAD generated different gas compounds within the produced bioethanol when compared with the MAD. In addition, all chemometric methods had high accuracy for classification and prediction. [ABSTRACT FROM AUTHOR] |
Full text from SpringerLink