Research On LIBS Spectral Clustering Recognition Based On Tungsten Alloy And Aluminum Alloy

The recycling of secondary resources is a booster for China's industrial green and sustainable development.With the development of industry,metal materials are widely used in various fields due to their excellent performance.In the processing and application of metal alloys,a large amount of industrial scrap products,processing scraps and various processing corner scraps are produced every year.How to make reasonable use of these secondary resources has become a key task for China to promote industrial green development and build a green manufacturing system.At present,the scale of metal scrap produced and the proportion of scrap recycling in China are not detailed and accurate statistical reports;in addition,due to weight and other reasons,metal alloy scrap often stacks metal alloys of different types and different compositions,resulting in resources.Difficult to recycle.Therefore,the rapid and simple identification and classification of a wide variety of metal alloy scraps has become a prerequisite for the recycling of secondary resources.Laser induced breakdown spectroscopy(LIBS)is an analysis technique that has developed rapidly in recent years.It has the advantages of rapid,full-element analysis,real-time,in-situ,and long-distance detection.It has been widely used in plastics,soil,meat,steel,etc.Most of the identification researches use the least squares discriminant analysis method,cluster independent soft mode,artificial neural network,support vector machine,random forest and other algorithms to establish the model.LIBS uses a laser energy of 200mJ to ensure full excitation of the sample and minimize the impact of particle splashing.The signal collection parameters are set by ICCD,and the acquisition delay is set to 5?s according to the time evolution characteristics of the plasma to avoid the strong continuous background generated by bremsstrahlung and composite radiation,in order to obtain a LIBS spectrum with a high signal-to-background ratio to achieve LIBS signals.Time-resolved measurement.In this paper,this article designs experiments to simulate clustering when the metal grades are not known during the recycling process of waste alloys.Unsupervised cluster analysis was performed using three tungsten and tungsten alloy samples,and spectral pre-processing was used to normalize the intensity of 465 sets of spectral data to reduce the influence of experimental conditions and matrix effects on data stability,and 29 related spectral lines were screened out.The optimization of model input data is realized.Cluster analysis of the data based on DBSCAN and K-Means yielded correct results clustered into three categories.By comparing the clustering process,it can be seen that DBSCAN clustering according to density can more appropriately reflect the true situation of spectral clustering and better adapt to clusters of any shape.At the same time,parameter settings can be used to optimize the shape of the clusters to achieve the most Good classification effect;noise points can also be reasonably clear location,which is convenient for further analysis and judgment.Experiments show that clustering of scrap metal can be performed through unsupervised learning when the sample is unknown.Next,based on the recycling of waste metal,the XGBoost algorithm based on iterative tree is used to conduct supervised identification and analysis on aluminum alloy samples.Based on 600 sets of spectral data of six aluminum alloy samples,the spectral characteristics were extracted through machine learning to determine the classification basis of spectral characteristic lines.The XGBoost algorithm based on decision tree is used for automatic classification and sorting,and the processed spectral data is randomly divided into a training set and a test set.The training model is used to construct the algorithm model and its classification features are extracted.The test set is used to check the stability and usability of the model to prevent overfitting.The model obtained by XGBoost under fixed parameters has a certain degree of adaptability,is less affected by the data set,and the overall accuracy rate can reach 96.67%.Its classification characteristics are consistent with the known element content information,which proves that the spectral-based characteristic spectral line data can provide a reference for classification identification.Experimental results show that LIBS can be used for rapid identification of different types of aluminum alloys,and provides a new technology for the classification and recovery of waste metals.