The deep learning object detection model can well extract the coordinate information of the separating point of the concentrate ore belt boundary line. However, with the drawbacks of acquired zonal image features, it is difficult to obtain the perfect prediction of the concentrate grade and recovery rate. In this study, by adapting deep learning semantic segmentation technique with DeepLab V3+, which can effectively extract multi-dimensional image features. The mapping relationship between image features and attributes of the Tabling equipment is achieved by constructing a multi-output support vector regression model optimized of a sparrow search algorithm (SSA-MSVR). The beneficiation indicators and operating parameters of the Tabling can be continually detected and optimized by an intelligent system that mainly includes image recognition softwares, automatic control units, data processing and communication workstations, and matching intelligent equipment, which accomplished the intelligent deployment solution of the Tabling beneficiation process.

This work studies the mechanism and optimization methods of the filter press dehydration process to better improve the efficiency of the concentrate filter press dehydration operation. Machine learning (ML) models of radial basis function (RBF)–OLS, RBF-generalized regression neural network, and support vector regression (SVR) are constructed, and laboratory and industrial simulations are performed separately, finally, optimization methods for the filtration dewatering process are designed and applied. In laboratory, all the machine learning models have obvious mistakes, but it can be seen that SVR has the best simulation effect. In order to achieve the optimization of the entire filtration and dewatering process, we obtained enough data from the industrial filtration and dewatering system, and in the industrial simulation results all the machine learning models performed considerably, SVR achieves the best accuracy in industrial simulation, and the simulated mean relative error of moisture and processing capacity are 1.57% and 3.81%, the model was tested with newly collected industrial data to verify the credibility. The optimal simulation results are obtained by optimization method based on control variables. Results show that the ML method of SVR and optimization methods of control variables applied to the industry not only can save energy consumption and cost but also can improves the efficiency of filter press operation fundamentally, which will provide some options for intelligent dewatering process and other industrial production optimization.

Injected Direct Current-Magnetic Field Measurement (IDC-MFM) is a promising nondestructive technique for crack evaluation of aluminum alloy due to its high sensitivity to defect states. Finite element simulation and experiment were introduced in this research to reveal the relationship between the magnetic field and the crack size. The finite element simulation showed that the magnetic field at the defect increased with the currents, and the peak-to-peak value of the tangential component magnetic field B_x (ΔB_xpp) correlated with both the width and depth of the crack very well. The high-sensitivity tunnel magnetoresistance probe was used for crack detection of aluminum alloy specimens with different crack widths and depths, and the results show that the relationship between the ΔB_xpp and the crack was consistent with the simulation results, and ΔB_xpp has a nonlinear positive correlation with crack width and a linear positive correlation with crack depth. The results of the present work show that IDC-MFM has high sensitivity in crack size detection and is a feasible solution for the quantitative evaluation of cracks in aluminum alloy.

Spiral concentrator is a kind of fluid-film gravity concentrator, and the flow pattern of the slurry film in the spiral chute is closely related to the geometry of the spiral chute, especially the cross-sectional shape. In order to explore the influence of the cross-section shape of the spiral concentrator on the sorting flow pattern, this paper utilizes the CFD technology to construct three kinds of flow film with the cross-section shapes of “ellipse”, “cubic parabola” and “composite curve”, which are the same as that of the spiral concentrator. This paper uses CFD technology to construct three kinds of CFD models of “ellipse”, “cubic parabola” and “composite curve”, and with the help of the constructed model, the sorting flow state of the three kinds of cross-section shapes of spiral concentrator was analyzed and studied comparatively. The results show that the thickness and Reynolds number of the flow film at the inner edge of the spiral chute with elliptical cross-section are larger, which is more suitable for the loose and stratified sorting of coarse-grained minerals. The spiral chute with composite curve cross-section has a larger laminar flow area and a wider secondary circulation area, which is conducive to the sorting and recovery of fine-grained minerals.

To investigate the relationship between metal magnetic memory testing (MMMT) signals and stress concentration factors (SCFs), four-level sinusoidal constant-amplitude load tension–tension fatigue tests were carried out on 45CrNiMoVA steel specimens with different SCFs. The normal component of MMMT signals, H_p(y), was collected during the fatigue tests, and three characteristics were extracted and analyzed during the tests, including the peak-to-peak value of abnormal peaks (ΔH_p(y)), the slope coefficient of the fitting line of H_p(y) (K₁), and the slope coefficient of the fitting line of H_p(y) between abnormal peaks (K₂), and a back-propagation (BP) neural network was developed to differentiate the SCF of the specimens. The results showed that both fatigue load and fatigue cycle number influenced MMMT signals, and the characteristics remained stable as the fatigue cycle number increased for the same fatigue load but increased significantly as fatigue load increased. In addition, all the characteristics increased as the distance between the scan line and the center line increased, but none of them could be used to differentiate the SCF of the specimens. With properly selected input vector and hidden nodes, the established BP neural network can quantitatively recognize the SCF of specimens.

Surface defect detection is a crucial method for controlling product quality, and convolutional neural networks(CNN) have shown great potential in this field. However, the researchers found that adversarial examples can cause the CNN to produce inaccurate results. It remains to be studied whether CNN can be applied in important industries. In order to improve the adversarial defense capability of neural networks, this paper proposes a neural network training framework based on cross-data space adversarial defense(CDSAD) for classified neural networks. The framework CDSAD trains a network that can adapt to multiple data distributions. During the inference process, the network employs multi-perspective evaluation methods to defend, warn and detect adversarial examples. The defensive capability of the CDSAD was tested against multiple adversarial attack algorithms using the CIFAR 10 and triode surface defect datasets. Experiment results demonstrate that the CDSAD defends against multiple adversarial attack algorithms without requiring any information about the opponent. The defensive effect is further improved when the framework acquires opponent information for adversarial identification module(AIM) training.