Real‐time detection of saponin content during the fermentation process of Tremella aurantialba using a homemade artificial olfaction system

Chunxia Dai,Xingyi Huang,Daming Huang,Riqin Lv,Jun Sun,Zhicai Zhang,Joshua H. Aheto
DOI: https://doi.org/10.1111/jfpe.13101
2019-01-01
Journal of Food Process Engineering
Abstract:Saponin is one of the most important products of Tremella aurantialba fermentation. The purpose of this research is to develop a portable artificial visualization system for convenient and nondestructive monitoring the saponin content during the fermentation process of T. aurantialba in real time. This system consisted of a colorimetric sensor array (CSA) and image acquisition program. Support vector machine regression (SVR) model was utilized to establish the quantitative model for predicting saponin content. Principal component analysis was applied to reduce the input dimension of SVR model and predigest model complexity. The reaction time was optimized with the predicted results as the main concern. The results indicated good performance in the correlation between reference measurement and CSA predicted when the reaction time was 2 min. Additionally, genetic algorithm was used to select colorimetric sensors that resulted in the best-fitted for models. The determination coefficient (R-p) was 0.8829 and the root-mean-square error of prediction was 1.7783 mg/mL in the testing set. The results validated that the designed system is feasible to quantify saponin content, and it could serve as a useful technology for quality control of submerged fermentation of T. aurantialba. Practical applications The fermentation process is susceptive and even small changes can result in unacceptable final product quality. However, the classical analysis methods need off-line extraction techniques and complex preparation of the sample and lengthy analysis time, which cannot satisfy the demand of in-situ and rapid test. This study developed a portable artificial visualization system for convenient and rapid detection during the submerged fermentation of Tremella aurantialba, and tested its reliability and accuracy. The results obtained from this research sufficiently showed that the developed system has broad application prospects in the real-time detection of the fermentation process.
What problem does this paper attempt to address?