20 September 2025, Volume 32 Issue 9

    

• Select all
  |

• Select
  Improved Fisherman Fishing Algorithm for Solving the Chiller Load Distribution Problem

  YU Junqi, ZONG Yue, ZHAO Anjun, GAO Zhikun

  Control Engineering of China. 2025, 32(9): 1537-1547.

  Abstract ( )   Knowledge map   Save

  An improved fisherman fishing algorithm is proposed for the load distribution optimization problem of parallel chillers. The optimization objective is to reduce the total power consumption of parallel chillers, and the partial load rate of each chiller is used as the optimization variable to solve the problem. In the improved algorithm, the Sobol sequence is used to initialize the position so that the initial solution completely and uniformly covers the whole feasible domain. In the two search strategies, different position updating methods are taken according to the environment in which the fisherman is located, including nonlinear inertia weight and Gaussian variation to improve the convergence speed and optimization accuracy of the algorithm. Finally, the improved algorithm is tested and verified by system simulation through classical cases. The results show that, compared to the unimproved algorithm and other mainstream algorithms, the improved fisherman fishing optimization algorithm can obtain a better operation strategy, and it exhibits strong convergence ability, fast search speed, and good robustness.
• Select
  Fractional-order PIDμ Controller Based on Ziegler-Nichols Tuning Rule

  TANG Mingzheng, LI Chuang, TANG Rongnian

  Control Engineering of China. 2025, 32(9): 1548-1555.

  Abstract ( )   Knowledge map   Save

  In order to reduce the design difficulties of fractional-order PID controllers and solve the problems of auto-tuning, a simple but effective auto-tuning fractional-order PIDμ controller is proposed. Firstly, according to the tuning formulas of the Ziegler-Nichols type PID controller, the constrained-point coordinates of the amplitude-phase curves of the open-loop systems are obtained. Then, the amplitude-phase curves of PIDμ control systems are traversed through the same constrained-point and a set of equations can be established. All parameters of the PIDμ controller are obtained by solving the equations. Finally, a tuning-formulas table for PIDμ controller is compiled, and a graphical user interface (GUI) is designed to facilitate human-computer interaction and engineering applications. The simulation results show that the PIDμ controller has the characteristics of short adjustment time, small overshoot, strong anti-interference ability, and simple enough but effective.
• Select
  Multi Energy Medium Allocation in Steel Industry Based on Improved MOPSO Algorithm

  LIU Yuecheng, WU Dinghui, LU Shenxin, WANG Jing

  Control Engineering of China. 2025, 32(9): 1556-1562.

  Abstract ( )   Knowledge map   Save

  In view of the fact that there are many variables in the multi medium energy allocation model of iron and steel enterprises and the particle swarm optimization algorithm is easy to fall into the local optimum, an improved MOPSO algorithm is proposed to achieve the optimization of multi medium energy allocation. Firstly, based on the changes of the real-time operating efficiency of each equipment, an energy optimal allocation model is established for gas, steam and electric power with the objective function of minimizing cost and energy consumption. The model considers the fluctuation penalty cost of boiler and steam turbine respectively. Then, on the basis of MOPSO algorithm, adaptive inertia weight strategy and Gaussian mutation strategy are introduced to improve the convergence of the algorithm and the diversity of the initial population. Finally, the production data of Baosteel was used for example analysis. The experimental results show that the improved MOPSO algorithm can effectively realize the optimization and allocation of steel energy plan, and the cost and energy consumption of energy operation are reduced by 0.8% and 0.5%, respectively.
• Select
  A Control Allocation Strategy of High-speed Helicopter in Transition Phase

  DUAN Biao, LIU Yehua, WANG Zhaoji, YUAN Tinghui, SHENG Shouzhao

  Control Engineering of China. 2025, 32(9): 1563-1568.

  Abstract ( )   Knowledge map   Save

  A multilevel control allocation architecture combining internal and external loops is designed to solve the problem of nonlinear control redundancy with state-constrained in transition phase of high-speed helicopter, and a control allocation strategy based on improved iterative learning algorithm is proposed. The convergence of iterative learning algorithm is improved by introducing the correction factor to reduce the variance of adaptive learning rate. The test results show that the proposed method can significantly improve the learning rate of algorithm, and the obtained control allocation strategy can fully consider the change of control efficiency of each control surface with the flight speed, which significantly improves the transition flight safety.
• Select
  A Method for Optimizing Lateral Control in Autonomous Driving of Intelligent Vehicle

  LV Xue, WANG Meng, XIONG Wei, CHEN Cheng, MA Yanwei

  Control Engineering of China. 2025, 32(9): 1569-1577.

  Abstract ( )   Knowledge map   Save

  Currently, autonomous driving systems seldom consider individualized differences among users at the control level, leading to a lack of adaptive mechanisms for vehicle lateral control styles to user habits. For the conflict between autonomous steering control styles and individualized driving habits of users, a personalized lateral control method for autonomous driving based on steering styles is proposed. Firstly, by collecting and processing natural driving data from users, a personalized steering style curve for each user is generated through least squares fitting. Subsequently, with the increment of the control sequence at future time points as the decision variable, a mathematical model for optimizing the lateral control of autonomous driving is constructed. Finally, a decoding method based on steering styles is proposed, and a genetic algorithm is employed to solve the optimization problem of lateral control, obtaining the optimal steering control input that aligns with the user’s style for vehicle steering control. Experimental data have validated the personalized differences in steering style curves among different users. Simulation results demonstrate the adaptability of the proposed method to user-specific driving styles, achieving the resolution of driving style conflicts between user and autonomous vehicle. This contributes to enhancing consumer acceptance of intelligent vehicles.
• Select
  Adaptive PID Sliding Mode Attitude Control of Quadrotor Based on Disturbance Observer

  XIA Peng, ZHENG Bochao, LIU Xiaoguang

  Control Engineering of China. 2025, 32(9): 1578-1585.

  Abstract ( )   Knowledge map   Save

  Due to the problems of parameter uncertainty, actuator failure and external disturbance in the quadrotor UAV, in the case of limited input, the use of adaptive PID sliding mode control alone will not fully compensate the total disturbance, resulting in poor control effect. An adaptive sliding mode controller based on disturbance observer is proposed. Firstly, the adaptive sliding mode control is used to compensate the internal disturbance of the system such as parameter uncertainty, and the external disturbance is estimated and compensated by the disturbance observer to reduce the controller output; Then, the RBF neural network is used to optimize the control parameters by self-learning and self-adaptive ability, so that the tracking effect is better; the stability of the closed-loop system is finally proved. The simulation results show that the proposed method can reduce the controller output and have higher control quality when the input is limited.
• Select
  Synchronous Tracking Control of Four-motor System Based on Command Filtered Backstepping Method

  ZHENG Zhichao, CAI Mingjie, WANG Baofang

  Control Engineering of China. 2025, 32(9): 1586-1593.

  Abstract ( )   Knowledge map   Save

  The synchronization tracking problem of four-motor servo system is studied, and the synchronous tracking controller is designed based on the backstepping method combined with the command filtering control method. Firstly, according to the mechanical structure of the four-motor servo system, the speed and torque synchronization control signals are designed to ensure that the system has a high synchronization performance. Secondly, in the process of controller design, the virtual control signal is filtered by the command to avoid the “calculation explosion” problem caused by repeated derivation, and an error compensation system is established to reduce the filtering error. Finally, Lyapunov theory is used to analyze the stability of the closed-loop system, and the simulation results verify the effectiveness of the proposed method.
• Select
  Distributed Secondary Voltage Resilient Control for Islanded Microgrids Subject to FDI Attacks#br#

  ZHAO Tianbao, HAN Fei, LIU Qingqiang, DONG Hongli, SONG Yanhua, WANG Mei

  Control Engineering of China. 2025, 32(9): 1594-1602.

  Abstract ( )   Knowledge map   Save

  The false data injection (FDI) attack would make the secondary voltage of islanded microgrids deviate from the nominal value. To address this problem, a distributed secondary resilient control scheme is proposed based on an auxiliary system to restore the voltage to the nominal value. Owing to the communication connection between the auxiliary system and the cyber layer, the internal variables of the auxiliary system are employed to adjust the synchronization error to restore the voltage of the islanded microgrid to the nominal value. The advantage of this design is to reduce the adverse impact of the FDI attacks on different locations and resist the fluctuation caused by the FID attacks by means of the Lyapunov stability method. The effectiveness and applicability of the proposed distributed secondary voltage resilient control scheme are demonstrated by a MATLAB/Simulink simulation.
• Select
  Bearing Fault Diagnosis with Multi-scale Feature Fusion Under Self-attention Mechanism#br#

  SHI Haojin, QIU Jier, TAO Hongfeng, TANG Jinlin, JIN Guanghu

  Control Engineering of China. 2025, 32(9): 1603-1610.

  Abstract ( )   Knowledge map   Save

  A fault diagnosis model with multi-scale feature fusion under self-attention mechanism is proposed for the problem that the rolling bearing vibration signals under different operating conditions have multi-level, non-linear and non-smooth characteristics, which leads to the difficulty of cross-operating condition fault diagnosis of bearings. Firstly, the low-frequency features and local time-domain features of the original vibration signal are extracted by using convolutional kernels of different scales. Secondly, a multi-scale feature fusion module with embedded multi-head self-attention (MHSA) and squeeze-and-excitation (SE) self-attention networks is constructed instead of the traditional concat method to further mine the intrinsic linkage of time-frequency features of vibration signals to improve the robustness of variable condition diagnosis. Also, batch normalization is introduced in the network to reduce the internal variable bias and improve the training performance. The experiments show that the end-to-end fault diagnosis method can fully combine different scale features, and the average diagnosis accuracy can reach more than 97% under variable working conditions.
• Select
  Visual Odometry of Mobile Robot Based on Deep Learning

  LIU Haitao, DAI Juan, ZHU Shengtao, LI Jianfeng

  Control Engineering of China. 2025, 32(9): 1611-1618.

  Abstract ( )   Knowledge map   Save

  In order to solve the problem of pose estimation accuracy degradation caused by error accumulation in mobile robot localization, a deep learning-based visual odometry method is proposed. Firstly, a convolutional neural network (CNN) is designed to extract more detailed features of the image sequences by optimizing the size of the convolutional kernel layer by layer. Then, an adaptive memory network records historical poses, while a bi-directional long short-term memory (Bi-LSTM) predicts future poses. By fusing both past and future information, the method reduces error accumulation in pose estimation. Finally, experiments on KITTI and TUM datasets show the method outperforms existing approaches in pose accuracy, absolute and relative trajectory error.
• Select
  Domain Adaptation for Underwater Image Enhancement

  BING Xuewen, REN Wenqi, TANG Yang

  Control Engineering of China. 2025, 32(9): 1619-1625.

  Abstract ( )   Knowledge map   Save

  Underwater image enhancement is crucial for marine exploration, but real-world data is difficult to obtain, so most current methods rely on synthetic data for training. However, synthetic datasets often fail to describe the natural appearance and show poor capabilities of the generalization. An in-air to underwater image enhancement framework that overcomes the limitations of underwater synthetic datasets is proposed. Specifically, the presented framework consists of a color correction step and a domain adaptation step. To better extract global information, we introduce Transformer module as an encoder to extract features of the input image in the domain adaptation step. We also propose a feature enhancement module to accommodate spatially varying textures and edges. Experimental results indicate that the framework not only obtains remarkable UIQM scores than the previous methods do, but also achieves superior results in visual perception.
• Select
  Remaining Useful Life Prediction of Grinding Wheel Based on Copula Entropy and Improved AM-LSTM

  CHENG Yi, WANG Cheng, YANG Guifeng

  Control Engineering of China. 2025, 32(9): 1626-1633.

  Abstract ( )   Knowledge map   Save

  The Grinding wheel is the core grinding component of a CNC grinding machine. The prediction of remaining useful life is a key part of the predictive maintenance of the grinding machines. For the problem of the high dimension of parameters, lack of useful information, and the difficulty in describing grinding wheel degradation mechanism, a prediction method of remaining useful life of grinding wheel is proposed based on Copula entropy and improved Attention Mechanism Long Short-term Memory (AM-LSTM) network. In order to reduce the input dimension of the model, the correlation between each parameter and the target is evaluated based on the Copula entropy method to select valid variables. For the difference of time series features of data on multiple time scales, a multi-scale learning strategy and attention mechanism are introduced to improve the feature extraction ability of the model for multi-parameter long-term series data. Compared with the AM-LSTM network, the root mean square error and mean absolute percentage error of the improved AM-LSTM model in the remaining useful life prediction experiment of the grinding wheel are reduced by 21.39% and 16.98%, which verifies the effectiveness of the model.
• Select
  Fault Diagnosis of Shipboard Power System Based on IGA Optimized BPNN

  LIU Xing, SHI Weifeng, XIE Jialing, SONG Tiewei

  Control Engineering of China. 2025, 32(9): 1634-1642.

  Abstract ( )   Knowledge map   Save

  Considering the non-stationary and aperiodic nature of fault current signals within shipboard electrical systems, a fault diagnosis method is proposed for ship power system based on immune mechanism based immune genetic algorithm (IGA) optimized BP neural network on the basis of BP algorithm. Firstly, wavelet packet decomposition (WPD) is employed to extract feature vectors from fault current signals. The subsequent dimensionality reduction of these feature vectors decreases the input node count in the Backpropagation Neural Network, thereby streamlining its architecture. Then, to address the slow convergence and susceptibility to local minima in the Backpropagation algorithm, an Improved Genetic Algorithm (IGA) is incorporated to optimize the weights and thresholds of the BP neural network. Simulation and case testing results for marine power systems demonstrate that, compared to Genetic Algorithms (GA), the Improved Genetic Algorithm (IGA) achieves a 56% reduction in error convergence iterations. The optimized Backpropagation Neural Network attains a 95.68% fault diagnosis accuracy rate, confirming superior diagnostic precision.
• Select
  Air Quality Prediction Method and Its Application of Elman Neural Network Based on DE-WOA

  QIAO Yinwei, JIA Xinchun, GUAN Yanpeng, HAO Jianhua

  Control Engineering of China. 2025, 32(9): 1643-1651.

  Abstract ( )   Knowledge map   Save

  To address the slow convergence and local - optimum trapping issues of the Whale Optimization Algorithm (WOA), an improved WOA integrating the Differential Evolution Algorithm and Cauchy-t perturbation is proposed. The algorithm optimizes the Elman Neural Network (ENN) for more accurate Air Quality Index (AQI) prediction. The Tent chaotic mapping and elite opposition - based learning initialize the population to enhance diversity. The Differential Evolution Algorithm boosts global search, while the Cauchy-t perturbation strengthens late - stage local search. The improved DE-WOA is then used to optimize the ENN, verified with Taiyuan City’s air quality data. Results show the model’s root - mean - square error is 5% lower than other models on average, dem
• Select
  Fault Diagnosis of Bearing Based on Multi-scale ResNet Feature Fusion

  XUE Bin, LI Yingshun, GUO Zhannan, LU Jilin

  Control Engineering of China. 2025, 32(9): 1652-1658.

  Abstract ( )   Knowledge map   Save

  For the problem that the shallow feature network can’t make full use of the feature information in the fault diagnosis of rolling bearings, a fault diagnosis model of rolling bearings based on multi-scale residual network feature fusion is proposed. Firstly, the original data is used as input of the model, and the first layer of the network employs convolutional kernels of varying scales to extract shallow features, thereby enhancing the model’s receptive field. Subsequently, Long Short-Term Memory network is used to mine time series feature information. Then the obtained feature information is input into the residual network for deep feature extraction, and an attention mechanism is introduced to screen useful information. Finally, Softmax is used as a classifier to realize fault diagnosis. Experimental results show that the proposed method has faster convergence speed and better robustness in cross-working condition identification than single-channel convolutional neural network, double-channel convolutional neural network and single-channel residual network.
• Select
  Bearing Fault Diagnosis of Wind Turbine Based on MSACNN-GMNR

  ZHU Xiaohui, REN Yanju, WEI Wei

  Control Engineering of China. 2025, 32(9): 1659-1665.

  Abstract ( )   Knowledge map   Save

  For the problem that the vibration signal of wind turbine rolling bearing is affected by complex transmission path and vibration coupling of other components, which leads to a large amount of noise affecting the fault diagnosis accuracy, a multi-scale attention convolutional neural network (MSACNN-GMNR) with noise reduction effect was proposed. Firstly, the Gram noise reduction module (GMNR) was used to reduce the original signal of the generator bearing. Then, a multi-scale convolution module (MSCNN) was used to extract multi-scale features, and attention mechanism (AM) was used to assign different weights to different scale features to realize fault diagnosis of wind turbine bearings. Finally, the input was sent to the softmax classifier for sorting. The proposed method was verified by the actual collected data of wind turbine bearings. The results show that the proposed algorithm still has a high diagnostic effect under strong background noise, and has a certain research value for the actual fault diagnosis of wind turbine bearings.
• Select
  Tower Crane Load Positioning Based on Least Squares D-S Evidence Theory and Kalman Filter

  WEN Jiakai, CHEN Zhimei, SHAO Xuejuan, ZHANG Jinggang

  Control Engineering of China. 2025, 32(9): 1666-1672.

  Abstract ( )   Knowledge map   Save

  In the tower crane positioning system, the load carries out three-dimensional movement, and its motion system is more complex and positioning is more difficult. A tower crane load location algorithm based on double least square method, D-S evidence theory and Kalman filter is established. Firstly, the least squares method is introduced to locate the coordinates of the load in a fixed three-dimensional space, and the least squares method is improved to a double-layer least squares method to roughly estimate the three-dimensional coordinates of the load. Then, the estimated information is fused with the D-S evidence theory to obtain the corresponding weights of the estimated information. Finally, the fused data is transferred to the Kalman filtering algorithm for a higher level of optimal estimation, so as to achieve accurate positioning of tower crane load. Simulation results demonstrate that the proposed algorithm outperforms comparative algorithms in both positioning accuracy and three-dimensional dynamic error, enabling precise positioning of tower crane loads.
• Select
  Dynamic Path Planning and Navigation Control Based on Improved TEB and PLOS Algorithms

  DU Bin, YANG Hai

  Control Engineering of China. 2025, 32(9): 1673-1680.

  Abstract ( )   Knowledge map   Save

  Currently, most robot path planning methods lack environmental awareness and are difficult to adapt to autonomous navigation control in complex environments. A solution is proposed to address this type of problem by integrating an extended state estimator, an improved time elastic band (TEB) algorithm, and an improved pure pursuit and LOS based (PLOS) algorithm. Firstly, an extended state estimator is introduced to estimate the distance and attitude changes between the robot and dynamic obstacles in real time, enhancing the ability to perceive the environment. Then, the TEB algorithm based on third-order spline interpolation is used for dynamic path planning, which reduces computational costs by smoothing the sequence of path points and generates local paths that balance safety and optimality. Finally, in the path tracking stage, a PLOS algorithm based on target point heading angle deviation control is proposed, which introduces heading angle deviation parameters to achieve adaptive control and avoid oscillation and path crossing phenomena. The simulation results indicate that the TEB algorithm based on third-order spline interpolation can adjust the path planning scheme according to the dynamic obstacle position, which can provide reliable technical support for applications such as unmanned driving and robot delivery.
• Select
  Mechanism Modeling of Steering and Reversing Mechanism for Double Pump Water Jet Propulsion Ship

  LI Huaqing, XU Zijing, YUAN Jingqi

  Control Engineering of China. 2025, 32(9): 1681-1686.

  Abstract ( )   Knowledge map   Save

  Dual-pump waterjet-propelled vessels feature independently adjustable steering and reversing mechanisms, whose excellent manoeuvrability ensures the successful completion of various complex navigation tasks. Taking the dual-pump waterjet propulsion system equipped on an 8.5 t waterjet vessel as the research object, a mechanistic analysis was conducted on the velocity vector distribution of the external jet flow from the steering and reversing mechanisms. A waterjet flow loss coefficient was introduced to describe the phenomena of jet flow loss and thrust loss caused by fluid kinetic energy loss under different operating conditions of these mechanisms. An identification method for the waterjet flow loss coefficient based on computational fluid dynamics (CFD) simulation was proposed. It refined the mechanistic model used for calculating the vector thrust of the waterjet propulsion device, and presented the open-loop simulation results of the waterjet propulsion device model.
• Select
  Simulation of Force/Position Hybrid Control of Mechanical Arm Based on Key Joint

  WANG Xiao, LU Zhiguo, LI Wenqiao

  Control Engineering of China. 2025, 32(9): 1687-1692.

  Abstract ( )   Knowledge map   Save

  Taking a six-degree-of-freedom (6-DoF) serial manipulator as the research object, the concept of key joints is first introduced. By implementing torque control at these key force-controlled joints while simultaneously incorporating their output angles into the manipulator’s position control, a hybrid force/position control algorithm based on active force-controlled joints is proposed. Leveraging the theories of manipulator dynamics and inverse kinematics, the proposed algorithm achieves force control along a predefined positional trajectory. Finally, for the RM65-B dexterous 6-DoF manipulator, a MATLAB-based simulation platform is established to conduct experiments and analyze the tracking performance of the hybrid force/position control scheme. The favorable simulation results validate the feasibility of the proposed control method.
• Select
  Anti-interference Filtering Algorithm for Linear Discrete Systems with Dual-unknown-inputs

  DONG Shigui, WANG Na, ZHAO Keyou, LU Zihao

  Control Engineering of China. 2025, 32(9): 1693-1699.

  Abstract ( )   Knowledge map   Save

  For linear discrete systems with dual-unknown-inputs, the classical recursive three-step filter cannot be applied. A new anti-interference filtering algorithm is proposed to estimate the state of the system and dual-unknown-inputs. Firstly, the innovation is designed to obtain the unbiased estimate of the unknown input in the measurement equation, and then the unbiased estimate of the unknown input in the state equation is obtained. Then, the two unknown input estimates are used to modify the one-step estimate of the state, and finally the unbiased estimate of the system state is obtained. The simulation results show that the filter can still estimate the system state and dual-unknown-inputs when the unknown input signals in the state equation an
• Select
  Composite Control of Active Heave Compensation for Deep-sea Cranes

  ZHOU Qing, CHEN Zhimei, SHAO Xuejuan, ZHANG Jinggang

  Control Engineering of China. 2025, 32(9): 1700-1708.

  Abstract ( )   Knowledge map   Save

  In order to solve the problem of poor control effect caused by phase lag and system nonlinearity in the active heave compensation system of deep-sea cranes, a composite control combining model predictive control and sliding mode control with high gain observer is proposed. To this end, the mathematical model of the active heave compensation system of the deep-sea crane is firstly established; then the sliding mode observer is designed according to the transfer function of the load to the mother ship to track the heave speed of the mother ship in real time and eliminate the phase difference between the mother ship and the load; A model predictive controller is designed to actively compensate the heave speed of the load after tracking to improve the control performance of the system. Finally, a Simulink simulation model of the control system is built for simulation experiments. The results show that the composite control can not only solve the system phase. The problem of lag can also improve the robustness and control accuracy of the system, and achieve the purpose of real-time response of the control system, providing a reference for future system development.
• Select
  A Nonlinear Multi-condition Soft Sensing Method Based on Fuzzy Domain Adaptive Regression

  YE Zefu, HAN Pengdong, ZHU Zhujun, REN Mifeng, YAN Gaowei

  Control Engineering of China. 2025, 32(9): 1709-1717.

  Abstract ( )   Knowledge map   Save

  Most of the existing approaches using transfer learning to solve soft sensor modeling for multiple conditions rely on domain-adaptation partial least squares regression modeling frameworks, but such frameworks cannot cope with the nonlinearity and uncertainty of data in complex industrial processes. To improve the prediction accuracy of the soft sensor model under cross-condition, a nonlinear multi-condition soft sensor method based on fuzzy domain adaptation regression is investigated. Firstly, the condition of the fuzzy rule in the T-S fuzzy model is regarded as a feature extractor, and the clustering prototype in the historical working condition is transferred to the current working condition by the transfer C-means clustering method to realize the conditional alignment of the fuzzy rule. Then, a partial least squares regression method based on the transfer subspace is introduced to replace the least squares method for calculating the optimal regression coefficient of the T-S fuzzy model and achieving the conclusion alignment of the fuzzy rule. Finally, the specific steps of the multi-condition fuzzy soft sensor modeling system are given. The effectiveness of the algorithm is verified by a numerical case and simulation experiments with Tennessee Eastman process data.
• Select
  Research on Fractional Order Automatic Generation Control Based on NAFPSO Algorithm

  DENG Qichen, RONG Na, LI Hao

  Control Engineering of China. 2025, 32(9): 1718-1728.

  Abstract ( )   Knowledge map   Save

  In order to improve the image segmentation accuracy, and solve the problems in the traditional fuzzy clustering algorithm, an image segmentation algorithm based on improved fruit fly algorithm optimizing fuzzy means clustering algorithm is proposed. Firstly, the evolutionary step of the fruit fly algorithm is adjusted according to the best flavor value and the current iteration number, which can avoid falling into local optimum, obtain higher convergence precision, and accelerate the convergence speed. Then, the initial cluster centers of fuzzy k-means clustering algorithm are selected by using the improved fruit fly algorithm to realize image segmentation. Finally, the performance is tested by simulation experiments. The experimental results show that, compared with other images segmentation algorithms, the proposed algorithm is better in segmentation accuracy rate, segmentation speed and robustness, and has wider application scope.