Abstract:NdFeB material has excellent comprehensive magnetic properties, which plays a crucial role in the field of rare earth magnetic materials, and is one of the important basic materials to support modern high-tech industries. In the production of NdFeB alloys, the control of rare earth elements (REEs) is directly related to the quality and resource utilization efficiency of NdFeB materials. Therefore, the prediction of REEs content in NdFeB alloys is of great research significance and application value for the quality control of products, the enhancement of production efficiency and reduction of energy consumption by the rare earth material industry. In this study, a combination of laser-induced breakdown spectroscopy (LIBS) and random forest (RF) was used to investigate the quantitative analysis of four REEs (Nd, Pr, Tb and Dy) in NdFeB alloys. Firstly, the original LIBS spectra were screened by principal component analysis-mahalanobis distance (PCA-MD). Then, the effects of different data processing methods on the screened LIBS spectra were explored, and next, the feature variables were extracted from the preprocessed spectral data by the variable importance measurement (VIM). In order to further verify the prediction performance of the model, the prediction results of the RF models based on the different methods were compared. Finally, a PCA-VIM-RF calibration model was established on the basis of the optimized spectra, selected feature variables and parameters. Leave-one-out cross validation (LOOCV) was used to optimize the parameters of PCA-MD method, spectral preprocessing method and variable importance thresholds during the construction of the calibration model. The results show that the PCA-VIM-RF model has better prediction performance than the RF calibration model based on the raw spectra. For the PCA-VIM-RF calibration method of Nd, Pr, Tb and Dy elements, the values of R 2 CV are 0.9991, 0.9998, 0.9986, and 0.9984, respectively, the values of RMSE CV are 0.06296%, 0.02788%, 0.04647% and 0.05252%, respectively. The values of R 2 P are 0.9508, 0.9975, 0.9691 and 0.9457, respectively, and the values of RMSE P are 0.6082%, 0.09205%, 0.5776% and 0.2631%, respectively.The above results indicate that LIBS combined with PCA-VIM-RF algorithm is a promising method for rapid quantitative analysis of REEs in NdFeB alloys without complicated sample preparation, which can provide some new ideas or strategies for the future research, development and quality control of rare earth materials.

Soft Measurement of Rare Earth Multi-Element Component Content Based on Multi-LightVGG Modeling

Multiple models soft-sensing technique based on online clustering arithmetic

Modeling Task Relationships in Multivariate Soft Sensor with Balanced Mixture-of-Experts

Component Content Soft-Sensor Based on Hybrid Models in Countercurrent Rare Earth Extraction Process

A Model Combining Seq2Seq Network and LightGBM Algorithm for Industrial Soft Sensor

Enhancing model robustness through different optimization methods and 1-D CNN to eliminate the variations in size and detection position for apple SSC determination

Component Content Soft-sensor Based on RBF Neural Network in Rare Earth Countercurrent Extraction Process

Predicting Rare Earth Elements Concentration in Coal Ashes with Multi-Task Neural Networks

An Improved GrabCut Method Based on a Visual Attention Model for Rare-Earth Ore Mining Area Recognition with High-Resolution Remote Sensing Images

Rapid quantitative analysis of multiple rare earth elements in NdFeB alloys based on laser-induced breakdown spectroscopy (LIBS) and random forest (RF)

Multi-element Quantitative Extraction of Mining Area of Laterite Nickel Mine Combined Deep Learning with Object-Oriented Method

Deep multimodal fusion model for moisture content measurement of sand gravel using images, NIR spectra, and dielectric data

Modeling Task Relationships in Multi-variate Soft Sensor with Balanced Mixture-of-Experts

Advancing Rare-Earth Separation by Machine Learning

Prediction of Mechanical Properties of Rare-Earth Magnesium Alloys Based on Convolutional Neural Networks

Deep Learning Application for Predicting Soil Organic Matter Content by VIS-NIR Spectroscopy

Efficient image segmentation based on deep learning for mineral image classification

Predicting element concentrations by machine learning models in neutron activation analysis

An integrated CBLA-Net with fractional discrete wavelet transform and frequency-based CARS to predict heavy metal elements by XRF

A Lightweight Model of VGG-16 for Remote Sensing Image Classification

Rapid detection of Molybdenum Ore Grade Based on Visible-Infrared Spectroscopy and MTSVD-TGJO-ELM