An electro-hydraulic drive system is essential for the stable operation of tunnel drilling rigs in underground coal mines. However, components such as pumps, valves, and controllers inevitably experience gradual degradation under long-term and high-load conditions. Conventional monitoring approaches often rely on labeled fault data or suffer from limited interpretability, restricting their applicability in real engineering environments. To overcome these limitations, this study proposes an unsupervised degradation trend analysis method that does not use labeled samples. A sliding-window strategy was adopted to extract key statistical features. Principal component analysis was then employed to construct a unified health index, and Z-score normalization enabled the interpretable detection of abnormal tendencies in individual features. Validation on real drilling data revealed clear degradation behaviors, such as main pump leakage and control current drift, demonstrating that the proposed method offered a lightweight and interpretable solution for trend-based condition monitoring and provided practical support for the intelligent maintenance of electro-hydraulic drive systems.

The operational performance of the coking process reflects its energy utilization and production efficiency, and its evaluation is a prerequisite for achieving optimal operation. However, the coking process faces several challenges, including multi-source data integration, reliance on manual performance evaluation, and low efficiency in model deployment. To address these issues, this paper proposes an intelligent evaluation method for coking process operational performance, based on a practical cloud-edge collaborative framework. First, an improved temporal fusion transformer is developed to accurately predict comprehensive production indicators. Then, based on the prediction results, a performance indices system is constructed, and a dual-scale fuzzy performance evaluation mechanism is introduced to assess the operational performance. Finally, a cloud-edge deployment architecture is established, where the cloud layer is responsible for model training and updating, and the edge layer enables accurate prediction and evaluation. The proposed method is validated on real industrial data, demonstrating its effectiveness and accuracy, and providing strong support for subsequent optimization and control.

Planar two-link rigid-flexible manipulators face significant control challenges when experiencing complete actuator failure, particularly in aerospace and precision industrial applications where traditional control methods prove inadequate. This paper presents a control strategy that combines infinite-dimensional distributed parameter modeling, motion optimization, and adaptive tracking control for a planar two-link rigid-flexible manipulator experiencing complete failure of its first joint actuator. The study establishes a partial differential equation (PDE) model using Hamiltonian principle and analyzes the dynamic coupling relationship between the active flexible link and passive rigid link at the distributed-parameter level. A genetic algorithm optimizes the trajectory of the active link to achieve simultaneous vibration suppression and indirect control of the passive link through dynamic coupling exploitation. Furthermore, the research develops a radial basis function neural network-based sliding mode controller that tracks the optimized trajectory under model uncertainties using only tip measurements rather than full-state feedback. The simulation results demonstrate the accuracy of the proposed dynamic coupling analysis and the effectiveness of the control strategy.

The lining temperature of a blast furnace is a key indicator of its health. Predicting the furnace lining temperature helps identify unreasonable operations and improves control of the thermal state of the blast furnace. However, due to the varying reaction times and the spatial distribution of the blast furnace ironmaking process, the influence of process parameters on the lining temperature exhibits dynamic time delay. To address this issue, we present a lining temperature prediction model that incorporates dynamic feature extraction across numerous time series, as well as an innovative attention mechanism that captures the dynamic features of time series. Finally, a convolutional neural network is employed to enhance the ability of the prediction model to perceive local temporal patterns. The originality lies in integrating gated recurrent units with attention mechanisms to address varying time delays among process variables. This integration results in improved prediction accuracy compared to existing methods. Based on the actual production data, we demonstrate the feasibility and effectiveness of the presented model.

The tunnel drilling rig is core equipment for underground coal mine drilling, and the electro-hydraulic circuit is one of its most critical subsystems. However, due to the harsh underground environment and the prolonged operation of drilling rigs under heavy loads, abnormalities such as hydraulic oil leakage frequently occur in the electro-hydraulic circuit, resulting in reduced operational efficiency and even severe accidents, including borehole wall collapse. To address this problem, in this paper, a mechanism-aided data-driven early warning method is proposed for the electro-hydraulic circuit of a tunnel drilling rig. First, the operating mechanism of the electro-hydraulic circuit is analyzed to investigate the leakage propagation pathway and related variables. Subsequently, three sets of features are constructed from the related variables to reflect the circuit state. Then, an early leakage warning model based on long short-term memory neural network is established, three sensitive indicators are developed, and on this basis, a monitoring statistic is constructed, and the corresponding threshold is estimated in real-time for early warning. Experiments based on actual drilling data demonstrate that the proposed method can reduce the false alarms and extend the warning lead time, and it is beneficial to improve the reliability and interpretability of early leakage warning for electro-hydraulic circuit.

Accurately predicting carbon efficiency is crucial for optimizing sintering operations in iron and steel production. This process is highly energy-intensive and has significant environmental and economic implications. However, the sintering process exhibits complex characteristics, including strong nonlinearity, time-varying dynamics, and operational uncertainties. These characteristics severely limit the performance of conventional modeling approaches. Existing data-driven methods often fail to capture temporal dependencies and handle uncertainty simultaneously, resulting in inaccurate and unreliable predictions. To address these challenges, this article introduces a novel type-2 fuzzy broad echo state learning system (T2FBESLS). This system integrates an interval type-2 fuzzy neural network into the feature layer of a broad learning system to improve uncertainty modeling. The enhancement nodes are replaced with echo state reservoirs that effectively encode the temporal dynamics of the sintering process. When evaluated using real-world industrial sintering data, the T2FBESLS achieved a reduction in root-mean-square error of at least about 15% compared to several state-of-the-art models. This demonstrates its superior prediction accuracy and stability. The key innovation lies in the fusion of type-2 fuzzy logic for handling uncertainty, broad learning for efficient structure expansion, and an echo state network for temporal modeling, offering a new approach to intelligent modeling in complex metallurgical processes.

Accurately and promptly detecting and diagnosing downhole faults is essential to ensure the safe and efficient operation of geological drilling processes. In the drilling circulation system, the incipient fault response signal is difficult to observe, while the kick (KK) and lost circulation (LC) faults share similar signal change patterns. Considering that faulty conditions can lead to significant deviations in the data distribution, this article presents a systematic fault detection and diagnosis method for the drilling circulation system based on slow feature analysis (SFA) and Riemannian metric (RM) clustering. The contributions are twofold: 1) a drilling fault detection index is proposed by calculating the RM between slow feature matrices and 2) a fault diagnosis scheme is proposed for the drilling circulation system based on RM clustering. Experimental results validate the effectiveness and practicality of the proposed method and demonstrate that the proposed method has superior performance to other methods.

Continuum robots (CRs) show great potential in complex environments due to their excellent deformability. For practical applications, the position control of the CRs is an important research field. The common drives of the CRs are rigid motors with mature control schemes. However, the driving force of the rigid motors is often impactive, which may cause safety concerns during interaction. Soft drives based on pneumatic soft actuators (PSAs) can provide compliant driving force for the CRs robot bodies to solve this problem, but it is necessary to comprehensively consider the control of the soft drives and the robot bodies. This article takes a CR with soft drives as the research objective, and proposes a double closed-loop adaptive position control method to achieve the endpoint position control of the CR. This CR includes a length-variable robot body with millimeter-scale diameter and pneumatic soft drives. The kinematic model of the robot body is built based on the piecewise constant curvature (PCC) method, and the static models of the soft drives are derived from the three-element model. Based on these models, we propose a double closed-loop adaptive position control method. The inner loop is used to control the displacements of the soft drives, and the outer loop combines the endpoint position control with the nonsingular fast terminal sliding mode function to adaptively control the endpoint position based on the inner loop. By Lyapunov method, we prove the convergence of the endpoint position error. The effectiveness of the proposed control method is verified through experiments.

A multimodal emotion recognition method based on a multi-scale vision Transformer and cross attention mechanism (MSFCA) is proposed. The proposed method integrates a multi-scale vision Transformer and a cross-attention mechanism to fully exploit the complementary information between facial expressions and speech modalities, thereby enabling more effective feature extraction and fusion. By incorporating L1 regularization, the sparsity and robustness of the fused features are enhanced, thereby improving the accuracy and generalization capability of multi-modal emotion recognition. Experimental results on the eNTERFACE’05 and RAVDESS multi-modal emotion recognition datasets demonstrate that the proposed MSFCA method achieves recognition accuracies of 86.16% and 87.14%, respectively, which satisfy the requirements for reliable multi-modal emotion recognition in practical applications.

As a closed-loop learning control method, repetitive control has been widely used in a variety of areas from appliances to aviation. A repetitive control system features perfect reference tracking and disturbance rejection in the steady state for periodic signals with a fixed period. This characteristic is important not only for conventional technologies and conventional industries but also for advanced technologies and emerging industries. This paper first explains the concept of repetitive control from its original idea. Next, it describes the structure of a repetitive controller as an internal model and shows the respective points of continuous- and discrete-time repetitive control. It presents a categorized list of practical applications of repetitive control. Moreover, two concrete applications, namely the control of a robotic manipulator and a rotating system, demonstrate the validity of the method with experimental results. Several current studies in this field are also reviewed, and some challenges and future studies for repetitive control are provided.

Flatness is an important quality indicator of steel plates in the roller quenching process. To obtain high-quality steel plates, an online flatness prediction model with high precision is urgently needed. However, during long-term online operation, the emergence of new type samples for model updating may lead to uneven activation, increased network complexity, and reduction of prediction accuracy. To address these issues, an accuracy-enhanced flatness prediction model based on the broad learning system is presented. First, combining the fuzzy comprehensive evaluation and broad learning system, the bending resistance of the steel plate is evaluated, and the flatness prediction model is established. Then, according to the distribution of node values and changes in prediction accuracy during online updates, the network nodes are adjusted dynamically. Finally, the comparative experiments are conducted with the actual data from a factory, and the designed model is applied to an industrial site. The results indicate that the model can maintain high and stable prediction accuracy during the online updating process, thereby achieving flatness prediction effectively. Therefore, the presented method provides a valid reference for the production of high-quality steel plates.

During drilling process, variations in formation hardness and frictional resistance often cause a mismatch between drilling operating parameters and actual conditions, reducing efficiency. A fuzzy decision-making strategy for drilling operating parameters is developed that enables feed speed and rotational speed to adapt to actual conditions. The formation hardness is accurately characterized using only drilling data through the integration of fuzzy C-means clustering method and defuzzification method. Meanwhile, the computational model for frictional resistance is established. They serve as inputs for subsequent decision-making models. The decision-making models for feed speed and rotational speed are developed based on the Mamdani fuzzy inference method, respectively. The effectiveness of the approach is demonstrated through an industrial case study based on actual drilling data.

Tuning of shape-performance coupling devices in many fields is essential but challenging due to their com plex characteristics. Manual tuning is inefficient and costly, and automatic tuning is urgently needed due to the growing demand. Although intelligent optimization-based tuning is effective, it faces two challenges: 1) high computational cost due to the large design space, and 2) significant complexity arising from the multi-solution and multi-modality characteristics. These challenges can be mitigated by focusing on a local yet effective design space. However, identifying such space is difficult. To tackle this, a novel optimization tuning approach driven by spatial multiple attentions (SMAs) is proposed. First, a data-driven space attention generation method is pre sented to select the space with a high likelihood of containing feasible solutions. Second, an SMAs decomposition strategy is devised to divide the whole space into several sub-spaces for reducing the optimization burden. Third, an SMAs-particle swarm optimization (SMAs-PSO) algorithm is designed to efficiently search for feasible solu tions in all spaces determined by SMAs. Simulation results demonstrate that the proposed method achieved a high tuning success rate while reducing the average iteration number by 53.4% compared to conventional PSO. Unlike existing intelligent optimization or hybrid methods that rely on global search, the proposed method identifies effective local spaces within the global space, thereby simultaneously enhancing accuracy and efficiency.

Drilling blast holes in open-pit coal mines is frequent, and the process data contain rich lithological information that provides a basis for bench blasting design. However, most existing studies rely on offline modeling using logging data and lack lithology identification models that can utilize drilling data to achieve parameter self-updating and adapt to changing drilling conditions. Therefore, in this paper, we propose a method for lithology identification with drilling based on time series feature learning. Interval constraints are introduced to optimize the distribution of Perceptually Important Points (PIP) in the drilling data sequence, the process of perceptually important point identification is transformed into a multi-objective optimization problem, and the strategy of priority gradient is adopted to optimize the solution. A segmented feature representation learning method is further designed, connecting adjacent perceptually important points (PIPs) to form subsequences of the original drilling data and extracting their temporal, numerical, and statistical features. This incremental learning approach enables rapid feature extraction based on drilling data collected in real time.Adaboost Extreme Learning Machine is used as the classifier, and recursive least squares method is introduced to update the model parameters. The proposed lithology recognition model can be integrated into the drilling system as a real-time control module, continuously sensing formation changes. This allows adaptive adjustment of drilling parameters, optimizing drilling efficiency and safety while effectively responding to complex strata variations. Therefore, based on the actual engineering needs of open-pit coal mines, this provides an important basis for the design of blasting parameters and the intelligent drilling of roller cone drilling rigs.

Geological hazards exhibit strong abruptness and severe destructive effects, frequently occur worldwide, and pose a serious threat to human life and property. Conventional methods for geological hazard monitoring are inadequate for fulfilling the requirements of accurate and real-time monitoring under complex geological conditions. This study provides an intelligent system for geological hazard monitoring based on multi-source information fusion. The characteristics of space-based, airborne, and ground-based multi-source information were analyzed. Multi-source information fusion methods at different scales were also discussed. Subsequently, intelligent algorithms for hazard identification, spatiotemporal prediction, and real-time warnings in the field of geological hazard monitoring were explored. Furthermore, practical application cases of the system were presented, laying a foundation for research on the application of artificial intelligence to geological hazards.

This article presents a control strategy that combines a disturbance observer-based adaptive integral sliding mode controller (DOB-AISMC) with a modified generalized repetitive controller (MGRC) to address the periodic reference tracking problem in nonlinear systems. The AISMC compensates only for the estimation error of the disturbance observer, not the unknown nonlinearity; therefore, the switching term of AISMC is reduced and chattering is effectively suppressed. The MGRC uses a dual-gain adjustment mechanism to adjust control and learning in repetitive control, thus enhancing the tracking accuracy for periodic signals. It also compensates for magnitude distortion and phase shift caused by the low-pass filter, significantly enhancing steady-state tracking performance. To improve gain tuning efficiency and increase flexibility in controller design, two error-independent performance indices are designed. These indices, derived through recursive iteration and the solution of matrix differential equations, provide direct insights into the influence of controller gains on system performance, eliminating the need for simulation or experimental error evaluation. Simulation and experimental results validate the superior control performance of the presented method.

Accurate prediction of rate of penetration is of great significance to improving the efficiency of deep drilling industrial process and reducing the cost of resource and energy exploration engineering. However, the deep drilling industrial process is accompanied by complex and variable formations, which make it difficult for offline models to capture the dynamic characteristics of rate of penetration, and the nonlinear characteristics of the drilling process also need to be considered. To solve the difficulties in rate of penetration modeling, this paper proposes a novel online modeling method, which includes two stages, offline model construction and online update of model. In the first stage, the wavelet denoising and amplitude limiting filter are adopted to eliminate noises and outliers, and then the support vector regression method and the improved snow ablation optimization algorithm are combined to construct an offline model of rate of penetration, which can solve nonlinear problem and non-convex optimization problem of model parameter selection. In the critical second stage, an improved sliding window method is developed to update the offline model by selecting data in the window to the training set, ensuring data diversity and information integrity while avoiding excessive similar data. Finally, the developed modeling method is verified by the actual drilling industrial site and the micro-drilling system. Experiment results demonstrate that the developed modeling method outperforms five modeling methods, including two online methods and three offline methods, highlighting its effectiveness in the online prediction of the rate of penetration. The developed modeling method has important application prospects for improving drilling efficiency and realizing intelligent control of the continuous drilling industrial process.

ObjectiveIn coal mines, harsh environments and complex geological conditions underground lead to frequent failures of tunnel drilling rigs. However, conventional techniques for failure diagnosis and early warning suffer from limitations in diagnostic accuracy, real-time performance, and interpretability. Therefore, these techniques are insufficient for both the rapid identification of the anomalous status of tunnel drilling rigs and the early warning of their potential failures. MethodsConsidering the complex underground environments in coal mines, this study developed an intelligent failure diagnosis and early warning system for tunnel drilling rigs in coal mines, placing a particular emphasis on key technologies for failure diagnosis and early warning. For failure diagnosis, the coupling relationships among variables were quantitatively characterized by combining the operating mechanisms of key circuits and cross-correlation function (CCF) analysis. Accordingly, the failure propagation patterns were revealed, and a diagnosis strategy based on node characteristics was established. For early warning of potential failures, the operational status of drilling rigs was quantified using time domain features such as mean and standard deviation. Then, a health index was constructed using principal component analysis (PCA). Finally, a hierarchical early warning mechanism was established in combination with Z-scores and the three-sigma rule. Based on these, an intelligent failure diagnosis and early warning system for tunnel drilling rigs was established through software development based on the Qt platform and the MySQL database.Results and ConclusionsUsing field tests at the Zhashui test base, the performance of the developed system was verified. The results indicate that the developed system achieved an accuracy of 96.8% and a recall of 90.5% for failure diagnosis. For early warning, this system yielded a failure detection rate of 94.6% and a false alarm rate of 1.5%, indicating a small number of false alarms. Compared to conventional passive protection technologies, the intelligent failure diagnosis and early warning system for tunnel drilling rigs developed in this study can accurately identify the operational status of tunnel drilling rigs and assess their health degrees in real time, thereby enabling timely early warning of potential faults. This study provides a theoretical basis and engineering solution for the stable operation and predictive maintenance of tunnel drilling rigs.

In coal mine tunnel drilling, the hydraulic rotary system of the drilling rig is often subjected to severe and rapidly varying load disturbances, posing significant challenges to system stability. To address this problem, this study develops a hydraulic rotary speed control system for tunnel drilling rigs based on the equivalent input disturbance (EID) method. First, a second-order dynamic model of the hydraulic rotary system is established, in which external load disturbances are uniformly modelled as a lumped disturbance. Then, an EID estimator is designed to provide real-time estimation of the total disturbance. To optimise controller performance, a hybrid bat algorithm is introduced to tune the adjustable parameters within the linear quadratic regulator and linear matrix inequality frameworks during gain synthesis. The effectiveness of the developed control system is evaluated through simulations, and the results demonstrate that the EID-based control system significantly enhances both transient response and steady-state performance under complex disturbance conditions. Furthermore, field application at a coal mining site validates the practicality and effectiveness of the developed control system.