Anti-swing control method for bridge crane based on neural network algorithm

Abstract

The invention discloses a bridge crane anti-swing control method based on a neural network algorithm, and relates to the technical field of crane control, and the method comprises the following steps: collecting suspended load swing state variables of a bridge crane and establishing a dynamic model; the dynamic model is used for predicting the suspended load swing trend of the bridge crane, and a swing trend prediction result is obtained; feed-forward compensation control is carried out according to the swing trend prediction result, and a feed-forward control signal is obtained; performing real-time feedback correction on the movement of the bridge crane trolley according to the feedforward control signal to obtain a real-time correction control signal; and performing nonlinear compensation calculation according to the real-time correction control signal to obtain a final anti-swing control result. According to the invention, the method achieves the precise estimation of the future swing state, improves the prediction accuracy of the future swing state, achieves the advanced compensation, improves the response speed of a control system, enables a compensation strategy to adapt to different operation environments, and improves the control precision and robustness.

Description

Technical Field

[0001] The present invention relates to the technical field of crane control, and in particular to an anti-sway control method for a bridge crane based on a neural network algorithm. Background Art

[0002] Bridge cranes are widely used in industrial production, logistics, and large-scale construction projects. Their primary mission is to smoothly transport loads from one location to the desired destination. However, because the load is suspended beneath a trolley, it can swing during movement. This swing not only reduces crane efficiency but can also damage equipment and pose a threat to personnel safety. Therefore, effectively suppressing load swing and improving the control accuracy of bridge cranes is a key technical challenge.

[0003] In the existing technology, feedforward compensation control has shortcomings: traditional prediction methods (such as state-space models and Kalman filters) are difficult to accurately model the swing characteristics under actual working conditions, resulting in limited feedforward compensation effects. In addition, existing methods usually only use the current swing state to calculate the control input, but do not effectively predict future swing trends, which causes the control signal to lag easily and cannot effectively suppress large swings.

[0004] In existing technologies, nonlinear compensation calculations have shortcomings: traditional nonlinear compensation is mostly based on static calculations with fixed model parameters. However, in actual applications, the operating environment of the crane often changes, such as changes in load mass, wind interference, track friction, etc., and dynamic adjustment is impossible. The compensation effect is easily limited, resulting in a decrease in control accuracy. Summary of the Invention

[0005] In view of the deficiencies in the prior art, the present invention provides a bridge crane anti-sway control method based on a neural network algorithm to solve the problems raised in the above background technology.

[0006] To achieve the above object, the present invention provides the following technical solutions: In a first aspect, an embodiment of the present invention provides a method for controlling an anti-sway of a bridge crane based on a neural network algorithm, comprising the following steps: S1. Collection of state variables of bridge crane load swing and establishment of dynamic model; S2. Use the dynamic model to predict the swing trend of the bridge crane load and obtain the swing trend prediction result; S3. Perform feedforward compensation control based on the swing trend prediction result to obtain a feedforward control signal; S4, performing real-time feedback correction on the bridge crane trolley motion according to the feedforward control signal to obtain a real-time correction control signal; S5. Perform nonlinear compensation calculation based on the real-time corrected control signal to obtain the final anti-sway control result.

[0007] To further optimize this technical solution, the feedforward control signal calculation in S3 includes: According to the obtained swing trend prediction result, it is converted into a feedforward control signal, and feedforward compensation control is performed using a feedforward compensation control model.

[0008] To further optimize this technical solution, the feedforward compensation control model includes: ; in: : feedforward gain matrix; : Predicted Hidden variables at the moment; : nonlinear transformation function; : disturbance compensation gain; : Nonlinear disturbance estimation term.

[0009] To further optimize this technical solution, the nonlinear transformation function includes: ; in: , : weight matrix; , : Bias term.

[0010] Further optimizing this technical solution, the nonlinear disturbance estimation term includes: ; in: : disturbance compensation coefficient; : The difference between the current latent variable and the predicted latent variable.

[0011] To further optimize this technical solution, the real-time feedback correction of the car motion in S4 includes: By adopting the linear quadratic regulator (LQR) control strategy, the error is calculated based on the predicted value of the neural network and the actual measured swing angle, and real-time correction is performed in combination with feedback control to adjust the motion trajectory of the trolley in real time and perform anti-sway control.

[0012] To further optimize this technical solution, the nonlinear compensation calculation in S5 includes: A nonlinear compensation model is used to perform nonlinear compensation calculations on the control of the trolley motion to further suppress the swing of the load.

[0013] To further optimize this technical solution, the nonlinear compensation model includes: ; in: : The final control signal of the car; : adaptive gain; : Nonlinear compensation function.

[0014] Further optimizing this technical solution, the adaptive gain includes: ; in: : Gain adjustment parameter; : compensation force control coefficient; : Desired horizontal position of the cart.

[0015] To further optimize this technical solution, the nonlinear compensation function includes: ; in: , : weight matrix; , : bias term; : nonlinear activation function; : Input variable.

[0016] In a second aspect, an embodiment of the present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program instructions are executed by the processor, the steps of a bridge crane anti-sway control method based on a neural network algorithm as described in the first aspect of the present invention are implemented.

[0017] In a third aspect, an embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program instructions are executed by a processor, the steps of a bridge crane anti-sway control method based on a neural network algorithm as described in the first aspect of the present invention are implemented.

[0018] Compared with the prior art, the present invention provides a bridge crane anti-sway control method based on a neural network algorithm, which has the following beneficial effects: This anti-sway control method for bridge cranes based on a neural network algorithm introduces a latent variable prediction method based on a neural network through a feedforward compensation control model. It uses a variational autoencoder (VAE) combined with time series prediction to achieve accurate estimation of future swing states, improve the prediction accuracy of future swing states, realize advance compensation, and improve the response speed of the control system.

[0019] Through the nonlinear compensation model, a nonlinear mapping function based on a neural network is introduced, which can be dynamically adjusted in combination with historical data, so that the compensation strategy can adapt to different operating environments, such as load changes, wind interference, etc., thereby improving control accuracy and robustness. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0021] Figure 1 This is a flow chart of a bridge crane anti-sway control method based on a neural network algorithm proposed in the present invention; Figure 2 This is a flow chart of the swing trend prediction of the anti-sway control method for a bridge crane based on a neural network algorithm proposed by the present invention; Figure 3 This is a flow chart of a feedforward compensation control model of a bridge crane anti-sway control method based on a neural network algorithm proposed in the present invention; Figure 4 This is a flow chart of a nonlinear compensation model of a bridge crane anti-sway control method based on a neural network algorithm proposed in the present invention. DETAILED DESCRIPTION

[0022] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0023] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0024] Secondly, the term "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in various places throughout this specification does not necessarily refer to the same embodiment, nor does it designate a separate or selective embodiment that is mutually exclusive with other embodiments.

[0025] Example 1: Reference Figures 1 to 4 , which is the first embodiment of the present invention, provides a bridge crane anti-sway control method based on a neural network algorithm, comprising the following steps: S1. Collection of state variables of bridge crane load swing and establishment of dynamic model.

[0026] In this embodiment, the swing state variable acquisition and dynamic model establishment include: The main structure of a bridge crane includes the bridge (the main load-bearing structure spanning above the working area), the lifting mechanism (used to lift and lower heavy objects), the trolley running mechanism (used to drive the bridge to move along the track), the trolley running mechanism (which enables the lifting mechanism to move along the track on the main beam to achieve lateral transportation), and the electrical control system (including the control cabinet, control buttons, limit switches, safety protection devices, etc.).

[0027] A Kalman filter is used to extract the load's sway state variables from sensor data, providing data on the crane's load mass, cable length, gravitational acceleration, and cable tension. To implement neural network-based anti-sway control for a bridge crane, an accurate description of the load's motion is required. Lagrange dynamics is used to construct a mathematical description of the load's motion, establishing a dynamic model based on the actual operating environment of the bridge crane to provide reference data for control.

[0028] The bridge crane system can be abstracted as a mobile suspension system with a swinging degree of freedom. That is, the trolley moves in the horizontal direction, and the load swings like a simple pendulum under the action of gravity. It is a coupled second-order nonlinear dynamic system. The dynamic model of this system is established using the Lagrangian method, resulting in: Lagrangian function: ; in: : Kinetic energy, including the translational kinetic energy of the trolley and the resultant kinetic energy of the load; : potential energy, gravitational potential energy of the load; Kinetic energy calculation: ; in: :Car mass: : load mass; : horizontal speed of the car; : angular velocity of the load; : sling length; : Swing angle of the load; Potential energy calculation: ; in: : acceleration due to gravity; Taking derivatives and establishing the Euler-Lagrange equations, we obtain the dynamic coupling equations of the bridge crane: The equation of motion of the car direction: ; Motion equation of the load in the swing direction: ; in: : Car acceleration; : swing angular acceleration; : The control force of the bridge crane trolley in the horizontal direction.

[0029] S2. Use the dynamic model to predict the swing trend of the bridge crane load and obtain the swing trend prediction result.

[0030] In this embodiment, the swing trend prediction includes: Based on the obtained dynamic model, variational autoencoder (VAE) is used for feature extraction to obtain key swing features, which are mapped to the latent variable space so that data from different working conditions can be processed in the same latent space. Recurrent dynamic mapping (RDM) is used for time series mapping, and latent variables are used to model the time series to predict the future swing state, thereby designing and training the neural network, improving the generalization ability of the model and the dynamic prediction accuracy.

[0031] Furthermore, the variational autoencoder VAE performs feature extraction including: Definition of collected state variables: ; in: : Input state vector, which represents the motion state and physical parameters of the bridge crane at time t; : horizontal position of the trolley; : horizontal speed of the car; : Swing angle of the load; : angular velocity of the load; : load mass; : sling length; State variable mapping: ; in: : Hidden variable; : The mean of the latent variable, indicating the center position of the swing feature; : The standard deviation of the latent variable, indicating the uncertainty of the swing characteristics; : Random noise, used to implement the reparameterization technique; Swing state reconstruction: ; in: :By latent variables The resulting swing state; :Decoder, the hidden variables Transition back to physical state; Objective function construction: ; in: : Reconstruction error, representing the error from latent variables Generated swing state and the real state the differences between; : KL divergence, ensuring latent variables Obey the standard normal distribution and improve the generalization ability of the model.

[0032] Furthermore, the cyclic dynamics mapping RDM performs time series mapping including: Nonlinear recursive prediction: ; in: : Hidden variables at the next moment; : The current car feedforward control signal, reflecting the car control input, that is, the car speed adjustment; : Nonlinear activation function to improve control accuracy; , : The state transfer matrix of latent variables and the nonlinear transformation weight matrix are obtained through model training; , : The influence matrix of the control input and the nonlinear transformation weight matrix are obtained through model training; : bias term; Swing state prediction value output: ; in: : The predicted future swing state.

[0033] This model describes how to combine variational autoencoders (VAEs) and recurrent dynamics maps (RDMs) to design neural networks to predict future swing states.

[0034] Traditional neural networks have poor adaptability to different load conditions in the anti-sway control of bridge cranes and have difficulty handling complex nonlinear dynamic relationships. However, this model, through VAE latent variable modeling, can handle different loads in a unified latent variable space, improving generalization ability. Combining VAE and RDM, it can simultaneously consider historical trajectories and physical characteristics, improving prediction accuracy. In combination with time series prediction, it can adjust the trajectory before swing occurs, thereby reducing swing and improving system stability.

[0035] The steps for using the above model include: Data acquisition: The swing state variables of the bridge crane are acquired through step S1; Neural network design: Using acquired data to train VAE and RDM, we design a neural network that can extract swing characteristics under different loads and trajectories, ensuring the model can predict future swing trends based on the current state. Swing state prediction: Use the designed and trained neural network to predict the swing state and obtain the swing trend prediction results.

[0036] S3. Perform feedforward compensation control according to the swing trend prediction result to obtain a feedforward control signal.

[0037] In this embodiment, the feedforward control signal calculation includes: According to the obtained swing trend prediction results, they are converted into feedforward control signals, and feedforward compensation control is performed using a feedforward compensation control model to make the swing angle and angular velocity of the load as close to zero as possible, thereby reducing the occurrence of swing.

[0038] Furthermore, the feedforward compensation control model includes: ; in: : feedforward gain matrix, obtained through model training; : Predicted Hidden variables at the moment; : nonlinear transformation function that maps latent variables to desired car control inputs; : disturbance compensation gain, obtained through model training; : Nonlinear disturbance estimation term, used to compensate for external disturbances.

[0039] Furthermore, the nonlinear transformation function includes: ; in: , : Weight matrix, obtained through model training; , : Bias term.

[0040] Furthermore, the nonlinear disturbance estimation term includes: ; in: : disturbance compensation coefficient, obtained through model training; : The difference between the current latent variable and the predicted latent variable, used to estimate the unmodeled disturbance.

[0041] The model describes how to convert the latent variables of the swing trend prediction results into feedforward control signals to perform feedforward compensation control.

[0042] Traditional feedforward control methods usually calculate compensation inputs based on simple pendulum models or analytical solutions to differential equations. They have limited accuracy when dealing with complex nonlinear systems, especially when the load mass changes, the sling length changes, or non-ideal swing conditions, which can easily lead to control errors. This model uses the results predicted in step S2 for feedforward compensation, improves generalization capabilities, and designs nonlinear disturbance estimation terms so that the control input can adapt to complex working conditions and reduce control errors.

[0043] Uses of this model include: Data acquisition: According to step S2, the swing trend prediction results are obtained and converted into latent variables ; Parameter calculation: Calculate model parameters based on the acquired data, including nonlinear transformation functions and nonlinear disturbance estimates , and obtain the weight matrix of the model during the training process; Feedforward control signal calculation: According to the calculated model parameters, the feedforward control signal is calculated to obtain the feedforward control signal of the car. , thereby adjusting the trolley speed according to the control signal to reduce the swing of the load.

[0044] S4. Perform real-time feedback correction on the bridge crane trolley motion according to the feedforward control signal to obtain a real-time correction control signal.

[0045] In this embodiment, the real-time feedback correction of the vehicle motion includes: Because load swing is affected by external disturbances, including wind and ground vibration, feedforward control alone may not be able to completely eliminate all swings. Therefore, feedback control is required for real-time correction. By employing a linear quadratic regulator (LQR) control strategy, the error between the neural network's predicted value and the actual measured swing angle is calculated, and the motion trajectory of the bridge crane trolley is adjusted in real time to minimize errors. This ensures that the system maintains load stability even in complex environments, improving the reliability of anti-sway control.

[0046] S5. Perform nonlinear compensation calculation based on the real-time corrected control signal to obtain the final anti-sway control result.

[0047] In this embodiment, the nonlinear compensation calculation includes: Even after implementing neural network-based predictive feedforward control and LQR feedback correction, some nonlinear factors, including rope elasticity and air resistance, still affect load swing. Using a nonlinear compensation model, the trolley motion is controlled through nonlinear compensation calculations. Based on the calculated final control signal, the operating parameters of the bridge crane are dynamically and precisely adjusted to further suppress load swing, minimize load swing errors, and achieve faster load stabilization.

[0048] Furthermore, the nonlinear compensation model includes: ; in: : The final control signal of the car; : Adaptive gain, used to dynamically adjust the compensation strength; : Nonlinear compensation function.

[0049] Furthermore, the adaptive gain includes: ; in: : Gain adjustment parameter, obtained through model training; : Compensation force control coefficient, so that it is weakened when the error is small and strengthened when the error is large; : Desired horizontal position of the cart.

[0050] Furthermore, the nonlinear compensation function includes: ; in: , : Weight matrix, obtained through model training; , : bias term; : Non-linear activation function, ReLU or Sigmoid activation function can be used; : Input variables, including latent variables, car position, swing angle and swing angular velocity.

[0051] This model describes how to perform nonlinear compensation to make up for the errors caused by nonlinear factors.

[0052] Traditional nonlinear compensation methods are mostly based on PID control or sliding mode control, but these methods often require precise system models when dealing with complex nonlinear disturbances and are difficult to adapt to the dynamic changes of the actual environment. This model uses adaptive gain control to control the compensation strength, ensuring that the system can achieve the best control effect under different loads and different working environments, improving the compensation accuracy, enabling the compensation calculation to adapt to unknown nonlinear interference, and improving the robustness of the system.

[0053] Uses of the model include: Data input: The result of the feedforward compensation control adjustment of the trolley and the desired trolley horizontal position after adjustment are obtained through step S3, and the latent variables and the operating status parameters of the bridge crane are obtained through step S2; Nonlinear compensation calculation: Based on the obtained data, the nonlinear compensation function and adaptive gain are calculated, the nonlinear impact is calculated and the compensation strength is adjusted to ensure that the control effect remains optimal under different working conditions; Perform final control: Based on the calculated nonlinear compensation result, a control force for compensation is obtained. Combined with the feedforward compensation control obtained in step S3 (i.e., the expected control force obtained by predicting the future latent state), the final control signal of the trolley is obtained by superposition of forces, i.e., the horizontal control force that ultimately needs to be applied to the trolley. The control signal is converted into the corresponding motor voltage, current, or PWM signal through the control system of the bridge crane, and the motor is driven to make the trolley generate actual acceleration, thereby adjusting the trolley speed and further suppressing the swing.

[0054] Example 2: This embodiment also provides a computer device, which is suitable for a bridge crane anti-sway control method based on a neural network algorithm, including a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute computer-executable instructions to implement a bridge crane anti-sway control method based on a neural network algorithm as proposed in the above embodiment.

[0055] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, an anti-sway control method for a bridge crane based on a neural network algorithm as proposed in the above embodiment is implemented.

[0056] The computer device may be a terminal, comprising a processor, memory, a communication interface, a display, and an input device connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores an operating system and computer programs. The internal memory provides an environment for the operating system and computer programs stored in the non-volatile storage media. The communication interface of the computer device is used to communicate with external terminals via wired or wireless communication. Wireless communication may be achieved via Wi-Fi, a carrier network, NFC (near-field communication), or other technologies. The display of the computer device may be a liquid crystal display or an electronic ink display. The input device may be a touchscreen overlay on the display, buttons, a trackball, or a touchpad on the computer device housing, or an external keyboard, touchpad, or mouse.

[0057] If a function is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the portion that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the various embodiments of the method of the present invention. The aforementioned storage medium includes various media that can store program code, such as USB flash drives, mobile hard drives, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical disks.

[0058] The logic and / or steps represented in the flowcharts or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing the logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (e.g., a computer-based system, a system including a processor, or other system that can fetch and execute instructions from an instruction execution system, apparatus, or device). For purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport a program for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0059] More specific examples (a non-exhaustive list) of computer-readable media include the following: an electrical connection with one or more wires (electronic devices), a portable computer disk cartridge (magnetic devices), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), a fiber optic device, and a portable compact disc read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium and then editing, interpreting, or processing it in another suitable manner as necessary, and then storing it in a computer memory.

[0060] It should be understood that various parts of the present invention can be implemented using hardware, software, firmware, or a combination thereof. In the above-described embodiments, multiple steps or methods can be implemented using software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented using hardware, as in another embodiment, any one of the following technologies known in the art or a combination thereof can be used: a discrete logic circuit having logic gate circuits for implementing logic functions on data signals, an application-specific integrated circuit having suitable combinational logic gate circuits, a programmable gate array (PGA), a field-programmable gate array (FPGA), etc.

[0061] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A bridge crane anti-sway control method based on a neural network algorithm, characterized in that: The following steps are involved: S1. Collect the state variables of the bridge crane load swing and establish a dynamic model; S2. Use the dynamic model to predict the swing trend of the bridge crane load and obtain the swing trend prediction result; S3. Perform feedforward compensation control based on the swing trend prediction result to obtain a feedforward control signal; S4, performing real-time feedback correction on the bridge crane trolley motion according to the feedforward control signal to obtain a real-time correction control signal; S5. Perform nonlinear compensation calculation based on the real-time corrected control signal to obtain a final anti-sway control result, and perform control based on the final anti-sway control result.

2. The anti-sway control method for a bridge crane based on a neural network algorithm according to claim 1, characterized in that: The feedforward control signal calculation in S3 includes: According to the obtained swing trend prediction result, it is converted into a feedforward control signal, and feedforward compensation control is performed using a feedforward compensation control model.

3. The anti-sway control method for a bridge crane based on a neural network algorithm according to claim 2, characterized in that: The feedforward compensation control model includes: ； in: : feedforward gain matrix; : Predicted Hidden variables at the moment; : nonlinear transformation function; : disturbance compensation gain; : Nonlinear disturbance estimation term.

4. The anti-sway control method for a bridge crane based on a neural network algorithm according to claim 3, characterized in that: The nonlinear transformation function includes: ； in: , : weight matrix; , : Bias term.

5. The anti-sway control method for a bridge crane based on a neural network algorithm according to claim 3, characterized in that: The nonlinear disturbance estimation term includes: ； in: : disturbance compensation coefficient; : The difference between the current latent variable and the predicted latent variable.

6. The anti-sway control method for a bridge crane based on a neural network algorithm according to claim 1, characterized in that: The real-time feedback correction of the car motion in S4 includes: By adopting the linear quadratic regulator (LQR) control strategy, the error is calculated based on the predicted value of the neural network and the actual measured swing angle, and real-time correction is performed in combination with feedback control to adjust the motion trajectory of the trolley in real time and perform anti-sway control.

7. The anti-sway control method for a bridge crane based on a neural network algorithm according to claim 1, characterized in that: The nonlinear compensation calculation in S5 includes: A nonlinear compensation model is used to perform nonlinear compensation calculations on the control of the trolley motion to further suppress the swing of the load.

8. The anti-sway control method for a bridge crane based on a neural network algorithm according to claim 7, characterized in that: The nonlinear compensation model includes: ； in: : The final control signal of the car; : adaptive gain; : Nonlinear compensation function.

9. The anti-sway control method for a bridge crane based on a neural network algorithm according to claim 8, characterized in that: The adaptive gain includes: ； in: : Gain adjustment parameter; : compensation force control coefficient; : Desired horizontal position of the cart.

10. The anti-sway control method for a bridge crane based on a neural network algorithm according to claim 8, characterized in that: The nonlinear compensation function includes: ； in: , : weight matrix; , : bias term; : nonlinear activation function; : Input variable.

Images