Building Near Infrared Prediction Models Of Nutrient Compositions In Corn Silage Based On BP-neural Network

In this study,corn samples of different varieties in different regions of China were collected to correlation analysis of agronomic traits,yield and nutritional value.Some samples were selected for fermentation to study the nutritional value of corn silage of different varieties in different regions and rumen digestibility of dairy.The aim of this study was to investigate the effects of region and variety on agronomic traits,yield,nutritional value and rumen digestibility of corn silage.At the same time,corn silage samples were collected from more than 400 farms in China to establish the NIR prediction model of corn silage.Partial least squares method and BP neural network model were used to establish the optimal NIR prediction model of corn silage nutrient composition.Experiment 1.Analysis of corn varieties in different regions of China.A total of 43 corn samples of different varieties were collected,which were distributed in five regions of China.The agronomic traits,yield and quality information of all samples were collected.The correlation degree of nine agronomic traits with yield and nutritional quality was analyzed by grey correlation degree method,and the correlation matrix was established.The results showed that the agronomic traits of corn populations in Northeast China,North China,Central China,East China and Northwest China were significantly similar,and plant weight and ear weight were the main similar agronomic traits.The analysis of agronomic traits that determine the nutritional quality of maize populations showed that the agronomic traits of maize populations in northeast China and central China were similar,and the main agronomic traits that affected the population were the number of ears of maize.The agronomic traits of maize populations in North China,East China and Northwest China are similar,and the main agronomic traits with greater influence are the number of grains.Experiment 2.Study on the nutritional value of corn silage and Rumen degradation of Dairy.Eight corns of different varieties were selected from the sample of experiment 1 for fermentation.The analysis of the conventional nutrients and the degradation rates of nutrients in the rumen at different time points of the 8different varieties of corn silage,The results showed that there were no significant differences in the contents of CP and EE among different varieties,and the differences mainly focused on starch,NDF and ADF,and the higher the NDF,The lower the starch level,the lower the degradation rate of rumen DM.Silage No.6and No.7 have high starch content,low NDF content and high degradation rate of DM,starch and NDF.Experiment 3.Establishment of NIR prediction model for corn silage.For 974 corn silage samples,Nonlinear iterative partial least-squares and BP neural network were used to establish near-infrared quantitative analysis models for 12 nutritional indicators.The results obtained by the two algorithms were compared.The results showed that DM,CP,NDF,ADF and Starch were better than those of partial least squares.BP neural network model has better calibration effect and higher prediction accuracy.For in vitro digestibility analysis IVTD30 h and IVTD48 h BP neural network prediction model is better than partial least squares prediction model.However,the prediction models obtained by Ca and P modeling methods cannot achieve the prediction.