Online Detection Of LIBS Molten Steel Based On Neural Network

Today,under the national policy of "energy-saving and emission reduction,reducing energy consumption".Many steel companies are facing the awkward situation of excessive energy consumption and overcapacity.In the situation,steel companies geting out of their predicament,not only can be effectively lower the cost of human resources,but also can improve the quality of steel production.Currently the world's most advanced detection of a component is laser induced breakdown spectroscopy,LIBS.LIBS uses a high-energy laser to irradiate the surface of an object,The surface of the object to generate plasma,the use of a spectrometer to analyze the emission spectra of the plasma.It has its own features,remote non-contact detection,less damage,such as steel composition testing has a strong prospect.Although LIBS technology has many advantages,its method of analysis is not mature,In this paper,the practicality of LIBS technology as a target to build LIBS material composition analysis platform,the main study of quantitative analysis of substances.Firstly,set up a set of LIBS experimental platform to understand the experimental equipment performance,The influence of several key parameters during the experiment on LIBS results was studied and selected the more appropriate parameters because the production of liquid steel line with a large number of overlapping lines,self-absorption between lines,and deformity and other issues,the results of a certain impact.In order to denoise the spectral signal,the wavelet threshold method can be used to achieve the purpose of denoising simply and effectively,after the de-noising processing of the spectral data,the use of genetic algorithms and neural networks combined quantitative analysis method,and the spectral peak intensity to do the normalization,the quantitative analysis of Mn content in molten steel.Compared with the traditional LIBS calibration results,the accuracy has been greatly improved.