| Laser-induced breakdown spectroscopy(LIBS)technology has the advantages of rapidness,slight loss,and diversity,and has become an important method for substance detection and analysis.However,there are still defects such as low sensitivity and poor precision.In order to improve the above-mentioned shortcomings,a self-made 80ns solid-state laser(wavelength:1064nm,single-pulse energy:20~200mJ)was used in this paper to investigate the long-pulse width(hundred nanosecond level)laser-induced breakdown copper alloy(BYG19431 tin bronze)plasma spectrum Characteristics and spectral characteristics at low pressure(9.6×10~4,9.2×10~4,8.8×10~4 and 8.4×10~4Pa);The data analysis algorithm of LIBS spectrum was discussed.1.The spectral characteristics of copper alloy induced breakdown by 80ns pulse were studied.The study found that at the same laser energy(20mJ),the pulse width 80ns laser has higher sensitivity and spectral intensity than the traditional 8ns laser LIBS spectrum:compared with 8ns,the lower content of elements in copper alloys under the action of80ns laser And the more difficult to excite element spectral lines(Zn I 328.23nm,Ni I331.56nm and Sn I 333.05nm),the net signal intensity increased by 40.23,41.84,43.22times.In order to further explore the experimental mechanism,the time-resolved characteristics of copper alloy spectra induced by 80ns laser were studied by comparing with traditional 8ns laser.The results showed that the LIBS background decreased when the 80ns laser pulse was used.During the whole evolution period,the maximum spectral intensity increased by 4.6 times compared with 8ns,the signal-to-back ratio(SNR)increased by 4 times,the plasma lifetime was extended by 40%,and the electron temperature was higher than 8ns during the whole period.2.The plasma spectral characteristics of copper alloy induced breakdown by 80ns pulse at low pressure were studied.The results showed that the self-absorption degree of matrix element Cu decreased significantly,the plasma temperature increased,and the spectral line widened and narrowed with the decrease of ambient air pressure.When the air pressure is 8.4×10~4Pa,the signal-to-back ratio of Cu I 324.75nm and Fe I 330.82nm lines is 5.31 times and 2.43 times higher than normal pressure;The electron temperature is increased by 21.6%;The broadening of Fe I 330.82nm line decreased from 0.29nm to0.21nm.3.The LIBS spectrum preprocessing,feature extraction and quantitative analysis algorithms were explored.The characteristic spectral data of Pb and matrix element Cu obtained from copper alloy samples induced by laser(Nd:YAG solid laser:wavelength1064nm,energy 100mJ,pulse width 8ns)were extracted for principal component spectral characteristics,and the content of Pb in copper alloy was quantitatively analyzed by the artificial neural network(ANN).The results show that the correlation coefficient of Pb element in the alloy calibration model obtained by two kinds of artificial neural network method is good,reaching above 0.987;Compared with the peak intensity feature extraction method,the average relative error of Pb element content in the alloy obtained by the principal component extraction artificial neural network method was significantly reduced from 2.3%to 0.3%.The principal component feature extraction method effectively improved the accuracy and stability of quantitative analysis of LIBS spectrum.4.A one-dimensional convolutional neural network(CNN)model with 8 hidden layers and 3 convolutional flows is constructed for feature extraction and analysis during quantitative analysis.The model has strong feature location and extraction capabilities.At the same time,the transparency of the convolutional neural network model allows researchers to visually observe its operation process in the form of feature maps,which makes up for the shortcomings of many intelligent algorithm gray boxes and even black boxes.This makes the construction of the model less blind and improves the ability of researchers to judge the rationality of the model by using knowledge and experience. |
PDF Full Text Request