Lactobacillus Fermentation Prediciton Using Convolution Nerual Network

• Published in: 2022 IEEE International Conference on Consumer Electronics - Taiwan pp. 19 - 20
• Main Authors: Wu, Jain-Shing, Wu, Chien-Chang, Liao, Chien-Sen
• Format: Conference Proceeding
• Language: English
• Published: IEEE 06.07.2022
• Subjects: Convolution; Costs; Lactobacillus Fermentation; Long Shorten-Memory; Mass production; Microorganisms; Neural networks; Prediction; Production; Training
• ISSN: 2575-8284
• DOI: 10.1109/ICCE-Taiwan55306.2022.9869168

For the advantages of lactic acid bacteria, many researchers have developed to improve the mass production of the lactic acid bacteria. Among all the producing process, the key process is fermentation process. However, it is hard to control the amount of production because the lactic acid bacteria is organism and there is manual errors occurred easily during fermentation. Hence, if we can predict the fermentation results as earlier as possible, we can stop the fermentation process to reduce more costs. In this paper, we use the Convolution Neural Network (CNN) to predict the final bacteria count of lactic acid bacteria. We collect 29 cases of fermentation parameters and pH values during the fermentation process. Among them, there are 5 cases are failure cases, and 12 semi-success cases, 12 success cases. We first examine the first 40 check points to exclude the failure cases. And then, we separate the rest cases in to training dataset and testing dataset for 10 different trials. All datasets are sent to CNN to classify. The average training and testing accuracies are 99.41% and 80% respectively. The experimental results show that our proposed method not only can predict the final bacteria counts but also can be used to exclude the failure cases to reduce the failure cost.