OBJECTIVE To explore the microwave vacuum drying characteristics of Rosa roxburghii Tratt. and to establish a method for predicting the moisture ratio during the drying process.

METHODS Microwave power, vacuum degree, and loading capacity were investigated as factors affecting the moisture ratio and drying rate during the microwave vacuum drying process of Rosa roxburghii Tratt.. Four drying kinetics models and a back propagation artificial neural network(BP-ANN) were employed to develop a predictive model for the relationship between moisture ratio and drying time, which was subsequently validated and compared for prediction accuracy.

RESULTS Within the range of experimental parameters, increasing microwave power and decreasing loading capacity were more effective in shortening the drying time of Rosa roxburghii Tratt. than adjustments to the vacuum degree. Among the 4 mathematical models, the Page model exhibited the best fit, with R² values ranging from 0.99599 to 0.99976, and χ² and SSE values ranging from 0.00003 to 0.00044 and from 0.00064 to 0.00697, respectively. The optimal number of hidden layer nodes for the BP-ANN model was determined to be 8, with a network topology of 4-8-1, with R²=0.99756 and MSE=0.00018. Experimental validation revealed that the prediction mean relative errors for the moisture ratio were 3.63% for the Page model and 1.15% for the BP-ANN model.

CONCLUSION The investigation of drying characteristics offers valuable data to optimize the microwave vacuum drying process of Rosa roxburghii Tratt.. Furthermore, the BP-ANN model serves as a scientific foundation for the online prediction of moisture ratio during the drying process of Rosa roxburghii Tratt..