Iron Ore Classification Research Articles

Interpretable deep learning-assisted laser-induced breakdown spectroscopy for brand classification of iron ores

Brand classification of iron ores using laser-induced breakdown spectroscopy (LIBS) combined with artificial neural networks can quickly realize the compliance verification and guarantee the interests of both trading partners. However, its practical application is impeded by complex pretreatments and unexplained feature learning problems. According to the LIBS data characteristics of iron ores, a convolutional neural network (CNN) is designed to predict 16 types of brand iron ores from Australia, Brazil, and South Africa. The accuracies of the calibration set and the prediction set with five-fold cross-validation (5-CV) were 99.86% and 99.88%, and the value of loss function was 0.0356. Meanwhile, the established CNN method was also compared with common machine learning methods using raw spectra as input variables, and it outperformed other methods. For the first time, this work interprets the CNN's effectiveness layer by layer in self-adaptively extracting LIBS features through t-distributed stochastic neighbor embedding (t-SNE) and the quantitative data of major chemical components in iron ores. Our approach shows that deep learning assisted LIBS is able to significantly reduce manual factors in preprocessing and feature selection and has broad application prospects in the brand classification of iron ores.

Classification of iron ore based on machine learning and laser induced breakdown spectroscopy

Iron ore is a very important mineral resource. Its development and utilization have a great impact on the development of the iron and steel industry. The selection and classification of iron ore is an indispensable link in the metallurgical industry. Different types of iron ores and its grade will directly affect the ratio of other substances, so the research on the selection and classification of iron ore is of great significance in the metallurgical industry. Laser-induced breakdown spectroscopy (LIBS) is a recently developed component detection technology. It has the advantages of non-destructive, fast, in-situ online detection, etc., and has certain advantages in the field of chemical composition detection and sample classification. In order to study the method of improving the classification accuracy of iron ores, 10 kinds of natural iron ores, including hematite, limonite, siderite, mica hematite, magnetite, maghmite, oolitic hematite, pyrite, cobalt-bearing magnetite, pyrrhotine, were classified with LIBS and machine study. In this study, 10 kinds of natural iron ores, were ablated by LIBS to obtain their corresponding spectral data; then the 10 features corresponding to the maximum spectral intensity were obtained by setting a threshold; the classification training and testing on selected feature spectra were performed with KNN, RF, and SVM models. The results show that the classification accuracy of the three machine learning models: KNN, RF and SVM are 83.0%, 80.7%, and 90.3%, respectively. It can be seen from the classification accuracy that combination of LIBS and machine learning can achieve rapid and accurate classification of iron ores, which will provide a new method for classification of iron ores in the metallurgical industry.

Metamorphic disturbances of magnetite chemistry and the Sm-Nd isotopic system of reworked Archean iron formations from NE Brazil

Iron formations are valuable archives of sedimentary conditions and post-depositional events. However, geochemical proxies commonly used to determine genetic characteristics can be variably modified during metamorphism and deformation, hampering their use as records of regional geological events. This work focuses on strongly reworked magnetite-quartz-rich rocks from the São José do Campestre Massif, one of the oldest fragments of preserved crust in South America. The genetic classification of these magnetite-quartz-rich rocks is not straightforward because primary assemblages and textures were variably modified by granulite facies metamorphism during a regional Paleoproterozoic migmatization event. To address genetic ambiguities, we analyzed their magnetite and pyroxene chemistry, whole-rock geochemistry, and Sm-Nd isotopes. Magnetite chemistry indicates that pyroxene-poor iron formations (Type B) are low in trace elements such as Ti, Al, V, and Mn, suggesting a chemical similarity to iron formations elsewhere. In contrast, magnetites from pyroxene-enriched Type A iron formations are rich in trace elements and more akin to magnetite crystallized from higher temperature systems, such as skarn and IOCG. The 147Sm/144Nd of these rocks show substantial variation even at the outcrop scale, indicating a locally-controlled, highly heterogeneous mixture of Archean, Paleoproterozoic, and Neoproterozoic sources. Therefore, our geochemical tools point out to heterogenous signatures of these magnetite-quartz rocks and proxies compatible with both low and high-temperature conditions and age of deposition spanning sources from the Archean to the Neoproterozoic. We interpret that the studied São José do Campestre magnetite-quartz rocks represent Archean iron formations with original magnetite chemistry and isotopic signatures variably modified by metamorphism and by at least one deformation-related hydrothermal event. These results contrast with similar examples from China and Greenland where iron formations either preserved the magnetite chemistry or the primary isotopic signatures. Our study indicates that metamorphism can selectively affect chemical proxies used to study iron formations and undermine the genetic classification of iron ores. Thus, these proxies should be carefully applied in the interpretation of syn-depositional environments of polydeformed belts.

Iron ore identification method using reflectance spectrometer and a deep neural network framework

In the first selection stage of iron ore, the ore classification accuracy plays a decisive role in subsequent work. Therefore, how to identify iron ore quickly and accurately is an important task. Traditional chemical, physical and manual identification methods have the disadvantages of high costs and high time consumption. This research proposes a new iron ore identification method, that combines deep learning with visible-infrared reflectance spectroscopy to establish an iron ore classification model. We collected iron ore samples from the Anshan iron ore area and measured the spectral data with a spectrometer. Then, a deep neural network framework is proposed based on the convolution neural network and the improved extreme learning machine algorithm, and an iron ore classification model is established based on the framework. The results show that the proposed model can effectively identify the types of iron ore, and the overall accuracy reaches 98.11%.

Application of Scikit and Keras Libraries for the Classification of Iron Ore Data Acquired by Laser-Induced Breakdown Spectroscopy (LIBS).

Due to the complexity of, and low accuracy in, iron ore classification, a method of Laser-Induced Breakdown Spectroscopy (LIBS) combined with machine learning is proposed. In the research, we collected LIBS spectra of 10 iron ore samples. At the beginning, principal component analysis algorithm was employed to reduce the dimensionality of spectral data, then we applied k-nearest neighbor model, neural network model, and support vector machine model to the classification. The results showed that the accuracy of three models were 82.96%, 93.33%, and 94.07% respectively. The results also demonstrated that LIBS with machine learning model exhibits an excellent classification performance. Therefore, LIBS technique combined with machine learning can achieve a rapid, precise classification of iron ores, and can provide a completely new method for iron ores’ selection in the metallurgical industry.

Development of an expert system for iron ore classification

The accurate classification of iron ores is one of the challenging tasks in the mining industry, and at the same time, it is essential for the future destination of the ores. The present study attempts to develop a machine-vision-based system, combining feature extraction, feature selection, and image classification, in order to predict the class of the iron ores. A total of 280 image features were extracted from each of the captured images of iron ores. The feature includes 10 statistical features for each component of the six colour spaces (RGB, HSI, XYZ, CMYK, Lab, and Grey) and four frequency-transformed components (DCT, DFT, DWT, and Gabor filter). In order to select the optimum feature subset for the model development, a sequential forward floating selection (SFFS) algorithm was used. The optimum feature contains only 3% of the total number of features. The support vector machine (SVM) algorithm was used for the development classification model. The four confusion matrix parameters (sensitivity, specificity, accuracy, and misclassification) along with the Q-statistics and the correlation were used as performance criteria for the model classification and were found to be 0.9792, 0.9949, 0.9918, 0.0082, 0.9999, and 0.9695, respectively. The high value of sensitivity, specificity, and accuracy and the low value of the misclassification indicate a good performance of the model. The performance of the model was also compared with other classification algorithms (k-nearest neighbours, classification tree, discriminant classifier, naive Bayes). It was observed that the proposed algorithm performs reasonably well in terms of categorising the classes of iron ore.

Development of machine vision-based ore classification model using support vector machine (SVM) algorithm

The product of the mining industry (ore) is considered to be the raw material for the metal industry. The destination policy of the raw materials of iron mine is highly dependent on the class of iron ores. Thus, regular monitoring of iron ore class is the urgent need at the mine for accurately assigning the destination policy of raw materials. In most of the iron ore mines, decisions on ore class are made based on either visual inspection by the geologist or laboratory analyses of the ores. This process of ore class estimation is time consuming and also challenging for continuous monitoring. Thus, the present study attempts to develop an online vision-based technology for classification of iron ores. A laboratory-scale transportation system is designed using conveyor belt for online image acquisition. A multiclass support vector machine (SVM) model was developed to classify the iron ores. A total of 2200 images were captured for developing the ore classification model. A set of 18 features (9-histogram-based colour features in red, green and blue (RGB) colour space and 9-texture features based on intensity (I) component of hue, saturation and intensity (HSI) colour space) were extracted from each image. The performance of the SVM model was evaluated using four confusion matrix parameters (sensitivity, accuracy, misclassification and specificity). The SVM model performance was also compared with the other methods like K-nearest neighbour, classification discriminant, Naive Bayes, classification tree and probabilistic neural network. It was observed that the SVM classification model performs better than the other classification methods.

Iron ore classification by XRD-Rietveld and cluster analysis

This study establishes a routine for quantitative phase analysis by X-ray diffraction (XRD-Rietveld) in iron ore samples. Studied samples came from Quadrilátero Ferrífero, Southeastern Brazil, and are mainly composed by hematite, magnetite, goethite, quartz and kaolinite, being divided in three clusters: cluster 1 (n = 8), characterized by the presence of hydrated phases goethite (25 to 72 wt %), kaolinite (< 25 wt %) and gibbsite (< 10 wt %), plus hematite (7 to 54 wt %) and quartz (2 to 19 wt %); cluster 2 (n = 16), which consists of hematite (22 to 81 wt %) and quartz (2 to 58 wt %) and smaller proportions of goethite (< 22 wt %) and kaolinite (< 13 wt %); and cluster 3 (n = 3), that has a predominance of quartz (62 to 74 wt %) and smaller amounts of hematite (17 to 33 wt %) and kaolinite (< 5 wt %). Classification of ores by XRD-Rietveld and cluster analysis is a fast, low-cost and reliable option for substituting ore classification with chemical analysis.

Iron ore classification by XRD-Rietveld and cluster analysis

Iron ore classification by XRD-Rietveld and cluster analysis

Classification of iron ores by laser-induced breakdown spectroscopy (LIBS) combined with random forest (RF)

Laser-induced breakdown spectroscopy combined with the random forest (RF) algorithm was proposed for the classification of ten iron ore samples.

Assessing the metallurgical value of iron ore

Given the lack of a generally accepted method for fast determination of the metallurgical value or iron ore in terms of its chemical composition, it is necessary to develop a method of determining a general (or integral) quality index of iron ore. In the iron-ore classification adopted in the sintering department at AO ArcelorMittal Temirtau, the Harrington desirability function is employed. This method permits integral assessment of iron ore—on the basis of the iron content, the composition of the barren rock, and the presence of useful components and harmful impurities—and comparison of any iron ore with the traditional ore used in sintering, in terms of chemical composition. The method also permits prediction of the enrichment of regular iron ore.

Improved blast pattern design on Mt Whaleback based on the Pilbara iron ore classification (PIOC) : Bellairs, P G Proc Large Open Pit Mining Conference, Newman, October 1986P61–71. Publ Parkville: Aus IMM, 1986

Improved blast pattern design on Mt Whaleback based on the Pilbara iron ore classification (PIOC) : Bellairs, P G Proc Large Open Pit Mining Conference, Newman, October 1986P61–71. Publ Parkville: Aus IMM, 1986

鉄鋼原料の鉄鉱石，ペレットおよび石炭の被粉砕性

The grindability of various types of ore and coal was studied and a new classification of iron ores is proposed on the basis of the ore-forming temperature of the deposits. The work indices of magnetite and hematite minerals ranged between 12.5-24.1kWh/t and 6.5-23.8kWh/t, respectively, and, in both cases, increased with the ore-forming temperature. It was also found that iron ores were mainly ground according to Bond's law, and coals were ground according to Rittinger's law in batch-type ball mill tests. It is proposed that the grindability of iron ores is characterized by the ore-forming temperature of deposits and that of coal, by its coking ability.