Study Of Process Extracting Ursolic Acid From Loquat Leaf By Using Visual Analysis And Neural Network

The extraction of Ursolic acid from Loquat Leaf with the extracting equipment of ultrasonic wave was systematically investigated. Loquat Leaf was cleaned and unhaired and then dried in an oven and cooled at room temperature and reserved after smashing and Sieved. Weighing samples of different amount of loquat powder, samples were added different appropriate amount of ethanol, and extracted with the ultrasonic wave. The experiments were arranged with the uniform design method of U10*(108) and five experimental factors (ethanol concentration, extraction time, ratio of liquid to solid, times of extraction and particle size of loquat leaf powder) were chosen in the study. The content of Ursolic acid was determined with HPLC method.The experimental target index is the extracting yield. This paper used Multi-factor & Multi-level Visual Analysis (m2VA) built by our research group to analyze experimental data. Finally the best extracting process parameters were found: the ethanol concentration range is 80~100%, the extraction time ranges are 50~65min and 80~100min, the times of extraction once. It provides a set of better process parameters for the manufacture in pilot scale. The conclusion from m2VA can robustly and macroscopicly guide the improvement of production process. The m2VA of experimental data suggested that the influences from the ratio of liquid to solid and the particle size of loquat leaf powder are complicated.This paper built the recognition model of extraction process of Ursolic acid by BP model of 5-3-1 topology structure of Artificial Neural Networks (ANN) based on the test with uniform design method.The leave-two-out method was presented in training BP networks, in which the arbitrary eight groups of experimental data were used for training and left two for verification. When the training times were 29, the error reached 10-6. By using the trained model, the method called Multi-factor and Multi-level Neighboring Analysis, short form: m2NA) created ourselves was put in practice, to study systematically the time characteristic in extraction process.Through the model instruction a number of verifying tests were done and the results were consistent well with the model calculation. A set of optimized process parameters was obtained: the concentration of ethanol is 65%, extraction time 15min, times of extraction 2, solid-liquid ratio 10, and particle size 40 meshes, and in this case productivity is 91.40% (value of model calculation is 91.38%). The model of ANN analyzing the quantitative relationship within the system can accurately guide further research and customize production process. The study not only calculated and predicted the better range of process by the model of ANN but also provide a large number of process sets according to customers. The study offered the scientific basis for the extraction of ursolic acid from Loquat Leaf and provided important theoretical basis, and process design blueprints for industrialized production. The study will make lower cost and efficient production possible.