| The rapid and in situ detection of compound fertilizer components is of great significance to the production process control and product quality control for fertilizer production enterprises.Compared with the traditional methods of detecting the composition of compound fertilizers,laser-induced breakdown spectroscopy(LIBS)has the advantages of fast detection speed,simultaneous detection of multiple elements in real-time field,and simple sample preparation.Therefore,it is very suitable for rapid on-site detection of compound fertilizer components.Due to the LIBS technique,which is affected by the matrix effect of compound fertilizer samples,the weak spectral line of phosphorus element,the in-situ detection environment,and the inability to accurately quantify problems.In order to realize the on-site detection and application of this technology in the process of compound fertilizer production,in this dissertation,the following research work is carried out by the optimization of system parameters,the characteristics of samples,the influence of environmental parameters and the application of different quantitative analysis methods for LIBS of compound fertilizer components:(1)According to the components of compound fertilizers,the analytical lines of N,P and K elements were 746.8nm,214.9nm and 769.9nm,respectively.The research on the temporal and spatial evolution characteristics of compound fertilizers was carried out.According to the delay characteristics of different elements in compound fertilizers,the delay time of simultaneous detection of multiple elements was determined as 1.28 ?s and the integration time was 1.05 ms.The influence of plasma collection angle and detection height on the spectral characteristics of compound fertilizer was also studied.The best plasma collection angle was 30 with the horizontal direction,and the detection height was 2mm from the surface of the compound fertilizer sample.(2)The effects of laser parameters on the plasma signal of compound fertilizer components were studied,including laser energy,repetition frequency and focal position.When the laser energy is less than 110mJ,the spectral line intensity and the signal-to-background ratio of the compound fertilizer have a linear relationship with the laser pulse energy.When the laser energy exceeds 110mJ,the line intensity is nonlinear and the signal-to-background ratio remains basically unchanged.The laser repetition frequency was adjusted,and the maximum spectral line intensity was got when the laser repetition frequency was 1 Hz.The stability of the laser induced breakdown spectrum first increased and then decreased with repetition frequency and at 7 Hz was the most stability.The characteristics of LIBS spectrum were analyzed when the laser focus position was located at different locations below the compound fertilizer sample.The largest spectral intensity and the most stable signal were obtained,when the laser focus was 3mm below the surface of the compound fertilizer sample.Finally,the laser energy of the LIBS detection system for compound fertilizer was 100 mJ,the repetition frequency was 1 Hz,the laser focus was 3 mm below the compound fertilizer sample,and the average measurement was performed 20 times for each measurement.(3)The effects of particle size,porosity of complex fertilizer and background gas on LIBS were systematically studied and analyzed theoretically.By simple pretreatment of compound fertilizer samples to change their physical properties,the compound fertilizer samples became even and compact.Seven kinds of compound fertilizer samples with different particle sizes were prepared.The spectral line intensity,signal-to-background ratio and stability under different particle sizes were studied.When the particle size of the sample was between 0.18-0.25mm,the signal-to-background ratio and relative standard deviations were best.It was because each pulse of laser only act on one particle of composite fertilizer,making the signal more stable and reproducible.By applying 9 different pressures on the compound fertilizer samples,the elemental characteristic line and the signal-to-background ratio gradually increased when the applied pressure was less than 8 MPa.When the pressure was between 8 MPa and 20 MPa,the line intensity and the signal-to-background ratio were tend to be stable.Therefore,it was determined that the pressure of 8 MPa was applied to the sample of compound fertilizer during the actual application.The spectral characteristics of laser plasma of compound fertilizer were studied under different atmospheres of air and argon for compound fertilizer samples.When argon was used as the background gas,the spectral line intensity of phosphorus element increased nearly two times,and the signal-to-background ratio increased about five times.At the same time,the correlation between plasma lifetime and element calibration curve has been improved to some extent.(4)The components of compound fertilizers were analyzed by using quantitative methods such as calibration curve method,internal standard method and multivariate nonlinear regression.The calibration curves of total nitrogen,phosphorus pentoxide and potassium oxide in compound fertilizer samples were quantitatively analyzed.The detection limits of the three components were 0.02%,0.13%and 0.25%,respectively.The internal standard curve for the analysis of phosphorus pentoxide concentration in compound fertilizer with silicon as internal standard element was established.The linear correlation was increased to 0.975.The concentrations of P2O5 in the four validated samples were calculated under the established internal curve.The average relative error was 1.03%,which showed that the internal standard method can improve the accuracy of LIBS.The application of multivariate nonlinear regression in LIBS was also studied.Potassium oxide concentration of compound fertilizer was taken as the target of the analysis.Ten samples were selected for modeling and four samples were used for validation.The correlation coefficient between the predicted value of LIBS and the true value for modeling and verification were increased to 0.981 and 0.978,respectively.The relative error of the prediction of multivariate nonlinear regression was 0.51%.The above results showed that multivariate nonlinear regression method was greatly improved the prediction accuracy of quantitative analysis of LIBS.(5)Focusing on the LIBS data analysis based on support vector machine regression.Combined with LIBS and nuclear transformation techniques,an adaptive hybrid kernel support vector machine regression model suitable for LIBS detection was established.The regression model of support vector machine,which was based on the analysis of phosphorus pentoxide concentration in compound fertilizer,was studied.The parameters of grid search,particle swarm optimization and genetic algorithm were optimized and the models were established,respectively.The results showed that the correlation coefficients of training samples with three parameters optimization methods were 0.980,0.987 and 0.985,respectively.The correlation coefficients of the test samples were 0.991,0.976 and 0.993,respectively.The average absolute errors were 0.043%,0.38%and 0.81%,and the maximum absolute error is 0.1%,0.1%,0.16%,respectively.Experimental results showed that support vector machine regression algorithm can be applied to the quantitative analysis of LIBS of compound fertilizer components,and the overall effect was better.The conclusion of the study provided data support and method support of LIBS for the rapid on-site detection of compound fertilizer components. |
PDF Full Text Request