Optimization of extraction technology of total flavonoids from lotus leaf by BP neural network and exploration of flavonoids stability

Box-Behnken response surface design coupled with neural network model was employed as a new method to optimize the conditions for ultrasonic-diacolation-assisted extraction of total flavonoids from lotus leaf. And the paper also studied the stability of flavonoids. The optimal extraction conditions were ethanol concentration of 60.2%, material-to-liquid ratio of 1∶39.7, extraction duration of 80 min, and collecting 9.2 folds of percolate at a rate of 3mL/min. The extraction yield of total flavonoids was 6.87%, the deviation between observed and predicted values of yield was 1.89%, which indicated the reliability and practicability in the optimized conditions. It also suggested that the extracted total flavonoids were more stable at low temperature, low pH value and in dark condition, β-cyclodextrin, vitamin C or D-glucose acid lactone also improved the stability of the total flavonoids.