In large coal preparation plants with a capacity of 30 million tons/year, the belt speed can reach 7 m/s and the thickness of the material layer can reach 500 mm. Therefore, in high-throughput and complex environments, the problem exists that harmful feeding materials such as iron and gangue are not easily detected, and thus fault diagnosis in the crushers lags behind. Therefore, it is necessary to extract the equipment operation signals from the noisy production environment and identify the feeding materials. Currently, there is no systematic research on signal processing and image classification of crusher feeding materials, while the convolutional neural network (CNN) is outstanding in computer vision. In this paper, sound and vibration signals of the feeding materials are denoised by spectral subtraction and transformed into feature images by continuous wavelet transforms. Then, an image classification model based on CNN is built for these feature images to study its classification mechanism and performance. The results show that the model classification accuracy is respectively 84.0%, 93.5% and 80.1% in coal–iron–wood classification, coal–iron classification, and coal–wood classification. The good classification performance for coal, iron and wood can satisfy the practical demands to remove the harmful feeding materials, which provides the core technical support for the establishment of operating status monitoring and fault diagnosis system of crushing equipment.

To investigate the effect of crack angle on the fracture performance of brittle materials under tensile load, a molecular dynamics simulation method based on ReaxFF is used to establish an amorphous silica model through the high-temperature melting and annealing process. Under the simulation environment of 300 K, 1.013 × 10⁵ Pa and 5 × 10⁹ s⁻¹, the impact of crack angle on the fracture performance of the model from three perspectives is analyzed: material mechanical properties, micro fracture process, and energy evolution. The result indicates that as the crack angle increases, the ultimate strain and stress of the model decrease accordingly. The crack propagation path of the model will exhibit a “Z” shape due to the coupling effect of tensile and shear stress. The elastic energy efficiency and new surface energy efficiency of the model increase with the increase in crack angle, and the most new surface is generated at 45° crack angle. The linear regression model and asymptotic regression model are used to fit the trends of elastic energy efficiency and new surface energy efficiency with crack angle, respectively, with correlation coefficients R² of 0.986 and 0.994. In the actual comminution process, the input energy required for crushing as well as the surface area and morphology of the material after crushing can be changed by adjusting the angle between the load and the main crack of the material being broken.

The seepage and diffusion of acid mine drainage (AMD) generated from self-heating coal gangue tailings caused acid pollution to the surrounding soil and groundwater. Red mud derived from the alumina smelting process has a high alkali content. To explore the feasibility of co-disposal of coal gangue and red mud for prevention of AMD, coal gangue and red mud were sampled from Yangquan (Shanxi Province, China), and dynamic leaching tests were carried out through the automatic temperature-controlled leaching system under the conditions of different temperatures, mass ratios, and storage methods. Our findings indicated that the heating temperature had a significant effect on the release characteristics of acidic pollutants derived from coal gangue, and that the fastest rate of acid production corresponding to temperature was 150 °C. The co-disposal dynamic leaching tests indicated that red mud not only significantly alleviated the release of AMD but also that it had a long-term effect on the treatment of acid pollution. The mass ratio and stacking method were selected to be 12:1 (coal gangue: red mud) and one layer was alternated (coal gangue covered with red mud), respectively, to ensure that the acid-base pollution indices of leachate reached the WHO drinking-water quality for long-term discharge. The results of this study provided a theoretical basis and data support for the industrial field application of solid waste co-treatment.

Profound knowledge of the movement characteristics and spatial distribution of the particles under compression during the crushing of rocks and ores is essential to further understanding kinetic energy release law. Various experimental methods such as high-speed camera technology, the coordinate method, and the color tracking method were adopted to improve the understanding of particles’ movement characteristics and spatial distribution in rock comminution. The average horizontal velocities of the four size particles α, β, γ, and δ are statistically calculated. The descending order of the particles’ average velocity is γ, β, α, and δ. In comparison, the descending order of the particles’ kinetic energy is α, β, γ, and δ. Moreover, the contribution of α particles to the total kinetic energy exceeds 70%. The spatial distribution characteristics of coarse and fine particles show different results. The probability of fine particles appearing in the range closer to the center area is greater, while the position of large particles appears to be more random. The color tracking results show that super-large particles generated by crushing are on the specimen’s surface, while small particles are generally produced from inside. The above results indicate a connection between the particle generation mechanism, movement characteristics, and spatial distribution in the comminution process.

In mineral engineering, harmful foreign objects mixed in the feed can cause faults in the sizer crusher (Sizer). The traditional feed identification methods are inefficient and manual, which fail to adapt to the present production conditions of high belt speed, thick material layer and large throughput. In previous studies, Sizer feed identification focused on single-component rather than mixed feeds. Therefore, we proposed a new Sizer mixed-feed identification method based on continuous wavelet transform (CWT) and squeeze-and-excitation residual network (SE-ResNet). Firstly, the three-channel sensor signals acquired during the Sizer feeding process were subjected to CWT, and the resulting scalograms were fused as the image dataset. Secondly, a Sizer feed identification model based on SE-ResNet was designed for single-component and mixed-feed classification. Ablation studies and method comparisons were conducted to validate the performance of the proposed model. Finally, the method of this paper was tested with industrial datasets to demonstrate its effectiveness and generalizability. The results show that the proposed method can achieve an accuracy of 85.99±0.16% for Sizer mixed-feed identification. This method also significantly improves the identification ability of wood, which solves the problem of distinguishing between coal and wood in the feed. Besides, industrial tests suggest that our method has good Sizer mixed-feed identification performance with an average accuracy of 95.64±1.77%, which can meet the demand of industrial production for the safe and smooth operation of the Sizer.