| As an important metal material,alloy steel is widely used in all aspects of national economy,especially in national defense projects.There are many varieties of alloy steel,in spite of its main contents being iron,alloy steel contains silicon,manganese,chromium,nickel,copper and the other elements.Whether the alloy steel contains these elements or how much it contains these elements,impacts on the performance and quality of alloy steel directly.Therefore,how to control the liquid steel in real time during the smelting of the alloy steel and add the raw materials timely according to the actual content of the detected elements so as to ensure that the content of the alloy steel finished in the final smelting are all in line with the standards becomes obviously important problem.At present,most of the detection methods are to test the composition after smelting the finished product.If the test fails,it will cause a lot of manpower and material waste in the production process of the alloy steel,which is hard to adapt to the current process requirements.Laser-Induced Breakdown Spectroscopy(LIBS)is a kind of material composition and concentration analysis based on atomic emission spectroscopy.It has the advantages of fast analysis,small sample destruction,simultaneous analysis of multiple elements,long-distance and non-contact type analysis.It has great potential for online analysis and is very suitable for on-line real-time monitoring of major elements in the process of smelting alloy steels.However,due to the influence of environmental factors and mechanism noise,the LIBS spectrum contains a lot of noise in the process of obtaining the actual spectrum,which affects the accuracy of the spectral data analysis in the later stage and thus affects the online presence of the final elemental composition and concentration measuring.Based on the analysis of LIBS mechanism noise with the working principle of LIBS,the noise of LIBS is divided into two kinds: linear spectral noise and non-linear spectral noise.And the corresponding noise reduction is proposed for each kind of noise method.The specific research contents and research results are as follows:Aiming at the linear spectral noise emitted by atoms and ions in LIBS noise,and considering the weight of different elements in the alloy steel is different and the large amount of data in the spectrum which has correlation,the weighted principal component analysis(WPCA)is applied to reduce the linear spectral noise.The LIBS spectrum which collected from 20#?30#?Q195?T10 liquid steel is used to make a test and compared with the PCA method currently used by the industry.Experimental results show that WPCA method can effectively reduce the linear noise in LIBS spectra and has better performs on the noise reduction and execution time than PCA.Aiming at the existence of electrons and the continuous spectral noise emittedby electrons and ions in LIBS noise,and considering the continuous spectral noise contains many characteristic lines that may be submerged and the large amount of data in the spectrum,the sparse self-coding method is applied to reduce the continuous spectral noise.In this paper,the LIBS spectrum which collected from20#?30#?Q195?T10 liquid steel are used to make a test and the network structure and the number of hidden layers were adjusted according to the value of signal to noise ratio Coding neural network model.The above neural network model is used to experiment on the above spectrum.Experimental results show that the sparse self-coding method can effectively reduce the continuous background noise in the LIBS spectrum. |
PDF Full Text Request