Research On XRF Detection Method For Assessment Of Soil Heavy Metal Pollution

In recent years,due to the rapid development of industry and agriculture in China,especially the uncontrolled mining of mineral deposits,"three wastes" discharge,pesticide input and abuse,resulting in the accumulation of a large number of heavy metal elements in the soil caused by the continuous deterioration of soil environmental quality.Therefore,the detection of heavy metals in soil is becoming more and more important not only in agricultural production activities,but also in environmental management and soil pollution remediation.At present,the standard method of heavy metal detection in soil is still based on laboratory analysis,with large sampling workload,complex pretreatment,long testing cycle,poor timeliness and high cost,which cannot meet the requirements of rapid analysis and detection on site.It is urgent to develop rapid analysis and detection methods and technologies of soil heavy metals.Energy dispersive X-ray fluorescence spectrometer(EDXRF)has the advantages of fast and efficient analysis,high accuracy,convenient operation and on-site detection,and is used in many fields of non-destructive analysis of elements.Due to the complex soil composition,some of the heavy metals(such as cadmium)have low content,which is vulnerable to the interference of background and overlapping peaks between elements,which cannot be accurately measured during XRF analysis.Therefore,it is a meaningful work to reduce the detection limit of XRF in soil trace heavy metal elements and to establish a better qualitative and quantitative inversion model.In this paper,the qualitative and quantitative methods for soil heavy metal detection are mainly studied in four aspects:improving the signal-to-noise ratio of raw spectra,accurate correction of matrix effects,reasonable selection of spectral components and optimization of intelligent inversion algorithms.The results are as follows:(1)The XRF spectra pre-processing method was studied.An X-ray fluorescence spectral signal denoising method based on fractional-order wavelet transform(FWT)is proposed.The method takes advantage of the fractional order based on the wavelet transform denoising to retain the details of the characteristic peaks of the noisy signal effectively,and the denoising effect is better.The accuracy and reliability of the method are verified by Monka(MC)simulations and real spectra.Results of MC simulations and real spectra show that the Iair PLS method outperforms the asymmetric weighted penalized least squares baseline correction(ASLS),multi-constrained asymmetric least squares(mca LS),and adaptive reweighted penalized least squares(air PLS)methods.In the Iair PLS method,the improved exponential function can effectively reduce the risk of baseline underestimation and speed up the weight iteration process.Practical experimental results of X-ray fluorescence spectroscopy show that the Iair PLS method can be maintained at a low root mean square error(RMSE)when the smoothing factor is set to 5.5,and the method can provide the best baseline estimate for real soil samples.(2)The optimal selection of different elemental feature variables and the optimized quantitative inversion model of the intelligent algorithm for soil heavy metal elements were investigated.The feature variable selection for the spectral interval of the soil using the Least Angular Competitive Adaptive Re-weighting Variable Feature Selection(LAR-CARS)algorithm reduces the size of the feature variables,and the effectiveness of the method is verified by comparing it with conventional feature selection methods.The Gray Wolf algorithm was then used to optimize the support vector machine(GWO-SVR)model,which improved the accuracy,stability and reliability of quantitative analysis of soil heavy metals such as Cu,Cr,Pb,Zn,Ni,As and Cd compared with the partial least squares(PLS)and support vector machine(SVR)models.Finally,the quality control factors and error sources between soil XRF analysis and laboratory inductively coupled plasma mass spectrometry(ICP-MS)analysis data were analyzed,and the performance of XRF analysis models,such as prediction accuracy,was compared and analyzed,and comparable results to laboratory ICP-MS detection were obtained.Meanwhile,based on the proposed algorithm,a simple qualitative and quantitative analysis software was developed,which is convenient for users to use.(3)The application of XRF spectral analysis in the assessment of soil heavy metal contamination was studied.The XRF nondestructive and rapid analytical detection technique combined with Kriging interpolation(OK)and positive definite matrix factor(PMF)analysis technique was applied to the ecological risk assessment of soil heavy metal pollution in Poyang Lake area,achieving good results,which provides important technical support for the rapid identification of soil pollution sources and soil remediation treatment in the future.In summary,this study established a spectral preprocessing method based on fractional-order wavelet variation denoising and improved adaptive iterative reweighting penalizing least-squares spectral background deduction to eliminate the effects of noise and background on the spectra;investigated the effects of least angle The effects of competitive adaptive reweighting variable feature selection and gray wolf algorithm to optimize the support vector machine quantitative analysis model on the quantitative analysis of soil trace heavy metals were investigated,and the prediction accuracy of the model was compared with the ICP-MS detection results using the model in a cross-sectional manner.Finally,the new model was applied to soil heavy metal measurements in the Poyang Lake area to qualitatively and quantitatively identify soil heavy metal pollution sources in the study area from a relatively objective perspective,providing some technical basis for further validation that XRF can replace ICP-MS analysis techniques.