| This experiment was conducted to evaluate the feasibility which predicted the nutrition value of cottonseed meal by near- infrared reflectance spectroscopy technology. 76 normal cottonseed meal samples were collected from different sources which included different planting regions, such as Xinjiang, Shandong, Hubei and so on. 56 samples were selected as calibration set to build calibration models of Moisture, crude protein(CP), ether extract(EE), crude fiber(CF), ash, gross energy(GE), apparent metabolizable energy(AME), true metabolizable energy(TME), and amino acid by near- infrared reflectance spectroscopy(NIRS), and conducted internal cross validation furtherly. Another 20 samples were used as external validation for prediction models. The stability and applicability of calibration and validation models for FOSS XDS and Nir Smart Eye 1700 were investigated.The first experiment aimed at exploring the possibility of rapidly quantitative determination of proximate nutrients and metabolizable energy in cottonseed meal by NIRS with FOSS XDS and N ir Smart Eye 1700 instruments. A total of 76 cottonseed meal samples with different varieties, producing areas, and processing methods were collected from all over C hina. The proximate nutrients(including moisture, CP, EE, CF, ash and GE) and metabolizable energy were analyzed. All the samples were divided into calibration set(N=56) and the external validation set(N=20) randomly. The near- infrared calibration models were established by using the modified partial least squares(MPLS) which belonged to multivariate calibration method.The results showed that:(1) The variations in proximate nutrients were large between different sources of the cottonseed meal samples. Coefficient of variation for proximate nutrients and metabolizable energy ranged from 2.52% to 84.75%. Coefficient of variation for moisture, EE, CF exceeded 10%; coefficients of variation for CP, ash and GE were 9.58%, 9.81% and 2.52% respectively.(2) The coefficient of determination for calibration(RSQcal), coefficient of determination for cross-validation(1-VR) and coefficient of determination for external validation(RSQv) for moisture, CP, EE, CF, ash and GE ranges were from 0.924 to 0.976, 0.8247 to 0.9303 and 0.879 to 0.896, respectively, which indicates that these NIRS models established by FOSS near-infrared instrument can be used in routine analysis.(3) The RSQcal, standard error of calibration(SEC) and RSQ v for moisture, CP, EE, CF and GE range from 0.905 to 0.951, 0.169 to 1.456 and 0.883 to 0.959, respectively, which indicates that these NIRS models established by N ir mininature near- infrared instrument can be used in routine analysis; The RSQv for ash is 0.524, so the model can not be used in routine analysis.The ranges of AME and TME were 4.63 MJ/kg~11.90 MJ/kg and 5.39 MJ/kg~13.20 MJ/kg. The calibration models of metabolizable energy were established by FOSS near- infrared instrument and Nri miniature near- infrared instrument respectively. The RSQcal, 1-VR and RSQv of AME and TME with FOSS near-infrared instrument were 0.969 and 0.927, 0.9170 and 0.9057, 0.911 and 0.892 respectively. The RSQcal, SEC and RSQv of AME and TME with Nri miniature near-infrared instrument were 0.954 and 0.949, 0.400 and 0.475, 0.915 and 0.907 respectively. These prediction equations can be used in rout ine analysis. The models reach the practical level.The possibility of using NIRS for quantitative determination of amino acids of cottonseed was investigated in the second experiment. A total of 76 cottonseed meal samples with different varieties, producing areas, and processing methods were collected from all over China, and the 16 amino acids contents were analyzed. 56 samples were selected as calibration set to build calibration models. Another 20 samples were used as external validation for prediction models. The results showed that: the ranges of variations in 16 amino acids were large between different sources of the collected cottonseed meal samples. The RSQcal, 1-VR and RSQ v of Asp, Thr, Glu, Gly, Lys, His, Arg and Trp with FOSS near- infrared instrument were from 0.872 to 0.953, 0.7813 to 0.9504, 0.840 and 0.887, respectively. The RSQcal, SEC and RSQv of Asp, Thr, Ser, Glu, Gly, Leu, Phe, Lys, His, Arg and Trp with Nri miniature near-infrared instrument were from 0.865 to 0.970, 0.016 to 0.537, 0.845 to 0.899, respectively. These prediction equations can be used in routine analysis. The models reached the practical level. The RSQ v<0.84 of other amino acids can not be used in routine analysis.The optimal method of scatter was different among the nutrients, metabo lizable energy and amino acids. |
PDF Full Text Request