Dovetail groove type sliding table air cylinder intelligent control system for surge protector manufacturing

Abstract

The invention discloses a dovetail groove type sliding table air cylinder intelligent control system for surge protector manufacturing, and belongs to the technical field of fluid pressure system control, and the dovetail groove type sliding table air cylinder intelligent control system comprises a data acquisition module, a state fusion module, a risk generation module, a mapping calculation module, a loading control calculation module and an execution feedback module. Risks are perceived through multi-source information fusion, closed-loop control is carried out by cooperatively driving pneumatic flexible damping and mechanical rigid preloading, and active prediction and stable intervention of falling risks can be realized while high response speed is guaranteed.

Description

Technical Field

[0001] This invention relates to the field of fluid pressure system control technology, and in particular to an intelligent control system for a dovetail groove slide cylinder used in the manufacture of surge protectors. Background Technology

[0002] Dovetail-groove slide cylinders, as integrated linear motion units, are widely used in surge protector manufacturing equipment for positioning, transferring, and assembling surge protectors. They compactly combine a cylinder with a precision guide rail structure, using compressed air to drive the slide for reciprocating linear motion. In vertical or inclined installations, the slide and its load are subject to gravity. If the air supply is interrupted or the control system malfunctions, the slide risks stalling and falling, potentially causing equipment damage or accidents. Therefore, equipping it with an effective fall protection mechanism is crucial.

[0003] In existing technologies, fall protection for such cylinders typically employs purely mechanical or pneumatic locking solutions. Mechanical solutions often involve passive braking by using spring-driven brake pads or wedges to clamp the guide rails in the event of a power or air supply failure. Pneumatic locking solutions primarily utilize a pilot-operated one-way valve installed in the cylinder's air circuit to seal the gas in the working chamber when position maintenance is required, using gas pressure to balance the load. Both methods are triggered upon detection of a clear fault signal.

[0004] However, mechanical braking devices suffer from delayed triggering, and the braking process is extremely rigid and violent, generating significant impact and vibration that can easily damage the slide or the precision workpiece it carries. Pneumatic locking methods, on the other hand, suffer from slow downward movement due to gas leakage, and their locking force is greatly affected by air pressure fluctuations, making it difficult to guarantee locking reliability under load changes or external impacts. More importantly, these solutions are all passive responses, only able to remedy after a failure occurs, unable to anticipate risks and intervene in advance, and lacking dynamic adaptability to the system's operating state. Summary of the Invention

[0005] To address the aforementioned issues, this invention provides an intelligent control system for a dovetail-groove slide cylinder used in surge protector manufacturing. This system uses multi-source information fusion to perceive risks and coordinates pneumatic flexible damping and mechanical rigid preloading for closed-loop control. This system can proactively predict and smoothly intervene in the risk of fall while ensuring high response speed.

[0006] The above objectives can be achieved through the following approach: The intelligent control system for a dovetail slide cylinder used in surge protector manufacturing includes a data acquisition module for acquiring the slide's stroke position parameters, real-time speed parameters, and pressure difference parameters between the upper and lower chambers of the cylinder to obtain motion state parameters; a state fusion module for performing multi-source information fusion calculations on the motion state parameters to generate a state fusion vector; a risk generation module for extracting and constructing dynamic risk features from the state fusion vector to generate a risk field strength; a mapping calculation module for performing asymmetric damping mapping calculations based on the risk field strength to generate an asymmetric damping curve; and a loading control calculation module. The module is used to synthesize micro-periodic pulsations of the air path based on the asymmetric damping curve and to perform preload calculations of the mechanical fall arrestor based on the risk field strength, thereby obtaining the air pulsation control quantity and the mechanical preload quantity respectively; the execution feedback module is used to generate a coordinated execution command based on the air pulsation control quantity and the mechanical preload quantity, and to execute the coordinated execution command to control the slide operation, while collecting deviation data after execution; the deviation data and the risk field strength are used to construct a time-series trajectory to generate a state memory trajectory; and the asymmetric damping mapping calculation is calibrated and corrected based on the state memory trajectory.

[0007] Furthermore, the data acquisition module includes: acquiring the stroke position parameters of the slide table through a displacement sensor; processing the stroke position parameters through a speed calculation unit to obtain real-time speed parameters; acquiring the upper chamber pressure parameters and lower chamber pressure parameters of the cylinder through a pressure sensor, and calculating the difference between the two to obtain pressure difference parameters; and integrating the stroke position parameters, the real-time speed parameters, and the pressure difference parameters to obtain motion state parameters.

[0008] Furthermore, the state fusion module includes: standardizing the motion state parameters to eliminate the influence of dimensions and obtain standardized state parameters; performing principal component analysis on the standardized state parameters to extract key feature dimensions and obtain a dimensionality-reduced feature set; and inputting the dimensionality-reduced feature set into a state observer for estimating the system state for data fusion to generate a state fusion vector.

[0009] Furthermore, the risk generation module includes: extracting velocity fluctuation components and pressure fluctuation components from the state fusion vector to obtain a dynamic fluctuation sequence; calculating the time-frequency domain characteristics of the dynamic fluctuation sequence to obtain a fluctuation feature matrix; inputting the fluctuation feature matrix into a risk assessment model for evaluating system risk, and outputting the risk field strength.

[0010] Furthermore, the mapping calculation module includes: querying a damping mapping table that defines the upper cavity pressure response relationship based on the risk field strength to obtain the upper cavity back pressure control curve; querying a damping mapping table that defines the lower cavity exhaust response relationship based on the risk field strength to obtain the lower cavity exhaust control curve; and performing asymmetric coupling between the upper cavity back pressure control curve and the lower cavity exhaust control curve to generate an asymmetric damping curve.

[0011] Furthermore, the loading control calculation module includes: performing micro-period discretization processing on the asymmetric damping curve to obtain the basic damping control quantity; superimposing a high-frequency pulsating signal with a specific frequency and amplitude on the basic damping control quantity to obtain the pneumatic pulsation control quantity; and querying a preload force mapping relationship that defines the preload force correspondence based on the amplitude characteristics of the risk field strength to obtain the mechanical preload quantity.

[0012] Furthermore, the execution feedback module includes: generating a coordinated execution command based on the pneumatic pulse control quantity and the mechanical preload quantity; sending the coordinated execution command to the actuator and collecting the actual running trajectory of the slide to obtain execution deviation data; associating and storing the execution deviation data with the corresponding risk field strength in a time series to generate a state memory trajectory; performing pattern recognition on the state memory trajectory to extract system drift features and obtain a calibration factor; and using the calibration factor to correct the damping mapping table used in the asymmetric damping mapping calculation.

[0013] Furthermore, the system also includes: identifying over- or under-loading mechanical preload patterns in historical control processes based on the state memory trajectory; dynamically adjusting the calculation method of the mechanical preload amount according to the over- or under-loading mechanical preload patterns; and generating a preload strategy that is co-optimized with the pneumatic pulse control quantity.

[0014] Furthermore, the system also includes: identifying the prediction bias of the risk assessment model based on the execution deviation data in the state memory trajectory; extracting new risk feature samples using the prediction bias; and updating the risk assessment model online using the new risk feature samples.

[0015] Based on the same inventive concept, this invention also provides an intelligent control method for a dovetail-groove slide cylinder used in surge protector manufacturing. The method includes: acquiring the slide's stroke position parameters, real-time speed parameters, and pressure difference parameters between the upper and lower chambers of the cylinder to obtain motion state parameters; performing multi-source information fusion calculation on the motion state parameters to generate a state fusion vector; extracting and constructing dynamic risk features from the state fusion vector to generate a risk field strength; performing asymmetric damping mapping calculation based on the risk field strength to generate an asymmetric damping curve; synthesizing micro-periodic pulsations in the airflow based on the asymmetric damping curve, and performing pre-loading calculation on the mechanical fall arrestor based on the risk field strength to obtain pneumatic pulsation control quantities and mechanical pre-loading quantities, respectively; generating a collaborative execution command based on the pneumatic pulsation control quantities and the mechanical pre-loading quantities, and executing the collaborative execution command to control the slide's operation, while simultaneously collecting deviation data after execution; constructing a time-series trajectory based on the deviation data and the risk field strength to generate a state memory trajectory; and calibrating and correcting the asymmetric damping mapping calculation based on the state memory trajectory.

[0016] Compared with the prior art, the present invention has the following advantages: This invention achieves forward-looking prediction of fall risk by real-time sensing and fusion of the motion state of the slide table and constructing a dynamic risk field strength. It transforms the traditional passive triggering safety protection into active preventive intervention, which can intervene at the incipient stage of system instability, thereby improving the inherent safety of equipment operation.

[0017] This invention employs a synergistic control strategy that combines pneumatic pulse damping with mechanical fall arrest preloading. When the system is running smoothly, precise pneumatic circuit adjustment ensures smooth and efficient movement. When the risk increases, the preloading force of the mechanical mechanism can be seamlessly introduced, achieving an organic unity of flexible control and rigid protection. This balances the system's high performance and high safety, avoiding the performance sacrifices or safety deficiencies caused by traditional single protection methods.

[0018] This invention establishes a closed-loop adaptive learning mechanism based on state memory trajectories. The system can continuously analyze the correlation between historical execution deviations and risk data, calibrate its control model online, and dynamically optimize preloading strategies. This compensates for system characteristic drift caused by factors such as mechanical wear, load changes, or environmental disturbances, ensuring the long-term stability and high precision of the control system throughout its entire lifecycle.

[0019] This invention achieves graded control of intervention intensity by quantifying risk assessment and generating an asymmetric damping curve. The system can apply appropriate control based on the risk level, avoiding the shocks, vibrations, and unnecessary shutdowns caused by the overreaction of traditional safety devices under low-risk disturbances, thus ensuring the continuity and stability of the production process. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 This is a schematic diagram of the intelligent control system for a dovetail groove slide cylinder used in the manufacture of surge protectors, according to an embodiment of the present invention.

[0022] Figure 2 This is a schematic diagram illustrating the mapping relationship between risk field strength and damping in an embodiment of the present invention.

[0023] Figure 3 This is a flowchart illustrating the intelligent control method of a dovetail groove slide cylinder for manufacturing surge protectors according to an embodiment of the present invention. Detailed Implementation

[0024] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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, 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.

[0025] Reference Figure 1 One embodiment of the present invention proposes an intelligent control system for a dovetail slide cylinder used in the manufacture of surge protectors. By fusion of multi-source information to perceive risks, and by coordinating pneumatic flexible damping and mechanical rigid preloading for closed-loop control, it can achieve proactive prediction and smooth intervention of fall risks while ensuring high response speed.

[0026] The system described in this embodiment specifically includes: The data acquisition module is used to acquire the stroke position parameters, real-time speed parameters, and pressure difference parameters of the upper and lower chambers of the cylinder of the slide table, so as to obtain the motion state parameters. The state fusion module is used to perform multi-source information fusion calculation on the motion state parameters to generate a state fusion vector. The risk generation module is used to dynamically extract and construct risk features from the state fusion vector to generate a risk field strength. The mapping calculation module is used to perform asymmetric damping mapping calculation based on the risk field strength and generate an asymmetric damping curve. The loading control calculation module is used to synthesize the micro-periodic pulsation of the air path based on the asymmetric damping curve, and to perform the preload calculation of the mechanical fall arrest mechanism based on the risk field strength, so as to obtain the air pulsation control quantity and the mechanical preload quantity respectively. The execution feedback module is used to generate a coordinated execution command based on the pneumatic pulse control quantity and the mechanical preload quantity, and execute the coordinated execution command to control the slide table operation, while collecting deviation data after execution; constructing a time-series trajectory for the deviation data and the risk field strength to generate a state memory trajectory; and calibrating and correcting the asymmetric damping mapping calculation based on the state memory trajectory.

[0027] Furthermore, the data acquisition module includes: The travel position parameters of the slide are collected by a displacement sensor; The travel position parameters are processed by the speed calculation unit to obtain the real-time speed parameters; The pressure difference parameter is obtained by collecting the pressure parameters of the upper chamber and the lower chamber of the cylinder through a pressure sensor and calculating the difference between the two. By integrating the stroke position parameters, the real-time speed parameters, and the pressure difference parameters, motion state parameters are obtained.

[0028] The data acquisition module uses high-frequency synchronous sampling to obtain key physical quantities characterizing the operating posture of the dovetail-groove slide cylinder, providing a raw data foundation for subsequent state fusion and risk assessment. This module is implemented by first deploying a magnetic grating displacement sensor with a measurement range of 0 to 500 mm and an accuracy better than 0.01 mm on the fixed base of the slide. This sensor is arranged along the slide's movement direction to continuously acquire the slide's stroke position parameter P(t). The sensor's data output interface is connected to the control system's I / O unit, and the sampling frequency is set between 1 and 5 kHz to ensure the capture of high-speed motion details.

[0029] The acquisition of real-time velocity parameters is accomplished by a velocity calculation unit integrated within the main controller. This unit receives a discrete travel position parameter sequence from the displacement sensor and processes it using a differential algorithm with low-pass filtering to suppress high-frequency noise introduced by mechanical vibration or electrical interference in the displacement signal. The real-time velocity parameter V(t) is calculated using the following formula: , Where V(t) is the real-time velocity parameter at the current moment, P(t) is the travel position parameter at the current moment, t-Δt represents the previous sampling moment, and a is a filter coefficient with a value between 0.8 and 0.95, used to smooth the velocity calculation result.

[0030] The pressure difference parameter is obtained by installing piezoelectric pressure sensors at the inlet and outlet ports of the upper and lower chambers of the cylinder, respectively, with a measurement range covering 0 to 1.6 MPa. These two sensors measure the pressure parameter of the upper chamber in real time. With lower chamber pressure parameters Pressure difference parameters It directly reflects the direction and magnitude of the net driving force acting on the piston, and is obtained by performing a real-time difference calculation on the pressure parameters of the two chambers: , The calculation process is completed inside the controller, and the positive or negative value of the result directly indicates the direction of the force on the cylinder.

[0031] Within each sampling period, the controller integrates the collected and calculated stroke position parameter P(t), real-time speed parameter V(t), and pressure difference parameter ΔP(t) into a three-dimensional vector, which serves as the motion state parameter S(t) for that moment. This motion state parameter S(t) is sent to the state fusion module in the form of a data packet as the standardized output of the data acquisition module for further processing, thereby completing a comprehensive digital representation of the physical state of the slide cylinder.

[0032] Furthermore, the state fusion module includes: The motion state parameters are standardized to eliminate the influence of dimensions and obtain standardized state parameters. Principal component analysis was performed on the standardized state parameters to extract key feature dimensions and obtain a dimensionality-reduced feature set. The reduced feature set is input into a state observer used to estimate the system state for data fusion to generate a state fusion vector.

[0033] The state fusion module transforms the original motion state parameters from different physical dimensions and with varying dimensions into a state fusion vector with lower dimensionality, higher signal-to-noise ratio, and the ability to comprehensively characterize the core dynamic properties of the system. The first step in this module is standardization, which aims to eliminate calculation biases caused by differences in units and numerical ranges among the three parameters: travel position, real-time velocity, and pressure difference. For each component of the motion state parameter S(t) obtained in each sampling period, the Z-score normalization method is used to calculate and generate standardized state parameters. : , in It is standardized The i-th component, It is the original i-th component. and These are the mean and standard deviation of the component, calculated over a recent sliding time window of 1 to 2 seconds. This dynamic calculation method allows the standardization process to adapt to changes in operating conditions.

[0034] Next, the module standardizes the state parameters. Principal component analysis (PCA) is performed on the sequence to achieve data dimensionality reduction and feature condensation. By constructing the covariance matrix of standardized state parameters and performing eigenvalue decomposition, principal components with a cumulative contribution rate exceeding 95% are extracted. Under normal operating conditions, the first two principal components can effectively summarize most of the dynamic information of the system. This step transforms the three-dimensional standardized state parameters... Projecting onto a two-dimensional feature space composed of principal components forms a dimension-reduced feature set Z(t) that can capture the coupling relationship between the main motion and the force.

[0035] To estimate the true internal state of the system from a noisy, reduced-dimensional feature set and incorporate the dynamic correlations over time into the analysis, the reduced-dimensional feature set Z(t) is input into a state observer for data fusion. In this embodiment, a Kalman filter is used as the state observer. The Kalman filter uses a pre-established system dynamic model, combined with the current reduced-dimensional feature set Z(t) as the observation, to perform an optimal estimate of the system state. Its core update equation is: , Where X(t) is the state fusion vector output at the current time. The current state is estimated prior to the state predicted at the previous time step. K(t) is the Kalman gain dynamically calculated based on the covariance of system noise and observation noise, and H is the observation matrix connecting the state space and the observation space. After iterative computation by the Kalman filter, the final generated state fusion vector X(t) not only fuses information from multiple sensors but also smooths out measurement noise through filtering, providing a more realistic estimate of the system state and offering high-quality input for the subsequent risk generation module.

[0036] Furthermore, the risk generation module includes: The velocity fluctuation component and pressure fluctuation component are extracted from the state fusion vector to obtain the dynamic fluctuation sequence; Calculate the time-frequency domain characteristics of the dynamic fluctuation sequence to obtain the fluctuation feature matrix; The fluctuation characteristic matrix is ​​input into a risk assessment model used to evaluate system risk, and the risk field strength is output.

[0037] The risk generation module transforms the state fusion vector, which indirectly reflects the system state, into a physical metric that directly and quantitatively characterizes the tendency of the slide to become unstable, namely the risk field strength, thus providing a basis for subsequent preventive control decisions. The first step in this process is to separate the core components representing the dynamic anomalies of the system from the state fusion vector X(t). By applying a digital high-pass filter with a cutoff frequency set at approximately 5 Hz to the components of the state fusion vector containing velocity and pressure information, the low-frequency command follow-up portion is filtered out, thereby extracting the high-frequency dynamic fluctuations, resulting in velocity fluctuation components and pressure fluctuation components. These two time series together constitute the dynamic fluctuation sequence.

[0038] Next, to comprehensively describe the characteristics of the dynamic fluctuation sequence, the module employs time-frequency domain analysis to calculate its features. The system continuously extracts data segments from the dynamic fluctuation sequence using a sliding time window with 256 sampling points and a 50% overlap. For each data window, the system uses wavelet packet transform to decompose it into 3 to 5 layers, thereby obtaining the energy distribution information of the signal in different frequency bands. Based on this, a series of time-frequency domain features are calculated, such as the total energy E, energy concentration C, and energy entropy Hs. These features can characterize the uncertainty and intensity of the fluctuation from multiple dimensions, including energy magnitude and the concentration and dispersion of energy distribution.

[0039] , , Where E is the total energy of the signal within the window. It is the energy of the j-th frequency band. It represents the proportion of energy in the j-th frequency band to the total energy. These calculated features are combined into a multidimensional wave characteristic matrix, which is dynamically updated in each calculation cycle, accurately describing the microscopic morphology of the system's vibration and pressure shock at the current moment.

[0040] Finally, the real-time generated fluctuation feature matrix is ​​used as input and fed into a risk assessment model pre-trained with offline data. This embodiment employs a regression model based on a radial basis function (RBF) neural network. This network structure includes an input layer with the same number of nodes as the fluctuation feature matrix, a hidden layer with 10 to 20 neurons, and an output layer with a single output node. The network has been trained using a large amount of historical data, including fluctuation feature matrices and their corresponding expert-assessed risk levels, ranging from normal operation to the occurrence of various failure precursors such as stall, jitter, and crash. Therefore, the model can non-linearly map a continuous value between 0 and 1 based on the input real-time fluctuation features. This output value is defined as the risk field strength. When the risk field strength is low, it indicates that the system is operating smoothly and the risk is controllable; when the value increases and exceeds a preset threshold, such as 0.7, it indicates that the system has entered a high-risk state, with a greater possibility of instability or crash.

[0041] Furthermore, the mapping calculation module includes: Based on the risk field strength, a damping mapping table defining the upper cavity pressure response relationship is queried to obtain the upper cavity back pressure control curve; Based on the risk field strength, a damping mapping table defining the lower chamber exhaust response relationship is queried to obtain the lower chamber exhaust control curve; The upper cavity back pressure control curve and the lower cavity exhaust control curve are asymmetrically coupled to generate an asymmetric damping curve.

[0042] like Figure 2 As shown, the mapping calculation module converts the abstract risk field strength scalar into a specific physical quantity target curve that can be directly used for pneumatic path control in real time, providing a dynamic and asymmetric basis for damping adjustment of the system. This process begins by querying a pre-set upper cavity damping mapping table in the controller's storage unit based on the real-time risk field strength output by the risk generation module. This mapping table is essentially a nonlinear function or discrete data lookup table obtained through extensive experimental calibration, defining the quantitative relationship between the risk field strength and the upper cavity target back pressure. When the risk field strength increases, the table will output a higher target back pressure value to form a stronger aerodynamic buffer layer above the slide. This sequence of target back pressure values ​​that dynamically changes with the risk field strength constitutes the upper cavity back pressure control curve.

[0043] Simultaneously, the controller uses the same risk field strength value to query another independent lower chamber damping mapping table. This mapping table defines the response relationship between the risk field strength and the degree of lower chamber exhaust throttling, typically outputting a signal value that controls the opening of the exhaust proportional valve. Its mapping logic is the opposite of that of the upper chamber; that is, as the risk field strength increases, the target opening of the exhaust valve obtained from the table decreases, thereby limiting the exhaust rate of the lower chamber gas and enhancing the back pressure braking effect. This series of dynamic exhaust control target values ​​forms the lower chamber exhaust control curve. The mapping relationship within these two damping mapping tables is established based on the dynamic characteristics of the cylinder under different loads and speeds through a system identification method.

[0044] Finally, the module asymmetrically couples the upper chamber back pressure control curve and the lower chamber exhaust control curve to generate a unified asymmetric damping curve. This asymmetric coupling is an engineering concept, not a simple mathematical addition, but rather refers to combining two independent control targets with different response characteristics into a single set of coordinated action instructions. The asymmetry is reflected in the fact that the same increase in risk field strength may lead to a significant increase in the upper chamber pressure target value, while the lower chamber exhaust valve opening may only slightly decrease. This strategy is specifically optimized for the anti-fall scenario of vertically mounted cylinders. The final generated asymmetric damping curve is formally a two-dimensional vector, where the two components represent the target pressure of the upper chamber and the target opening of the lower chamber exhaust valve at the current moment, respectively. This vector is transmitted as a whole instruction to the loading control calculation module.

[0045] , in, Let be the vector of the asymmetric damping curve at time t. The target value for upper cavity back pressure control is obtained at time t by querying the upper cavity damping mapping table. The target value for lower chamber exhaust control is obtained by querying the lower chamber damping mapping table at time t.

[0046] Furthermore, the loading control calculation module includes: The asymmetric damping curve is discretized using micro-period processing to obtain the basic damping control quantity; A high-frequency pulsating signal with a specific frequency and amplitude is superimposed on the basic damping control quantity to obtain the pneumatic pulse control quantity. Based on the amplitude characteristics of the risk field strength, a preload force mapping relationship that defines the preload force correspondence is queried to obtain the mechanical preload amount.

[0047] The loading control calculation module transforms the macroscopic control strategy generated in the previous step, namely the asymmetric damping curve, into a microscopic control signal that can be directly driven by the actuator, and simultaneously calculates the preload force of the mechanical fall arrestor coordinated with the pneumatic control. This module first performs micro-period discretization processing on the asymmetric damping curve output by the mapping calculation module. This process involves sampling the asymmetric damping curve at a fixed time step, such as 1 to 2 milliseconds, within the controller's main control cycle to obtain a series of discrete target values. These target values ​​constitute the basic damping control quantity. This basic damping control quantity is the fundamental command for driving the pneumatic proportional valve to achieve smooth pressure regulation.

[0048] To improve the dynamic responsiveness and control accuracy of the pneumatic system, especially to overcome the static friction and hysteresis effects of the valve core during minute pressure regulation, the module superimposes a high-frequency pulsating signal onto the basic damping control quantity to synthesize the final pneumatic pulsating control quantity. This high-frequency pulsating signal is typically a sine wave or a triangular wave, with its frequency set outside the valve's mechanical response frequency and the system's dominant vibration frequency, generally between 100 and 300 Hz. The amplitude is set according to the valve characteristics, typically 1% to 5% of its control voltage range. The final pneumatic pulsating control quantity... Generated by the following formula: , in, It is the pneumatic pulse control quantity in the k-th control cycle. This is the corresponding basic damping control quantity, where A is the set pulsation amplitude, f is the pulsation frequency, and t is the current time.

[0049] Simultaneously, the module uses the risk field strength output by the risk generation module to calculate the preload of the mechanical fall arrestor. The core of this step is to look up a preset preload force mapping relationship. This mapping relationship is a non-linear function or lookup table that maps a risk field strength ranging from 0 to 1 to a specific preload force value, such as 0 to 50 Newtons. The mapping logic is as follows: when the risk field strength is low, the preload is zero or a very small value to avoid unnecessary friction and energy consumption; as the risk field strength increases, the preload increases non-linearly, causing the friction plates or wedges of the mechanical fall arrestor to make slight contact with the slide rail in advance, eliminating transmission gaps and entering a standby state. When the risk field strength reaches a high-risk threshold, the preload also approaches its set upper limit, ensuring that in the event of stall or fall, the mechanical mechanism can instantly and without delay generate full braking force. This calculation process directly outputs the mechanical preload based on the amplitude characteristics of the risk field strength, realizing active, feedforward loading for mechanical safety redundancy.

[0050] Furthermore, the execution feedback module includes: Generate a coordinated execution command based on the pneumatic pulse control quantity and the mechanical preload quantity; The collaborative execution command is sent to the actuator, and the actual running trajectory of the slide is collected to obtain execution deviation data; The execution deviation data and the corresponding risk field strength are associated and stored in a time series to generate a state memory trajectory; Pattern recognition is performed on the state memory trajectory to extract system drift features and obtain calibration factors; The damping mapping table used in the asymmetric damping mapping calculation is corrected using the calibration factor.

[0051] The execution feedback module constructs a closed-loop adaptive learning mechanism. By continuously monitoring the execution effect of control commands and using execution deviation data to correct the core control model of the system, it addresses characteristic drift caused by wear, load changes, or environmental disturbances. The first step of this module is to integrate the pneumatic pulse control quantity output from the loading control calculation module with the mechanical preload quantity into a coordinated execution command. This command is a structured data packet containing pulsating voltage signals for driving the upper and lower chamber pneumatic proportional valves, as well as the target force or displacement signal for driving the mechanical preload mechanism. This signal is synchronously sent to the corresponding actuator via a digital-to-analog conversion channel and the drive circuit.

[0052] While executing the coordinated execution instructions, the module uses displacement sensors in the data acquisition module to collect the actual running trajectory of the slide table at the same sampling frequency. By comparing the actual running trajectory with the expected running trajectory generated internally by the instructions, the execution deviation data is calculated.

[0053] , in, These are the execution deviation data at time t. These are the desired travel position parameters. This is the actual stroke position parameter fed back by the displacement sensor. This deviation sequence directly reflects the control accuracy. The module will process the execution deviation data at each sampling moment. Risk field strength corresponding to that moment The data are correlated and timestamped to form a data triple. These data triples are then sequentially stored in a first-in, first-out (FIFO) circular buffer, sized to store data from the most recent 1000 to 5000 sampling points, thus constructing a dynamically updated state memory trajectory. This trajectory records the system's control performance under different risk levels, serving as the data foundation for the system's self-evaluation and optimization.

[0054] The module periodically or under specific event triggers performs pattern recognition on the state memory trajectory. Its engineering implementation involves analyzing the statistical characteristics of execution deviation data, such as mean and variance, within a specific risk field strength range. If it is found that the mean deviation continuously deviates from zero within a certain risk field strength range, exceeding a set threshold, such as 0.05 mm, the system determines that control model drift exists under this operating condition. This continuous mean deviation is considered the system drift characteristic. Based on the magnitude and direction of this characteristic, the system calculates a calibration factor K.

[0055] , Where K is the value of the risk field strength. and The calibration factor for the interval, It is a proportional gain coefficient. This is the average value of the deviation data executed within the risk range. The system uses this calibration factor to locally correct the damping mapping table used in the asymmetric damping mapping calculation, such as fine-tuning the target value of pressure or valve opening in the corresponding risk range, thereby compensating for system drift and achieving online self-calibration of control accuracy.

[0056] Furthermore, the system also includes: Based on the state memory trajectory, identify the mechanical preload overload or underload patterns in the historical control process; The calculation method for dynamically adjusting the mechanical preload amount is based on the aforementioned excessive or insufficient mechanical preload mode. Generate a preloading strategy that is co-optimized with the pneumatic pulse control quantity.

[0057] The additional optimization function in this embodiment is based on long-term historical operating data of the system to self-optimize the mechanical preloading strategy. This ensures safety redundancy while minimizing interference with normal motion, thereby improving the energy efficiency and responsiveness of pneumatic-mechanical coordinated control. This function firstly executes a pattern recognition algorithm periodically based on the complete state memory trajectory generated by the execution feedback module. This algorithm focuses on analyzing the effectiveness of the mechanical preloading amount.

[0058] The identification criteria for the insufficient mechanical preload mode are as follows: when the risk field strength recorded in the state memory trajectory instantaneously jumps to a high level, such as exceeding 0.8, and the subsequent execution deviation data shows a spike or oscillation exceeding the dynamic stability threshold, it indicates that even with preload applied, the mechanical mechanism has failed to effectively suppress initial instability. Conversely, the excessive mechanical preload mode is identified when the risk field strength is in the low to medium range, such as 0.2 to 0.6, and a certain amount of mechanical preload is applied to the system, but the execution deviation data in the state memory trajectory shows a continuous tracking lag, while the cylinder pressure difference parameter is significantly increased compared to the case without preload. This indicates that the frictional force generated by preload unnecessarily increases the resistance to system motion.

[0059] Once the above patterns are identified, the system will dynamically adjust the calculation method for the mechanical preload, which refers to the preload force mapping relationship used in the loading control calculation module. If an insufficient preload pattern is identified, the system will apply a gain coefficient to the output value in the high-risk range of the preload force mapping relationship, for example, increasing the target preload force value corresponding to a risk field strength greater than 0.7 by 5% to 15%. If an overload pattern is identified, a decay coefficient will be applied to the output value in the medium-to-low-risk range, for example, decreasing the target preload force value corresponding to a risk field strength between 0.2 and 0.6 by 10% to 20%. This adjustment is not a one-time event, but rather an iterative accumulation based on the frequency and severity of the identified patterns, thereby generating a preload strategy that is co-optimized with the pneumatic dynamic control quantity. This new strategy enables the mechanical fall arrestor to provide stronger and more timely support when truly needed, while remaining as inconspicuous as possible during stable system operation, avoiding the negative impact of safety functions on system performance.

[0060] Furthermore, the system also includes: Based on the execution deviation data in the state memory trajectory, the prediction deviation of the risk assessment model is identified; Using the prediction bias, extract new risk feature samples; The risk assessment model is updated online using the new risk feature samples.

[0061] The added self-learning function in this embodiment enables the risk assessment model to self-correct and evolve online, ensuring that its prediction accuracy for fall risk does not decrease due to long-term evolution of the system state or the occurrence of unforeseen operating conditions. This function first identifies prediction biases in the risk assessment model based on the continuously updated state memory trajectory in the execution feedback module. The system monitors a specific mismatch pattern in real time: the risk field strength output by the risk assessment model remains within a low, safe range, for example, below 0.3, but the actual execution bias data shows a sharp peak exceeding the normal fluctuation range at this moment or shortly thereafter, for example, an absolute deviation exceeding 0.5 mm. This indicates a missed risk event, i.e., a prediction bias.

[0062] Once this prediction bias is identified, the system marks the event point. Then, the system backtracks to a short time window before the event, such as 50 to 100 milliseconds prior, and extracts a continuous fluctuation feature matrix from the historical data buffer for that period. This sequence, containing the original feature data that led to the missed event, along with the relabeled high-risk event outcomes, constitutes a new risk feature sample. The key to this new sample is that it provides the model with a dangerous pattern that was previously misidentified, serving as valuable counterexample learning material.

[0063] Finally, the system uses this newly extracted risk feature sample to update the original risk assessment model, namely the radial basis function neural network, online. This update process employs an incremental learning algorithm, rather than complete retraining, to ensure the system's real-time performance. Specifically, the new risk feature sample is input into the network, and the error between its current output and our expected high-risk label is calculated. Then, based on this error, gradient descent is used to fine-tune the weights and center points associated with the few hidden layer neurons with the highest activation levels of that sample. The update step size, i.e., the learning rate, is set to a small value, such as 0.01 to 0.05, to avoid a single sample having an excessive impact on the model and to ensure stable convergence of the learning process. By continuously repeating this cycle of identification, extraction, and updating, the risk assessment model can continuously learn from its own errors, gradually improving its knowledge base, thereby possessing increasingly stronger generalization ability and warning accuracy for various potential fall risks.

[0064] Based on the same inventive concept, such as Figure 3 As shown, the present invention also provides an intelligent control method for a dovetail-groove slide cylinder used in the manufacture of surge protectors, the method comprising: The motion state parameters are obtained by acquiring the stroke position parameters of the slide, the real-time speed parameters, and the pressure difference parameters of the upper and lower chambers of the cylinder. The motion state parameters are fused from multiple sources to generate a state fusion vector. Dynamic risk feature extraction and construction are performed on the state fusion vector to generate a risk field strength; Based on the risk field strength, perform asymmetric damping mapping calculation to generate an asymmetric damping curve; Based on the asymmetric damping curve, the micro-periodic pulsation of the air path is synthesized, and based on the risk field strength, the preload calculation of the mechanical fall arrest mechanism is performed to obtain the air pulsation control quantity and the mechanical preload quantity, respectively. Based on the pneumatic pulse control quantity and the mechanical preload quantity, a coordinated execution command is generated and executed to control the slide table operation, while simultaneously collecting deviation data after execution; A time-series trajectory is constructed from the deviation data and the risk field strength to generate a state memory trajectory; the asymmetric damping mapping calculation is calibrated and corrected based on the state memory trajectory.

[0065] This invention is applied to a vertical transport unit for surge protectors in electrical equipment manufacturing equipment. This unit uses a dovetail-groove slide cylinder to rapidly and smoothly raise and lower a robotic arm carrying a precision surge protector in the vertical direction. Because surge protectors require high stability and positioning accuracy from the slide, and absolutely no stalling or falling due to air pressure fluctuations or sudden load changes is permissible.

[0066] A dovetail-groove type slide cylinder with a total stroke of 500 mm was selected, with a slide load of 5 kg, and equipped with a high-response proportional control valve. The core processor of the control system operates on a 2-millisecond (500 Hz) cycle. The test procedure was set as follows: the slide moved downwards from the top starting position (position = 500 mm) to the target position (position = 100 mm), and when it reached the middle of the stroke, a disturbance of 20% instantaneous pressure drop in the air source was artificially introduced for 100 milliseconds to simulate a sudden abnormal working condition and to test the risk identification and fall prevention control capabilities of the present invention.

[0067] The system uses a magnetic grating displacement sensor with a precision of 0.01 mm, a piezoelectric pressure sensor, and an internal processing unit to collect and calculate the slide's stroke position P(t), real-time speed V(t), and the pressure difference ΔP(t) between the upper and lower chambers of the cylinder in real time. At t=2.500 seconds, the slide runs smoothly, and the collected motion state parameters S(2.500s)=[P:350.2mm,V:-0.40m / s,ΔP:0.15MPa]. After Z-score normalization and principal component analysis (PCA) dimensionality reduction, the parameters are input into a Kalman filter to generate a stable state fusion vector X(2.500s).

[0068] Under steady-state conditions, the state fusion vector exhibits minimal fluctuations. The system calculates the fluctuation feature matrix through high-pass filtering and time-frequency domain analysis, inputs it into a trained radial basis function (RBF) neural network risk assessment model, and obtains a risk field strength of 0.15, which is far below the warning threshold of 0.7.

[0069] Based on a risk field strength of 0.15, the mapping calculation module queries its internal mapping table to obtain a target back pressure of 0.18 MPa in the upper chamber and an exhaust valve opening of 90% in the lower chamber. These are the basic damping control values. The loading control calculation module then superimposes a high-frequency pulsating signal with a frequency of 200 Hz and an amplitude of 2% of the control range onto this signal to generate a pneumatic pulse control value. Simultaneously, the preload force mapping relationship is queried, revealing a mechanical preload of 0 Newtons. The system executes the coordinated command, and the slide descends smoothly.

[0070] At t=2.600 seconds, a sudden drop in air source pressure began. The data acquisition module detected the early abnormal signal through pressure fluctuation at t=2.605 seconds, and the risk generation module responded quickly, with the risk field strength climbing to 0.65 within 10 milliseconds. At this time, the slide speed fluctuated slightly, briefly dropping from -0.40 m / s to -0.38 m / s. Based on the risk field strength of 0.65, the mapping calculation module immediately performed asymmetric damping mapping calculations, consulting the upper chamber damping mapping table to find that the target back pressure in the upper chamber increased to 0.55 MPa (enhanced buffering), and the opening of the lower chamber exhaust valve narrowed to 40% (enhanced braking). The loading control calculation module generated the pneumatic pulse control quantity accordingly, while the mechanical preload was 30 Newtons. The coordinated execution command was issued, the pneumatic proportional valve responded quickly to adjust the pressure, and the friction plate of the mechanical anti-fall mechanism was preloaded with a force of 30 Newtons, making slight contact with the guide rail and entering a standby state.

[0071] Thanks to the active intervention of the pneumatic-mechanical coordination, the descent speed of the slide table stabilized at approximately -0.39 m / s after a brief fluctuation, preventing further stalling or descent. At t=2.612 seconds, the desired position was 349.250 mm, and the actual acquired position was 349.260 mm, with an execution deviation of -0.010 mm. The controller stored the data triplet in the state memory trajectory buffer. After the disturbance ended, the risk field strength decreased, the mechanical preload returned to zero, the pneumatic damping returned to normal, and the slide table continued to run smoothly to the target point without any stalling.

[0072] All equivalent changes and modifications made in accordance with the teachings of this invention are still within the scope of this invention. Those skilled in the art will readily conceive of other embodiments of this invention upon considering the specification and the disclosure of practical truth. This application is intended to cover any variations, uses, or adaptations of this invention that follow the general principles of this invention and include common knowledge or conventional techniques in the art not described herein.

Claims

1. An intelligent control system for a dovetail groove type slide cylinder used in the manufacture of surge protectors, characterized in that, The system includes: The data acquisition module is used to acquire the stroke position parameters, real-time speed parameters, and pressure difference parameters of the upper and lower chambers of the cylinder of the slide table, so as to obtain the motion state parameters. The state fusion module is used to perform multi-source information fusion calculation on the motion state parameters to generate a state fusion vector. The risk generation module is used to dynamically extract and construct risk features from the state fusion vector to generate a risk field strength. The mapping calculation module is used to perform asymmetric damping mapping calculation based on the risk field strength and generate an asymmetric damping curve. The loading control calculation module is used to synthesize the micro-periodic pulsation of the air path based on the asymmetric damping curve, and to perform the preload calculation of the mechanical fall arrest mechanism based on the risk field strength, so as to obtain the air pulsation control quantity and the mechanical preload quantity respectively. The execution feedback module is used to generate a coordinated execution command based on the pneumatic pulse control quantity and the mechanical preload quantity, and execute the coordinated execution command to control the slide table operation, while collecting deviation data after execution; constructing a time-series trajectory for the deviation data and the risk field strength to generate a state memory trajectory; and calibrating and correcting the asymmetric damping mapping calculation based on the state memory trajectory.

2. The intelligent control system for the dovetail groove type slide cylinder used in the manufacture of surge protectors according to claim 1, characterized in that, The data acquisition module includes: The travel position parameters of the slide are collected by a displacement sensor; The travel position parameters are processed by the speed calculation unit to obtain the real-time speed parameters; The pressure difference parameter is obtained by collecting the pressure parameters of the upper chamber and the lower chamber of the cylinder through a pressure sensor and calculating the difference between the two. By integrating the stroke position parameters, the real-time speed parameters, and the pressure difference parameters, motion state parameters are obtained.

3. The intelligent control system for the dovetail groove type slide cylinder used in the manufacture of surge protectors according to claim 2, characterized in that, The state fusion module includes: The motion state parameters are standardized to eliminate the influence of dimensions and obtain standardized state parameters. Principal component analysis was performed on the standardized state parameters to extract key feature dimensions and obtain a dimensionality-reduced feature set. The reduced feature set is input into a state observer used to estimate the system state for data fusion to generate a state fusion vector.

4. The intelligent control system for the dovetail groove type slide cylinder used in the manufacture of surge protectors according to claim 3, characterized in that, The risk generation module includes: The velocity fluctuation component and pressure fluctuation component are extracted from the state fusion vector to obtain the dynamic fluctuation sequence; Calculate the time-frequency domain characteristics of the dynamic fluctuation sequence to obtain the fluctuation feature matrix; The fluctuation characteristic matrix is ​​input into a risk assessment model used to evaluate system risk, and the risk field strength is output.

5. The intelligent control system for the dovetail groove type slide cylinder used in the manufacture of surge protectors according to claim 4, characterized in that, The mapping calculation module includes: Based on the risk field strength, a damping mapping table defining the upper cavity pressure response relationship is queried to obtain the upper cavity back pressure control curve; Based on the risk field strength, a damping mapping table defining the lower chamber exhaust response relationship is queried to obtain the lower chamber exhaust control curve; The upper cavity back pressure control curve and the lower cavity exhaust control curve are asymmetrically coupled to generate an asymmetric damping curve.

6. The intelligent control system for the dovetail groove type slide cylinder used in the manufacture of surge protectors according to claim 5, characterized in that, The loading control calculation module includes: The asymmetric damping curve is discretized using micro-period processing to obtain the basic damping control quantity; A high-frequency pulsating signal with a specific frequency and amplitude is superimposed on the basic damping control quantity to obtain the pneumatic pulse control quantity. Based on the amplitude characteristics of the risk field strength, a preload force mapping relationship that defines the preload force correspondence is queried to obtain the mechanical preload amount.

7. The intelligent control system for the dovetail groove type slide cylinder used in the manufacture of surge protectors according to claim 6, characterized in that, The execution feedback module includes: Generate a coordinated execution command based on the pneumatic pulse control quantity and the mechanical preload quantity; The collaborative execution command is sent to the actuator, and the actual running trajectory of the slide is collected to obtain execution deviation data; The execution deviation data and the corresponding risk field strength are associated and stored in a time series to generate a state memory trajectory; Pattern recognition is performed on the state memory trajectory to extract system drift features and obtain calibration factors; The damping mapping table used in the asymmetric damping mapping calculation is corrected using the calibration factor.

8. The intelligent control system for the dovetail groove type slide cylinder used in the manufacture of surge protectors according to claim 7, characterized in that, The system also includes: Based on the state memory trajectory, identify the mechanical preload overload or underload patterns in the historical control process; The calculation method for dynamically adjusting the mechanical preload amount is based on the aforementioned excessive or insufficient mechanical preload mode. Generate a preloading strategy that is co-optimized with the pneumatic pulse control quantity.

9. The intelligent control system for the dovetail groove type slide cylinder used in the manufacture of surge protectors according to claim 8, characterized in that, The system also includes: Based on the execution deviation data in the state memory trajectory, the prediction deviation of the risk assessment model is identified; Using the prediction bias, extract new risk feature samples; The risk assessment model is updated online using the new risk feature samples.

10. A method for intelligent control of a dovetail groove type slide cylinder used in the manufacture of surge protectors, characterized in that, The method includes: The motion state parameters are obtained by acquiring the stroke position parameters of the slide, the real-time speed parameters, and the pressure difference parameters of the upper and lower chambers of the cylinder. The motion state parameters are fused from multiple sources to generate a state fusion vector. Dynamic risk feature extraction and construction are performed on the state fusion vector to generate a risk field strength; Based on the risk field strength, perform asymmetric damping mapping calculation to generate an asymmetric damping curve; Based on the asymmetric damping curve, the micro-periodic pulsation of the air path is synthesized, and based on the risk field strength, the preload calculation of the mechanical fall arrest mechanism is performed to obtain the air pulsation control quantity and the mechanical preload quantity, respectively. Based on the pneumatic pulse control quantity and the mechanical preload quantity, a coordinated execution command is generated and executed to control the slide table operation, while simultaneously collecting deviation data after execution; A time-series trajectory is constructed from the deviation data and the risk field strength to generate a state memory trajectory; the asymmetric damping mapping calculation is calibrated and corrected based on the state memory trajectory.

Images