| Summary: | 碩士 === 國立中興大學 === 化學工程學系所 === 103 === Piceid is used in cardiovascular disease in recently years; however, its low water solubility causes a problem for human body absorption. Therefore, encapsulated piceid into liposome is a feasible way to improve piceid water solubility, we observe the encapsulation efficiency on piceid liposome under sonication process. This study is divided into two parts.
In the first part, L-α-phosphatidylcholine and cholesterol were used as lipid content and are added to piceid dissolved in organic solvent, The conditions of encapsulation were under vaccum and hydration, experiment condition under lipid content (120~180 mg), sonication power(90~150 W) and sonication time(10~50 min) with effect of encapsulation efficiency(E.E%), absolute loading(A.L%) and particle size(PS). The 3-level-3factor Box-Behnken design was applied to optimize the response. The result indicate lipid content, sonication power and sonication time have significant effect.In second part, Response surface methodology (RSM) and artificial neural network (ANN). First step is establish a model of ANN, BBD data for further analysis. The learning cycle number is ranging from 1,000~100,000 times, and transfer function including Gaussian, Tanh and Sigmoid. Hidden layer including 3~6 nodes. To find minimum root mean square error (RMSE) and absolute average deviation(AAD) and maximum R2. As result the model with the condition of 10,000 learning cycle number, sigmoid transfer function, Levenberg-Marquardt algorithm and 6 nodes in hidden layer were the best combination for ANN.
Comparing with RSM, the R2 of ANN is 0.997, better than the R2 of RSM is 0.955. The RMSE and AAD in ANN model are 0.744% and 0.595% respectively, and RMSE and AAD in RSM model are 1.842% and 2.476% respectively. The lower RMSE and AAD values represent the more precise prediction of E.E%. Therefore the ANN model is more suitable than the RSM model to explain the data of this study.
|
|---|
