Research On Deep Learning Modeling Method Based On Laser-induced Br Eakdown Spectroscopy

Laser induced breakdown spectroscopy is a kind of scattering spectrum that can detect molecular vibration information,and can give "fingerprint" information at the level of material molecules.It is an extremely important analysis technology in the scientific research community.Laser induced breakdown spectroscopy analysis has the advantages of lossless,rapid characteristics and no pollution,and is widely used in various research fields as an analysis tool.Aiming at the problem that the analysis procedure of laser induced breakdown spectroscopy is complex and the feature selection process depends heavily on manual operation in the past,in view of the advantages of deep learning algorithm,such as high accuracy,high efficiency and easy operation,the analysis method of deep learning model is introduced,which can simplify the analysis process and improve the efficiency and accuracy of analysis results.It provides help for the application of LIBS to automation,greatly improves the accuracy of material analysis,and expands the scope of application of this technology in practice.In order to improve the stability of the deep learning model and the accuracy of the prediction results,and avoid the decrease of the analysis efficiency caused by the inefficient operation in the algorithm,the feature engineering processing should be adopted for different feature types of data,the integration algorithm idea should be combined,the appropriate deep learning algorithm should be selected,the model structure should be reasonably designed,and the algorithm parameters should be appropriately optimized.It is important to improve the accuracy and stability of the analysis results.Based on the advantages of accuracy,efficiency and stability of deep learning algorithm,this paper studies the deep learning classification model method based on laser induced breakdown spectroscopy.The main research contents are as follows:Based on the LIBS data set of coal samples,classification models based on machine learning and deep learning algorithms are built respectively,and the LIBS data set of coal samples is tested.Through the analysis of the experimental results,each machine learning classification model is evaluated and summarized,and each deep learning classification model is also evaluated and summarized.Finally,the best performing model is selected,and based on this model,an improved PCA-Resnet model is proposed and optimized.The coal sample LIBS data were input into the model for training,and finally verified on the test set.The results show that the convolutional neural network model can fuse multi-scale features while retaining the feature information of the time series.In the original data set,the average accuracy,precision,sensitivity and specificity of the model are very high.It shows that the deep learning model can retain more discriminative information when optimizing the analysis process.The classification accuracy of the model is more than 98.0%.Finally,the stability of the model classification results is compared,which proves that the deep learning method has strong LIBS data processing ability,so that the method can have good performance in the classification of complex LIBS data at a low consistency.Finally,the results of classification experiments verify that the proposed method is reliable.