Quantitative Analysis Of Trace Elements In Soil Using Laser-induced Breakdown Spectroscopy

In recent years,as the per capita cultivated area further decreases,the country has paid more and more attention to the development of "precision agriculture"."Precision agriculture" not only requires growers to achieve "accurate fertilization",but also "accurate seeding".In the actual production process,it is necessary to know the types and contents of nutrient elements in compound fertilizers,and there is also a clear concept of the type and content of nutrients contained in soil.The premise of precision fertilization is to have a thorough understanding of the growing environment of the crops,so obtaining the distribution of trace element content in soil plays a vital role in precision agriculture.It is the key to guiding the production of crops in China.At present,the detection of trace elements in soil is mainly based on laboratory testing,which requires a lot of manpower and material resources,and takes a long time,including soil sample sampling,pre-processing,national standard method testing,post-processing and analysis of data.At the same time,these methods have high requirements on operators and require professionals to perform operations.The detection costs are large,time-consuming,and secondary pollution exists.Laser-induced breakdown spectroscopy(LIBS)is a new method for the detection of material analysis.It has the characteristics of short detection time,simultaneous detection of multiple elements,less pre-processing,and less pollution.It is used by majority of researchers for the rapid detection of trace elements in soil,providing a data basis for "precise fertilization" in China.This paper firstly analyzes the conventional detection methods of trace element in soil.Then,the problems of the existing detection methods are proposed.This paper summarizes the current research status of trace elements in soil by LIBS technology in China and abroad,and the shortcomings of LIBS technology in quantitative analysis are pointed out,which leads to the necessity of research work in this paper.In this paper,the formation principle of laser-induced plasma and the theoretical knowledge of support vector machine algorithm are introduced in detail.The optimalparameters of the system were determined by building a LIBS experimental platform,including delay time,detection gate width,laser pulse energy,repetition frequency,and soil porosity,etc.Based on the comprehensive experimental results,the system parameters were set as,10 MPa for pressure,pulse energy at 100 mJ,and delay time at 1.75 us,and detector gate width is 1.05 ms.The soil is a complex system,which has many types of elements.The matrix effect is a more serious problem when using LIBS to detect soil samples,and a nonlinear relationship is required for quantitative analysis.In this paper,the quantitative analysis of the trace elements manganese(Mn),strontium(Sr),and vanadium(V)in soil were carried out.The soil used in the experiment was collected in the Huaibei region of Anhui.The actual contents of the three elements were measured in the laboratory using atomic absorption spectroscopy(AAS).Under the conditions of the best system parameters,the LIBS spectrum of the soil sample was measured.The characteristic spectral lines of the three elements were selected for establishing the quantitative analysis,and the intensity of the fitted spectral lines was normalized.According to the acquired spectral data,the soil samples were divided into training set(33)and test set(11)by Random selection.The quantitative analysis models were used the grid search method(GSM),genetic algorithm(GA),particle swarm algorithm(PSO)and least squares method(LS)for parameter optimization to establish the SVM regression model.The quantitative analysis results of the three elements given by different optimization methods are compared and analyzed.The quantitative analysis results of the four optimization methods are different.Among them,LS-SVM has the best effect and the shortest training time.The correlation coefficients of the three element training sets of Mn,Sr,and V are: 0.9987,0.9998,and 0.9830,and the correlation coefficients of the test set are: 0.9670,0.9694,and0.9460.At the same time,the absolute error graphs of the four optimization methods also show that the LS-SVM error is the smallest.These results show that the LS-SVM quantitative model has good generalization ability and predictive ability,and the fitting effect is best,which is suitable for quantitative analysis of rapid detection of trace elements in soil.