Study On Shelf Life Prediction Of Packaged Baking Food

This paper presents a thorough study of packaged cookies called Master Kong. Determing main factors influence on shelf life by analyzing quality changes of packaged cookies during the process of manufacture, packaging, storage and sales, then established shelf life prediction models using three different methods and compared their performance.Based on the theory of chemical kinetics, POV first-order equation of packaged cookies was obtained under three storage temperature (35℃,45℃,55℃), then the relationship between the storage temperature and reaction rate of POV was bulided using Arrhenius equation. Finaly, mathematic model of shelf life prediction was established which reflect the relationship between shelf life of packaged cookies and the storage temperature.Data of organoleptic evaluation was collected under three storage conditions above, then established the relationship between the risk rate (∑H) and storage time using Weibull hazard analysis. After that, the shelf life prediction equations were derived combinated kinetics theory and weibull method.Using BP artifical neural network to predict the shelf life of packaged cookies by MATLAB, the architecture was three layers. There were one input layer, one output layer and one hidden layer; The input layer had6neurons、including the temperature and relative humidity of storage environment、initial water ratio、initial Peroxide Value、oxygen permeability and Water Vapour Permeability of packaging material; The output layer was a single unite, that was the shelf life of packaged cookies. The hidden layer had five neurons: transfer function of input layer was tansig, transfer function of output layer was purelin, study function was trainlm.Comparing the prediction value of each mathematical model with actual shelf life under different storage conditions, the error rate was0.0985、0.1015、0.077respectively. The result shows that BP model won the minimum error rate and more accurate predictive value. In addition, it can reflect influnence of various factors, including the intial water ratio and Peroxide Value, oxygen permeability and Water Vapour Permeability of packaging material, temperature and relative humidity of storage environment. So BP model is more excellent than dynamic model and Weibull hazard analysis model. In brief, application of BP neural network on food shelf life prediction provided a new idea and novel attacking approach for food science and packaging engineering, also promoted computer science application to the food shelf life research.