To address the no-uniform distribution of alumina concentration in large-scale aluminum reduction cells, a feeding strategy based on adaptive distributed subspace prediction control is proposed. Firstly, the aluminum reduction cell is divided into multiple interconnected subsystems according to the spatial distribution of feeding ports, where adaptive distributed subspace predictive models for the alumina concentration of each subsystem are directly constructed from online data. Secondly, a residual-driven model switching mechanism is designed to trigger parameter updates only when model deviations exceed thresholds, thereby balancing computational efficiency and model accuracy. Finally, a distributed predictive controller based on Nash optimality theory is developed to achieve global coordinated feeding optimization. Simulation tests using real production data from an aluminum plant demonstrated that the proposed method significantly improved the spatiotemporal uniformity of alumina concentration in large-scale aluminum reduction cells.

An online disturbance feedback compensation structure is proposed to suppress the adverse effects of uncertainties such as sensor noise disturbance, external disturbance and model parameter perturbation on the inverter output voltage. Firstly, the state space model and residual generator model of the inverter under the influence of model parameter perturbation are established, and the influence of the disturbance and parameter perturbation problems on the system is analyzed based on the output from the residual generator. Secondly, considering that the sensor noise disturbance during operation will affect the observation accuracy, the Kalman filter is introduced to improve the observation effect. Next, for the external disturbance, compensation based on the model matching principle is designed. Then, the compensation controller Q(z) based on the model matching principle is designed for external disturbances, and the compensation controller is optimized online using a gradient descent algorithm to enhance the suppression capability of the compensation controller for parameter perturbation. Finally, the effectiveness of the method is verified using a PXI-based semi-physical simulation experiment platform.

To improve the efficiency of wet flue gas desulfurization in thermal power plants, an extended fictitious reference iterative tuning algorithm based on minimum variance index is designed to optimize the control performance for the pH control process of desulfurization slurry. Firstly, the input and output data of pH control process are used to calculate the control performance index of the system, and it decides whether to optimize the control parameters or not after comparing with specified performance index threshold. Secondly, when parameter optimization is needed, let the actual system output index track the reference model output index through data-driven technology, and the optimal control parameters are obtained by minimizing the objective function. Finally, MATLAB simulation analysis is carried out on the pH control system of desulfurization slurry. The simulation result shows that the algorithm can ensure stable control of pH value, and effectively suppresses random interference and improves the control performance of this system.

Aiming at the problem of positioning and anti-swing control of bridge crane system, a parallel control method based on data driven model free adaptive control and nonlinear PID control with tracking differentiator (TD) is proposed. The dynamic linearized data model of the bridge crane system is given based on the displacement output and the control input signals of the system. A model-free adaptive control (MFAC) method is designed on the base of the virtual model. Considering that the model-free adaptive control method is sensitive to system disturbance data and requires high data accuracy, the dual TD nonlinear PID control is paralleled with the model-free adaptive control and the system stability and error convergence are proved. The simulation results show that this compound control method not only realizes the accurate positioning of the trolley and the suppression of the load swing angle, but also reduces the influence of the disturbance on the system and improves the robustness of the system.

A split-range control strategy based on feedforward decoupling to address the problem of large online computation complexity in the muti-variable predictive control algorithm is proposed for suspension density in dense medium separation process. Firstly, the operation valve is selected based on the error between suspension density and the setpoint to reduce the number of inputs in the predictive model. Secondly, the inputs having large influence of the controlled variables are retained while unimportant inputs can be regarded as measurable disturbances, and a feedforward compensation mechanism can be introduced to decouple the multivariate control model. Then, the multivariable predictive model in standard predictive control is simplified into a combination of multiple univariate predictive models, based on which a decentralized rolling optimization policy is designed. Finally, MATLAB simulation shows that the control policy can reduce computational complexity, shorten calculation time, and thereby meet the requirements of rapid response in industrial sites.

In order to formulate reasonable carbon emission reduction strategies, it is necessary to predict such emissions accurately. Currently, there are few studies on carbon emissions from power plants, and traditional predictive models require excessive training time to attain a prescribed accuracy. Therefore, a component- augmented-input model is proposed, denoted WPD-ISSA-CA-CNN, for carbon-emission prediction. The model implements a “decomposition-augmented fusion” prediction framework. First, a wavelet packet decomposition (WPD) is used to decompose the signal into frequency-based sub-sequences. And these components are augmented as model inputs to reduce the model training time. Second, to address the challenge of hyper-parameter selection, an improved multi-strategy fusion sparrow search algorithm (ISSA) is employed to optimize the parameters of convolutional neural networks (CNNs). Using historical data from 2×25 MW boilers at a Shanxi power plant, the proposed model was compared with BP, LSTM, CNN, and their hybrid models using five evaluation indicators. The results show that the proposed hybrid model has a higher accuracy in predicting carbon emissions from thermal power generation while markedly reducing training time.

For the attitude control and target navigation problems of quadrotor, a control and target navigation algorithm based on offline reinforcement learning is proposed. Firstly, the TD3BC (twin delayed deep deterministic behavioral cloning) algorithm is improved. The Q value of the action sampled by the variational autoencoder is taken as the benchmark, and only the actions with higher Q values in the dataset are added as constraints to the loss function of the policy network, reducing the constraint strength of the behavioral cloning term and avoiding the deviation of the optimal action caused by training with low-quality data. Then, a simulation environment for quadrotor is constructed based on the Mujoco simulation platform to train the Q network and policy network under offline conditions, and the effects of three reinforcement learning algorithms in attitude control and target navigation tasks are compared. The simulation results show that the proposed algorithm can effectively control the attitude of the aircraft, and reach the navigation target with a shorter flight trajectory, and has better fault tolerance.

In response to the challenges of applying multi-source heterogeneous knowledge in robot development scenarios, a knowledge graph construction method based on large models is proposed. By employing long-text processing algorithms, the completeness and accuracy of complex long-text knowledge extraction are enhanced. A comprehensive similarity-based knowledge fusion method is used to fully leverage the structural correlation information in texts. Moreover, knowledge graph technology is applied to the field of robot development, enabling effective integration of multi-modal knowledge. The quality of the knowledge graph is improved significantly, endowing it with the capability for generalized understanding and reasoning of robot-specific knowledge. Firstly, the theoretical basis and explains the technical background has been introduced. Then, the overall construction process of knowledge graph has been proposed, analyzing the key technologies of knowledge extraction and knowledge fusion. Finally, experimental results are validated the accuracy of knowledge extraction, the efficiency of rapid multi-hop queries, and the correctness and interpretability of knowledge reasoning performed

To solve the problem of low diagnosis efficiency caused by noise interference and changing working conditions in bearing operation, a dual flow bearing fault diagnosis model based on improved depth residual shrinkage network and adaptive combination of multi countermeasure domains is proposed. The variable soft threshold function is embedded into the residual block, and the sub network is used to set the threshold adaptively. The improved residual shrinkage network is established as the backbone structure of the diagnosis model. The backbone network is introduced into the DropBlock layer to reduce redundancy and complex coadaptation between neurons, and enhance the orthogonality between features of the convolution layer. The maximum mean difference of multiple cores is used to measure the distribution between domains and classes, and a dynamic adjustment factor is established. Combined with a multi domain discriminator, the importance of marginal distribution and conditional distribution is adjusted. Finally, the migration experiment under noise environment is carried out on two types of bearing datasets, and the results show that the propose

The amount of data information space has grown vigorously. Intelligent information processing technology is driven by scenarios and data. With the improvement of information processing ability, it has achieved successful applications in multiple fields. We take information security as the research background and sorts out and analyses the current research hotspots and main achievements according to the types of data modalities. Then, the important role and further influence of intelligent information processing technology in the field of network and social security are analyzed. An important reference for the further research and application of this technology in the field of information security in the future is provided.

For the consensus control problem of second-order multi-agent systems with external disturbances in undirected topology, a distributed fixed-time control protocol based on sliding mode technology is proposed. Firstly, a second-order distributed fixed-time observer is designed to realize the accurate estimation of the leader’s state information in fixed time for each follower. Then, based on a novel distributed fixed-time sliding surface designed with system state errors, a distributed fixed-time consensus control protocol considering external disturbances is proposed. The control protocol not only effectively improves the convergence rate of the system, but also ensure that the system state tracking error reaches zero within the fixed time. At the same time, the stability of the closed-loop system is proved by Lyapunov theory. Besides, the upper bound of the stability time can be clearly estimated according to the control protocol parameters within unknown initial state of the system. Finally, the correctness and validity of the theoretical analysis results are verified by numerical simulation.

In order to improve the accuracy and stability of the navigation on unmanned ship, a track fusion filtering algorithm is applied, and the adaptive filtering factor and fading factor are introduced to improve the estimation accuracy and overcome the dependence of the filtering algorithm on the statistical characteristics of noise. In the process of filtering, the statistical characteristics of noise are estimated dynamically and the divergence is judged and suppressed. Simulation experiments are conducted on the established random movement state model of the unmanned ship. The simulation results show that compared with the standard Kalman filter algorithm and the traditional track fusion filter algorithm, the proposed algorithm has better accuracy and stability, and still has better track estimation performance in the case of unknown statistical noise.

A predictive sliding mode control method based on Kalman filter is proposed to improve control performance of predictive sliding mode control method for systems with model mismatch, random interference and control input saturation. The unmatched information expressed in the form of a parametric covariance matrix by treating the model mismatch as interference is regarded as the process noise of system together with the external white noise interference. The sensor measurement error is taken as a part of measurement noise of the system. Kalman filter is used to estimate the disturbed state variables and applied to controller design. Finally, the experimental results of the inverted pendulum demonstrate the effectiveness of the proposed method.

External torque estimation of cooperative robots is one of the key problems in human-computer cooperation. Therefore, a scheme of external torque estimation based on low-speed dynamic model is proposed. First of all, most cooperative robots operate at a slow and uniform speed. For the complex problem of parameter identification of the complete dynamic model, the dynamic model is simplified. Then, the joint friction torque and heavy torque of the cooperative robot are analyzed. The least square method is used to identify the friction and gravity of the robot, and the empirical friction model is used to fit the friction characteristics of the robot joint. Finally, the joint trajectory is designed on the experimental platform of the Xarm6 robot to verify the accuracy of friction and gravity identification. The experimental results show that the fitting effect of the robot shutdown friction characteristics using empirical models is ideal. The proposed scheme can quickly identify the parameters of the dynamic model and effectively estimate the external torque of the cooperative robot, which has certain reference significance for efficient cooperation and safety problems in the process of human-computer cooperation.

A practical prescribed time adaptive control algorithm is proposed for the dynamics of excavators. Firstly, based on the design of a time-varying constraining and compound function, the algorithm ensures that the tracking errors are limited to pre-set bounds in the prescribed time and removes the initial condition- dependent restriction, which is more general and flexible for actual constructions. In addition, an adaptive method is used to complete the robustness against external disturbances while the bucket tip is operating. Finally, the Lyapunov-based method proves the stability of the system and simulations verify the effectiveness of the proposed control algorithm.

For the trajectory tracking problem of two-joint mechanical leg, an adaptive backstepping sliding mode control scheme based on dynamic model simplification is proposed, which fully considered the unknown external disturbance and modeling errors. Firstly, a dynamic model of the double-joint mechanical leg is established and divided into two interrelated subsystems. Then, a decentralized adaptive control scheme is designed for the interaction between the joints, external interference and modeling errors. The controller of all subsystems is regarded as a whole control strategy, which can inhibit the interaction between subsystems. The proposed control scheme can accurately compensate the unknown effects of uncertainty, rather than treating most uncertainties as interference directly. Finally, in order to verify the effectiveness of the designed controller on the time-varying nonlinear system, a double joint mechanical leg platform driven by pneumatic muscles is simulated and verified in MATLAB/Simulink. The results show that the proposed control scheme is effective.

Process monitoring based on multivariate statistical analysis requires large amounts of historical data for modeling. The presence of outliers in the data can affect the accuracy of the modeling, thus reducing process monitoring performance, even increasing the false alarm rate and the missing alarm rate. A robust process monitoring method based on artificial neural network (ANN) and canonical correlation analysis (CCA) are proposed to deal with the outlier issue in the data. The method is divided into two phases: offline training and online monitoring. In the offline training phase, the expanded Kurtosis is chosen as the loss function of ANN according to the principle that the outliers obey the sub-Gaussian distribution to eliminate the influence of outliers; in the online monitoring phase, outliers are first rejected based on the difference in probability of consecutive occurrence of outliers and faults, and then process monitoring is performed using CCA. To verify the effectiveness of the proposed method, simulation tests are performed on a numerical example and continuous stirred-tank reactor. The results show that the proposed method can effectively eliminate the effect of outliers, improve the fault detection rate, and reduce the false alarm rate.

For the inconvenient placement of traditional motor temperature measurement sensors, taking the Prius motor as an example, a twin model with motor speed as the input and maximum temperature as the output is established through electromagnetic-steady-state thermal-stress deformation joint finite element simulation. Firstly, the total loss of the motor is calculated by electromagnetic simulation to be about 1,408.45 W. Secondly, the loss is imported into the temperature field to calculate the temperature distribution , and the maximum temperature is located in the winding at 74.855 ℃ and the minimum temperature is located in the outside of the stator at 69.406 ℃, which is in the normal operating range. The cooling performance of the motor is analyzed by changing the convection coefficient of the outer side of the motor. Then, the steady state thermal results are imported into the structural analysis, and the maximum deformation is calculated to be 0.062 924 mm, which meets the requirement of temperature rise of the motor on the deformation. Finally, the thermal circuit of the motor winding is plotted, and by comparing the results of the finite element simulation, the maximum error of the temperature is 7.001 ℃, which verifies the relative accuracy of the finite element method.

Intelligent underwater robots are prone to malfunctions, which can affect underwater operations. Propose a fault-tolerant control method based on adaptive reinforcement learning. This method introduces the Actor-Critic algorithm, which learns and formulates action strategies through the Actor network, while the Critic network evaluates the value of actions and adaptively adjusts strategies based on external environmental changes. Meanwhile, an improved extended state observer based on integral mechanism was designed, and an anti integral saturation algorithm was adopted to avoid integral saturation. The simulation results show that when the thruster of the intelligent underwater robot fails, the error values of the proposed fault-tolerant controller in the x-axis and y-axis directions gradually approach 0 after 15 seconds, proving that the designed fault-tolerant controller has excellent fault-tolerant performance and stability, and can provide effective technical support for the safe operation of underwater intelligent robots.

In order to accurately reflect the instantaneous driving characteristics of energy storage tram, a multi-dimensional Markov chain based condition construction method is proposed. Firstly, the data of a modern tram line is used to form the sample database, and the principal component analysis is used to reduce the dimension of the characteristic parameters, and the principal component score matrix is used for clustering analysis to divide the driving characteristics. Secondly, the traditional Markov chain is extended to the multi-dimensional state space, and the three-dimensional state transition probability matrix of train speed, average acceleration and road slope is constructed. The cumulative Poisson distribution function and Monte Carlo sampling method are used to generate random numbers to determine the next time state, so as to obtain the real-time condition. Finally, a comparative analysis of the different dimensions of the construction conditions. The results show that the deviation of each characteristic parameter from the average value of the sample and the transient error of the running speed are the lowest in the three-dimensional state space. The speed and power history have the strongest spatiotemporal correlation and robustness to the operation scenario. Compared with the traditional one-dimensional method and the improved two-dimensional method, the driving range of tram increased by 93.5% and 23.9% respectively. The effectiveness and superiority of improving the dimension of Markov chain to deduce the actual working conditions of trams are verified.

For the regional target search problem with partial prior information, a probability-driven multi-UAV collaborative regional target search method is proposed. Firstly, a mathematical description of the prior information in the task area is formed, and based on this description, a probability map model is constructed. Voronoi diagrams are utilized to segment the probability map, and an assignment index matrix is established. Then, the Kuhn-Munkres algorithm is applied to solve the flight direction vectors for each UAV. Subsequently, numerical simulation experiments are conducted, and the results validate that the designed coordination search strategy effectively enhances the search efficiency for regional targets. Finally, physical flight tests are performed using three quadrotor UAVs, demonstrating the feasibility of the proposed collaborative search method in engineering applications.

Barrel finishing process is a surface integrity processing technology to improve surface quality and usage performance of parts in advanced manufacturing technology. The abrasive media, the finishing equipment and the grinding fluid are the core process elements affecting the processing effect and efficiency in the barrel finishing process. Currently, the expert reasoning method can be used to achieve the decision of the abrasive media, but its accuracy is not good when it is applied to the decision of the finishing equipment and the grinding fluid, and the process elements cannot be decided when the information of the new problem is incomplete. Therefore, a decision model of process elements of barrel finishing based on subjective and objective empowerment expert reasoning (SOE-ER) is proposed. Firstly, the construction process of hierarchical classification rules is elaborated in detail. Secondly, subjective and objective empowerment in the expert reasoning is introduced. Finally, a decision system for barrel finishing process elements based on SOE-ER model is established, and experimental research is conducted. The results show that the SOE-ER model has a high decision accuracy for three core process elements, which can provide reasonable and effective guidance for the process elements decision of the new problem.