Abstract
離胺酸是一種必需胺基酸，在動物体內無法合成，必需靠添加於飼料及醫 藥中，以提供足夠之營養。在製程方面，為利用麩胺酸醱酵工廠之剩餘設 備，採用糖蜜為碳源之饋料醱酵方式進行離胺酸醱酵。在本研究中，主要 探討離胺酸饋料醱酵製程，使用狀態估算方式，進行醱酵狀態變數估計及 控制。當使用糖蜜為碳源，進行離胺酸醱酵時，發現約有總糖濃度之13% 為醱酵不可利用之糖。在糖控設定點選擇上，為避免基質濃度過高造成之 阻害作用，及基質濃度不足造成菌體之活性降低、產物生產速率降低之現 象，將饋料醱酵之糖控設定點定於不可醱酵糖濃度上0.5%。在醱酵程序的 線上狀態估計方面，分別以經驗式反應次域及類神經網路方法進行醱酵程 序中菌體濃度，基質及產物濃度的線上估計。結果顯示，兩種方法均能有 效且準確的估計出醱酵製程之線上狀態變數值。以B.flavum DB24 進行離 胺酸醱酵，殘糖控制點設定於不可醱酵糖濃度上0.5%。在14公升國產醱酵 槽中培養，以間接估算之還原糖量，進行加料控制。在72小時內可生產 達9.2 %(W/V)之離胺酸鹽酸鹽，其對總糖收率為35 %。 Lysine is one of the essential amino acids which cannot be synthesized by animal. Therefore, it must be added into the feedstuff or medicine for nutrient supplements. The lysine process uses molasses as the carbon source in fed-batch culture operation in order to simulate the production of glutamic acid for the sake of reusing the extra utilities of glutamic acid factory. In this study, on-line states estimation in lysine fermentation are investigated. Results obtained from the fed- batch culture indicate a non-fermentable sugar concentration level of 13% in the total sugar concentration of molasses. In the process, sugar concentration setpoint is selected to be 0.5% higher than the nonfermentable sugar concentration. Consequently, sugar repression as well as low activity and low productivity due to sugar shortage can be prevented. On-line states estimation by empirical reaction subspace and neural network methods are employed to predict the cell concentration, lysine concentration and substrate concentration in lysine fermentation. The two estimating systems reveal the same adequate results in the estimation applications. Lysine fermentation with the strain of B. flavum DB24 is performed in a 14 liter domestic-made fermentor. By applying the estimation and control methods in lysine fermentation, the results turn out to be a lysine concentration level of 9.2% (W/V) in 72 hours with 35% lysine yield for total sugar contents.