Predicting the ash content in coal flotation concentrate based on convolutional neural network

Determination of solids weight fraction and ash content of coal in slurries of variable voidage: Laboratory measurements

Investigation of oil agglomeration behaviour of Tuncbilek clean coal and separation of artificial mixture of coal–clay by oil agglomeration

Accurate real-time monitoring of the ash content in flotation clean coal is pivotal for intelligent optimization and closed-loop control of the flotation process, directly affecting product quality and the economic performance of coal preparation plants. To address the limitations of traditional approaches–namely response lag, insufficient accuracy, and inefficient fusion of multisource information–this study proposes an intelligent online sensing method based on multisource data fusion, with the prediction pipeline decoupled into three stages: alignment – representation – prediction. First, a multiscale, differentiable dynamic time-warping (MSSoftDTW) scheme is employed to precisely align asynchronous multisource time-series data, thereby enhancing cross-modal temporal consistency. Second, an interpretable Constructive algorithm with response-weight mechanism (ICA-RW) is introduced to enable feature learning and structural adaptation, suppressing redundancy and collinearity while improving feature robustness. Third, an ensemble regression model that combines a relevance vector machine with adaptive boosting (RVM-Adaboost) is developed to better accommodate nonlinear relationships and drifts in operating conditions. By fusing X-ray fluorescence (XRF) spectra, key process variables, and features extracted from tailings images, the method achieves high-accuracy, real-time prediction of clean-coal ash content. Validation on industrial-site data demonstrates significant gains in both accuracy and stability over conventional regression baselines, meeting the real-time requirements of online monitoring and control and providing deployable support for flotation process optimization and intelligent upgrading.

Flotation characteristics of oxidized coal slimes within low-rank metamorphic

Application of central composite design method to coal flotation: Modelling, optimization and verification

The article presents brief characteristics of the Kharbalakhs and Nadezhda deposits currently being mined. Exploration data on their specific structure, the variability of the geometric parameters of the seams, the statistical heterogeneity of the ash content in coal are reported. Frame and block models of the main coal seams are constructed based on the compiled database with use of mining and geological information systems. The paper considers a number of methods for interpolation of coal grade at different density of the exploration network. The spatial distribution of ash content in coal seams is mapped. The use of the obtained results can increase the validity of the decisions made at the planning, mining and coal preparation stages.

Optimization of the oil agglomeration for high-ash content coal slime based on design and analysis of response surface methodology (RSM)

Feed property and flotation operation parameters are critical factors influencing coal flotation performance. However, no previous studies have integrated machine learning SHAP (SHapley Additive exPlanations) analysis to simulate coal flotation outcomes and interpret the influence of relevant factors. To address this gap, several machine learning models are employed to predict coal flotation performance metrics (clean coal yield and ash content). Furthermore, the impact of excluding low clean coal yield data on prediction accuracy is systematically investigated. Additionally, Pearson correlation analysis and SHAP analysis are used to quantify the relative importance of feed properties and operational parameters on flotation performance, as well as their interactive effects. Results show that the linear model achieves the R2 of 0.2881 and 0.4348 for yield and ash content prediction, while the nonlinear models achieve the largest R2 of 0.6402 and 0.9312 for yield and ash content prediction, respectively. This proves that nonlinear models have higher accuracy than linear model. Importance analysis indicates that Ad of feed, fraction of −74 μm and Ad of −74 μm particles have a significant impact on both yield and ash content. The rest variables can only affect one output or have a minimal impact on results. This research can advance the mechanistic understanding of coal flotation processes and provide actionable insights for optimizing coal flotation.

Coal slurry ash content is an important variable in the control of a flotation process. It can be concluded that coal slurry image gray feature has a certain correlation with ash content by analyzing coal slurry images at different ash values in this article. Based on image gray features, coal slurry ash content soft sensor models are developed by using BP Neural Network and simple eigenvalue-based Least Square Regression Method (LS), respectively. The simulation results of the two models indicate that the soft sensor model of coal slurry ash content based on the BP Neural Network is more optimal than that based on LS. The model based on BP Neural Network has a high accuracy when coal slurry ash content is higher than 35%. Through this investigation, coal slurry ash can be detected rapidly online.

The analysis of Lakhra Coal to determine its total carbon content and ash content is reported by calibration-free laser-induced breakdown spectroscopy (CF-LIBS). The total carbon content and ash content in coal are key parameters for efficient energy u in sepower plants. The optical emission spectrum of coal contains spectral lines of Ca, Si, Fe, Ti, Mg, Na, K, Li, Al, and C together with traces of Ba and Sr. The excitation temperature and electron number density of the laser-produced plasma were estimated using a Boltzmann plot and Stark broadened Hα line, respectively. The self-absorption correction in the emission intensities of the spectral lines were performed using the internal reference self-absorption correction (IRSAC) and the density correction. The analysis of the coal was performed by CF-LIBS, CF-LIBS with IRSAC, and CF-LIBS with density correction which were then compared with the certified composition reported by the Lakhra Coal Development Company. The results revealed that IRSAC with a fixed slope is more easily applicable and reliable than the number density correction for analyzing the total carbon content, ash content, and trace elements.

On-line ash monitors have been in use for many years in coal industry. Their operation is usually based on the scattering or the absorption of incident gamma radiation, and the derived ash value is the result of a time-averaged measurement. The natural radiation method is also used in on-line ash content measurements. A review of the ash monitors applied in Polish coal preparation plants is presented in this paper. Most of monitor installations are designed to monitor trends of ash content in coals, rather than to react to rapid changes of ash content. These data averaged over the period of tens of seconds (or several minutes) are often used by an operator to adjust the operation of a plant manually. In the case of closed loop control systems with on-line ash monitors, the dynamic properties of these instruments are important. The concept of an ash monitor with a time constant (or a time of measurement) adapting to variations of an input signal (ash content) have been analyzed. The fuzzy logic has been applied to establish the structure of the adaptive filter. Such a system allows us to speed-up the reaction of the instrument to rapid variations of ash content and at the same time to achieve better statistical accuracy over a longer period of time. This is particularly important in closed loop control systems or in splitting of a coal stream to different products. Simulation analysis of the coal blending control system applied in the "Sosnica" plant is presented in this article. Application of an on-line ash monitor with an adaptive time constant is discussed and some results of practical experiments are presented.

Study on fluidization characteristics and drying characteristics of fine coal in thermal vibration separation fluidized bed

Triboelectrostatic separation of pulverized fuel of coal power plant based on mineralogical analyses

The research is aimed at assessing the possibility of studying the ash content of coal seams using neutron-neutron logging in coal exploration wells drilled at the outcrops of coal seams under loose deposits using the materials of experimental work in the South Yakutsk coal basin. The prospects of using neutron-neutron logging to study coal well sections, on the one hand, is determined by the fact that hydrogen content in coal seams normally exceeds its content in the host rocks, and, on the other, by the small cross-sections of neutron capture by carbon. Within the same coal grade, an increase in its ash content both means a decrease in its hydrogen content and an increase in the content of elements with a higher capture cross section. Experimental studies were carried out at the Syllakh coal deposit. An IBN-8-1 type plutonium-beryllium source with an output of 5*104 neutrons/s was used as a fast neutron source, and a highly efficient SNM-17 type helium gas-discharge counter was used as a slow neutron detector. The logging depth provided by the equipment was 10 - 30 cm. During the processing of the obtained field data, a correlation was established between the count rate of neutron-neutron logging and coal ash content. It should be noted that the functional dependence of the count rate of neutron-neutron logging on ash content is not continuous – in the range of ash content of 45 - 55% a certain discontinuity point is observed, after which the functional dependence changes. To simplify the assessment of the dependence, the range of rock ash content above 45% is neglected since bituminous coals with an ash content of more than 45% are classified as carbonaceous rock and are not of industrial interest. A close correlation is established between the count rate of neutron-neutron logging and coal ash content. According to the results of statistical processing, the correlation coefficient is 0.97, which makes it possible to quantitatively determine the ash content according to neutron-neutron logging data. The absolute errors in ash content determination by neutron-neutron logging over the entire dataset are up to 3.625 %. The degree of analytical moisture influence on the data of neutron-neutron logging in determining coal ash content is estimated. No regular changes in the count rate of neutron-neutron logging due to a change in the analytical moisture index have been established. It is noteworthy that, with a sufficiently large dispersion value of the analytical moisture index, the trend line of this parameter regularly changes synchronously with the trend lines of ash content and count rate.

Breast cancer remains one of the most common cancers amongst women globally. Early detection is crucial for improving survival rates. While mammography is widely used and an effective imaging technique, it can sometimes yield false positive or false negatives. Mammogram interpretation is highly operator-dependent, introducing variability and the potential for diagnostic errors. Additionally, mammographic images have limitations, such as low contrast in breast tissue and overlapping structures that can obscure lesions or mimic abnormalities. These limitations can lead to unnecessary biopsies or delayed diagnosis. These challenges highlight the needs for advanced and data driven diagnostic tools to support and enhance current screening practices.</p> <p>In recent years, machine learning has emerged as a transformative force in the health care industry offering innovative solution to complex medical challenges. One of the most promising applications of this technology is in the early and accurate detection of breast cancer particularly through the mammography image analysis. Machine learning algorithms, especially deep learning models, have shown significant potential in identifying patterns in mammography that may indicate abnormal cases. This study utilizes the power of deep learning particularly Convolutional Neural Networks (CNNs) to enhance the classification of breast ultrasound images.</p> <p>This research proposes a novel two-level classification approach. At the first level classification, the system distinguishes normal from abnormal cases. Next, the second level classification, it further classifies abnormal cases into benign or malignant tumors. To address challenges such as limited and imbalanced datasets, the study incorporates comprehensive preprocessing techniques including contrast enhancement via Contrast Limited Adaptive Histogram Equalization (CLAHE) and region of interest (ROI). The preprocessing step were evaluated separately and in combination to assess their individual contributions. Additionally, data augmentation is utilized to improve model generalization. To further enhance performance and robustness, multiple CNN architectures are combined using an ensemble mechanism. Soft voting and hard voting leverage the strengths of diverse CNNs reducing individual model biases and improving overall accuracy.</p> <p>The main contributions of this research are: (1) the design of a two-level classification system (2) the utilization of multiple preprocessing techniques to improve image quality, (3) the application of voting ensemble learning to improve predictive stability and accuracy, and (4) an extensive evaluation on the BUSI dataset. This breaks down the complex diagnostic task into smaller, manageable steps to allow for more accurate comparisons, analysis, and improvements at each stage.</p> <p>The first preprocessing step involved the application of CLAHE. In the first level classification (normal vs abnormal), performance of various models has been evaluated. The models achieved an 82.05% accuracy for AlexNet, 84.62% for DenseNet, 85.26% for Inception, 82.69% for MobileNet, 82.05% for ResNet and 82.69% for VGG16. The accuracy of soft voting is 96.20% and accuracy of hard voting 92.09%. In the second level classification (benign vs malignant), the models achieved an 75.00% accuracy for AlexNet, 80.47% for DenseNet, 35.94% for Inception, 41.41% for MobileNet, 35.94% for ResNet and 82.81% for VGG16. The accuracy of soft voting is 90.84% and hard voting is 85.76%.</p> <p>The second preprocessing step involved the ROI was implemented. This step helped isolate the regions containing potential abnormalities, such as tumors, while filtering out irrelevant areas. After the application of ROI (normal vs abnormal), the system evaluated the performance of various models. In the first level classification (normal vs abnormal), the models achieved an 82.69% accuracy for AlexNet, 82.69% for DenseNet, 83.97% for Inception, 87.18% for MobileNet, 82.69% for ResNet and 82.05% for VGG16. The accuracy of soft voting is 100.00% and hard voting is 98.08%. In the second level classification (benign vs malignant), the system evaluated the performance of various models. The models achieved an 75.00% accuracy for AlexNet, 80.47% for DenseNet, 35.94% for Inception, 41.41% for MobileNet, 35.94% for ResNet and 82.81% for VGG16. Soft voting achieved 89.15% and hard voting achieved 80.95%.</p> <p>The final step in the preprocessing stage combines both CLAHE and ROI. The combination led to an improvement in classification performance. In the first level classification (normal vs abnormal), the models achieved an 91.77% accuracy for AlexNet, 92.72% for DenseNet, 95.57% for Inception, 96.52% for MobileNet, 76.90% for ResNet and 91.14% for VGG16. The accuracy of soft voting is 90.88% and hard voting is 93.99%. In the second level classification (benign vs malignant), the system evaluated the performance of various models. The models achieved an 87.69% accuracy for AlexNet, 99.23% for DenseNet, 96.92% for Inception, 93.08% for MobileNet, 94.62% for ResNet and 97.69% for VGG16. Soft voting achieved 96.92% and hard voting achieved 97.69%.</p> <p>These results indicate that every preprocessing step contributed to an improvement in classification accuracy across all models. This study demonstrates that a two-level, ensemble-based deep learning system can significantly improve the diagnostic accuracy of breast ultrasound images, offering a promising tool to assist radiologists and support early detection efforts, particularly in resource-constrained healthcare environments.