LIBS Combined With Feature Fusion Method Can Improve The Accuracy Of Coal Classification

When laser-induced breakdown spectroscopy(LIBS)is used for coal detection,the characteristic spectral lines of main metal elements in coal are selected as input characteristics.Because the full spectral characteristics cannot be taken into account in the selection of characteristic spectral lines,spectral information is likely to be lost.This affects the qualitative analysis.Therefore,in view of the deficiency of feature selection,this paper adopts the feature fusion method to analyze the LIBS spectral data.The feature fusion method can make use of multiple spectral features in coal samples to realize the complementary advantages of multiple spectral features and obtain more accurate recognition results.The main contents include:(1)When LIBS is used for coal quality detection,standard anthracite samples are selected and spectral data of anthracite samples are collected under the best experimental conditions.By observing the LIBS spectrograph of coal samples and combining with NIST database,15 spectral features related to major and minor components of coal samples are selected.Through the feature fusion method,the selected 15 spectral features are fused into915 features as input features of the classification model.Based on the high dimension and complexity of coal spectral data,KNN(K-nearest neighbor,KNN)and SVM(Support Vector Machine)are selected respectively.SVM)and Random Forest(RF)can classify and predict coal spectral data before and after feature fusion.(2)In the process of constructing classification model,KNN is insensitive to anomalies compared with naive Bayes and logistic regression algorithms,and can more accurately solve the classification problem of coal data with abnormal characteristics.Therefore,firstly,KNN algorithm is used to classify and predict coal spectral data,and the accuracy rate is88.3%.Then,after the data features are fused,the newly generated features are used as the input of the KNN algorithm for classification,and a high accuracy of 92.9% is achieved.By comparing the classification results before and after feature fusion,it is found that the classification accuracy of KNN coal after feature fusion is significantly improved by 4.6%.(3)Due to the influence of coal noise data,KNN classification results are not ideal,the introduction of SVM algorithm to classify coal data,SVM is a supervised learning algorithm,SVM has the ability to overcome noise,and kernel function is introduced to solve nonlinear problems,can be used to solve the problem of data classification in the field of pattern recognition.The coal classification accuracy of SVM was 91.7%,higher than that of KNN88.3%.When the features after feature fusion are input as new features,the SVM coal classification accuracy after feature fusion is 96.3%,4.6% higher.(4)For coal classification prediction,SVM is difficult to train large-scale data set when solving coal classification problem,so SVM is not ideal for coal classification prediction.RF is an integrated learning algorithm based on decision tree learning,with strong antioverfitting ability and good performance in processing large-scale data sets.When RF is introduced to classify coal,the classification accuracy is 94.6% higher than that of SVM(91.7%),and features after feature fusion are input as new features.The coal classification accuracy of RF after feature fusion is 99.6%,which is improved by 5.0%.In this paper,laser induced breakdown spectroscopy combined with feature fusion method is used to classify and predict coal species.Select suitable features according to coal spectral lines,and then introduce the feature fusion method to fuse the selected features into new features as input features.KNN,SVM and RF were used to classify and compare the data before and after feature fusion.It was found that the coal classification accuracy of KNN,SVM and RF after feature fusion was higher than that before feature fusion,indicating that the combination of feature fusion method can improve the accuracy of coal classification.