A hydraulic lifting system synchronization control system and method suitable for complex situations

Abstract

The application relates to the technical field of control systems, in particular to a hydraulic lifting system synchronous control system and method suitable for complex conditions, which comprises a period error monitoring module, which collects the reference position of a hydraulic cylinder and the position fed back by a displacement sensor at multiple time points in a period based on a preset lifting task period, and makes a difference between the reference position collected at each time point and the position. In the application, a synchronous error time domain curve is collected and constructed, periodic deviations of the hydraulic lifting system caused by mechanical characteristics and load changes in repetitive tasks can be captured, and a correction feedforward control sequence is generated based on the control input of the last period and the error curve. The processing logic changes the control system from pure real-time feedback to a prediction-compensation mode with learning ability, which does not depend on a perfect system model, but through continuous trial and error learning, historical error information is converted into pre-compensation for future control instructions.

Description

Technical Field

[0001] This invention relates to the field of control system technology, and in particular to a synchronous control system and method for hydraulic lifting systems applicable to complex situations. Background Technology

[0002] Control system technology is a core engineering discipline that studies how to operate and influence a dynamic system to achieve a desired output or state. This field manages, instructs, or regulates the behavior of a controlled object by constructing controllers and applying specific input signals to it.

[0003] Traditional control systems, when handling tasks like hydraulic lifting synchronization, suffer from a core deficiency in their inherent responsiveness and limited ability to handle unstructured uncertainties. These systems operate based on a fixed mathematical model and real-time error feedback, with control commands determined entirely by the system's deviation at the current moment. This purely reactive approach makes it difficult to anticipate and proactively address disturbances that recur in each task cycle, such as periodic synchronization errors caused by mechanical friction at specific locations, hydraulic circuit asymmetry, or load eccentricity. Consequently, even with well-tuned controller parameters, the system repeats the same errors in each cycle, resulting in persistent periodic tracking errors. Therefore, improvements are necessary. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of the existing technology and to propose a synchronous control system and method for hydraulic lifting systems applicable to complex situations.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: a synchronous control system for a hydraulic lifting system suitable for complex situations includes:

[0006] The cycle error monitoring module, based on the preset lifting task cycle, collects the reference position of the hydraulic cylinder and the position fed back by the displacement sensor at multiple time points within the cycle. It calculates the difference between the reference position and the position collected at each time point, and arranges all the differences in chronological order to obtain the synchronization error time domain curve.

[0007] The feedforward instruction iteration module, based on the control input of the previous cycle and the synchronization error time-domain curve, multiplies each data point in the synchronization error time-domain curve by the learning gain coefficient to generate an error correction amount, adds the error correction amount to the control input of the previous cycle at the corresponding time point to generate an update control signal, and then filters out signal components in the update control signal that are higher than a preset frequency threshold to establish a corrected feedforward control sequence.

[0008] Preferably, the system further includes:

[0009] The system state observation module, based on the modified feedforward control sequence, the system state space model, and the hydraulic oil flow pulsation characteristics, predicts the current state according to the system state space model and the state estimate of the previous moment. Then, it fuses the measured values ​​from the displacement sensor and the pressure sensor, calculates the difference between the predicted value and the measured value, and uses the difference to correct the predicted state to obtain the optimal state estimation vector.

[0010] The real-time feedback compensation module, based on the optimal state estimation vector and the modified feedforward control sequence, sets a state error weight matrix and a control input weight matrix, obtains the optimal feedback gain by solving the algebraic Riccati equation, performs matrix multiplication between the optimal feedback gain and the optimal state estimation vector to generate a feedback compensation amount, and adds the feedback compensation amount to the modified feedforward control sequence to generate a composite control command to drive the servo valve.

[0011] Preferably, the periodic error monitoring module includes:

[0012] The position deviation acquisition submodule, based on a preset lifting task cycle, synchronously records the reference position of the hydraulic cylinder and the position fed back by the displacement sensor at multiple time points within the cycle. The instantaneous position deviation sequence is obtained by subtracting the corresponding position from the reference position at each time point.

[0013] The error curve construction submodule, based on the instantaneous position deviation sequence, pairs each deviation value in the sequence with the acquisition time point, and sorts all paired data according to the chronological order to generate a synchronization error time domain curve.

[0014] Preferably, the feedforward instruction iteration module includes:

[0015] The error signal weighting submodule, based on the control input and synchronization error time-domain curve of the previous cycle, multiplies each data point in the synchronization error time-domain curve by a preset learning gain coefficient to generate a time-domain error correction signal.

[0016] The feedforward instruction update submodule, based on the time-domain error correction signal and the control input of the previous cycle, adds the values ​​of the time-domain error correction signal and the control input of the previous cycle at the same time point point to obtain the unfiltered feedforward instruction.

[0017] The high-frequency noise filtering submodule, based on the unfiltered feedforward instruction, performs a Fourier transform on the instruction sequence to the frequency domain, sets the amplitude of all frequency components in the frequency domain that are higher than the preset frequency threshold to zero, and then performs an inverse Fourier transform to convert back to the time domain, thereby establishing a corrected feedforward control sequence.

[0018] Preferably, the system status observation module includes:

[0019] The state prediction submodule, based on the modified feedforward control sequence, the system state space model, and the hydraulic oil flow pulsation characteristics, substitutes the state estimate of the previous moment into the state equation of the system state space model, and combines the modified feedforward control sequence as input to calculate the prior state prediction value at the current moment.

[0020] The observation gain calculation submodule, based on the prior state prediction value, and combined with the process noise covariance matrix and the measurement noise covariance matrix, obtains the state observation gain matrix by solving.

[0021] Preferably, the system status observation module further includes:

[0022] The state correction submodule, based on the prior state prediction value and the state observation gain matrix, subtracts the corresponding terms in the prior state prediction value from the measured values ​​from the displacement sensor and pressure sensor to obtain the measurement residual, multiplies the measurement residual with the state observation gain matrix, and adds the product result to the prior state prediction value to obtain the optimal state estimation vector.

[0023] Preferably, the real-time feedback compensation module includes:

[0024] The feedback gain solution submodule, based on the optimal state estimation vector and the modified feedforward control sequence, sets the state error weight matrix and the control input weight matrix, and obtains the optimal feedback gain matrix by solving the algebraic Riccati equation associated with these two weight matrices;

[0025] The compensation signal generation submodule generates a feedback compensation signal by performing matrix multiplication on the negative value of the optimal feedback gain matrix and the optimal state estimation vector, based on the optimal state estimation vector and the optimal feedback gain matrix.

[0026] Preferably, the real-time feedback compensation module further includes:

[0027] The control command fusion submodule, based on the feedback compensation signal and the modified feedforward control sequence, adds the values ​​of the feedback compensation signal and the modified feedforward control sequence at each time point to generate a composite control command that drives the servo valve.

[0028] This invention also provides a synchronous control method for hydraulic lifting systems applicable to complex situations, comprising the following steps:

[0029] Based on the preset lifting task cycle, the reference position of the hydraulic cylinder and the position fed back by the displacement sensor are collected at multiple time points within the cycle. The difference between the reference position collected at each time point and the position is calculated, and all the differences are arranged in chronological order to obtain the synchronization error time domain curve.

[0030] Based on the control input of the previous cycle and the synchronization error time-domain curve, each data point in the synchronization error time-domain curve is multiplied by the learning gain coefficient to generate an error correction amount. The error correction amount is added to the control input of the previous cycle at the corresponding time point to generate an update control signal. Then, signal components higher than the preset frequency threshold in the update control signal are filtered out to establish a corrected feedforward control sequence.

[0031] Based on the modified feedforward control sequence, the system state space model, and the hydraulic oil flow pulsation characteristics, the current state is predicted according to the system state space model and the state estimate of the previous moment. Then, the measured values ​​from the displacement sensor and the pressure sensor are fused, the difference between the predicted value and the measured value is calculated, and the predicted state is corrected using the difference to obtain the optimal state estimation vector.

[0032] Based on the optimal state estimation vector and the modified feedforward control sequence, a state error weight matrix and a control input weight matrix are set. The optimal feedback gain is obtained by solving the algebraic Riccati equation. The optimal feedback gain and the optimal state estimation vector are multiplied by a matrix to generate a feedback compensation amount. The feedback compensation amount is added to the modified feedforward control sequence to generate a composite control command that drives the servo valve.

[0033] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0034] In this invention, by acquiring and constructing a synchronization error time-domain curve, the periodic deviations of a hydraulic lifting system caused by mechanical characteristics and load changes during repetitive tasks can be captured. Then, a corrected feedforward control sequence is generated based on the control input of the previous cycle and this error curve. This processing logic transforms the control system from a purely real-time feedback model to a predictive-compensation mode with learning capabilities. It does not rely on a perfect system model but learns through continuous trial and error, transforming historical error information into pre-compensation for future control commands. This proactively suppresses repeatable disturbances in each subsequent task cycle, gradually reducing the synchronization error and achieving near-tracking of the target trajectory. Based on this, a system state-space model incorporating hydraulic oil flow pulsation characteristics is established, and measurements from displacement and pressure sensors are integrated to correct and obtain the optimal state estimation vector, overcoming the interference of sensor noise and internal random disturbances on state observation. The feedback compensation amount generated based on the optimal state estimation vector is added to the learned and corrected feedforward control sequence to form a composite control command. This allows the system to eliminate periodic errors through feedforward learning and suppress non-periodic and random disturbances in real time through optimal feedback. While ensuring high synchronization accuracy, it also takes into account the consumption of control energy and the stability of system operation, thus improving the control robustness and overall performance of the hydraulic lifting system under complex working conditions. Attached Figure Description

[0035] Figure 1 This is a system flowchart of the present invention. Detailed Implementation

[0036] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0037] Please see Figure 1 This invention provides a technical solution: a synchronous control system for a hydraulic lifting system suitable for complex situations, comprising:

[0038] The cycle error monitoring module, based on the preset lifting task cycle, collects the reference position of the hydraulic cylinder and the position fed back by the displacement sensor at multiple time points within the cycle. It calculates the difference between the reference position and the position collected at each time point, and arranges all the differences in chronological order to obtain the synchronization error time domain curve.

[0039] The feedforward instruction iteration module, based on the control input and synchronization error time-domain curve of the previous cycle, multiplies each data point in the synchronization error time-domain curve by the learning gain coefficient to generate an error correction amount. The error correction amount is added to the control input of the previous cycle at the corresponding time point to generate an update control signal. Then, the signal components in the update control signal that are higher than the preset frequency threshold are filtered out to establish a corrected feedforward control sequence.

[0040] The system state observation module, based on the modified feedforward control sequence, the system state space model, and the hydraulic oil flow pulsation characteristics, predicts the current state according to the system state space model and the state estimate of the previous moment. Then, it fuses the measured values ​​from the displacement sensor and the pressure sensor, calculates the difference between the predicted value and the measured value, and uses the difference to correct the predicted state to obtain the optimal state estimation vector.

[0041] The real-time feedback compensation module, based on the optimal state estimation vector and the modified feedforward control sequence, sets the state error weight matrix and the control input weight matrix, obtains the optimal feedback gain by solving the algebraic Riccati equation, performs matrix multiplication between the optimal feedback gain and the optimal state estimation vector to generate the feedback compensation amount, and adds the feedback compensation amount to the modified feedforward control sequence to generate a composite control command to drive the servo valve.

[0042] The periodic error monitoring module includes:

[0043] The position deviation acquisition submodule, based on a preset lifting task cycle, synchronously records the reference position of the hydraulic cylinder and the position fed back by the displacement sensor at multiple time points within the cycle. The instantaneous position deviation sequence is obtained by subtracting the corresponding position from the reference position at each time point.

[0044] The error curve construction submodule, based on the instantaneous position deviation sequence, pairs each deviation value in the sequence with the acquisition time point, and sorts all paired data according to the chronological order to generate a synchronization error time domain curve.

[0045] Specifically, based on a preset lifting and lowering task cycle, for example, a complete lifting and lowering process is defined as 10 seconds. Within this 10-second cycle, at a fixed sampling frequency, such as 100 times per second (i.e., every 0.01 seconds), the reference position of the hydraulic cylinder and the actual position fed back by the displacement sensor are recorded synchronously at that moment. The reference position is an ideal trajectory point planned according to the preset lifting speed and acceleration, while the actual position is measured in real time by the displacement sensor (such as a magnetostrictive displacement sensor or a wire encoder) installed on the hydraulic cylinder. Synchronous recording is achieved by triggering data acquisition commands through a unified system clock, ensuring that at each sampling time point... (in , This represents the total number of sampling points within the period, in this example... Both simultaneously acquired the reference position. and actual feedback location Next, for each sampling time point Calculate the instantaneous position deviation For example, in If the reference position is 100.5mm and the sensor feedback position is 100.2mm, then the instantaneous position deviation is: All these calculated instantaneous position deviation values ​​are arranged in chronological order of their occurrence to form a list containing... The time series of each deviation value is used to obtain the instantaneous position deviation sequence.

[0046] Based on the instantaneous position deviation sequence, which is a series of deviation values ​​obtained in the previous step arranged in chronological order, for example... ,in At a certain point in time The instantaneous position deviation is used to construct the synchronization error time-domain curve, and each deviation value in this sequence is... Its corresponding collection time point Explicit pairing is performed to form a series of data point pairs. For example, if the first three values ​​of the instantaneous position deviation sequence are 0.1mm, 0.3mm, and 0.2mm, and the corresponding acquisition time points are 0.01s, 0.02s, and 0.03s, then the resulting data point pairs are (0.01s, 0.1mm), (0.02s, 0.3mm), and (0.03s, 0.2mm). Since the instantaneous position deviation sequence itself is acquired and arranged in chronological order, these data point pairs are logically arranged according to chronological order. In actual construction, it is crucial to ensure the strict preservation of this chronological order and to pair all these data points... When combined, they form a discrete time series data. This ordered sequence, consisting of timestamps and their corresponding instantaneous position deviations, can intuitively represent the complete situation of the synchronization error changing over time within a task cycle, generating a synchronization error time-domain curve.

[0047] The feedforward instruction iteration module includes:

[0048] The error signal weighting submodule, based on the control input and synchronization error time-domain curve of the previous cycle, multiplies each data point in the synchronization error time-domain curve by a preset learning gain coefficient to generate a time-domain error correction signal.

[0049] The feedforward instruction update submodule, based on the time-domain error correction signal and the control input of the previous cycle, adds the values ​​of the time-domain error correction signal and the control input of the previous cycle at the same time point to obtain the unfiltered feedforward instruction.

[0050] The high-frequency noise filtering submodule, based on the unfiltered feedforward command, performs a Fourier transform on the command sequence to convert it to the frequency domain, sets the amplitude of all frequency components in the frequency domain that are higher than the preset frequency threshold to zero, and then performs an inverse Fourier transform to convert it back to the time domain, thus establishing a corrected feedforward control sequence.

[0051] Specifically, based on the control input of the previous cycle, this is a record of every time point in the previous complete lifting and lowering task cycle. Corresponding control command value The time series, and the synchronization error time-domain curve calculated for the current period, which represents a series of data point pairs. ,in At a certain point in time The core of handling synchronization errors lies in processing each data point in the synchronization error time-domain curve. Multiply by a preset learning gain coefficient This learning gain coefficient This is a key parameter, and its setting directly affects the system's learning speed and stability. The value of is usually determined experimentally. Initially, a small value, such as 0.2, can be set based on experience. Then, the value is adjusted by observing the convergence of the synchronization error after several cycles. If the error decreases slowly, the value can be increased appropriately. For example, if the value increases by 0.1 each time, and oscillations or divergences occur, then the value is decreased. This continues until an optimal value is found that can quickly reduce errors without causing system instability; for example, it is ultimately determined to be... For each data point in the synchronization error time-domain curve Calculate the corresponding error correction amount for each. For example, if at a certain point in time Synchronization error The error unit is 0.5mm (the unit of error here needs to be adjusted through subsequent processing or the gain itself needs to include a unit conversion factor to match the unit of the control input; for example, this has already been processed or the gain is dimensionless, and the correction amount directly applies to the value of the control signal). Then the error correction amount at that point is The error correction values ​​calculated at all time points are arranged in chronological order to generate a time-domain error correction signal.

[0052] Based on the time-domain error correction signal, this is a signal generated from various time points. Error correction amount The time series, along with the control input from the previous cycle, is composed of various time points. Previous cycle control command value To update the feedforward command, the time-domain error correction signal is added point-by-point to the control input value of the previous cycle at the same time point, forming a time series. Specifically, for each discrete time point within the cycle... (from arrive ,in (This is the total number of sampling points within the period), extract the time-domain error correction signal value corresponding to that time point. and the control input value of the previous cycle Then add these two values ​​together to get the current time point. Unfiltered feedforward command value ,Right now For example, at a point in time If the control input of the previous cycle The corresponding time-domain error correction signal value is 5.2 (units such as volts or milliamperes, depending on the servo valve drive requirements). If the value is 0.325 (consistent with or converted to the control input unit), then the unfiltered feedforward command value at that time point is... For all throughout the entire task cycle The same addition operation is performed at each time point to obtain a result consisting of... indivual The complete sequence of values ​​is used to obtain the unfiltered feedforward instruction.

[0053] Based on the unfiltered feedforward instruction, this is a collection of time points. Updated control command values The time series is first processed by performing a Fourier transform on the instruction sequence, typically using the Fast Fourier Transform (FFT) algorithm, to convert it from the time domain to the frequency domain, obtaining the amplitude and phase information corresponding to each frequency component, i.e., the frequency domain representation. Then, all frequencies in the frequency domain that are higher than a preset frequency threshold are... The amplitude of the frequency component is set to zero; this preset frequency threshold... The setting is based on the dynamic response characteristics of the hydraulic system and the expected operating frequency range. For example, if the hydraulic system can no longer effectively follow control signal changes above 30Hz, and the main operating command frequency is concentrated in the 0-20Hz range, then... The frequency is set to 25Hz. This value can be determined by analyzing the spectrum of the system's open-loop step response to find the effective bandwidth of the system, or by setting an empirical value that is slightly higher than the highest frequency of the desired signal but lower than the known high-frequency noise onset frequency. For example, if the system's desired response frequency does not exceed 20Hz, and the sensor noise is mainly above 40Hz, then... It can be set to To retain most of the useful signal and filter out some noise, for all frequencies. The corresponding frequency domain amplitude Set it to 0 (or directly represent the complex number) (Set to 0), and finally, perform an inverse Fourier transform on the modified frequency domain signal. Usually, the inverse fast Fourier transform (IFFT) algorithm is used to transform it back to the time domain and establish the corrected feedforward control sequence.

[0054] The system status observation module includes:

[0055] The state prediction submodule, based on the modified feedforward control sequence, the system state space model, and the hydraulic oil flow pulsation characteristics, substitutes the state estimate of the previous moment into the state equation of the system state space model, and combines the modified feedforward control sequence as input to calculate the prior state prediction value at the current moment.

[0056] The observation gain calculation submodule, based on the prior state prediction value, combines the process noise covariance matrix and the measurement noise covariance matrix to obtain the state observation gain matrix by solving.

[0057] The state correction submodule, based on the prior state prediction value and the state observation gain matrix, subtracts the corresponding terms from the measured values ​​from the displacement sensor and pressure sensor and the prior state prediction value to obtain the measurement residual. The measurement residual is multiplied by the state observation gain matrix, and the product result is added to the prior state prediction value to obtain the optimal state estimation vector.

[0058] Specifically, based on the modified feedforward control sequence The system state-space model and the hydraulic oil flow pulsation characteristics are typically obtained through mechanistic modeling or system identification of the hydraulic system. For example, the state-space model of a hydraulic cylinder system can be represented in discrete-time form. ,in yes The system state vector at any given time includes the position, velocity, and intracavity pressure of the hydraulic cylinder. It is a control input. It is the state transition matrix. It is the input matrix. It is process noise, for example, if the state vector is ,but It might be a matrix, It is The specific values ​​of these matrices are obtained by linearizing and discretizing the dynamic equations of the hydraulic system (such as the flow continuity equation and Newton's second law), or by estimating parameters from experimental data, such as by using the least squares method to identify input and output data. The hydraulic oil flow pulsation characteristics mainly refer to the flow non-uniformity caused by components such as hydraulic pumps and servo valves. This characteristic can be modeled as a response to system input... A disturbance, or directly manifested in process noise. In the statistical characteristics, or through more refined models in Nonlinear terms or time-varying parameters are introduced into the matrix or state equation to represent this. In the prediction step, the previous time step is used as the basis for the prediction. Optimal state estimate Substituting the state transition part of the discrete state equation above, and combining it with the current time... The corresponding value in the corrected feedforward control sequence As a control input, if the hydraulic oil flow pulsation characteristics are modeled as a disturbance to the input... The input item is then corrected to ,in The characteristics (such as amplitude and frequency) are estimated based on known pulsation characteristics; for example, if the flow pulsation is mainly caused by the number of pump plungers. and rotational speed The fundamental frequency is determined by [the specific frequency]. ,but It may contain this frequency and its harmonic components, if the pulsation characteristics are already incorporated. and If the model is accurate or considered as part of the process noise, then directly use Therefore, the state prediction equation is: ,in This is the predicted value of the prior state at the current moment.

[0059] Based on prior state prediction values Combined with the process noise covariance matrix and measurement noise covariance matrix Perform state observation gain matrix The calculation of the process noise covariance matrix. This reflects the uncertainty of the system model and the impact of unmodeled dynamics. Its diagonal elements represent the variance of the process noise of the corresponding state variables, and the off-diagonal elements represent the correlation between the process noise of different state variables. Settings are typically based on experience and debugging; for example, for a state containing location... and speed In a system where velocity is considered to be significantly affected by unmodeled friction and other factors, while position is primarily obtained by velocity integral, then... The variance term corresponding to the velocity It will be set to be more than the variance term at the corresponding position. Larger, for example, if the range of speed variation is The random fluctuations within the range, the variance of which can be initially set as ,but The process noise of positioning may be relatively small, such as Measure the noise covariance matrix This reflects the statistical characteristics of the sensor's measurement error. Its diagonal element is the variance of the measurement noise of each sensor, which can usually be obtained from the sensor's datasheet or through statistical analysis of a large amount of data collected by the sensor under static conditions. For example, if the accuracy of a displacement sensor is... For example, the error follows a normal distribution and If this range is covered, its variance can be set as... If there is also a pressure sensor, its noise variance is calculated and filled in similarly. The corresponding diagonal position, for example, the standard deviation of the pressure sensor measurement noise is The corresponding variance is State observation gain matrix (i.e., Kalman gain) is obtained by solving the following standard Kalman filter equations: First, calculate the prior error covariance matrix. Then calculate the state observation gain matrix. ,in It is the observation matrix, which maps the state vector to the measurement vector. Its form is determined by the sensor configuration and the measured state variables. For example, if position and pressure are being measured directly, then... The corresponding rows of the matrix select the position and pressure components in the state vector, and the state observation gain matrix is ​​obtained by calculating using the above formula.

[0060] Based on prior state prediction values The state observation gain matrix calculated in the previous step First, the current time is obtained from the displacement sensor and the pressure sensor. The actual measured value is denoted as the measurement vector. ,For example, Next, these actual measured values With observation matrix Predicted values ​​from prior states Extracted corresponding predicted measurement values By performing subtraction, the measurement residual (or new information) is obtained. For example, if the predicted position is 100.2 mm and the predicted pressure is 5.0 MPa in the prior state prediction, while the actual sensor measurement is 100.1 mm for position and 5.1 MPa for pressure, and the observation matrix... If the matrix is ​​an identity matrix (e.g., the state is directly measurable), then the measurement residual is... Then, the calculated measurement residuals With state observation gain matrix Perform matrix multiplication to obtain the correction amount. This correction amount represents the magnitude of the adjustment to the prior state prediction based on the actual measurement. Finally, this correction amount is compared with the prior state prediction value. Adding, that is Thus, the current time is obtained. Optimal state estimation vector Simultaneously, update the posterior error covariance matrix. ,in It is the identity matrix, and the optimal state estimation vector is obtained.

[0061] Specifically, firstly, the feedforward command iteration module achieves adaptive compensation for system repeatability errors through the core idea of ​​iterative learning control. It utilizes the synchronization error of the previous cycle to correct the feedforward control command of the current cycle point by point. This mechanism effectively addresses periodic disturbances in hydraulic systems, such as nonlinear friction and changes in oil parameters, which are difficult to model precisely. This allows the system to approximate the ideal trajectory through continuous "practice," thereby improving the synchronization control accuracy cycle by cycle. Simultaneously, by filtering high-frequency noise from the updated command in the frequency domain, the stability of the iterative learning process is ensured, avoiding system oscillations caused by noise amplification. This results in a final corrected feedforward control sequence that is both accurate and smooth, laying the foundation for achieving high dynamic response and stable operation.

[0062] Secondly, the system state observation module, by introducing a Kalman filter, provides accurate and complete real-time state information for the system, which is crucial for dealing with complex aperiodic disturbances. This module optimally fuses predicted values ​​based on the system model with actual measurements from multiple sensors, such as displacement and pressure. This not only effectively filters out measurement noise from the sensors themselves but also accurately estimates state variables that are crucial to control but cannot be directly measured, such as velocity. In particular, by considering characteristics such as hydraulic oil flow pulsation in the model, the accuracy of state prediction is further improved. The final optimal state estimation vector provides a high-quality, low-noise basis for subsequent real-time feedback compensation, enabling the feedback controller to respond more sensitively and accurately to sudden disturbances and model uncertainties, thereby enhancing the robustness and dynamic stability of the entire control system.

[0063] The real-time feedback compensation module includes:

[0064] The feedback gain solution submodule, based on the optimal state estimation vector and the modified feedforward control sequence, sets the state error weight matrix and the control input weight matrix, and obtains the optimal feedback gain matrix by solving the algebraic Riccati equation associated with these two weight matrices.

[0065] The compensation signal generation submodule generates a feedback compensation signal by performing matrix multiplication on the negative value of the optimal feedback gain matrix and the optimal state estimation vector, based on the optimal state estimation vector and the optimal feedback gain matrix.

[0066] The control command fusion submodule, based on the feedback compensation signal and the modified feedforward control sequence, adds the values ​​of the feedback compensation signal and the modified feedforward control sequence at each time point to generate a composite control command that drives the servo valve.

[0067] Specifically, based on the optimal state estimation vector and modified feedforward control sequence First, the state error weight matrix needs to be set. and control input weight matrix These two matrices are core parameters in the design of a linear quadratic regulator (LQR). They determine the relative importance the control system places on state error and control energy consumption. (State error weight matrix) It is usually a positive semi-definite symmetric matrix with diagonal elements Represents the state variable The penalty for deviation from the expected value (usually zero, which in tracking problems is the error of deviating from the reference trajectory). The larger the size, the harder the controller will work. To keep the position near the desired value, for example, in the position control of a hydraulic cylinder, if the state vector contains position error. and speed error Those who are more concerned with positional accuracy may... It can be set as a diagonal matrix ,in Much larger For example, if the allowable standard deviation of position error is 0.1 mm and the standard deviation of velocity error is 1 mm / s, then according to Bryson's rule, we can take... and Control input weight matrix It is usually a positive definite symmetric matrix with diagonal elements Indicates control input Punishment based on size The larger the value, the smaller the amplitude of the control signal generated by the controller, thereby saving control energy or avoiding actuator saturation. For example, if the control input is a servo valve current, its allowable range is... ,but Can be set to These weight values ​​typically need to be repeatedly tested and adjusted based on simulation or actual system performance to achieve satisfactory control performance. and Subsequently, by solving the state-space model of the discrete system... , Matrix and these two weight matrices , Related discrete algebraic Riccati equations: To calculate the steady-state solution Once obtained Optimal feedback gain matrix It can be calculated as To obtain the optimal feedback gain matrix.

[0068] Based on the optimal state estimation vector The optimal feedback gain matrix obtained in the previous step by solving the algebraic Riccati equation. To perform the specific matrix multiplication operation, first, the optimal feedback gain matrix... Taking the negative value, we get This negative sign indicates that the feedback control law is usually negative feedback, meaning the control action is opposite in direction to the state deviation. Then, this negative optimal feedback gain matrix is... The optimal state estimation vector at the current time Perform matrix multiplication, that is, calculate ,in It is a column vector whose dimensions are the same as the number of system state variables (for example, if the states are position and velocity, then it is...). (vector), optimal feedback gain matrix The dimension is the number of control inputs multiplied by the number of state variables (for example, if it's a single-input, two-state system, then it's...). Matrix multiplication follows standard matrix multiplication rules. Each line and The corresponding elements are multiplied and then summed to obtain the feedback compensation signal vector. The corresponding element in, for example, if and Then the feedback compensation signal ,this It involves calculating the control adjustment amount used to correct system behavior based on the current system state deviation, and then generating a feedback compensation signal.

[0069] Based on feedback compensation signal and the established modified feedforward control sequence At each discrete control time point The values ​​of the two signals are algebraically added together to correct the feedforward control sequence. It is obtained through pre-calculation or iterative learning based on the desired trajectory and system model, aiming to drive the system in an open-loop manner to roughly follow the target, while providing feedback compensation signals. It is calculated using the optimal feedback gain based on the deviation between the real-time optimal state estimation vector and the desired state (here, for example, the desired state is zero, meaning LQR is designed for regulation problems, or the state vector itself represents the error state). It is used to correct deviations and suppress disturbances in real time. The fusion process is as follows: for the time point of the current control cycle... Extract the corresponding corrected feedforward control sequence value and feedback compensation signal value Then, they are added together to obtain the composite control command value at that time point. For example, at a point in time If the value of the corrected feedforward control sequence is 5.525 (unit, e.g., volts), and the calculated feedback compensation signal is -0.125 (unit as before), then the composite control command at that time point is... (Units as before) This addition operation is performed in each control sampling period to form a time-varying composite control command sequence. This sequence combines the speed of feedforward control with the robustness and accuracy of feedback control to generate composite control commands that drive the servo valve.

Claims

1. A synchronous control system for a hydraulic lifting system suitable for complex situations, characterized in that, The system includes: The cycle error monitoring module, based on the preset lifting task cycle, collects the reference position of the hydraulic cylinder and the position fed back by the displacement sensor at multiple time points within the cycle. It calculates the difference between the reference position and the position collected at each time point, and arranges all the differences in chronological order to obtain the synchronization error time domain curve. The feedforward instruction iteration module, based on the control input of the previous cycle and the synchronization error time-domain curve, multiplies each data point in the synchronization error time-domain curve by the learning gain coefficient to generate an error correction amount, adds the error correction amount to the control input of the previous cycle at the corresponding time point to generate an update control signal, and then filters out signal components in the update control signal that are higher than the preset frequency threshold to establish a corrected feedforward control sequence. The system state observation module, based on the modified feedforward control sequence, the system state space model, and the hydraulic oil flow pulsation characteristics, predicts the current state according to the system state space model and the state estimate of the previous moment. Then, it fuses the measured values ​​from the displacement sensor and the pressure sensor, calculates the difference between the predicted value and the measured value, and uses the difference to correct the predicted state to obtain the optimal state estimation vector. The real-time feedback compensation module, based on the optimal state estimation vector and the modified feedforward control sequence, sets the state error weight matrix and the control input weight matrix, obtains the optimal feedback gain by solving the algebraic Riccati equation, performs matrix multiplication on the optimal feedback gain and the optimal state estimation vector to generate a feedback compensation amount, and adds the feedback compensation amount to the modified feedforward control sequence to generate a composite control command to drive the servo valve. The system status observation module includes: The state prediction submodule, based on the modified feedforward control sequence, the system state space model, and the hydraulic oil flow pulsation characteristics, substitutes the state estimate of the previous moment into the state equation of the system state space model, and combines the modified feedforward control sequence as input to calculate the prior state prediction value at the current moment. The observation gain calculation submodule, based on the prior state prediction value, and combined with the process noise covariance matrix and the measurement noise covariance matrix, obtains the state observation gain matrix by solving. The system status observation module also includes: The state correction submodule, based on the prior state prediction value and the state observation gain matrix, subtracts the corresponding terms in the prior state prediction value from the measured values ​​from the displacement sensor and pressure sensor to obtain the measurement residual, multiplies the measurement residual with the state observation gain matrix, and adds the product result to the prior state prediction value to obtain the optimal state estimation vector.

2. The synchronous control system for a hydraulic lifting system applicable to complex situations according to claim 1, characterized in that, The periodic error monitoring module includes: The position deviation acquisition submodule, based on a preset lifting task cycle, synchronously records the reference position of the hydraulic cylinder and the position fed back by the displacement sensor at multiple time points within the cycle. The instantaneous position deviation sequence is obtained by subtracting the corresponding position from the reference position at each time point. The error curve construction submodule, based on the instantaneous position deviation sequence, pairs each deviation value in the sequence with the acquisition time point, and sorts all paired data according to the chronological order to generate a synchronization error time domain curve.

3. The synchronous control system for a hydraulic lifting system applicable to complex situations according to claim 1, characterized in that, The feedforward instruction iteration module includes: The error signal weighting submodule, based on the control input and synchronization error time-domain curve of the previous cycle, multiplies each data point in the synchronization error time-domain curve by a preset learning gain coefficient to generate a time-domain error correction signal. The feedforward instruction update submodule, based on the time-domain error correction signal and the control input of the previous cycle, adds the values ​​of the time-domain error correction signal and the control input of the previous cycle at the same time point point to obtain the unfiltered feedforward instruction. The high-frequency noise filtering submodule, based on the unfiltered feedforward instruction, performs a Fourier transform on the instruction sequence to the frequency domain, sets the amplitude of all frequency components in the frequency domain that are higher than the preset frequency threshold to zero, and then performs an inverse Fourier transform to convert back to the time domain, thereby establishing a corrected feedforward control sequence.

4. The synchronous control system for a hydraulic lifting system applicable to complex situations according to claim 1, characterized in that, The real-time feedback compensation module includes: The feedback gain solution submodule, based on the optimal state estimation vector and the modified feedforward control sequence, sets the state error weight matrix and the control input weight matrix, and obtains the optimal feedback gain matrix by solving the algebraic Riccati equation associated with these two weight matrices; The compensation signal generation submodule generates a feedback compensation signal by performing matrix multiplication on the negative value of the optimal feedback gain matrix and the optimal state estimation vector, based on the optimal state estimation vector and the optimal feedback gain matrix.

5. The synchronous control system for a hydraulic lifting system applicable to complex situations according to claim 4, characterized in that, The real-time feedback compensation module also includes: The control command fusion submodule, based on the feedback compensation signal and the modified feedforward control sequence, adds the values ​​of the feedback compensation signal and the modified feedforward control sequence at each time point to generate a composite control command that drives the servo valve.

6. The synchronous control method for a hydraulic lifting system applicable to complex situations, as described in any one of claims 1-5, is characterized in that... Includes the following steps: Based on the preset lifting task cycle, the reference position of the hydraulic cylinder and the position fed back by the displacement sensor are collected at multiple time points within the cycle. The difference between the reference position collected at each time point and the position is calculated, and all the differences are arranged in chronological order to obtain the synchronization error time domain curve. Based on the control input of the previous cycle and the synchronization error time-domain curve, each data point in the synchronization error time-domain curve is multiplied by the learning gain coefficient to generate an error correction amount. The error correction amount is added to the control input of the previous cycle at the corresponding time point to generate an update control signal. Then, signal components higher than the preset frequency threshold in the update control signal are filtered out to establish a corrected feedforward control sequence. Based on the modified feedforward control sequence, the system state space model, and the hydraulic oil flow pulsation characteristics, the current state is predicted according to the system state space model and the state estimate of the previous moment. Then, the measured values ​​from the displacement sensor and the pressure sensor are fused, the difference between the predicted value and the measured value is calculated, and the predicted state is corrected using the difference to obtain the optimal state estimation vector. Based on the optimal state estimation vector and the modified feedforward control sequence, a state error weight matrix and a control input weight matrix are set. The optimal feedback gain is obtained by solving the algebraic Riccati equation. The optimal feedback gain and the optimal state estimation vector are multiplied by a matrix to generate a feedback compensation amount. The feedback compensation amount is added to the modified feedforward control sequence to generate a composite control command that drives the servo valve.

Images