Rapid Detection Of Heavy Metals In Tegillarca Granosa Based On LIBS

With the speeding up of the process of industrialization,a large number of poisonous and harmful heavy metals are poured into the water,which causes heavy metals pollution problems becoming more and more serious,so the detection of heavy metals is particularly important.Laser induced breakdown spectroscopy(LIBS)is a new kind of testing technique,which can analyze the sample ingredient,and it has the advantage of small destructiveness,simple operation,high resolution,wide application range and so on.However,due to the spectral noise and sample matrix effect,the detection limit and precision of LIBS are very poor,which need to be improved.The combination of chemometrics methods and spectra processing method can improve the spectral precision and stability.This paper regards Tegillarca granosa as the research object.Through constructing LIBS experiment system,the LIBS technique combined with chemometrics method is adopted to identify Tegillarca granosa polluted by heavy metals and detect Cu content in the Tegillarca granosa by quantitative analysis.The main contents are as followings:1.The pattern recognition model is established.Three kinds of recognition methods the partial least squares algorithm,support vector machine and random forest combined with wavelet transform method and information gain variable selection method are used for qualitative analysis of Tegillarca granosa samples polluted by heavy metals.Firstly,the wavelet transform method is adopted to obtain high signal-to-noise ratio of spectrum data;Then,information gain variable selection method is applied to extract characteristic variables;Finally,three kinds of discriminant method partial least squares algorithm,support vector machine and random forest are used for pattern recognition,and then the preprocessed spectral data of the Tegillarca granosa are analyzed for wavelet reconstruction.2.Multivariate calibration model is set up.In view of Cu content detection problem of Tegillarca granosa,first of all,the relevant LIBS spectra are analyzed to extract the characteristic bands including Cu characteristic spectral lines;Then,the spectral smoothing method is adopted to reduce random noise in the spectral data and Standard Normal Variate is used to reduce the baseline drift and high frequency noise effects;Finally,the partial least square regression algorithm and least squares support vector machine regression method on the characteristic spectral band are established for regression modeling,which can verify the feasibility and stability of the model.