| The calibration models to predict the chemical composition in wheat bran and cottonseed meal were developed by applying on Fourier near infrared spectroscopy, also the influences of the moisture on the near infrared spectroscopy assay were investigated in this research. Seventy-six wheat bran samples and Seventy-three cottonseed meal samples were collected and crush through 40 mesh size. Based on the moisture (H2O), crude protein (CP), ether extract (EE), neutral detergent fiber (NDF), acid detergent fiber (ADF) and crude ash (Ash) of these samples analysis, the samples were assayed by Fourier near infrared spectroscopy and Partial Least Squares (PSL) was applied to establish the calibration models of H2O, CP, EE, NDF, ADF and Ash in wheat bran and cottonseed meal. Meanwhile, 10 samples were chosen as the Validation group (Ⅴ) according to the content of CP, NDF, ADF and Ash in separately, and 50 samples constitute the Calibration group(C). Each sample was divided into three parts and the water content were adjusted to 3 different levels (8%, 10% and 12%). The calibration models of CP, NDF, ADF and Ash were established in separately under the condition of different moisture (8%, 10% and 12%), and at last the global calibration model based on different moisture contents (8～12%) was confirmed. The results are as follows:1. The coefficient of determination in validation (Rval2) and relative square error (RSD) between the chemical analysis values and the FNIS Prediction values of wheat bran were 0.9668 and 1.53%(H2O), 0.9311 and 1.58%(CP), 0.9662 and 3.78% (EE), 0.9764 and 2.61%(NDF),0.9900 and 1.98%(ADF),0.9904 and 1.65%(Ash), respectively; And the coefficient of determination in validation (Rval2) and relative square error (RSD) were 0.9421 and 1.70%(H20),0.8468 and 2.11%(CP),0.9506 and 3.71%(EE),0.9542 and 3.13%(NDF),0.9701 and 3.31%(ADF),0.9702 and 2.75%(Ash) in separately. 2. The coefficient of determination in validation (Rval2)) and relative square error (RSD) between the chemical analysis values and the FNIS Prediction values of cottonseed meal were 0.9972 and 0.53%(H20),0.9644 and 1.32%(CP),0.8992 and 11.83%(EE),0.993 and 1.01%(NDF),0.9778 and 1.86%(ADF),0.9721 and 1.32%(Ash), respectively; And the coefficient of determination in validation (Rval2) and relative square error (RSD) were 0.9952 and 0.64% (H2O),0.8959 and 1.86% (CP),0.6876 and 17.49%(EE), 0.9527 and 3.14% (NDF), 0.9187 and 2.56% (ADF), 0.8521 and 1.35% (Ash) in separately.3. The root mean square error of prediction (RMSEP) of those models developed on different moisture background were 0.665～4.611(CP), 0.546～6.586(NDF), 0.792～4.16(ADF) and 0.081～1.12(Ash), respectively.Conclusion: The calibration models to predict the H2O, CP, EE, NDF, ADF and Ash in wheat bran could meet the needs of quality evaluation. The results show the models could be well applied on chemical contents prediction in cottonseed meal except that of the EE. The moisture global model developed base on gradient moisture of samples could increase the application of the prediction models and decrease the error caused by the moisture.
|
PDF Full Text Request