| The content of nitrogen,phosphorus and potassium in compound fertilizer is the most important detection index for fertilizer manufacturers.The most common test method is to use chemical reaction to obtain the content of nutrients in the compound fertilizer after sampling on the production line.Due to the hysteresis of the chemical reaction,the problems in the production process cannot be found in time,and many unqualified products are produced,which brings economic losses to the enterprise,and the use of chemical reagents does not conform to the concept of environmental protection testing.In order to solve this problem,a fast,accurate and non-destructive detection method is urgently needed.Visible/near infrared spectroscopy has a natural advantage in this respect,and has been successfully applied to many agricultural related fields such as soil testing.However,there are few visible/near-infrared spectroscopy studies on nutrients in compound fertilizers.At the same time,there is no chemometric analysis software that is compatible with the rapid detection of compound fertilizers.Therefore,this article will use visible/near-infrared spectroscopy and laboratory-developed visible/near-infrared spectrometer to carry out visible/near-infrared spectroscopy compound fertilizer analysis,and develop a set of compound fertilizer nutrient online analysis application software,application software function module It mainly includes compound fertilizer spectral acquisition,data preprocessing,establishment of calibration model,and content prediction.In addition,when establishing the calibration model,it is easy to introduce redundant data.In view of this deficiency,this thesis adopts the sub-band to establish the regression model,and obtains the best representative band as the base band.For the wavelength data outside the base band,a preferred wavelength method is proposed.Preferably,a representative wavelength is used to establish a regression model in combination with the above two partial wavelengths.After that,the experiment proves the validity and practicability of the preferred wavelength algorithm by comparison with the successive projections algorithm.Firstly,after comparing the preferred wavelength algorithm and the successive projections algorithm,it is found that the RPD prediction index of the total nitrogen content prediction model in compound fertilizer is increased from 2.8772 to 10.6060,and the RMSEP prediction index is reduced from 1.4842 to 0.3938 using the preferred wavelength algorithm.The RPD prediction index of the model increased from 1.4950 to 2.0314,the RMSEP prediction index decreased from 0.9579 to 0.6849,the RPD prediction index of the total potassium content prediction model increased from 1.7996 to 2.8174,and the RMSEP prediction index decreased from 0.5640 to 0.3418.The algorithm is significantly better than the successive projections algorithm.Secondly,using the developed application software and the established quantitative model with visible/near-infrared spectroscopy instruments,16 samples of a chemical fertilizer production base were tested.The results showed that the maximum residual value of the prediction results of total nitrogen in the sample by the application software.The minimum and average values are 1.3768,0.1383,and 0.6358,respectively.The ratio of the residual to the true value is 5.79%,0.55%,and 3.57%,respectively.The maximum,minimum,and maximum residuals of the predicted results for total phosphorus in the sample.The average values are 2.2863,0.435,1.15,respectively.The ratio of residual to real value is 7.52%,2.6%,and 6.73%,respectively.The maximum,minimum,and average residuals of the prediction results for total potassium in the sample are 0.9596,0.1988,0.6812,the ratio of residual to real value is 8.72%,4.96%,7.57%.In summary,the application software has positive significance for on-line detection and analysis of compound fertilizer. |
PDF Full Text Request