Huang, L. ; Chen, J. Y. ; Chen, T. B. ; Liu, M. H. ; Rao, G. F. ; Yang, H.
DATA NORMALIZATION; DOMINANT FACTOR; CLASSIFICATION; MODEL; LIBS; REGRESSION; SAMPLES; STEEL; RAMAN; NIR
Laser-induced breakdown spectroscopy (LIBS) as a rapid and green method was used to detect heavy metals Cr and Pb in pork contaminated in the lab. The laser-induced plasma was generated by a Q-switched Nd:YAG laser, and the LIBS signal was collected by a spectrometer with a charge-coupled device detector. The traditional calibration curves (CC) and multivariate partial least squares (PLS) algorithm were applied and compared to validate the accuracy in predicting the content of heavy metals in samples. The results demonstrated that the correlation coefficient of CC is poor by the classical univariate calibration method, so the univariate calibration analysis cannot effectively serve the quantitative purpose in analyzing heavy metals' residue in pork with a complex matrix. The analysis accuracy was improved effectively by the PLS method, and the correlation coefficient is 0.9894 for Cr and 0.9908 for Pb. The concentration of Cr and Pb in samples from a prediction set was obtained using the PLS calibration method, and the average relative errors for the 21 samples in the prediction set are lower than 6.53% and 7.82% for Cr and Pb, respectively. The investigated results display that the matrix effect would be reduced effectively during the quantitative analysis of pork by a LIBS-combined PLS model, and the predictive accuracy would be improved greatly compared to traditional univariate analysis. (C) 2017 Optical Society of America