Optimization of Ultrasound-Assisted Extraction of Chlorogenic Acid from Lonicera japonica by Artificial Neural Network

忍冬(Lonicera japonica)為忍冬科的一種植物，其花卉為金銀花(Honeysuckle flower)，為亞洲常見中藥材之一。綠原酸(Chlorogenic acid)是金銀花的主要有效成分之一，在一些研究文獻中指出，綠原酸具有各種生物活性，包括抗氧化、抗菌、抗高血壓和抗糖尿病等有效生理活性。由於現代人開始注重養身和保養，因此保健食品與化妝品產業持續蓬勃發展，市場需求也不斷地增加。從近年綠原酸的相關研究文獻來看，綠原酸不論是從抗氧化、美白或者是抗菌活性等，皆顯示出綠原酸於未來有極大的發展性。 傳統天然物萃取方法主要是利用索式熱回流萃取配合醇類溶劑進行萃取，但利用此萃取方法通常耗時、耗能。相較於傳統方法，本實驗是以乙醇當溶劑，藉由超音波機輔助萃取金銀花中的綠原酸。目的為萃取出金銀花的活性成分，藉由超音波作用於液體所產生的空穴效應來提高萃取效率，超音波可幫助植物細胞壁的破壞，加速固相至液相的質能傳遞，促進活性成分能釋出至溶劑當中，進而增加綠原酸之萃取產率。 本論文主要使用超音波輔助萃取法同時以乙醇為溶劑進行萃取，比較超音波輔助萃取法與搖瓶萃取系統的萃取效率。實驗結果顯示，超音波輔助萃取法在萃取10分鐘後，萃取產率可達最大值且隨時間增加而趨於穩定；而搖瓶萃取法於萃取時間達60分鐘時仍未達穩定，且萃取產率不及超音波輔助萃取法，並利用質傳係數(KT)去作比較，計算結果顯示出超音波輔助萃取法質傳效率來得比搖瓶輔助輔助萃取法來得好。再以溫度(X1, 30–70℃)、乙醇濃度(X2, 55–95%)和液固比(X3, 10:1–50:1 mL/g)及超音波功率(X4, 90–150W)探討超音波萃取法對金銀花之綠原酸萃取產率之影響，接著以反應曲面法(Response surface methodlogy, RSM)配合五階層四變數(5-levels-4-factors)之中心混成實驗設計(Central composite design, CCD)，探討各實驗因子對萃取效率的影響，再利用反應曲面找出最適化之萃取條件及預測之萃取產率，再以類神經網路(Artificial neural network, ANN)資訊處理系統，對此實驗系統進一步預測分析，同時與RSM的預測結果進行比較。以RSM進行變異數分析(ANOVA)，RSM及ANN的R2分別為0.7913及0.9898，且ANN的RMSE、MAD及MAPE皆優於RSM；RSM最適化之萃取產率條件為溫度33.56℃、乙醇濃度65.88%、液固比46：1 mL/g和超音波功率150W，RSM所得到的最高預測萃取產率為45.63 mg/g；ANN得到的預測萃取產率為44.78 mg/g，以此最適化條件進行實驗所得到的產率為43.13±1.44 mg/g。綜合以上實驗結果顯示，ANN比起RSM有更高的預測能力。Lonicera japonica is a traditional Chinese herbal medicine in East Asia. Flower of buds of L. japonica, called honeysuckle, have been applied to anti-oxidant, anti-bacterial、anti-hypertensive effects and anti-diabetes, etc. Chlorogenic acid (CGA), the ester formed between caffeic acid and quinic acid, is a well-known phenolic compound. The known chlorogenic acid, a primary bioactive compound in flower buds, has more pharmacological activities. Many studies have indicated that chlorogenic acid possesses many benefits for health, such as anti-oxidant, whitening, anti-bacterial, antihypertensive effects, anti-obesity, anti-diabetes, neuroprotective properties and anti-tumor. Thus, chlorogenic acid is considerable potential for future development. To extract the bioactive compound in the plant, the traditional extraction method is conventionally performed by heat-reflux extraction with alcohol as the solvent. However, this method is very time-consuming for heat-reflux extraction. To improve the shortcomings of the traditional extraction method, ultrasonic-assisted extraction has been applied to extract the active ingredients of plants in recent years. The ultrasound could create cavitation in the liquid solution to increase the efficiency of extraction as well as reduce extraction time and solvent consumption. This study used ultrasonic-assisted extraction and ethanol as a solvent for extraction. The extraction efficiencies of the ultrasonic-assisted extraction and the shaking bath extraction system were compared. The experimental results showed that the ultrasonic-assisted extraction reached its maximum value after 10 minutes of extraction, followed by a steady state for extraction. However, the extraction by shaking bath for 60 minutes had not yet reached a maximum yield, which was inferior to ultrasound-assisted extraction. Using the mass transfer coefficient (KT) for comparison, the calculation results showed that the ultrasonic-assisted extraction method was better than the shaking extraction method. Moreover, we investigated parameter effects on extraction. The parameters are respectively temperature (X1: 30–70°C), ethanol concentration (X2: 55–95%), liquid to solid ratio (X3: 10:1–50:1 mL/g) and ultrasonic power (X4: 90–150W). This study was designed to optimize the ultrasound-assisted extraction of chlorogenic acid of Lonicera japonica using a response surface methodology (RSM) and central composite design (CCD; five-level-four-factors). According to the RSM, we found out the optimum extraction conditions and predicted extraction yield. Additionally, artificial neural network (ANN) is further performed to predict, analyze and compare with the predicted results of RSM. By analysis of variance (ANOVA), the R2 of RSM and ANN were 0.7913 and 0.9898, respectively, and the RMSE, MAD, and MAPE of ANN were better than RSM. The optimal conditions of extraction process were as follows: temperature of 33.56°C, ethanol concentration of 65.88%, liquid to solid ratio of 46:1 mL/g and ultrasonic power of 150W. The RSM showed that the predicted yield was 45.63 mg/g. The ANN showed that the predicted yield was 44.78 mg/g. The actual experimental value was 43.13±1.44 mg/g. Overall, the above experimental results show that ANN has higher prediction accuracy than RSM.1. 緒論 1 2. 文獻回顧 3 2.1. 忍冬簡介 3 2.1.1. 活性萃取物 3 2.1.2. 綠原酸簡介 3 2.2. 萃取技術 4 2.2.1. 萃取技術開發的目的 4 2.2.2. 傳統萃取技術 4 2.2.2.1.索式萃取法 4 2.2.2.2.搖瓶萃取法 5 2.2.3. 新型萃取技術 6 2.2.3.1.超音波輔助萃取法 6 2.2.3.2.超音波反應器種類 7 2.2.3.3.超音波輔助萃取法的應用 9 2.3. 實驗設計方法 10 2.3.1. 單因子實驗法(Single-factor experiments) 10 2.3.2. 全因子實驗設計(Full factorial design) 10 2.3.3. 反應曲面法(Response surface methodology, RSM) 11 2.3.3.1.Box-Behnken實驗設計法(Box-Behnken design, BBD) 13 2.3.3.2.中心混成實驗設計法(Central composite design, CCD) 14 2.3.3.3.反應曲面法的應用 15 2.3.4. 類神經網路(Artificial neural network, ANN) 16 2.3.4.1.訓練函數(Training function)介紹 18 2.3.4.2.轉移函數(Transfer function)介紹 20 2.3.4.3.類神經網路的應用 21 3. 材料與方法 22 3.1. 研究架構 22 3.2. 材料與藥品 23 3.3. 設備儀器 23 3.4. 高效能液相層析分析方法 24 3.5. 濃度定量 25 3.6. 萃取產率計算 26 3.7. 萃取系統 26 3.7.1. 超音波輔助萃取系統 26 3.7.2. 搖瓶萃取系統 26 3.8. 質傳效率評估方法 26 3.8.1. 萃取動力學 26 3.8.2. 質傳係數分析 27 3.9. 實驗設計 27 3.10. 模組之適切性評估 29 3.11. 類神經網路建立 30 4. 結果與討論 32 4.1. 預實驗 32 4.1.1. 時間對綠原酸萃取產率之影響 32 4.1.2. 溫度對綠原酸萃取產率之影響 33 4.1.3. 乙醇濃度對綠原酸萃取產率之影響 34 4.1.4. 液固比對綠原酸萃取產率之影響 35 4.1.5. 超音波功率對綠原酸萃取產率之影響 36 4.2. 超音波輔助萃取系統與搖瓶萃取系統之比較 37 4.2.1. 萃取動力學 37 4.2.2. 質傳係數分析 37 4.3. 反應曲面法 39 4.3.1. 變異數分析(ANOVA) 39 4.3.2. 反應曲面圖分析 43 4.4. 類神經網路法 49 4.5. 反應曲面法與類神經網路法之比較分析 50 4.6. 類神經網路之曲面分析 53 4.7. 驗證實驗 57 5. 結論與未來展望 58 5.1. 結論 58 5.2. 未來展望 59 6. 參考文獻 60 附錄 (APPENDIX) 67 附錄一 68 附錄二 7