Abstract: A BP neutral network model for predicting storage quality traits of ‘Hanfu’ apple including juice yield, soluble solids, total acid and solid/acid ratio based on texture properties such as hardness, resilience, adhesiveness, cohesiveness and chewiness was established. Quality changes of ‘Hanfu’ apple were measured during storage at 0 ℃ and 20 ℃. A topological network structure was constructed to train the established predictive model. The results of simulation demonstrated that the BP neutral network model allowed the prediction of apple quality based on texture properties. The model was validated using two sets of non-sample data and relative errors lower than 5% between the predicted and the observed values were obtained.

Key words: ‘Hanfu’apple, texture properties, quality properties, BP neural network, predictive model