Sihai Zhao

Displacement measurement in hydraulic support systems faces unique challenges from rock burst-induced vibrations that introduce non-smooth stochastic disturbances characterized by Wiener processes. This paper addresses measurement uncertainty in harsh underground environments through a fixed-time control framework. A second-order stochastic nonlinear model is established using Itô calculus to rigorously capture sensor-induced uncertainties. A fixed-time backstepping controller is developed, employing a self-organizing radial basis function neural network to approximate uncertain nonlinearities, with a novel power augmentation term introduced to eliminate singularities while preserving fixed-time convergence. Closed-loop probabilistic stability is proven via stochastic Lyapunov theory. Experimental validation is conducted on a Beckhoff-controlled hydraulic test bench, where Wiener noise is injected into sensor feedback signals to emulate underground stochastic disturbances. Results demonstrate tracking errors within ±2 mm and settling times independent of initial conditions, confirming the effectiveness of the proposed method under realistic measurement uncertainties.

As coal mine intelligentization advances, precise navigation for rapid excavation systems is critical for efficiency and safety. This study addresses several challenges, including GPS denial, a limited visual line-of-sight, and a lack of multiequipment collaboration in tunnels, by proposing an INS + Vision “mutual reference” collaborative positioning scheme. The method meets high-precision positioning requirements under complex working face conditions. Leveraging the alternating operation of the Bolter Miners (BMs) and the Bolt Transfer Units (BTUs), the scheme uses coordinate transformation and data fusion. When the BM moves while the BTU remains stationary, the BTU’s vision-inertial system acts as a local reference to update the BM’s pose in real-time. Conversely, when the BTU moves with the BM stationary, the BM’s fixed pose becomes the reference for reverse calculation of the BTU’s pose. This enables dynamic relative positioning between equipment, solving issues such as single-sensor error accumulation, visual occlusion, and long-range dependency. Experiments show that the BTU achieves maximum lateral/longitudinal errors of 34 mm/66 mm, whereas the BM achieves 28 mm/66 mm. These satisfy underground positioning requirements, confirming the feasibility and expected performance of the proposed integrated navigation method for rapid excavation systems.

Electro-hydraulic position servo systems are widely applied in aerospace, industrial automation, and robotics due to their high power density, rapid response, and precision. However, inherent nonlinearities, parametric uncertainties, and external disturbances pose significant challenges to accurate control. This paper proposes a robust extended Kalman filter (EKF)-based rolling horizon H-infinity control strategy for output feedback control of such systems. First, a state-space model of the electro-hydraulic position servo system is established, integrating unknown dynamics into a single disturbance term. Second, a robust EKF with adaptive structure is designed to dynamically adjust the covariance matrix by integrating motion model observations and sensor measurements, achieving accurate state and disturbance estimation with convergence proven through rigorous stability analysis. Then, based on locally linearized system models, the control problem is transformed into a linear matrix inequality (LMI) problem, reducing real-time computational complexity. Through comparative simulations and experiments, the proposed control method demonstrates superior performance over traditional PID, backstepping, and standard MPC controllers in position tracking accuracy, disturbance rejection capability, and control smoothness. Specifically, under measurement noise and external disturbances, the proposed method achieves smaller mean square tracking error (0.022) and maximum error (0.055) while maintaining acceptable computation time (27.3 ms). The research findings provide an effective solution for high-performance control of electro-hydraulic position servo systems and offer valuable reference for improving control performance of similar nonlinear uncertain systems.

During the high-yield continuous crushing operations in coal mines, the detection of feeding metallic foreign objects (MFOs) faces challenges such as complex noise interference and varying lift-off distances. This paper proposes a pulsed eddy current (PEC) data-driven interpretable neural network (PECDINet) based on optimized variational mode decomposition and improved wavelet threshold denoising for multi-feature classification and identification of MFOs. First, for industrial dataset collected at the crushing station, the variational mode decomposition is optimized using snow geese algorithm and envelope entropy function that incorporates environmental noise. Improved adaptive threshold function is then applied for wavelet threshold denoising of the intrinsic mode functions, effectively filtering out complex noise while preserving key characteristics of PEC signals corresponding to the time-domain solutions of PEC reflection fields in truncated regions of MFOs. Then, PECDINet with subtraction, square difference, and average layers is developed to extract start and end times, and time to peak of experimental signals, enabling characteristics identification of conductivity, thickness, and vertical width of MFOs. Finally, experiments demonstrate the proposed method’s advantages in accuracy and efficiency, addressing the challenges of multi-feature classification and identification of MFOs in coal mine field environments.

As the core component of digital hydraulic systems, the precise control of high-speed on/off valves (HSVs) largely depends on current sensor monitoring and feedback. However, sensors may fail due to aging, interference, and other factors, leading to degraded system performance. This article proposes a fault-tolerant control (FTC) method for HSV current sensors based on an unknown input sliding mode observer. First, a first-order low-pass filter is introduced to equivalently transform sensor faults into actuator faults, and coordinate transformation is employed to decouple faults and disturbances. Subsequently, an unknown input sliding mode state observer is designed to simultaneously achieve system state observation and fault disturbance reconstruction. Finally, a fault-tolerant controller is designed based on the reconstructed values. Simulation and experimental results demonstrate that the proposed method achieves precise control of HSVs in the presence of both sensor faults and external disturbances. Compared with traditional methods, the proposed approach effectively suppresses high-frequency chattering phenomena, conserves computational resources, and reduces position tracking error by 72.73% in hydraulic cylinder control applications. This research provides a new technical pathway for high-precision control of digital hydraulic systems.

The double‐tooth roll crusher, commonly driven by an induction motor, often faces sensor and actuator malfunctions. To address these, this study proposes a model predictive fault‐tolerant control framework to enhance system reliability and robustness. Initially, based on the system's mathematical model, a state and unknown input observer is developed. This observer effectively detects current sensor faults and isolates unknown disturbances, maintaining system functionality even with single‐sensor operation. Additionally, a velocity observer is designed using the estimated state data. Secondly, system faults are treated as uncertainties, leading to the development of a disturbance observer with fewer tunable parameters. This observer identifies actuator failures by analyzing disturbance values, with its convergence validated theoretically. This enables rapid fault detection and mitigation, reducing performance degradation. Finally, a model predictive torque controller is introduced to improve torque regulation. Experimental results demonstrate the method's strong fault‐tolerant capabilities and control effectiveness across various fault scenarios.

This study addresses the challenges of low accuracy and poor timeliness in feeding metallic foreign object detection during high-output continuous crushing operations in coal mines. Based on the PEC method and TREE method, the time-domain solution for coaxial excitation, receiving coils, and Hall sensors interacting with the truncated region reflection eddy current field of feeding metallic foreign objects is calculated. The effect of the lift-off effect on the characteristic classification and identification of feeding metallic foreign objects is analyzed in detail. The sensitivity of the lift-off distance to the peak value and time to peak in PEC signals is compared, and the relationship between the LOI coordinates and the characteristics of feeding metallic foreign objects is established. Simulation and experimental results show that the PEC signals from the metallic foreign object detection system, processed by the DWT-TLBO noise reduction algorithm, enable the classification and identification of the falling position, size, conductivity, and magnetic permeability of feeding metallic foreign objects based on the lift-off point, peak value, time to peak, and effective time range. Experimental results meet the practical requirements for metallic foreign object detection in coal mine sites.

The magnitude and management of temporary support forces in underground coal mine excavation tunnels are essential for ensuring maintaining tunnel safety and stability. We developed a coupling mechanical model of temporary support and roof, grounded in the working principles of temporary support in excavation tunnels, to ascertain the rated support force in the temporary support area. A three-dimensional numerical model was created utilizing the discrete element method to assess the efficacy of the rated support force. The results indicate that the maximum subsidence of the roof in the temporary support area is below 10.5 mm, with no yielding zone detected. We developed a mathematical model for the temporary support force control system and evaluated its performance. To overcome the limitations of conventional control strategies, we proposed two novel methods: (1) a proportional–integral–derivative (PID) controller utilizing the Newton–Raphson optimizer (NRO-PID) demonstrated a convergence speed 89.7% faster than the conventional particle swarm optimization (PSO)-PID during the initial stabilization phase, with an overshoot of less than 1.2%; and (2) an adaptive sliding mode controller (ASMC) effectively reduced the support force tracking error to below 2.5% under dynamic disturbances, in contrast to 8.3% for the standard PID. Simulation and experimental results indicate that the NRO-PID + ASMC support force control strategy markedly improves the effectiveness and safety of temporary support, thus facilitating efficient and secure tunneling operations.

In the open-pit coal mine operations involving semi-continuous or continuous mining, the coal crusher station plays a pivotal role. Its operational status directly impacts the continuity and stability of the entire production process. To achieve real-time monitoring for fault diagnosis and maintenance of the crusher station, this study constructs a fault repository for the crushing system, encompassing 13 typical vibration faults and two types of electrical signal faults. Based on the digital twin theory, a high-fidelity virtual model is established, accurately replicating the signal characteristics of the real system. The model is dynamically calibrated using differential evolution algorithms to adjust critical fault parameters in real-time, ensuring consistency between the virtual model and the physical entity. Furthermore, combining multi-head attention mechanisms with one-dimensional convolutional neural networks, the proposed approach extracts features from the virtual model’s output signals and performs fault classification diagnosis. Extensive experimental validations demonstrate that the proposed scheme offers significant advantages in terms of diagnostic accuracy and real-time performance (accuracy exceeding 97%, adjustment time less than 1 s), providing a reliable technical support for the intelligent maintenance of open-pit coal mine crushing systems.