Output feedback control method of single-rod hydraulic cylinder position servo system

Abstract

The invention relates to the technical field of hydraulic servo control, and discloses an output feedback control method of a single-rod hydraulic cylinder position servo system. The method comprises the following steps: synchronously acquiring real-time data of load pressure and piston displacement, identifying a current dynamic working mode of a system by analyzing a sequence evolution trend, and calling a corresponding pre-stored control gain. During mode switching, a smooth transition parameter is generated through gain fusion interpolation, and sudden change of a control instruction is avoided. A multivariable control framework is constructed based on pressure and displacement deviation, channel weights are dynamically distributed, and a comprehensive control instruction is synthesized to drive an execution element. According to the method, rapid adaptive matching of the controller to complex working conditions is realized, the stability of dynamic response of the system is ensured through smooth gain transition, and the control precision and robustness are improved.

Description

Technical Field

[0001] This invention relates to the field of hydraulic servo control technology, specifically to an output feedback control method for a single-rod hydraulic cylinder position servo system. Background Technology

[0002] In the field of hydraulic position servo control, traditional methods generally rely on fixed control parameters or adjustment mechanisms based on a single position error signal. Faced with complex operating conditions such as varying load forces, nonlinear friction forces, and time-varying system parameters, these methods struggle to guarantee consistently high-performance control. Advanced techniques such as gain scheduling or model reference adaptive control have been used to address changing operating conditions, but their designs are complex, requiring high accuracy of the system model and real-time parameter identification. Furthermore, they often exhibit response lag during rapid changes in operating conditions or necessitate a trade-off between stability and speed.

[0003] Existing technical solutions have significant drawbacks. Using fixed control parameters leads to decreased control performance when system operating conditions deviate from the design conditions, resulting in slow response, increased tracking errors, or overshoot oscillations. While using gain-switching strategies involves parameter jumps between different operating points or control modes, the lack of a necessary transition process means that abrupt changes in control output directly translate into impacts on hydraulic actuators, easily triggering system mechanical resonance and causing unstable motion, decreased positioning accuracy, or even instability. A key challenge in complex, time-varying applications is how to enable the control system to quickly match the current optimal control strategy while maintaining smoothness and stability during strategy switching. Summary of the Invention

[0004] The purpose of this invention is to provide an output feedback control method for a single-rod hydraulic cylinder position servo system to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides an output feedback control method for a single-rod hydraulic cylinder position servo system, the method comprising:

[0006] Real-time measurements of load pressure and piston rod displacement are simultaneously captured from a detection unit deployed in the hydraulic cylinder.

[0007] The instantaneous measurements are preprocessed to generate a load pressure observation sequence and a piston displacement observation sequence that contain time-series correlations;

[0008] Based on the evolution trends of the load pressure observation sequence and the piston displacement observation sequence, the typical mode to which the current working state belongs is identified.

[0009] Retrieve the set of control gains corresponding to the typical mode from the parameter memory;

[0010] When the typical mode transition occurs, the currently effective control gain and the target control gain are fused and interpolated to generate a transition control gain;

[0011] A multivariable control architecture is constructed with pressure deviation and displacement deviation as inputs. The multivariable control architecture includes pressure compensation channels and displacement compensation channels that run in parallel.

[0012] Based on the real-time characteristics of the working state, appropriate influence factors are assigned to the pressure compensation channel and the displacement compensation channel respectively;

[0013] Based on the output of the pressure compensation channel, the output of the displacement compensation channel, and their respective influence factors, a comprehensive control command for driving the hydraulic actuator is synthesized.

[0014] Preferably, the preprocessing of the instantaneous measurements to generate a load pressure observation sequence and a piston displacement observation sequence containing time-series correlation includes:

[0015] Outlier removal and missing value filling are performed on the captured raw pressure measurements to form a continuous load pressure data stream;

[0016] The captured raw displacement measurements are low-pass filtered to form a smooth piston displacement data stream;

[0017] A fixed-length recent data segment is extracted from the continuous load pressure data stream, and the first-order difference of the values ​​within the segment is calculated to obtain an approximate load pressure change rate curve.

[0018] The displacement increment per unit time is calculated from the smoothed piston displacement data stream to generate a piston movement speed curve;

[0019] The zero-crossing positions and displacement extreme values ​​between adjacent zero-crossings are extracted from the smooth piston displacement data stream to calculate the oscillation characteristic quantities of the system.

[0020] The load pressure change rate curve, the piston movement speed curve, and the oscillation characteristic quantity are normalized and spliced ​​together to form a multi-dimensional feature vector describing the dynamic characteristics of the system.

[0021] Preferably, identifying the typical pattern of the current operating state based on the evolution trends of the load pressure observation sequence and the piston displacement observation sequence includes:

[0022] The multidimensional feature vector is input into a pattern classifier that has been pre-trained using historical data;

[0023] The pattern classifier parses the multidimensional feature vector and outputs a pattern identifier that represents the category to which the current dynamic behavior belongs.

[0024] The parsing process of the pattern classifier includes: calculating the matching degree between the multidimensional feature vector and multiple standard pattern features stored in the template library;

[0025] The matching degree calculation involves comparing the similarity between the statistical distribution of each component in the multidimensional feature vector and the statistical distribution of the corresponding component in each standard pattern feature;

[0026] The identifier associated with the standard pattern feature that has the highest matching degree with the multidimensional feature vector is determined as the typical pattern to which the current working state belongs.

[0027] Preferably, when the typical mode transition occurs, the process of fusing and interpolating the currently effective control gain and the target control gain to generate a transition control gain includes:

[0028] When a change in the pattern identifier is detected, the start time of the pattern migration is recorded;

[0029] Read the complete set of control gains corresponding to the new mode identifier from the parameter memory and mark it as the target set of control gains;

[0030] Maintain the currently used set of control gains and mark it as the source set of control gains;

[0031] Define a transition window that starts at the initial time and lasts for a specific duration;

[0032] Within the transition window, a mixing coefficient is generated, increasing from zero to one, according to a linear or nonlinear function of time.

[0033] For each gain in the source control gain set and the corresponding gain in the target control gain set, the weighted sum is calculated according to the mixing coefficient at the current time, and the result is the transition control gain that takes effect at that time.

[0034] At the end of the transition window, the transition control gain is fully transitioned to the target control gain set.

[0035] Preferably, the construction of a multivariable control architecture with pressure deviation and displacement deviation as inputs includes a pressure compensation channel and a displacement compensation channel operating in parallel, comprising:

[0036] Set a desired load pressure trajectory and a desired piston displacement trajectory;

[0037] The difference between the real-time load pressure observation sequence and the expected load pressure trajectory is calculated as the real-time pressure deviation.

[0038] The difference between the real-time piston displacement observation sequence and the expected piston displacement trajectory is calculated as the real-time displacement deviation.

[0039] The real-time pressure deviation is input into a proportional-integral-derivative controller, and the output of the controller constitutes the initial output signal of the pressure compensation channel.

[0040] The real-time displacement deviation is input into another proportional-integral-derivative controller, and the output of this controller constitutes the initial output signal of the displacement compensation channel.

[0041] The coefficients of the integral and derivative terms of the proportional-integral-derivative controller are corrected online using the multidimensional feature vector.

[0042] Preferably, the method further includes:

[0043] During the control process, the operating status of multiple hydraulic cylinders is compared and analyzed. Based on the results of the comparison and analysis, the control gain set or the influencing factor of each hydraulic cylinder is fine-tuned to achieve coordinated action.

[0044] The intensity of local fluctuations in the load pressure observation sequence is continuously monitored. When the intensity of local fluctuations exceeds the warning line, the pressure suppression procedure is activated.

[0045] The operation iteration interval of the output feedback control method is adaptively adjusted according to the dynamic level of the operating state.

[0046] Preferably, during the control process, the operating states of multiple hydraulic cylinders are compared and analyzed, and fine-tuning is applied to the control gain set or the influencing factors of each hydraulic cylinder based on the results of the comparison and analysis to achieve coordinated action, including:

[0047] Periodically collect real-time piston displacement data of all controlled hydraulic cylinders;

[0048] The displacement of one of the hydraulic cylinders can be used as a reference, or the average displacement of all hydraulic cylinders can be calculated as a reference.

[0049] Calculate the difference between the real-time displacement of each other hydraulic cylinder and the reference benchmark to form a set of displacement synchronization errors;

[0050] Analyze the statistical characteristics of the displacement synchronization error, including the mean, variance, and trend of change;

[0051] Based on the statistical characteristics of the displacement synchronization error, a fine-tuning amount is generated for each hydraulic cylinder, which is used to slightly correct the control gain set or influencing factor currently used by the hydraulic cylinder.

[0052] The generated fine-tuning amount is superimposed on the original control gain set or influence factor of the corresponding hydraulic cylinder to form a new local control parameter.

[0053] Preferably, the continuous monitoring of the local fluctuation intensity of the load pressure observation sequence, and the activation of the pressure suppression procedure when the local fluctuation intensity exceeds the warning line, includes:

[0054] On the load pressure observation sequence, the standard deviation of the data within the window is calculated in a sliding time window manner as an indicator to measure the intensity of local fluctuations;

[0055] The calculated local fluctuation intensity index is compared with the pre-set multi-level warning threshold.

[0056] Once the local fluctuation intensity index exceeds the lowest level warning threshold, the primary pressure suppression procedure is activated, which includes limiting the maximum output change rate of the control command.

[0057] If the local fluctuation intensity index continues to rise and exceeds a higher level of warning threshold, an advanced pressure suppression procedure is initiated. The advanced pressure suppression procedure includes superimposing a reverse damping component opposite to the direction of the pressure fluctuation into the control command.

[0058] Once the local fluctuation intensity index falls below all warning thresholds and remains stable, the various measures of the pressure suppression procedure shall be gradually withdrawn.

[0059] Preferably, the step of adaptively adjusting the computational iteration interval of the output feedback control method according to the dynamic level of the working state includes:

[0060] Define a basic iteration interval and a maximum and minimum allowed iteration interval;

[0061] Calculate the rate of change of the magnitude of the multidimensional feature vector and use it as a quantitative indicator to measure the dynamic level of the working state;

[0062] Establish a mapping relationship between the quantitative index of the dynamic level and the expected iteration interval, such that the higher the dynamic level, the shorter the expected iteration interval;

[0063] The calculated expected iteration interval is subjected to a limiting process to ensure that it is between the minimum iteration interval and the maximum iteration interval;

[0064] The adjusted iteration interval is used as the sampling and control operation time length for the next control cycle.

[0065] Preferably, the method further includes:

[0066] After each fusion interpolation process is completed, the overall control performance index of the system within the transition window is evaluated.

[0067] The comprehensive control performance index obtained from the evaluation is associated with and stored in relation to the functional relationship of the mixing coefficient over time;

[0068] When a pattern migration to the same target pattern occurs again, retrieve the historically stored functional relationships related to the combination of the current source pattern and the target pattern;

[0069] Based on the retrieved historical function relationships, the mixing coefficient change curve during this mode migration process is optimized in order to obtain better transition performance.

[0070] Compared with the prior art, the beneficial effects of the present invention are:

[0071] By synchronously analyzing the evolution trend of the time-series observation sequence of load pressure and piston displacement, the dynamic operating mode of the system is identified in real time, and the optimal control gain set pre-tuned for that mode is directly invoked. This allows the control law to immediately match the characteristics of the current operating condition, improving the response speed and control accuracy to load disturbances and changes in operating state. The system can achieve near-optimal transient and steady-state performance without a complex online parameter tuning process, enhancing its adaptability to complex and changing operating environments.

[0072] When the system's operating mode shifts, the currently effective control gain and the target mode's control gain are fused and interpolated to generate a series of continuously changing transitional control gains, rather than executing direct switching commands. This suppresses abrupt changes in control force caused by step changes in control parameters and eliminates system shocks, oscillations, and overshoot that may occur during mode switching. The output of control commands remains smooth and continuous, ensuring the motion stability and dynamic quality of the hydraulic actuator during condition transitions, maintaining the stable operation of the entire system, and broadening the application range of servo systems in high-quality, high-dynamic-requirement applications. Attached Figure Description

[0073] Figure 1 This is a schematic diagram illustrating the working principle of the output feedback control method of the single-rod hydraulic cylinder position servo system described in this invention.

[0074] Figure 2 A flowchart for generating observation sequences through preprocessing;

[0075] Figure 3A flowchart for generating transition control gain for fusion interpolation;

[0076] Figure 4 A graph showing the performance index changes of multi-cylinder coordinated control of hydraulic cylinders;

[0077] Figure 5 This is a graph showing the trend of the mixing coefficient during the hydraulic cylinder mode migration process. Detailed Implementation

[0078] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0079] Please see Figure 1 This invention provides an output feedback control method for a single-rod hydraulic cylinder position servo system. The method includes: synchronously capturing real-time measurements of load pressure and piston rod displacement from a detection unit deployed in the hydraulic cylinder; preprocessing the real-time measurements to generate a load pressure observation sequence and a piston displacement observation sequence containing time-series correlations; identifying the typical mode to which the current working state belongs based on the evolution trends of the load pressure observation sequence and the piston displacement observation sequence; retrieving the control gain set corresponding to the typical mode from a parameter memory; when a typical mode transition occurs, performing fusion interpolation processing on the currently effective control gain and the target control gain to generate a transition control gain; constructing a multivariable control architecture with pressure deviation and displacement deviation as inputs, the multivariable control architecture including a pressure compensation channel and a displacement compensation channel running in parallel; assigning appropriate influence factors to the pressure compensation channel and the displacement compensation channel respectively according to the real-time characteristics of the working state; and synthesizing a comprehensive control command for driving the hydraulic actuator based on the output of the pressure compensation channel, the output of the displacement compensation channel, and their respective influence factors.

[0080] Example 1: See Figure 2The system performs outlier removal and missing value filling on the captured raw pressure measurements to form a continuous load pressure data stream. It also performs low-pass filtering on the captured raw displacement measurements to form a smooth piston displacement data stream. A fixed-length recent data segment is extracted from the continuous load pressure data stream, and the first-order difference of the values ​​within that segment is calculated to obtain an approximate load pressure change rate curve. The displacement increment per unit time is calculated from the smooth piston displacement data stream to generate a piston velocity curve. Zero-crossing positions and displacement extrema between adjacent zero-crossing points are extracted from the smooth piston displacement data stream to calculate the system's oscillation characteristics. Finally, the load pressure change rate curve, piston velocity curve, and oscillation characteristics are normalized and concatenated to form a multi-dimensional feature vector describing the system's dynamic characteristics. A multidimensional feature vector is input into a pattern classifier that has been pre-trained using historical data. The pattern classifier parses the multidimensional feature vector and outputs a pattern identifier representing the category to which the current dynamic behavior belongs. The parsing process of the pattern classifier includes calculating the matching degree between the multidimensional feature vector and multiple standard pattern features stored in the template library. The matching degree calculation involves comparing the similarity between the statistical distribution of each component in the multidimensional feature vector and the statistical distribution of the corresponding component in each standard pattern feature. The identifier associated with the standard pattern feature that has the highest matching degree with the multidimensional feature vector is determined as the typical pattern to which the current working state belongs.

[0081] In practical implementation, pressure and displacement sensors deployed on the hydraulic cylinder synchronously capture real-time measurements of load pressure and piston rod displacement at a fixed sampling frequency. A field point discrimination method based on the Raida criterion is applied to the captured raw pressure measurements to identify and remove anomalous data points that significantly deviate from the normal range. For missing values ​​in the raw pressure measurement sequence caused by field point removal or temporary signal loss, linear interpolation is used to fill in the gaps, thus forming a continuous and numerically complete load pressure data stream over a period of time. The captured raw displacement measurements are passed through a second-order Butterworth low-pass digital filter with a cutoff frequency set according to the system's highest operating frequency to filter out high-frequency measurement noise, forming a smooth piston displacement data stream. From the continuous load pressure data stream, a recent data segment is extracted using a sliding time window of one hundred sampling points of fixed length. The first-order backward difference between adjacent sampling points within this window is calculated. These difference values ​​are arranged in chronological order to form an approximate load pressure change rate curve. From the smoothed piston displacement data stream, a discrete-time series of piston movement speed, i.e., the piston movement speed curve, is generated by calculating the difference between the displacement value at the current sampling time and the displacement value at the previous sampling time and dividing it by the sampling period. In the smoothed piston displacement data stream, points where the sign of the displacement value changes are identified as zero-crossing points, and the maximum absolute value of the displacement between two adjacent zero-crossing points is identified as the extreme value. By calculating the ratio of the time interval between three consecutive zero-crossing points to the extreme value of the displacement, an oscillation characteristic quantity describing the relationship between the damped oscillation frequency and amplitude of the system is obtained. The load pressure change rate curve, the piston movement speed curve, and the oscillation characteristic quantity are all normalized to unify their numerical range to between zero and one. Then, these three normalized data sequences are concatenated dimensionally to form a multi-dimensional feature vector with temporal correlation, used to comprehensively describe the dynamic characteristics of the system.

[0082] In some embodiments, the multidimensional feature vector is input in real time to a support vector machine-based pattern classifier pre-trained on historical data. The pattern classifier performs parsing operations on the received multidimensional feature vector and finally outputs a pattern identifier representing the category to which the current dynamic behavior belongs. The pattern identifier can be a category label such as "steady-state tracking," "acceleration start," "deceleration braking," or "load mutation." The internal parsing process of the pattern classifier specifically includes calculating the matching degree between the input multidimensional feature vector and multiple standard pattern features stored in a template library. The matching degree calculation process involves comparing the similarity between the statistical distribution of each component in the input multidimensional feature vector and the statistical distribution of the corresponding component of each standard pattern feature in the template library. An optional matching degree calculation formula can be expressed as:

[0083]

[0084] Where: symbol Indicates the relationship between the input feature vector and the first... The overall matching score between standard pattern features, denoted as follows: It is the standard mathematical notation for the natural exponential function. The symbol represents the total number of dimensions of a multidimensional feature vector. Representative bestowed upon the first Weight coefficients of dimensional features, sign It is a related to the first Positive parameters related to the scale of the feature distribution, sign This indicates that the input multidimensional feature vector is at the th... Statistical mean on dimension, sign Indicates the first The standard pattern feature in the first The statistical mean over the dimension. After calculating the matching score between the input feature vector and all standard pattern features, the pattern classifier determines the pattern identifier associated with the standard pattern feature that has the highest matching score with the input multidimensional feature vector as the typical pattern to which the current working state belongs.

[0085] It is understandable that the standard pattern features and their statistical distribution parameters in the template library are derived from cluster analysis and feature extraction of a large amount of historical operating data under different working conditions. The pattern classifier has already completed offline training and parameter calibration before deployment. In specific implementation, the process of extracting oscillation features from the smooth piston displacement data stream requires continuous monitoring of the sign change of the displacement value. When two consecutive sign changes are detected, the time difference between these two zero-crossing points is recorded as the half-cycle of the oscillation, and the maximum value of the absolute displacement within this half-cycle is also recorded. The oscillation features can be obtained by combining the average of multiple consecutive half-cycles with the average of the corresponding displacement extreme values. This combination operation aims to quantify the oscillation tendency of the system response. In some embodiments, the energy distribution within a specific frequency band can also be directly calculated from the piston displacement data stream as a supplement or substitute for the oscillation features. The maximum and minimum value ranges on which the normalization processing of the multidimensional feature vector is based are fixed values ​​preset according to the theoretical limit values ​​of the hydraulic cylinder system throughout the entire expected working range or the historically observed safe operating boundaries. It is understandable that normalizing the load pressure change rate curve and piston movement speed curve can eliminate the influence of different physical dimensions and numerical ranges on subsequent pattern classification calculations, making the contributions of each dimension of the multidimensional feature vector to the classification decision comparable in magnitude. An optional implementation detail is that before concatenating the load pressure change rate curve, piston movement speed curve, and oscillation feature quantity into a multidimensional feature vector, each curve can be resampled at equal intervals to ensure that all data sequences have the same length, thus facilitating point-by-point or statistical comparison with standard pattern features in the template library.

[0086] Example 2: See Figure 3 When a change in the mode identifier is detected, the start time of the mode migration is recorded; the complete set of control gains corresponding to the new mode identifier is read from the parameter memory and marked as the target control gain set; the currently used control gain set is maintained and marked as the source control gain set; a transition window is set from the start time and lasts for a specific duration; within the transition window, a mixing coefficient is generated from zero to one according to a linear or nonlinear function of time; for each gain in the source control gain set and the corresponding gain in the target control gain set, the weighted sum is calculated according to the mixing coefficient at the current time, and the result is the transition control gain that takes effect at that time; at the end of the transition window, the transition control gain is completely transitioned to the target control gain set. A desired load pressure trajectory and a desired piston displacement trajectory are defined; the difference between the real-time load pressure observation sequence and the desired load pressure trajectory is calculated as the real-time pressure deviation; the difference between the real-time piston displacement observation sequence and the desired piston displacement trajectory is calculated as the real-time displacement deviation; the real-time pressure deviation is input into a proportional-integral-derivative (PID) controller, and the output of this controller constitutes the initial output signal of the pressure compensation channel; the real-time displacement deviation is input into another PID controller, and the output of this controller constitutes the initial output signal of the displacement compensation channel; the integral and derivative coefficients of the PID controller are corrected online using multidimensional eigenvectors.

[0087] In practical implementation, when the mode identifier output by the mode classifier changes from "accelerated start" to "steady-state tracking," the system records the start time of the mode transition. The system reads the complete control gain set corresponding to the "steady-state tracking" mode identifier from the parameter memory and marks this set as the target control gain set. Simultaneously, the system maintains the currently used control gain set corresponding to the "accelerated start" mode and marks this set as the source control gain set. The system sets a transition window lasting 0.2 seconds from the start time. Within this transition window, the system generates a mixing coefficient that increases from zero to one over time according to a preset time nonlinear function. An optional nonlinear function formula for generating the mixing coefficient can be expressed as:

[0088]

[0089] Where: symbol Indicates the time within the transition window Variational mixing coefficient, sign It is a positive real number parameter used to adjust the shape of a curve, with the symbol... Indicates the time elapsed since the start of the pattern transition, symbol This represents the preset total duration of the transition window. For each gain in the source control gain set and its corresponding gain in the target control gain set, the system calculates the mixing coefficient based on the current time step. A weighted summation is performed, and the result is the transient control gain effective at that moment. At the end of the transient window, the mixing coefficient... Once one is reached, the transition control gain completely transitions into the target control gain set.

[0090] In some embodiments, the system sets a desired load pressure trajectory and a desired piston displacement trajectory given by a host computer. The desired load pressure trajectory defines a reference value for the change of pressure in the rodless chamber of the hydraulic cylinder over time, and the desired piston displacement trajectory defines a reference value for the change of the target position of the extended piston rod over time. The system calculates the difference between the real-time load pressure observation sequence and the desired load pressure trajectory at the current sampling time, and uses this difference as the real-time pressure deviation. Simultaneously, the system calculates the difference between the real-time piston displacement observation sequence and the desired piston displacement trajectory at the current sampling time, and uses this difference as the real-time displacement deviation. The real-time pressure deviation is input into a proportional-integral-derivative (PID) controller, and the output of the PID controller constitutes the initial output signal of the pressure compensation channel. The real-time displacement deviation is input into another independent PID controller, and the output of this PID controller constitutes the initial output signal of the displacement compensation channel. The system uses the generated multidimensional eigenvectors to perform online correction on the integral and derivative coefficients of the two proportional-integral-derivative (PID) controllers. It can be understood that the online correction process is based on the dynamic characteristics of the system reflected by the multidimensional eigenvectors. For example, when the multidimensional eigenvectors indicate that the system is in a high dynamic change stage, the derivative coefficients of the PID controllers will be appropriately enhanced to provide additional damping.

[0091] In practical implementation, the online correction of the integral and derivative coefficients of the proportional-integral-derivative (PID) controller is based on a preset rule table. This rule table establishes a mapping relationship between the numerical ranges of key components in the multidimensional feature vector and the coefficient adjustment factors. For example, when the component value corresponding to the piston movement speed curve exceeds a preset threshold, the integral coefficient of the PID controller in the displacement compensation channel is multiplied by an adjustment factor less than one to prevent integral saturation. Conversely, when the component value corresponding to the load pressure change rate curve increases rapidly, the derivative coefficient of the PID controller in the pressure compensation channel is multiplied by an adjustment factor greater than one to enhance the suppression response to pressure surges. This online correction allows the parameters of the PID controller to adapt to changes in the operating state, rather than remaining fixed.

[0092] In a specific scenario involving the transition from a "deceleration braking" mode to a "load holding" mode, the proportional gain value in the target control gain set is significantly lower than that in the source control gain set. The mixing coefficients generated by the nonlinear function cause the transition control gain to change slowly in the first half of the transition window and accelerate in the second half. This transition smooths the change in control commands, reducing the fluctuation amplitude of the piston displacement observation sequence during mode transition by approximately 30% compared to directly switching the control gain set. The system continuously sends pressure and displacement deviations to the corresponding proportional-integral-derivative (PID) controllers. The outputs of the PID controllers are then weighted according to the influence factors assigned to their respective channels, ultimately synthesizing a comprehensive control command to drive the hydraulic actuator. This comprehensive control command is then applied to the electro-hydraulic servo valve via a servo amplifier.

[0093] Example 3: During the control process, the operating states of multiple hydraulic cylinders are compared and analyzed. Based on the results of the comparison and analysis, the control gain set or influencing factors of each hydraulic cylinder are fine-tuned to achieve coordinated action. The local fluctuation intensity of the load pressure observation sequence is continuously monitored. When the local fluctuation intensity exceeds the warning line, the pressure suppression procedure is activated. The calculation iteration interval of the output feedback control method is adaptively adjusted according to the dynamic level of the working state. Real-time piston displacement data of all controlled hydraulic cylinders are collected periodically. The displacement of one hydraulic cylinder is used as a reference benchmark, or the average displacement of all hydraulic cylinders is calculated as a reference benchmark. The difference between the real-time displacement of each other hydraulic cylinder and the reference benchmark is calculated to form a set of displacement synchronization errors. The statistical characteristics of the displacement synchronization errors are analyzed, including the mean, variance, and trend. Based on the statistical characteristics of the displacement synchronization errors, a fine-tuning amount is generated for each hydraulic cylinder. The fine-tuning amount is used to slightly correct the control gain set or influencing factors currently used by the hydraulic cylinder. The generated fine-tuning amount is superimposed on the original control gain set or influencing factors of the corresponding hydraulic cylinder to form new local control parameters.

[0094] In practical implementation, when the system controls multiple hydraulic cylinders to perform coordinated lifting actions, the operating states of the multiple hydraulic cylinders are compared and analyzed during the control process. Based on the results of the comparison and analysis, the system fine-tunes the control gain set or influencing factors of each hydraulic cylinder to achieve coordinated action. The system periodically collects real-time piston displacement data of all controlled hydraulic cylinders, with the collection period set to be the same as the main control cycle. The system uses the piston displacement of one of the hydraulic cylinders designated as the master cylinder as a reference benchmark. The system calculates the difference between the real-time piston displacement of each of the other slave hydraulic cylinders and the master cylinder's piston displacement reference benchmark, forming a set of displacement synchronization errors. The system analyzes the statistical characteristics of the displacement synchronization error, including the mean, variance, and trend of the displacement synchronization error within a certain time window. Based on the statistical characteristics of the displacement synchronization error, the system generates a fine-tuning amount for each slave hydraulic cylinder. The fine-tuning amount is used to slightly correct the control gain set or influencing factors currently used by the slave hydraulic cylinder. An optional calculation relationship for generating the fine-tuning amount can be expressed as:

[0095]

[0096] Where: symbol Indicates that for the first The fine-tuning amount generated by a subordinate hydraulic cylinder, symbol , and These are pre-set weighting coefficients, with symbols... Indicates the first The average displacement synchronization error of each slave hydraulic cylinder over the most recent ten control cycles, with the sign... This represents the current rate of change of the average value, indicated by the symbol. This represents the standard deviation of the displacement synchronization error within the same time window. The system will generate fine-tuning amounts. The proportional gain term is superimposed on the original control gain set of the corresponding subordinate hydraulic cylinder, or superimposed on the influence factor of the displacement compensation channel, to form a new local control parameter for the hydraulic cylinder. In some embodiments, the system can also calculate the average value of the piston displacement of all hydraulic cylinders as a common reference benchmark. In this case, the displacement synchronization error of each hydraulic cylinder is the difference between its own piston displacement and this common average benchmark.

[0097] Understandably, the system continuously monitors the local fluctuation intensity of the load pressure observation sequence for each hydraulic cylinder. The system calculates the standard deviation of the data within a sliding time window containing fifty sampling points on the load pressure observation sequence. The calculated standard deviation of the sliding window serves as an indicator of the local fluctuation intensity. The system compares the calculated local fluctuation intensity indicator with pre-set multi-level warning thresholds, including a primary warning threshold and a high-level warning threshold. Once the local fluctuation intensity indicator of a hydraulic cylinder exceeds the primary warning threshold, the system initiates a primary pressure suppression procedure for that hydraulic cylinder. The primary pressure suppression procedure includes limiting the maximum output change rate of the control command sent to the electro-hydraulic servo valve of that hydraulic cylinder, for example, setting the upper limit of the change amplitude between adjacent cycles of the control command to 50% of the normal value. If the local fluctuation intensity indicator of that hydraulic cylinder continues to rise and exceeds the high-level warning threshold, the system initiates a high-level pressure suppression procedure. The high-level pressure suppression procedure includes superimposing a reverse damping component, opposite in direction to the current load pressure fluctuation, when calculating the comprehensive control command for that hydraulic cylinder. The amplitude of the reverse damping component is proportional to the degree to which the local fluctuation intensity indicator exceeds the high-level warning threshold. Once the local fluctuation intensity index falls below all warning thresholds and remains stable for more than one hundred control cycles, the system gradually removes various measures of the pressure suppression procedure, such as limiting the maximum output change rate of the linear recovery control command to the normal value and proportionally reducing the amplitude of the reverse damping component until it is zero.

[0098] In some embodiments, the system adaptively adjusts the iteration interval of the output feedback control method based on the dynamic level of the working state. The system defines a basic iteration interval and a maximum and minimum allowable iteration interval. The basic iteration interval is typically five milliseconds, the maximum iteration interval is twenty milliseconds, and the minimum iteration interval is one millisecond. The system calculates the rate of change of the magnitude of the multidimensional feature vector, using it as a quantitative indicator to measure the dynamic level of the working state. The rate of change of the magnitude of the multidimensional feature vector is obtained by calculating the absolute value of the difference between the magnitude of the multidimensional feature vector at the current moment and the magnitude at the previous moment, and then dividing it by the iteration interval. The system establishes a mapping relationship between the quantitative indicator of the dynamic level and the expected iteration interval. This mapping relationship is implemented in the form of a lookup table, which specifies an inverse relationship where a higher quantitative indicator of the dynamic level corresponds to a shorter expected iteration interval. The system performs amplitude limiting on the expected iteration interval calculated from the lookup table to ensure that it is between the minimum and maximum iteration interval. The system uses the adjusted iteration interval as the sampling and control operation time length for the next control cycle. It can be understood that when the system dynamic level is high, i.e. when the multidimensional feature vector changes drastically, a shorter iteration interval can improve the control response speed, while a longer iteration interval helps to reduce the burden on the computing unit when the system is close to steady state.

[0099] See Figure 4This is a graph showing the performance changes of multi-cylinder coordinated control of hydraulic cylinders. The first 20 cycles show significant fluctuations (up to 1.4 mm), followed by a gradual decrease and stabilization at around 0.1 mm, indicating a continuous improvement in multi-cylinder synchronization. The rapid stabilization (approaching 0) after the initial fluctuations aligns with the control logic of "increased iteration intervals and slower response speeds after a decrease in dynamic level." Consistent with the trend in coordinated accuracy, the gradual increase after initial fluctuations reflects the effectiveness of the pressure suppression procedure. The simultaneous improvement in coordinated accuracy and enhanced pressure stability demonstrates that the combination of multi-cylinder coordinated control and pressure suppression measures effectively improves the overall system performance.

[0100] Example 4: On the load pressure observation sequence, the standard deviation of the data within the window is calculated using a sliding time window method as an indicator to measure the intensity of local fluctuations. The calculated local fluctuation intensity indicator is compared with a pre-set multi-level warning threshold. Once the local fluctuation intensity indicator exceeds the lowest level warning threshold, the primary pressure suppression procedure is activated, which includes limiting the maximum output change rate of the control command. If the local fluctuation intensity indicator continues to rise and exceeds a higher level warning threshold, the advanced pressure suppression procedure is activated, which includes superimposing a reverse damping component opposite to the direction of pressure fluctuation into the control command. Once the local fluctuation intensity indicator falls back below all warning thresholds and remains stable, the various measures of the pressure suppression procedure are gradually withdrawn. Define a basic iteration interval and an allowed maximum and minimum iteration interval; calculate the rate of change of the magnitude of the multidimensional feature vector and use it as a quantitative indicator to measure the dynamic level of the working state; establish a mapping relationship between the quantitative indicator of the dynamic level and the expected iteration interval, such that the higher the dynamic level, the shorter the expected iteration interval; perform amplitude limiting on the calculated expected iteration interval to ensure that it is between the minimum and maximum iteration intervals; use the adjusted iteration interval as the sampling and control operation time length for the next control cycle.

[0101] In practice, the system continuously monitors the local fluctuation intensity of the load pressure observation sequence. The system calculates the standard deviation of the data within a sliding time window containing a fixed number of sampling points. The sliding time window contains 64 sampling points, and the window slides at a step size of one sampling point. The calculated standard deviation of the data within the window serves as an indicator of the local fluctuation intensity. The system compares the calculated local fluctuation intensity indicator with pre-set multi-level warning thresholds, including a primary warning threshold and a high-level warning threshold. The primary warning threshold is set to 1% of the system's rated operating pressure, and the high-level warning threshold is set to 2% of the system's rated operating pressure. Once the local fluctuation intensity indicator exceeds the primary warning threshold, the system activates a primary pressure suppression procedure. This procedure limits the maximum output change rate of the control command; specifically, it sets the upper limit of the allowable change in the control command between two adjacent control cycles to 60% of the upper limit of the allowable change under normal, unsuppressed conditions. If the local fluctuation intensity index continues to rise and exceeds the advanced warning threshold, the system activates the advanced pressure suppression procedure. This procedure includes superimposing a reverse damping component, opposite in direction to the pressure fluctuation, into the control command. The calculation of the reverse damping component is related to the instantaneous change trend of the current load pressure observation sequence. A formula for calculating the reverse damping component can be expressed as:

[0102]

[0103] Where: symbol Represents the reverse damping component to be superimposed, with the symbol... It is a positive damping gain coefficient, with the sign... It is a sign function whose output is the first-order backward difference of the load pressure observation sequence at the current time. The sign determines the sign. This represents the calculated local fluctuation intensity index, with the symbol... This represents the value of the advanced warning threshold. Once the local fluctuation intensity index falls below all warning thresholds and remains stable for more than fifty consecutive control cycles, the system gradually removes the measures of the pressure suppression procedure. For example, it linearly increases the maximum output rate of change limit of the control command until it returns to normal levels, and linearly decreases the gain coefficient of the reverse damping component. Until it reaches zero.

[0104] In some embodiments, the system adaptively adjusts the iteration interval of the output feedback control method based on the dynamic level of the working state. The system defines a basic iteration interval and a maximum and minimum allowable iteration interval. The basic iteration interval is 10 milliseconds, the maximum iteration interval is 50 milliseconds, and the minimum iteration interval is 2 milliseconds. The system calculates the rate of change of the magnitude of the multidimensional feature vector. The rate of change of the magnitude is obtained by calculating the absolute value of the difference between the Euclidean magnitude of the multidimensional feature vector in the current control cycle and the magnitude of the previous control cycle, and then dividing it by the actual iteration interval time used in the previous control cycle. This calculation result is used as a quantitative indicator to measure the dynamic level of the working state. The system establishes a mapping relationship between the quantitative indicator of the dynamic level and the expected iteration interval. The mapping relationship is implemented through a preset lookup table. This lookup table specifies an inverse relationship: the higher the quantitative indicator of the dynamic level, the shorter the expected iteration interval value obtained by mapping. Refer to Table 1 for an exemplary mapping relationship lookup table.

[0105] Table 1: Mapping Relationship between Dynamic Level and Iteration Interval

[0106]

[0107] It is understandable that the system performs a limiting process on the expected iteration interval calculated from the mapping lookup table. This limiting process ensures that the final iteration interval value is between the minimum and maximum iteration interval. The system uses the iteration interval value after limiting adjustment as the sampling and control calculation time length for the next control cycle. In one specific scenario, when the system is in a stable operating state, the calculated dynamic level quantization index is 1.5 seconds. Looking up the mapping table, the expected iteration interval is 50 milliseconds. After limiting, the actual maximum iteration interval of 50 milliseconds is used as the next control cycle. In another specific scenario, when the system encounters an external load shock, the dynamic level quantization index rapidly rises to 15 seconds. Looking up the mapping table, the expected iteration interval is 5 milliseconds. This value is between the minimum and maximum iteration interval limits, and the system immediately uses 5 milliseconds as the next control cycle, thereby accelerating the control response frequency.

[0108] Example 5: After each fusion interpolation process is completed, the overall control performance index of the system within the transition window is evaluated; the evaluated overall control performance index is associated with and stored as a function of the mixing coefficient changing over time; when the mode migration to the same target mode occurs again, the historically stored function relationship related to the combination of the current source mode and the target mode is retrieved; based on the retrieved historical function relationship, the mixing coefficient change curve during this mode migration process is optimized in order to obtain better transition performance.

[0109] In practice, after each fusion interpolation process, the system evaluates the overall control performance index within the transition window. The calculation of the overall control performance index occurs within a fixed evaluation time interval after the mode transition window ends, for example, within 100 milliseconds after the transition window ends. The overall control performance index is a scalar value, and its calculation integrates displacement tracking error, pressure tracking error, and control energy consumption within and after the transition window, and is used to calculate the overall control performance index. The formula can be expressed as:

[0110]

[0111] Where: symbol The calculated comprehensive control performance index is represented by the symbol. , and These are weighting coefficients used to weigh various costs; their symbols are... This represents the sum of the absolute values ​​of displacement deviations at all sampling times within the evaluation period, with the sign... This represents the sum of the absolute values ​​of pressure deviations at all sampling times within the evaluation period, with the sign... This represents the cumulative sum of the squared amplitudes of the integrated control commands at all sampling times within the evaluation period. The system associates and stores the evaluated integrated control performance index with the time-varying function of the mixing coefficients used in this mode migration process. The stored information includes the source mode identifier, the target mode identifier, the type and parameters of the mixing coefficient variation curve, and the final integrated control performance index. .

[0112] In some embodiments, when the system migrates to the same target mode identifier again, for example, from "accelerated start" mode to "steady-state tracking" mode, the system retrieves all historically stored records of functional relationships related to the combination of the current source mode identifier and target mode identifier, along with their corresponding comprehensive control performance indicators. Based on the retrieved historical functional relationship records and their corresponding comprehensive control performance indicators, the system optimizes the mixing coefficient change curve during this mode migration process. The system selects the curve with the minimum comprehensive control performance indicator from the historical records. The curve showing the change in the mixing coefficient associated with the value is used as the initial reference curve for this migration. Optionally, if multiple migration records exist in the historical records, the system can perform weighted fusion of the multiple historical mixing coefficient change curves to generate a new initial curve. The weights of the weighted fusion are related to the comprehensive control performance index corresponding to each historical record. It is directly proportional to the reciprocal of the number.

[0113] In practical implementation, the functional relationship of the mixing coefficient changing over time can have different predefined types, such as linear growth functions, sigmoid growth functions, or exponential growth functions. Each type is defined by the function form and one or more shape parameters. When storing the correlation, the system records the function type and its specific shape parameter values ​​used for calculating the mixing coefficient within this transition window. An example scenario is that when migrating from the "load mutation" mode to the "steady-state tracking" mode for the first time, the system uses a linear growth function as the mixing coefficient change curve, and the overall control performance index is evaluated after the transition window ends. Value When the system experienced its second transition from "load surge" to "steady-state tracking," it retrieved only one relevant record from its historical records, and its overall control performance index was [missing data]. This time, the system can choose to use an S-shaped growth function and try a new set of shape parameters in order to obtain results different from those of the previous system. The system calculates the comprehensive control performance index. The sigmoid growth function, a type of nonlinear function, is defined by its function form and one or more shape parameters. These shape parameters adjust the growth rate and smoothness of the curve, for example, by changing the inflection point or slope of the curve to influence the process of the mixing coefficient increasing from zero to one. The system attempts a new set of shape parameters, initialized based on the distribution of historical performance indicators or engineering experience, generating different curve variants by adjusting the parameter values. Within the transition window, the system calculates the mixing coefficient based on the selected sigmoid growth function and the new shape parameters, and applies it in real-time to the fusion interpolation processing of the control gain. After this migration, the system calculates the new comprehensive control performance index. Value And the new S-shaped function and its parameters are compared with Values ​​are stored in association. Subsequently, when the same pattern migration occurs a third time, the system retrieves two historical records corresponding to different metrics. and The system can be selected with smaller The S-shaped function curve associated with the values ​​is used as the curve for the change of the mixing coefficient in this study.

[0114] See Figure 5 This is a graph showing the trend of the mixing coefficient during the hydraulic cylinder mode migration process. A rapid approach to 0.8 within the first 20ms of the transition window is suitable for mode migration scenarios requiring rapid control gain switching; a uniform change is suitable for scenarios requiring a smooth transition; balancing initial smoothness with rapid mid-term switching makes it a compromise for most mode migrations. Different function types correspond to different control gain switching rhythms; for example, mode migrations involving sudden load changes require rapid response (selecting exponential growth), while steady-state switching requires a smooth transition.

[0115] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An output feedback control method for a single-rod hydraulic cylinder position servo system, characterized in that, The method includes: Real-time measurements of load pressure and piston rod displacement are simultaneously captured from a detection unit deployed in the hydraulic cylinder. The instantaneous measurements are preprocessed to generate a load pressure observation sequence and a piston displacement observation sequence that contain time-series correlations; Based on the evolution trends of the load pressure observation sequence and the piston displacement observation sequence, the typical mode to which the current working state belongs is identified. Retrieve the set of control gains corresponding to the typical mode from the parameter memory; When the typical mode transition occurs, the currently effective control gain and the target control gain are fused and interpolated to generate a transition control gain; A multivariable control architecture is constructed with pressure deviation and displacement deviation as inputs. The multivariable control architecture includes pressure compensation channels and displacement compensation channels that run in parallel. Based on the real-time characteristics of the working state, appropriate influence factors are assigned to the pressure compensation channel and the displacement compensation channel respectively; Based on the output of the pressure compensation channel, the output of the displacement compensation channel, and their respective influence factors, a comprehensive control command for driving the hydraulic actuator is synthesized.

2. The output feedback control method of the single-rod hydraulic cylinder position servo system according to claim 1, characterized in that, The preprocessing of the instantaneous measurements to generate a load pressure observation sequence and a piston displacement observation sequence containing time-series correlations includes: Outlier removal and missing value filling are performed on the captured raw pressure measurements to form a continuous load pressure data stream; The captured raw displacement measurements are low-pass filtered to form a smooth piston displacement data stream; A fixed-length recent data segment is extracted from the continuous load pressure data stream, and the first-order difference of the values ​​within the segment is calculated to obtain an approximate load pressure change rate curve. The displacement increment per unit time is calculated from the smoothed piston displacement data stream to generate a piston movement speed curve; The zero-crossing positions and displacement extreme values ​​between adjacent zero-crossings are extracted from the smooth piston displacement data stream to calculate the oscillation characteristic quantities of the system. The load pressure change rate curve, the piston movement speed curve, and the oscillation characteristic quantity are normalized and spliced ​​together to form a multi-dimensional feature vector describing the dynamic characteristics of the system.

3. The output feedback control method of the single-rod hydraulic cylinder position servo system according to claim 2, characterized in that, The process of identifying the typical pattern of the current operating state based on the evolution trends of the load pressure observation sequence and the piston displacement observation sequence includes: The multidimensional feature vector is input into a pattern classifier that has been pre-trained using historical data; The pattern classifier parses the multidimensional feature vector and outputs a pattern identifier that represents the category to which the current dynamic behavior belongs. The parsing process of the pattern classifier includes: calculating the matching degree between the multidimensional feature vector and multiple standard pattern features stored in the template library; The matching degree calculation involves comparing the similarity between the statistical distribution of each component in the multidimensional feature vector and the statistical distribution of the corresponding component in each standard pattern feature; The identifier associated with the standard pattern feature that has the highest matching degree with the multidimensional feature vector is determined as the typical pattern to which the current working state belongs.

4. The output feedback control method of the single-rod hydraulic cylinder position servo system according to claim 3, characterized in that, When the typical mode transition occurs, the currently effective control gain and the target control gain are fused and interpolated to generate a transition control gain, including: When a change in the pattern identifier is detected, the start time of the pattern migration is recorded; Read the complete set of control gains corresponding to the new mode identifier from the parameter memory and mark it as the target set of control gains; Maintain the currently used set of control gains and mark it as the source set of control gains; Define a transition window that starts at the initial time and lasts for a specific duration; Within the transition window, a mixing coefficient is generated, increasing from zero to one, according to a linear or nonlinear function of time. For each gain in the source control gain set and the corresponding gain in the target control gain set, the weighted sum is calculated according to the mixing coefficient at the current time, and the result is the transition control gain that takes effect at that time. At the end of the transition window, the transition control gain is fully transitioned to the target control gain set.

5. The output feedback control method of the single-rod hydraulic cylinder position servo system according to claim 2, characterized in that, The constructed multivariable control architecture uses pressure deviation and displacement deviation as inputs. This multivariable control architecture includes parallel-running pressure compensation channels and displacement compensation channels, including: Set a desired load pressure trajectory and a desired piston displacement trajectory; The difference between the real-time load pressure observation sequence and the expected load pressure trajectory is calculated as the real-time pressure deviation. The difference between the real-time piston displacement observation sequence and the expected piston displacement trajectory is calculated as the real-time displacement deviation. The real-time pressure deviation is input into a proportional-integral-derivative controller, and the output of the controller constitutes the initial output signal of the pressure compensation channel. The real-time displacement deviation is input into another proportional-integral-derivative controller, and the output of this controller constitutes the initial output signal of the displacement compensation channel. The coefficients of the integral and derivative terms of the proportional-integral-derivative controller are corrected online using the multidimensional feature vector.

6. The output feedback control method of the single-rod hydraulic cylinder position servo system according to claim 1, characterized in that, The method further includes: During the control process, the operating status of multiple hydraulic cylinders is compared and analyzed. Based on the results of the comparison and analysis, the control gain set or the influencing factor of each hydraulic cylinder is fine-tuned to achieve coordinated action. The intensity of local fluctuations in the load pressure observation sequence is continuously monitored. When the intensity of local fluctuations exceeds the warning line, the pressure suppression procedure is activated. The operation iteration interval of the output feedback control method is adaptively adjusted according to the dynamic level of the operating state.

7. The output feedback control method for the single-rod hydraulic cylinder position servo system according to claim 1, characterized in that, During the control process, the operating states of multiple hydraulic cylinders are compared and analyzed. Based on the results of the comparison and analysis, fine-tuning is applied to the control gain set or the influencing factors of each hydraulic cylinder to achieve coordinated action, including: Periodically collect real-time piston displacement data of all controlled hydraulic cylinders; The displacement of one of the hydraulic cylinders can be used as a reference, or the average displacement of all hydraulic cylinders can be calculated as a reference. Calculate the difference between the real-time displacement of each other hydraulic cylinder and the reference benchmark to form a set of displacement synchronization errors; Analyze the statistical characteristics of the displacement synchronization error, including the mean, variance, and trend of change; Based on the statistical characteristics of the displacement synchronization error, a fine-tuning amount is generated for each hydraulic cylinder, which is used to slightly correct the control gain set or influencing factor currently used by the hydraulic cylinder. The generated fine-tuning amount is superimposed on the original control gain set or influence factor of the corresponding hydraulic cylinder to form a new local control parameter.

8. The output feedback control method of the single-rod hydraulic cylinder position servo system according to claim 6, characterized in that, The continuous monitoring of the local fluctuation intensity of the load pressure observation sequence, and the activation of the pressure suppression procedure when the local fluctuation intensity exceeds the warning line, includes: On the load pressure observation sequence, the standard deviation of the data within the window is calculated in a sliding time window manner as an indicator to measure the intensity of local fluctuations; The calculated local fluctuation intensity index is compared with the pre-set multi-level warning threshold. Once the local fluctuation intensity index exceeds the lowest level warning threshold, the primary pressure suppression procedure is activated, which includes limiting the maximum output change rate of the control command. If the local fluctuation intensity index continues to rise and exceeds a higher level of warning threshold, an advanced pressure suppression procedure is initiated. The advanced pressure suppression procedure includes superimposing a reverse damping component opposite to the direction of the pressure fluctuation into the control command. Once the local fluctuation intensity index falls below all warning thresholds and remains stable, the various measures of the pressure suppression procedure shall be gradually withdrawn.

9. The output feedback control method of the single-rod hydraulic cylinder position servo system according to claim 6, characterized in that, The step of adaptively adjusting the computation iteration interval of the output feedback control method according to the dynamic level of the working state includes: Define a basic iteration interval and a maximum and minimum allowed iteration interval; Calculate the rate of change of the magnitude of the multidimensional feature vector and use it as a quantitative indicator to measure the dynamic level of the working state; Establish a mapping relationship between the quantitative index of the dynamic level and the expected iteration interval, such that the higher the dynamic level, the shorter the expected iteration interval; The calculated expected iteration interval is subjected to a limiting process to ensure that it is between the minimum iteration interval and the maximum iteration interval; The adjusted iteration interval is used as the sampling and control operation time length for the next control cycle.

10. The output feedback control method of the single-rod hydraulic cylinder position servo system according to claim 4, characterized in that, The method further includes: After each fusion interpolation process is completed, the overall control performance index of the system within the transition window is evaluated. The comprehensive control performance index obtained from the evaluation is associated with and stored in relation to the functional relationship of the mixing coefficient over time; When a pattern migration to the same target pattern occurs again, retrieve the historically stored functional relationships related to the combination of the current source pattern and the target pattern; Based on the retrieved historical function relationships, the mixing coefficient change curve during this mode migration process is optimized in order to obtain better transition performance.

Images