Research On Rapid Classification Of Nonferrous Metals Based On Laser Induced Breakdown Spectroscopy

Non-ferrous metals usually refer to all metals and alloys except iron,manganese and chromium.Compared with black metals,non-ferrous metals are scarce and costly.Abuse,misuse and disorderly abandonment of non-ferrous metals often cause huge economic losses and environmental pressures.The existing metal classification technology,although the test results are excellent,but generally has the defects of material damage,expensive equipment and difficult operation.Therefore,exploring a new type of nonferrous metal rapid classification and identification detection method has become a current research hotspot.Laser-induced Breakdown Spectroscopy(LIBS)is a new element analysis technology based on atomic emission spectrometry.It has the advantages of fast,real-time,in-situ,micro-damage,simultaneous analysis of multiple elements,and unlimited sample state.In recent years,it has been widely used in the chemical composition analysis of the metallurgical industry,environmental monitoring and protection,agriculture and food safety detection.Based on this,this paper takes common non-ferrous metal alloys as the experimental objects and uses LIBS technology for classification and identification.The specific research contents and results are as follows :(1)It is difficult to unify the experimental parameters and the optimal algorithm when the samples are metal materials of different types(different matrix elements and different main elements).Four common metal materials with different matrix elements(titanium alloy,aluminum alloy,aluminum bronze and microalloyed steel)and five aluminum alloy materials with different main elements(aluminum copper alloy,aluminum manganese alloy,aluminum magnesium alloy,aluminum silicon alloy and aluminum zinc alloy)were selected.The sequential analysis method was used to optimize the single laser pulse energy and acquisition delay in order to solve the problem of parameter unification.Then,six typical machine learning algorithms(supervised : linear discriminant analysis,decision tree,Naive Bayes classification,support vector machine,K nearest neighbor algorithm and unsupervised : K-means clustering)are used to train the data and make comparative analysis to solve the problem of algorithm unification.The results show that the recognition rate of the K-nearest neighbor classification algorithm model is excellent when the laser energy is 80 m J and the acquisition delay is 2.5 μs and 2 μs.The classification accuracy of the test set of the two groups of samples reaches 99.95 % and 100 %.(2)The classification results of various algorithms are compared.By using six different classical algorithms to classify and analyze the LIBS spectra of five different brands of aluminum alloy samples,it is concluded that the K-nearest neighbor classification algorithm model has high recognition rate,and as one of the most basic machine learning algorithms,it is simple and easy to optimize so that it is suitable for the study of this paper.(3)The rapid and accurate classification method of titanium alloy with different national standard numbers under the same brand was studied.Aiming at the problem that the spectral similarity caused by the similarity of element content is too high and the classification accuracy is too poor,LIBS technology combined with K nearest neighbor algorithm is used.Through spectral screening,data dimension reduction and other processing methods and constantly adjusting the kernel parameters of the algorithm,the final input variables are reduced,and the modeling efficiency and spectral classification accuracy of the algorithm are improved(the cross validation accuracy of the training set is increased from 81.40 % to 98.64 %,and the classification accuracy of the test set is increased from 84.20 % to 99.14 %).The AUC value of the evaluation model is also improved from 0.9643 to 0.9991.The rapid and accurate classification of the same grade titanium alloys is realized.This paper realizes the rapid and accurate classification of non-ferrous metal materials.The advantages of laser-induced breakdown spectroscopy in metal classification are proved,which is of great value to further promote the application of laser-induced technology in the detection,application,recovery and reuse of non-ferrous metals.