Cyber physical systems (CPS), in the event of a cyber attack, can have a serious impact on the operating physical equipment. In order to improve the attack detection capability of CPS, an support vector machine (SVM) attacks detection model based on particle swarm optimization (PSO) is proposed. First, the box plot anomaly detection method is used to detect the characteristic variables, and the characteristic variables with abnormal distribution are discretized. Secondly, the number of attack samples was increased by the SMOTE method to solve the problem of data imbalance, and the linear combination of characteristic variables was performed on the high-dimensional CPS network traffic data using principal component analysis (PCA). Then, the penalty coefficient and the hyperparameter of the kernel function in the SVM model are optimized by the PSO algorithm. Finally, Experiments on attack detection of CPS network traffic data show that the proposed model can detect different types of attack data and has higher detection accuracy compared with general detection models.

The particle size of pellets is an important parameter in steel big data, and the high density and high overlap rate of pellets bring a great challenge to particle size detection. To address this problem, a particle size intelligent detection algorithm with an improved watershed and a Gaussian mixture model (GMM) is proposed. First, the initial segmentation of the pellets and background is achieved by using adaptive binary segmentation, and then the secondary fine segmentation of the pellets and background is achieved by combining morphological operations such as skeleton extraction and marked watershed segmentation; then, the contour of the connected domain of pellets is calculated, and the non-overlapping pellets in the foreground and the overlapping pellets are filtered according to the roundness of their contours. Finally, the number of overlapping pellets is predicted by Gaussian reconstruction of the grayscale image of the overlapping pellets, and the number and granularity of the overlapping pellets are predicted by the Gaussian reconstruction of the overlapping pellets. The experimental results showed that the algorithm achieved a 91.98% segmentation accuracy in the experimental images. Compared with other algorithms, the algorithm can also effectively suppress the over-segmentation and under-segmentation problems, and it can effectively realize the pellet size detection of dense, overlapping pellets such as those on a pelletizing disk, which provides an effective technical means for the metallurgical performance analysis of pellet ore and intelligent pellet-making driven by big data.

With the rapid development of AI (artificial intelligence) in recent years, AI has been widely used in the image processing of iron tailings. The main component of iron tailings is silica, which is the most difficult part of the iron tailing composition to fuse. Therefore, the melting behavior of iron tailings can be characterized by the melting behavior of silica. First of all, in the actual production process, the temperature of the high-temperature furnace exceeds 1500 °C, which leads to a short service life of the conventional testing equipment. A line array CCD (charge coupled device) camera system with amplification effect was used to acquire data on silica in a high-temperature melt pool in a non-contact manner. The corundum crucible position is fixed, which can be used to establish a two-dimensional coordinate system to reproduce the object’s motion pattern to solve the problem of the short service life of conventional inspection equipment. The color and wobble problems generated by the Rift Valley CCD system when taking pictures need to be corrected for accurate image processing. Secondly, the change in the center-of-mass position of silica during melting was studied using the temporal images of silica in a high-temperature melt pool, and the edge contour features were extracted from the silica images and filled inside the contours, from which the mass, area and perimeter were obtained. Finally, the volume change of silica in the time series image is calculated after eliminating the effect of background factors, and the 2D image is constructed into a 3D image. We built a SFS (shape from shading) model according to the lighting conditions, and the actual melting rate of silica is calculated accordingly. According to the model calculation results, it can be seen that with the change in time, silica surface area under heat gradually decreases, then the silica melting rate gradually decreases; among them, the faster melting rate is about 0.007 cubic millimeter/second, and the slower melting rate is about 0.0015 cubic millimeter/second. Data support was provided for the blast furnace slag direct fiber formation process. The introduction of the algorithmic model into the actual production process has certain advantages. Therefore, the SFS model established in this paper has some practical value and is worth promoting in related enterprises.

To detect the state characteristics of abnormally large pellets in the disc pelletizer in real time, this paper proposes an abnormally large pellet target detection visual network based on improved YOLOv11 and a dedicated motion trajectory tracking algorithm for dynamic state detection. The original SPPF module in the backbone feature extraction network is improved by parallel integrating AvgPooling and MaxPooling layers, forming the Multi-SPPF module that balances global and local key information. The CBAM module is redesigned into a lightweight CSPCBAM module based on the CSP(Cross Stage Partial) structure, embedded in the front end of the feature fusion network to inject spatial-channel dual-dimensional attention mechanisms into multi-scale feature layers beforehand. A Det-CKF (Detection-dominated Kalman Filter) algorithm and a Dual Cost Matrix target matching strategy are designed, overcoming traditional Kalman Filter’s drawbacks of numerous hyperparameters and poor nonlinear scenario adaptability. Experimental results show that compared with mainstream target detection algorithms (SSD, Centernet, YOLOv3, DETR, Faster R-CNN) and original YOLOv11, the proposed network performs excellently, with mAP50 at 0.949, mAP50:95 at 0.834, and FPS up to 103. Grad-CAM heatmap experiments confirm its superior detection stability over the unimproved YOLOv11. Compared with Kalman Filter-based ByteTrack, the proposed tracking model achieves more accurate and longer-sequence trajectory tracking, with 83.99% average prediction-detection box matching rate and 8.62 average continuous tracking steps. Overall, it can real-timely detect the number, size, speed and trajectory changes of all abnormally large pellets, providing strong technical support for precise disc pelletizing regulation and promising prospects for promoting pelletizing technology intelligence.