The steam specific enthalpy and dryness of steam turbines in small-scale cogeneration units are important indicators for evaluating turbine safety and economy. However, they are usually not measurable online. An approach for online calculation of these parameters is proposed. First, the superheated state of the steam at the outlet of the steam turbine stage is determined. For the superheated steam, the specific enthalpy of the stage outlet steam is calculated by solving the isentropic enthalpy drop and internal efficiency of the stage. For the saturated steam, the specific enthalpy and dryness of the stage outlet steam are calculated by means of a comprehensive calculation model incorporating the isentropic enthalpy drop, internal efficiency and saturated steam specific enthalpy. Verification results for a 15 MW cogeneration steam turbine demonstrate that the proposed approach is feasible for online application and achieves high precision.
As an underactuated crane, the tower crane is prone to load swinging during operation. Addressing the issue that previous studies mainly focused on eliminating swing angles while paying little attention to the vibrations of the wire rope itself, a method is proposed based on the sliding mode anti-swing approach, which considers the vibration of the wire rope and drives the travel motor for vibration reduction. Firstly, a vibration model for the steel wire rope is established using the vibration theory of beam systems. Secondly, the transfer function between the steel wire rope and the travel motor is derived, thereby equating the angular velocity of the steel wire rope to a nonlinear vibration torque on the motor. Thirdly, an easily implementable anti-sway control method is employed, enabling the travel motor to drive an anti-sway torque. Additionally, the travel motor simultaneously drives an extra torque equal in magnitude but opposite in direction to the vibration torque, thereby achieving vibration reduction. Finally, simulations and physical experiments demonstrate that the proposed method maintains effective control performance even when accounting for wire rope vibration.
The automatic phase partition method of online fault monitoring depends on the prior knowledge of the process, which will affect the monitoring results. By introducing density-sensitive distance measure and considering time-series information, the affine propagation clustering algorithm is used to realize the automatic and ordered partition of the transition phase and stable phase in the batch process, and then the support vector data description algorithm is used to establish the fault monitoring model. By taking the performance index of batch process monitoring and the stability index of the clustering algorithm as the optimization goal, the proposed phase partition method uses the fruit fly optimization algorithm to achieve the optimal time period acquisition through iterative optimization. The method is applied to the simulation of batch fermentation process of penicillin feeding, and compared with step-wise sequential phase partition (SSPP) and iterative two-step sequential phase partition (ITSPP), the results show that the proposed method has high fault detection accuracy.
Temperature system of distillation column has the characteristics of strong nonlinear, multi-variable strong coupling, large time delay and time varying, etc. The use of a traditional PID control can lead to overshoot and system oscillation, making it difficult to achieve an optimal control effect. Therefore, an inverse decoupling linear active disturbance rejection control (ID-LADRC) strategy is proposed. Firstly, an inverse decoupler is used to eliminate the coupling between variables, and the linear active disturbance rejection control (LADRC) strategy is adopted to conduct research on the decoupled system. Secondly, four control strategies of traditional PID, ID-PID, LADRC and ID-LADRC were introduced to simulate and analyze the nonlinear coupled mathematical model of temperature system of distillation column. Then, for the problem that manual parameter adjustment of ID-LADRC controller is difficult to obtain the optimal effect, an improved sparrow search algorithm (ISSA) is proposed for parameter optimization. The simulation results show that the ID-LADRC strategy has better dynamic performance, anti-interference ability and robustness. Moreover, after optimization with ISSA, the system features a faster response speed and a shorter adjustment time, which further improves the control performance of the system.
For the problems of steam-water heat exchanger network control system with unknown model information and limited bandwidth, an event-triggered data-driven control algorithm is proposed. Firstly, the algorithm effectively reduces the communication network transmission burden through an event-triggered mechanism, initiating data transmission only when triggering conditions are met, which significantly conserves network resources. Secondly, without requiring prior model information of the system, this method ensures that the tracking error of the closed-loop system is ultimately uniformly bounded. Furthermore, based on a data-driven design approach, the system is transformed into an input-output mapping form, allowing for the simultaneous construction of the controller and the event-triggering conditions, ultimately realizing model-free adaptive control. Different from the traditional data-driven control methods, the presented algorithm is low computation burden and effectively saves communication resources. The effectiveness of the proposed algorithm is verified by MATLAB simulation.
In order to improve the effectiveness of satellite swarm in application, the problem where satellite swarm must perform observation tasks is addressed while engaging in orbital gaming, and proposes a task‑coordinated orbital gaming method. Firstly, by analyzing the reconnaissance gain, jamming gain and loss cost of the satellite swarm on both sides, the orbit game model of the satellite group confrontation of both sides is established by using the idea of the noncooperative game. Secondly, based on three indicator functions, a payoff matrix for the non‑cooperative dual satellite clusters is established. The bimatrix game problem is then solved to obtain its Nash equilibrium, based on which the optimal orbital configuration distribution for both clusters under equilibrium conditions is determined. At this point, neither of the non-cooperative dual-satellite clusters can gain a greater advantage by altering the orbital position of their own satellite cluster. Finally, simulation experiments show that the orbit placement strategy selected by using the game idea can improve the mission execution capability and the application performance of the satellite swarm compared with the general strategy.
To prevent fractures, stalls, wear, pitting, plastic deformation, and adhesive failure, a gearbox high-speed shaft failure prediction method that utilizes genetic algorithms (GA) to optimize the parameters of a back propagation (BP) neural network model is proposed. Firstly, Pearson and Spearman correlation coefficients are applied to analyze multi-source data, selecting correlated variables such as rotational speed, active power, wind speed, and torque. Secondly, a BP neural network is employed to establish a vibration prediction model for the high-speed shaft. The correlated variable data serve as input to the BP neural network, with GA used for optimization to determine weights and thresholds. Results indicate that the GA+BP neural network algorithm achieves higher vibration prediction accuracy than the traditional BP algorithm. Moreover, compared to the other two algorithms, the GA+BP neural network demonstrates superior predictive precision, verifies the feasibility and accuracy of the proposed algorithm, and provides a theoretical basis for fault prediction of the high-speed shaft in wind turbine gearboxes.
To address the problem of difficult PV power prediction caused by the strong randomness of PV power, a PV power combination prediction method based on optimal multivariate variational mode decomposition (OMVMD) and multi-strategy improved Harris hawk optimization (MHHO) optimized the deep extreme learning machine (DELM) for PV power combination prediction method is proposed, referred to as the POMD model. Firstly, the meteorological features with large contribution values to the original power are determined by feature selection, and the alignment entropy is used as the fitness function to solve the optimal parameters of the MVMD algorithm using the MHHO. Then, the important features and the actual power are decomposed simultaneously using the OMVMD algorithm to improve the multi-channel data fusion processing capability to obtain several subsequences. Finally, the MHHO algorithm is used to obtain the optimal weights and biases of the input layer of the DELM network, build a PV power prediction model, and use the feature components to predict the power components to achieve the goal of homogeneous frequency smooth prediction. The experimental results show that the POMD model has higher prediction accuracy and better fitting effect than other combined methods under the three weather conditions.
Real-time anomaly detection of nuclear reactor core temperature is crucial for ensuring the safe operation of nuclear power plants. Addressing the issues of high missed detection and false alarm rates caused by feature redundancy in deep learning methods, specifically autoencoders (AE), an incrementally feature disentangled autoencoder (IFDAE) for core temperature anomaly detection is proposed. Firstly, the concept of feature disentanglement is integrated into the AE framework, and a latent space disentanglement loss is defined to enhance the independence constraints of the learned features. Secondly, an incremental feature generation strategy is developed alongside a neuron-incremental network architecture. Through iterative training, the feature dimensions of the latent space are adaptively determined, effectively resolving the insufficient disentanglement problem inherent in traditional AEs. Finally, two monitoring statistics are constructed to characterize the feature space and the residual space of the core temperature data, respectively. Experimental results on real-world nuclear reactor core temperature data demonstrate the effectiveness of the proposed method.
Key words
To improve the accuracy of coal and gas outburst prediction, a method for predicting coal and gas outbursts using the kernel extreme learning machine (KELM) optimized by an improved spider wasp optimizer (ISWO) algorithm is proposed. Firstly, a multi-strategy fusion method is used to improve the spider wasp optimizer algorithm and verify the algorithm performance through simulation. The result showed that the improved algorithm accelerates convergence. Then, the ISWO algorithm is used to optimize the parameters of KELM. Finally, simulation experiments are conducted to verify the prediction ability of the ISWO-KELM model. The experimental results show that
In order to improve the stability of high-speed train operation and the accuracy of tracking control, and ensure the train can effectively overcome the influence of unknown disturbances during operation, a neural network backstepping sliding mode control method based on disturbance observer (DOB) is proposed. Firstly, to address the problem that the accuracy of the single point model is insufficient and the actual operating resistance variation is difficult to be accurately calculated by the empirical formula, mass point model for high-speed train with non-parametric resistance is established and the resistance part is approximated by a Radial Basis Function (RBF) neural network; Secondly, based on the train dynamics model, the backstepping control is combined with the sliding mode control and a DOB is designed to compensate the disturbance of the control output for the unknown external disturbance and the jitter problem of the sliding mode control; Finally, the results of comparative experiments demonstrate the superiority of the designed control method in terms of its performance in dithering suppression, tracking error reduction and robustness improvement.
An adaptive composite decoupling sliding mode model-free control strategy combining a switching differentiator (SD) type disturbance observer is proposed for the problem that the 3D overhead crane has the inability to obtain an accurate mathematical model and is affected by external disturbances. The strategy consists of three controllers, namely, a PD controller, an adaptive composite decoupling sliding mode controller and a SD type disturbance observer. Firstly, the PD controller is used to replace the equivalent controller of the sliding mode controller, making the controller design independent of the exact mathematical model of the system. Secondly, a new adaptive reaching law is designed in the composite decoupling sliding mode controller, which can speed up the convergence of the sliding mode variables and solve the controller chattering problem. Thirdly, the designed SD type disturbance observer can effectively compensate and suppress the effect of unknown disturbances on the system and improve the robustness of the system. Finally, the stability analysis of the designed controller is performed using Lyapunov stability theory and Barbalat’s lemma, and the simulation results illustrate the effectiveness of the suggested control scheme.
For the speed regulation problem of permanent magnet synchronous motor (PMSM) servo systems, a control strategy based on a super twisting extended state observer (STESO) for adaptive super twisting fast integral terminal sliding mode and adaptive backstepping is proposed. Firstly, a nonsingular fast integral terminal sliding mode control strategy is proposed to solve the problem that the existing integral terminal sliding mode control converges slowly in the region far from the equilibrium point. At the same time, in order to enhance the robustness of the controller and weaken the chatting, an adaptive super twisting (AST) algorithm is designed as reaching law. Secondly, an adaptive backstepping controller is designed for the current loop of the system, using adaptive laws to estimate parameter disturbances, and applying them to feedforward compensation. For unknown lumped disturbances in the system, the STESO is designed using a super twisting algorithm, and the estimated value is compensated to the speed controller. The Lyapunov method serves as the stability. Finally, the simulation and experiments prove the superiority of the designed control strategy.
A dynamic-memory event-triggered heading control method is proposed for unmanned surface vehicle (USV) systems subject to denial-of-service (DoS) attacks and limited communication resources. Firstly, to improve the dynamic performance of the system, an adaptive dynamic-memory event-triggered mechanism (ADMETM) is designed. Secondly, the control signal at the last successful triggering moment is used as the system input when the DoS attack is active to ensure the stability of the USV heading control system. A new closed-loop switched system model is established by comprehensively considering the impact of event-triggered mechanism and DoS attacks. Then, based on the piecewise Lyapunov functional (PLF) approach, the exponential stability criterion and co-design of event-triggered parameters and controller gains in the form of linear matrix inequality are derived. Simulation results show that the proposed method can effectively eliminate the influence of DoS attacks on the system performance, and make the USV quickly track the expected heading angle.
To address the safe operation issues arising from the large-scale integration of electric vehicles (EV) into distribution networks, a security optimization decision-making strategy for active support of virtual power plants (VPP) under EV clustering is proposed. Firstly, a multi-level security decision-making architecture is constructed, which consists of EV - electric vehicles aggregator (EVA) - VPP. Considering the varying dynamic response characteristics of EV in power regulation, a clustering method combining mean shift and K-means clustering is proposed. Additionally, a parameter weighting algorithm based on the gradient descent method and an effectiveness evaluation method based on the Silhouette index are employed to ensure the clustering accuracy of EV with different power regulation characteristics. Furthermore, a multi-chain Markov-based power capacity prediction method for EVA under EV clustering is proposed, and a security optimization decision-making strategy for distribution networks with the active support of VPP is constructed while considering the power flow security constraints of the distribution network. Finally, simulation results demonstrate that this strategy can effectively ensure the safe and stable operation of distribution networks under large-scale EV integration.
Matching problem is a key problem in the collaborative operation mode of fourth party logistics (4PL). However, in this process, it will face the challenge brought by incomplete evaluation value information to decision-making. To solve this challenge, a method of predicting incomplete evaluation value information based on the funk singular value decomposition (FunkSVD) technique is used. A 4PL bilateral matching model with the objective of maximizing the satisfaction of both the third-party logistics and enterprise customers and the constraints of one-to-one matching is established. Subsequently, a particle swarm optimization algorithm integrating hungarian method-based particle repairing mechanism (Hungary-PSO) is proposed. The example analysis shows that FunkSVD predicts the incomplete evaluation value information more closely to the real value, and Hungary-PSO is more suitable for solving large-scale 4PL bilateral matching problems than CPLEX. Finally, through parameter analysis, management suggestions on the marketing strength of 4PL platform are proposed.
Defect segmentation of petrochemical pipeline is very important to maintain the long-term stable operation of petrochemical pipeline. To solve the problems of small amount of defect samples and high real-time detection requirements, a real-time few-shot segmentation method is introduced. Firstly, an improved few-shot segmentation model, ASGNet-BG, is introduced to improve the accuracy of defect segmentation by adding background information. At the same time, a two-stage pruning method is introduced to compress the model and improve the inference speed. The feasibility of the method is verified by experiments on the petrochemical pipeline defect dataset. Experimental results show that the average segmentation accuracy of the improved ASGNet-BG reaches 64%, which is better than other few-shot segmentation models. After pruning, the memory consumption is reduced by 28%, the amount of floating point calculation is reduced by 50%, the inference speed is nearly 3 times faster, and the overall accuracy loss is only 2.19%. The results show that the proposed method can effectively achieve real-time and precise segmentation of petrochemical pipeline defects.
To improve the accuracy of bearing fault diagnosis and address the issue of multi-class data imbalance, a bearing fault diagnosis algorithm based on frequency-domain input and data augmentation modified convolutional neural network (DAMCNN) is proposed. Firstly, the sampled time-domain bearing data are converted into frequency-domain data to enhance feature distinctiveness. Secondly, an overlapping sliding window sampling strategy is applied to expand the dataset and mitigate the data imbalance. Finally, the performance of the proposed algorithm is evaluated through simulation experiments using the Case Western Reserve University (CWRU) bearing dataset. Experimental results demonstrate that, compared with other multi-class algorithms, the proposed DAMCNN algorithm significantly improves the precision and recall in imbalanced data multi-class classification tasks.
For the problem that labeled data are limited in practical batch processes while unlabeled data are relatively adequate, based on merely labeled data cannot realize accurate and effective modelling. A semi-supervised multi-source domain transfer learning strategy for batch process quality prediction is proposed to realize the effective modeling with scarce labeled data. Firstly, the Multi-task Least Squares Support Vector Regression is adopted to develop the model of target process based on the information contained in multiple source processes. Secondly, semi-supervised learning is introduced on the basis of the multi-source domain migration learning strategy and labeling the unlabeled data of the source and target domain processes with pseudo-labels respectively, which can effectively improve the initial accuracy of pseudo-labels. Thirdly, to further improve the prediction accuracy of the model, the pseudo-labels with high confidence are continuouslyadded to the modeling dataset for retraining by the confidence determination method and iterative learning. Finally, the effectiveness of the proposed method is verified by the simulation of the quality prediction of cobalt oxalate synthesis process.
In view of the problem of fabric defect detection, most of the existing methods focus on the extraction of defect features, and do not consider how to eliminate the interference of factors such as the texture structure and fabric folds of the fabric itself on the feature extraction. Therefore, a fabric defect detection method based on image main structure extraction and AFLBP is proposed. Firstly, the extraction method of texture image main structure based on total variation model is used, which can remove the interference of its own texture information in the fabric image. Secondly, the traditional LBP algorithm is improved, and the AFLBP algorithm is proposed to segment the picture to obtain a binary image, which improves the problem that the original LBP algorithm cannot distinguish the local features of the image, alleviates the vibration change of the obviously changing pixels, and improves the accuracy of image edge contour extraction to a certain extent. Finally, the SVM classifier is used to determine whether the segmented fabric binary image contains defects. Experimental results show that the proposed method has better performance in the accuracy of defect detection, and the proposed AFLBP algorithm has better segmentation effect compared with other image segmentation algorithms.
Addressing the issue that conventional deep learning-based time series prediction methods do not consider the varying importance of time series features, an attention-based convolutional neural network and long short-term memory (AB-CNN-LSTM) network is proposed. Firstly, this model combines a convolutional neural network with a long short-term memory network. Secondly, it introduces an attention mechanism into the network to extract significant features through parallel attention branches, thereby expanding the receptive field of the convolutional neural network. The long short-term memory network is then utilized to mine long term information. Finally, ablation studies confirm that the proposed AB-CNN-LSTM model exhibits the best prediction performance on large feature datasets.
In response to the low-latency and low-energy-consumption requirements for computational tasks in industrial manufacturing amid the development of the industrial internet, the characteristics of various computing devices is analyzed and a cloud-edge-end multi-edge collaborative computing model is designed. To optimize the task offloading strategy, the sparrow search algorithm is improved by introducing Tent chaotic mapping to construct the initial population, which enhances the uniformity and ergodicity of population distribution. A random walk strategy is adopted to perturb the optimal sparrow individual during iterations, balancing the algorithm’s global exploration and local exploitation capabilities. On this basis, a Tent random walk-sparrow search algorithm (TR-SSA) is proposed. Simulation results demonstrate that the TR-SSA algorithm improves the stability and flexibility of the offloading strategy, reduces latency and energy consumption costs, and validates the effectiveness of the multi-edge collaborative model.
