Accurate quantification of alkalinity of sintered ore by random forest model based on PCA and variable importance (PCA-VI-RF)

A Versatile Method for Quantitative Analysis of Total Iron Content in Iron Ore Using Laser-Induced Breakdown Spectroscopy

Focus in quality assessment of iron ore is the content of total iron (TFe). Laser-induced breakdown spectroscopy (LIBS) technology possesses the merits of rapid, in situ, real-time multielement analysis for iron ore, but its application to quantitative TFe content is subject to interference of the iron matrix effect and the lack of suitable data mining tools. Here, a new method of LIBS-based variable importance back propagation artificial neural network (VI-BP-ANN) for quantitative TFe content in iron ore was first proposed. After the LIBS spectra of 80 representative iron samples were obtained, random forest (RF) was optimized by out-of-bag (OOB) error and then used to measure and rank variable importance. The variable importance thresholds and the number of neurons were optimized with five-fold cross-validation (CV) with correlation coefficient (R²) and root mean square error (RMSE). With using only 1.40% of full spectral variables to construct BP-ANN model, the resulted R², the root mean squared error of prediction (RMSEP) and the modeling time of the final VI-BP-ANN model was 0.9450, 0.3174 wt%, and 24 s, respectively. Compared with full spectrum-based model, for example, BP-ANN, RF, support vector machine (SVM), and PLS and VI-RF model, the VI-BP-ANN model reduced overfitting and obtained the highest R² and the lowest RMSE both for calibration and prediction. Meanwhile, the characteristics of variables selected by VI were analyzed. In addition to the elemental emission lines of Ca, Al, Na, K, Mn, Si, Mg, Ti, Zr, and Li, partial spectral baselines of 540-610 nm and 820-970 nm were also selected as characteristic variables, which indicated that VI can take into full consideration the elemental interactions and the spectral baselines. Our approach shows that LIBS combined with VI-BP-ANN is able to quantify TFe content rapidly and accurately in iron ore.

Efficient weakly supervised LIBS feature selection method in quantitative analysis of iron ore slurry

On-stream analysis of the element content in ore slurry plays an important role in the control of the mineral flotation process. Therefore, our laboratory developed a LIBS-based slurry analyzer named LIBSlurry, which can monitor the iron content in slurries in real time. However, achieving high-precision quantitative analysis results of the slurries is challenging. In this paper, a weakly supervised feature selection method named spectral distance variable selection was proposed for the raw spectral data. This method utilizes the prior information that multiple spectra of the same slurry sample have the same reference concentration to assess the important weight of spectral features, and features selected by this prior can avoid over-fitting compared with a traditional wrapper method. The spectral data were collected on-stream of iron ore concentrate slurry samples during the mineral flotation process. The results show that the prediction accuracy is greatly improved compared with the full-spectrum input and other feature selection methods; the root mean square error of the prediction of iron content can be decreased to 0.75%, which helps to realize the successful application of the analyzer.

Accuracy improvement on quantitative analysis of the total iron content in branded iron ores by laser-induced breakdown spectroscopy combined with the double back propagation artificial neural network

The rapid and accurate quantitative analysis of the total iron (TFe) content in iron ores is extremely important in global iron ore trade. Due to the matrix effect among iron ores from different origins, it is a major challenge to accurately determine the TFe content of iron ores by laser-induced breakdown spectroscopy (LIBS). The double back propagation artificial neural network (DBP-ANN) proposed in this paper provides a solution to improve the accuracy of LIBS in determining the TFe content of branded iron ores, which is a combination of pattern recognition and regression analysis based on BP-ANN. In this study, LIBS spectra of 80 batches of representative iron ore samples from 4 brands were collected. The univariate regression methods based on brand-independent and brand-hybrid were analyzed and compared for determining the TFe content of branded iron ores, and the multivariate model based on DBP-ANN was constructed for the first time. BP-ANN was employed to establish different quantitative models of the TFe content of each type of brand after brand classification of iron ores based on the BP-ANN algorithm. Compared with the brand-hybrid BP-ANN, the coefficient of determination (R²) of the test samples using DBP-ANN increased from 0.972 to 0.996, and the root mean square error of prediction (RMSEP) and the average relative error of prediction (AREP) were reduced from 0.456 wt% and 0.584% to 0.177 wt% and 0.228% respectively. Moreover, the prediction error based on the DBP-ANN model was within the error range (<0.275 wt%) accepted by the traditional chemical analysis method GB/T 6730.5-2009. Meanwhile, the established DBP-ANN method was also compared with the common multivariate method, and it showed better analytical performance. The results showed that LIBS combined with DBP-ANN has the potential to achieve rapid and accurate analysis of the TFe content of branded iron ores.

Pollution risk estimation of the Cu element in atmospheric sedimentation samples by laser induced breakdown spectroscopy (LIBS) combined with random forest (RF)

Laser-induced breakdown spectroscopy (LIBS) combined with the random forest (RF) algorithm was proposed to predict three pollution indexes (geo-accumulation index, enrichment factor, and potential ecological risk index) of the Cu element in atmospheric sedimentation samples to evaluate the pollution risk. To begin with, the LIBS spectra of 15 atmospheric sedimentation samples from different locations were collected and the copper element was identified using the National Institute of Standards and Technology (NIST) database. Then, the influence of different spectral pretreatment methods (MSC, WT and D_1st) on the predictive performance of the RF was discussed according to the calibration set with the determination coefficient (R_c²) and mean relative error (MREC) as evaluation indexes. Next, in order to obtain a better RF calibration model, a variable importance (VI) measurement was applied to select input variables from LIBS spectral data based on the optimal spectral pretreatment method, and the optimal variable importance threshold was selected as the input variable to establish the RF calibration model. Finally, the predictive performance of the optimal RF calibration model was verified using the prediction set with the determination coefficient (R_p²) and the mean relative error (MREP). The results show that R_p² of the geo-accumulation index, enrichment factor and potential ecological risk index is up to 0.9971, 0.9919 and 0.9290, respectively, and MREP of the three indexes is 0.0234, 0.1173 and 0.0810, respectively; the average relative standard deviation (RSD) of the prediction set for the three indexes is 2.16%, 5.78% and 0.71%, respectively. Furthermore, it can be inferred that Cu was at levels corresponding to serious pollution primarily because of anthropogenic activities based on the predictive I_geo, E_r and RI values. Therefore, LIBS combined with the RF algorithm is a promising means which can achieve fast and simple estimation of the pollution risk degree of Cu in atmospheric sedimentation samples without complicated sample preparation to provide a basis for pollution prevention and control measures.

Island feature classification for single-wavelength airborne lidar bathymetry based on full-waveform parameters

Because it is lightweight, low cost, and has high sampling density, single-wavelength airborne lidar bathymetry (ALB) is an ideal choice for shallow water measurements. However, due to severe waveform mixing, waveform classification has become the key difficulty in the research of single-wavelength ALB signal detection. Generally, the interaction between a laser and a water column leads to energy attenuation, pulse delay, or broadening of the water waveform, which has a discernible difference between terrestrial laser echo. This work attempts to focus on the morphology features in different waveforms to classify isolated, supersaturated, land, and water waveforms, and obtain a water-land division. The generalized Gaussian model optimized by the Levenberg-Marquardt algorithm (LM-GGM) is driven to extract 38-dimensional waveform parameters, covering different echo signals and their relationships. Ten-dimensional dominant features are selected from the feature matrix based on the random forest feature selection (RFFS) model, and input to the random forest classification model. Experiments show that the overall classification accuracy of the waveform is 97%.