Machine Learning Allows Calibration Models to Predict Trace Element Concentration in Soils with Generalized LIBS Spectra - Archive ouverte HAL Accéder directement au contenu
Article Dans Une Revue Scientific Reports Année : 2019

Machine Learning Allows Calibration Models to Predict Trace Element Concentration in Soils with Generalized LIBS Spectra

Résumé

Determination of trace elements in soils with laser-induced breakdown spectroscopy is significantly affected by the matrix effect, due to large variations in chemical composition and physical property of different soils. Spectroscopic data treatment with univariate models often leads to poor analytical performances. We have developed in this work a multivariate model using machine learning algorithms based on a back-propagation neural network (BPNN). Beyond the classical chemometry approach, machine learning, with tremendous progresses the last years especially for image processing, is offering an ensemble of powerful and constantly renewed algorithms and tools efficient for the different steps in the construction of a spectroscopic data treatment model, including feature selection and neural network training. Considering the matrix effect as the focus of this work, we have developed the concept of generalized spectrum, where the information about the soil matrix is explicitly included in the input vector of the model as an additional dimension. After a brief presentation of the experimental procedure and the results of regression with a univariate model, the development of the multivariate model will be described in detail together with its analytical performances, showing average relative errors of calibration (REC) and of prediction (REP) within the range of 5-6%. Soil test occupies a particularly important place in environment-related activities, such as agriculture, horticulture , mining, geotechnical engineering, as well as geochemical or ecological investigations 1. It becomes also crucial when an area needs to be decontaminated with respect to human activity-caused pollutions 2. Such test may often concern elements, especially metals, since a number of them are considered as essential nutrients for plants and animals 3 and some others, heavy metals for example, are determined as toxic, even highly poisonous, in large amounts or certain forms for any living material 4. It is therefore of great importance to develop techniques and methods for an efficient access to the elemental composition of soils. Established atomic spectroscopy techniques often offer good performances for quantitative elemental analysis in soils. Atomic absorption spectroscopy (AAS) offers limit of quantification (LOQ) in the order of ppm for soil samples prepared in solution 5. Similar performances can be realized with inductively coupled plasma-optical emission spectrometry (ICP-OES) 6 , while inductively coupled plasma-mass spectrometry (ICP-MS) presents for digested soil solutions, lower LOQ below 100 ppb for most of the elements found in soils 7. Beside the abovementioned techniques which can rather be considered as laboratory-based ones characterized by the need of sample pretreatment with a certain degree of complexity , other techniques have been developed with significantly less requirement of sample preparation, so being better suited for in situ and online detections and analyses. Among them, X-ray fluorescence (XRF) allows determining concentrations of major and trace elements in soils 5,8. A better performance has been demonstrated with total reflection X-ray fluorescence spectroscopy (TXRF) 9. Techniques based on plasma emission spectroscopy,
Fichier principal
Vignette du fichier
2019_Scientific_Reports_Sun.pdf (3.02 Mo) Télécharger le fichier
Origine : Publication financée par une institution
Loading...

Dates et versions

hal-02276189 , version 1 (02-09-2019)

Licence

Paternité

Identifiants

Citer

Chen Sun, Ye Tian, Liang Gao, Yishuai Niu, Tianlong Zhang, et al.. Machine Learning Allows Calibration Models to Predict Trace Element Concentration in Soils with Generalized LIBS Spectra. Scientific Reports, 2019, 9 (1), ⟨10.1038/s41598-019-47751-y⟩. ⟨hal-02276189⟩
179 Consultations
42 Téléchargements

Altmetric

Partager

Gmail Facebook X LinkedIn More