The Modeling Of Spoilage Bacterial Growth In Chilled Beef With Modified Atmosphere Packaging

Predictive microbiology is a new subject applying to predict and control the living of bacteria in foods which based on microbiology, mathematics, statistics and applied computer science. The core of predictive microbiology is building the models which divided to primary, secondary and tertiary models. It has a highly requirement to experimental data, the more data, the more accurate of the model, and the more excellent applying ability of the predicting software. Through the developing in nearly half century, researchers has employed dozens of professional predictive software. However, some improvement is still necessary in the field of data collecting and common model.The instruction presented the development of predictive microbiology, it is fairly related to food safety in our daily life. Although it has been a mature subject in recent years, there’re some problems. For example, most of the models are developed according to pure culture, one model can only simulate one species of microorganism. Since there’re four major spoilage microbes in the beef, the Artificial Neural Networks is introduced into this research to predict the microbial growth beyond the beef stuff.Chapter One introduced the traditional modeling method. The chilled beef was stored under a target condition(temperature(-2,0,5and10℃) and Modified Atmosphere Packaging(MAP) air component(65%O2,35%CO2and80%O2,20%CO2)). Then the data of bacterial growth were fitted by a modified Gompertz equation. The growing parameters of the modified Gompertz equation were then fitted by a Response Surface Model to describe the relationship of parameters and environmental factors. The criteria indict the models are able to predict the bacterial grow precisely.In Chapter Two, the Artificial Neural Networks were employed to model the same dataset. The supervised back-propagation(BP) network is based on searching an error surface for points with minimum error. The ANN model applied in the study was a three-layer BP network. The first layer was input layer. It consisted of the input neurons which were bacterial species, storing time, storing temperature and MAP component. The second layer was called hidden layer, and used for processing the nonlinearity of the input information. The third layer(output layer) was with one output neuron, representing the corresponding microbial counts. When the counts under each storing condition were obtained, the MG equation was applied to describe the growing curve. And the criteria indicated the model was closely matching to the modeling datasets. Using this model, the bacterial counts and remaining shelf-time of beef can be rapidly predicted by filling the air component and temperature into input layer.The program language of Matlab was provided in the Chapter Three. Some subroutines used in the modeling was like "sim","train" and "newff’. The Matlab is a powerful software in modeling and mathematical statistics. It can also present the growing curves of bacterial in a direct way.