Voice coil motor driven valve controller based on neural network active disturbance rejection control

Abstract

The application provides a voice coil motor driving valve controller based on a neural network active disturbance rejection control and a control method thereof.The controller comprises a target control signal preprocessing module, a valve core displacement sensor, a real-time error calculation unit, a controller module, an extended state observer and an active disturbance rejection controller.The application pre-processes a target control signal to obtain a smooth target displacement control curve, thereby avoiding system response overshoot.The application adopts a control method combining a neural network, an active disturbance rejection control and signal preprocessing, and through real-time self-tuning of the active disturbance rejection algorithm parameters by the neural network, the environmental adaptability and robustness of high-precision and high-response position control of the voice coil motor driving valve are improved.The active disturbance rejection algorithm setting parameters are tuned in real time according to different working conditions, system disturbance is observed and compensated in real time through the active disturbance rejection control, and accurate position control of the voice coil motor driving valve main valve core is realized.

Description

Technical Field

[0001] This invention relates to the field of hydraulic control technology, and more particularly to a voice coil motor driven valve controller and its control method based on neural network active disturbance rejection control. This control technology is suitable for high-precision, high-dynamic-response voice coil motor driven proportional valves, and can significantly improve the response speed, control accuracy, and dynamic stability of voice coil motor driven proportional valves in complex and frequently fluctuating load environments. Background Technology

[0002] Hydraulic control systems, due to their high power density and rapid dynamic response, are widely used in engineering machinery, aerospace, and precision manufacturing. Voice coil motor-driven proportional valves are one of the core components for achieving high dynamic and high-precision control in hydraulic systems. However, the inherent nonlinear characteristics of voice coil motor-driven valves, load fluctuations, and external disturbances lead to problems such as lag in valve core motion control response, low control accuracy, and valve core displacement overshoot. Traditional linear control methods, such as PID control, struggle to balance stability and accuracy requirements in complex dynamic environments. Therefore, a new voice coil motor-driven valve controller and its algorithm are needed to improve the control accuracy of voice coil motor-driven valves under complex operating conditions.

[0003] Active disturbance rejection control (ADRC) achieves effective control of unknown external disturbances and internal nonlinearities of proportional valves by observing and compensating for system disturbances in real time. However, under rapid dynamic commands, ADRC may exhibit response lag or control overshoot. Neural networks have the ability to learn and model nonlinear systems and can effectively adapt to changes in proportional valve characteristics, but their independent application makes it difficult to achieve rapid adjustment of dynamic response. Signal preprocessing algorithms can smooth control commands, reduce the impact of command mutations on valve core control during the control process, and improve the stability of dynamic response. Summary of the Invention

[0004] To overcome the shortcomings of existing technologies, this invention proposes a voice coil motor driven valve controller and its control method based on neural network-based active disturbance rejection control. This controller and method can achieve high-precision valve core position control, fast dynamic response, and excellent disturbance rejection performance.

[0005] The objective of this invention is achieved through the following technical solution:

[0006] This invention provides a control method for a voice coil motor driven valve based on neural network active disturbance rejection control, which includes the following steps:

[0007] S1. Obtain the target control signal of the controlled voice coil motor driven valve, preprocess the target control signal, and process the target control signal into a smooth target displacement control curve that includes acceleration stage, constant speed stage and deceleration stage.

[0008] S2. Real-time acquisition of the valve core displacement of the valve driven by the controlled voice coil motor, and subtraction of the valve core displacement signal with the smoothed target displacement input signal obtained by S1 to calculate the real-time error signal.

[0009] S3. Based on the current working state of the voice coil motor driven valve, the controller module uses the trained neural network model to obtain the feedback gain parameters of the active disturbance rejection controller and the working parameters of the extended state observer.

[0010] S4. The extended state observer outputs the total disturbance estimate of the current system based on the operating parameters obtained in S3.

[0011] S5. Based on the real-time error signal, the total disturbance estimate and the feedback gain parameter, the controller module uses an active disturbance rejection controller to dynamically adjust the real-time error signal and outputs control commands to drive the voice coil motor drive valve.

[0012] The present invention also provides a voice coil motor driven valve control system for implementing the method, comprising:

[0013] The target control signal preprocessing module is used to preprocess the target control signal to obtain a smooth target displacement control curve;

[0014] A valve core displacement sensor is used to acquire the valve core displacement of the valve driven by the controlled voice coil motor in real time.

[0015] The real-time error calculation unit calculates the real-time error signal by subtracting the valve core displacement signal from the smoothed target displacement input signal.

[0016] The controller module uses a trained neural network model to obtain the feedback gain parameters of the active disturbance rejection controller and the operating parameters of the extended state observer.

[0017] The extended state observer outputs the total disturbance estimate of the current system based on the operating parameters obtained from the controller module.

[0018] The active disturbance rejection controller dynamically adjusts the real-time error signal based on the real-time error signal, the total disturbance estimate, and the feedback gain parameter, and outputs control commands to drive the voice coil motor drive valve.

[0019] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0020] (1) The present invention performs smoothing preprocessing on the target control signal, and the preprocessed target control signal avoids system response overshoot;

[0021] (2) The present invention adopts a control method that combines neural network, active disturbance rejection control and signal preprocessing. Through signal preprocessing, the control command is made smoother and the impact caused by command mutation is reduced. Through the real-time self-tuning of the active disturbance rejection algorithm parameters by the neural network, the environmental adaptability and robustness of the high-precision and high-response position control of the voice coil motor driven valve are improved. Attached Figure Description

[0022] Figure 1 This is a flowchart of a voice coil motor driven valve control method based on neural network active disturbance rejection control.

[0023] Figure 2 This is a schematic diagram of the controller in the embodiment. Detailed Implementation

[0024] The present invention will be further described and illustrated below with reference to specific embodiments. The technical features of each embodiment of the present invention can be combined accordingly, provided that there is no mutual conflict.

[0025] like Figure 1 and Figure 2 As shown, the voice coil motor driven valve control method based on neural network active disturbance rejection control of the present invention includes the following steps:

[0026] S1. Obtain the target control signal of the controlled voice coil motor driven valve, preprocess the target control signal, and process the target control signal into a smooth target displacement control curve that includes acceleration stage, constant speed stage and deceleration stage.

[0027] The target control signal is the control task given by the outside world, that is, the valve core needs to reach a certain position at a certain time. Therefore, the target control signal can usually be expressed as the relationship between the target displacement of the valve core and time.

[0028] Preprocessing the target control signal is crucial to prevent overshoot of the valve spool during control. Specifically, this invention employs a tracking differentiator to perform first or second-order differentiation on the target control signal, extracting the dynamic rate of change of the signal and generating a smooth target displacement control curve to avoid system response overshoot. This smooth target displacement control curve is a time-valve spool target displacement curve, including an acceleration phase, a constant speed phase, and a deceleration phase. In the acceleration phase, the valve spool speed smoothly accelerates from zero, and the valve spool displacement increases rapidly. In the constant speed phase, the valve spool speed remains constant, and the valve spool displacement gradually approaches the target displacement. In the deceleration phase, the valve spool speed smoothly decreases, and the valve spool decelerates and eventually reaches the target displacement. This three-stage variable speed curve improves the valve spool's response speed and maintains a low speed when the valve spool is close to the target displacement, thus avoiding valve spool overshoot and oscillation.

[0029] In one embodiment of the present invention, the smooth target displacement control curve is expressed by the following formula:

[0030] S(t) = A / 2(1-cos(2πt / T))

[0031] Where A is the amplitude of the target control signal, and T is the time of the acceleration and deceleration phases. Smooth control of the target control signal is achieved by adjusting A and T.

[0032] S2. Real-time acquisition of the valve core displacement of the valve driven by the controlled voice coil motor, and subtraction of the valve core displacement signal with the smoothed target displacement input signal obtained by S1 to calculate the real-time error signal.

[0033] The valve core displacement of the controlled voice coil motor driven valve can be obtained by any method in the prior art. In one embodiment of the present invention, the valve core displacement signal of the controlled voice coil motor driven valve is obtained by an LVDT sensor. The LVDT sensor is excited to work by a sine wave with fixed frequency and amplitude through a multivibrator circuit and an amplitude adjustment circuit. The rectifier amplifier circuit processes the sine signal of the LVDT and outputs a DC voltage signal proportional to the LVDT. The acquisition circuit uses an operational amplifier with adjustable amplification factor to finally obtain the main valve core displacement information of the voice coil motor.

[0034] The present invention further employs a filtering and denoising unit to perform data filtering and denoising processing on the obtained valve core displacement signal in order to obtain an accurate feedback signal; the filtering algorithm adopts adaptive Kalman filtering and moving mean filtering to reduce noise and transient interference and ensure high-precision transmission of the feedback signal.

[0035] The formula for adjusting the real-time error signal is as follows:

[0036] e(t) = y ref (t)-y(t)

[0037] Where e(t) is the error, y ref y(t) represents the target position command generated by the smooth target displacement control curve, and y(t) represents the actual position of the valve core.

[0038] S3. Based on the current working state of the voice coil motor driven valve, the controller module uses the trained neural network model to obtain the feedback gain parameters of the active disturbance rejection controller and the working parameters of the extended state observer.

[0039] A neural network model is used for nonlinear modeling and dynamic adjustment of a voice coil motor-driven valve, and is trained using historical operating data of the valve. In an embodiment of the invention, the neural network model adopts a three-layer feedforward neural network structure. The input layer contains valve spool displacement, valve spool velocity, and valve spool acceleration signals. The hidden layers in the middle use the ReLU activation function, and the output layer contains the feedback gain parameters of the active disturbance rejection controller and the operating parameters of the extended state observer. During training, the neural network model can be trained independently or as a whole control system with the active disturbance rejection controller and the extended state observer. The backpropagation (BP) algorithm is used during training to optimize the network parameters.

[0040] The training process updates parameters using the following formula:

[0041]

[0042] Where θ represents the network parameters, η is the learning rate, and L is the loss function. In real-time applications, neural network models can adaptively adjust their parameters according to the dynamic changes of the system.

[0043] S4. The Extended State Observer (ESO) outputs the total disturbance estimate of the current system based on the operating parameters obtained in S3.

[0044] The Extended State Observer (ESO) estimates the disturbance, nonlinear dynamic characteristics, and unmodeled errors of the voice coil motor-driven valve, and then achieves dynamic compensation through nonlinear feedback. The output of the Extended State Observer is the estimated total disturbance, defined as follows:

[0045] d^(t)=Lo·(yy^)

[0046] Where d^(t) is the total disturbance estimate, Lo is the observation gain matrix, i.e. the operating parameters of the extended state observer, y is the actual displacement of the valve core, and y^ is the estimated valve core displacement; by adjusting the value of Lo, the response speed of the extended state observer to the disturbance is controlled, thereby optimizing the control effect.

[0047] S5. Based on the real-time error signal, the total disturbance estimate and the feedback gain parameter, the controller module uses an active disturbance rejection controller to dynamically adjust the real-time error signal and outputs control commands to drive the voice coil motor drive valve.

[0048] The primary task of the Active Disturbance Rejection Control (ADRC) is to adjust the compensated error using a nonlinear feedback formula based on the disturbance estimation results of the ESO (Effective Disturbance Occurrence). The output signal of the ADRC drives the actuator to achieve precise control of the valve core.

[0049] The feedback formula for the active disturbance rejection controller is:

[0050] u(t)=-K·e(t)-Kd·e(t)+d^(t)

[0051] Where u(t) is the control input signal, e(t) is the error between the expected value and the actual value, d^(t) is the disturbance estimate, and K and Kd are the feedback gain coefficients, respectively.

[0052] The neural network model of this invention provides a nonlinear estimate of the system, which is used to adjust the feedback gain and ESO parameters of the Active Disturbance Rejection Controller (ADRC) in real time, enabling the ADRC to respond quickly under complex operating conditions. The working principle of this cooperative control mechanism is as follows: the neural network continuously updates the nonlinear modeling results of the system's dynamic characteristics and feeds them back to the gain adjustment unit of the ADRC. Based on the modeling results, the feedback gain and extended state observer (ESO) parameters of the ADRC adaptively change with the system state to optimize the dynamic response performance of the control system.

[0053] This invention makes control commands smoother by preprocessing control and feedback signals; it performs self-learning and modeling of the nonlinear system of the voice coil motor driven valve through a neural network algorithm, and sets parameters for the active disturbance rejection algorithm in real time for different operating conditions; and it achieves precise position control of the main valve core of the voice coil motor driven valve by real-time observation and compensation of system disturbances through active disturbance rejection control.

[0054] like Figure 2 As shown, the present invention also provides a voice coil motor driven valve control system for implementing the method, comprising:

[0055] The target control signal preprocessing module is used to preprocess the target control signal to obtain a smooth target displacement control curve;

[0056] A valve core displacement sensor is used to acquire the valve core displacement of the valve driven by the controlled voice coil motor in real time.

[0057] The real-time error calculation unit calculates the real-time error signal by subtracting the valve core displacement signal from the smoothed target displacement input signal.

[0058] The controller module uses a trained neural network model to obtain the feedback gain parameters of the active disturbance rejection controller and the operating parameters of the extended state observer.

[0059] The extended state observer outputs the total disturbance estimate of the current system based on the operating parameters obtained from the controller module.

[0060] The active disturbance rejection controller dynamically adjusts the real-time error signal based on the real-time error signal, the total disturbance estimate, and the feedback gain parameter, and outputs control commands to drive the voice coil motor drive valve.

[0061] Regarding the apparatus in the above embodiments, the specific manner in which each module performs its operation has been described in detail in the embodiments related to the method, and will not be elaborated upon here.

[0062] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to the description of the method embodiments. The device embodiments described above are merely illustrative; the various modules in the device represent a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules may be combined or integrated into another unit. Furthermore, the connections between the shown or discussed modules may be communication connections through some interfaces, which may be electrical or other forms. Some or all of the modules can be selected to achieve the purpose of this application according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0063] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the appended claims.

Claims

1. A control method for a voice coil motor driven valve based on neural network active disturbance rejection control, characterized in that, Includes the following steps: S1. Obtain the target control signal of the controlled voice coil motor driven valve, preprocess the target control signal, and process the target control signal into a smooth target displacement control curve that includes acceleration stage, constant speed stage and deceleration stage. S2. Real-time acquisition of the valve core displacement of the valve driven by the controlled voice coil motor, and subtraction of the valve core displacement signal with the smoothed target displacement input signal obtained by S1 to calculate the real-time error signal. S3. Based on the current operating state of the voice coil motor driven valve, the controller module uses a trained neural network model to obtain the feedback gain parameters of the active disturbance rejection controller and the operating parameters of the extended state observer. The neural network model adjusts the feedback gain of the active disturbance rejection controller and the parameters of the extended state observer in real time based on the current valve core displacement, valve core velocity, and valve core acceleration signals of the voice coil motor driven valve. The neural network model adopts a three-layer feedforward neural network structure. The input layer takes the valve core displacement, valve core velocity, and valve core acceleration signals as input, the hidden layers use the ReLU activation function, and the output layer outputs the system nonlinear estimate. During the training process, the neural network model uses an error backpropagation algorithm to optimize the model parameters. S4. The extended state observer outputs the total disturbance estimate of the current system based on the operating parameters obtained in S3. The extended state observer estimates the disturbance, nonlinear dynamic characteristics, and unmodeled error of the voice coil motor driven valve. The output of the extended state observer is the estimated total disturbance, defined as follows: d^(t) = Lo·(yy^); Where d^(t) is the total disturbance estimate, Lo is the observation gain matrix, i.e. the operating parameters of the extended state observer, y is the actual displacement of the valve core, and y^ is the estimated valve core displacement; by adjusting the value of Lo, the response speed of the extended state observer to the disturbance is controlled, thereby optimizing the control effect. S5. Based on the real-time error signal, the total disturbance estimate, and the feedback gain parameter, the controller module uses an active disturbance rejection controller to dynamically adjust the real-time error signal and outputs control commands to drive the voice coil motor-driven valve; the active disturbance rejection controller adjusts the real-time error signal using a nonlinear feedback formula based on the disturbance estimation results from the extended state observer. u(t)=-K·e(t)-Kd·e(t)+d^(t); Where u(t) is the control command, e(t) is the real-time error signal, d^(t) is the disturbance estimate, and K and Kd are the feedback gain coefficients.

2. The voice coil motor driven valve control method based on neural network active disturbance rejection control according to claim 1, characterized in that, The smooth target displacement control curve is a curve showing the relationship between time and valve core target displacement, including acceleration phase, constant speed phase, and deceleration phase. During the acceleration phase, the valve core speed accelerates smoothly from zero, and the valve core displacement increases rapidly. During the constant speed phase, the valve core speed remains constant, and the valve core displacement gradually reaches the vicinity of the target displacement. During the deceleration phase, the valve core speed decreases smoothly, and the valve core decelerates and eventually reaches the target displacement, avoiding valve core overshoot and oscillation.

3. The voice coil motor driven valve control method based on neural network active disturbance rejection control according to claim 1 or 2, characterized in that, The smooth target displacement control curve is expressed by the following formula: S(t) = A / 2(1-cos(2πt / T)); Where A is the amplitude of the target control signal, and T is the time of the acceleration and deceleration phases. Smooth control of the target control signal is achieved by adjusting A and T.

4. The voice coil motor driven valve control method based on neural network active disturbance rejection control according to claim 1, characterized in that, In S2, the valve core displacement signal of the controlled voice coil motor driven valve is obtained through the LVDT sensor. The obtained valve core displacement signal is then processed by adaptive Kalman filtering and sliding mean filtering to reduce noise and transient interference and ensure high-precision transmission of the feedback signal.

5. A voice coil motor driven valve control system implementing the method of any one of claims 1-4, characterized in that... include: The target control signal preprocessing module is used to preprocess the target control signal to obtain a smooth target displacement control curve; A valve core displacement sensor is used to acquire the valve core displacement of the valve driven by the controlled voice coil motor in real time. The real-time error calculation unit calculates the real-time error signal by subtracting the valve core displacement signal from the smoothed target displacement input signal. The controller module uses a trained neural network model to obtain the feedback gain parameters of the active disturbance rejection controller and the operating parameters of the extended state observer. The extended state observer outputs the total disturbance estimate of the current system based on the operating parameters obtained from the controller module. The active disturbance rejection controller dynamically adjusts the real-time error signal based on the real-time error signal, the total disturbance estimate, and the feedback gain parameter, and outputs control commands to drive the voice coil motor drive valve.

Images