Stainless steel bobbin winding pressure control method based on indentation online detection

Abstract

The application discloses a stainless steel winding drum winding pressure control method based on indentation online detection and relates to the field of stainless steel winding equipment control, which comprises the following steps: I, establishing an online detection environment and synchronously collecting indentation characteristic data; and II, identifying a dynamic friction coefficient and generating a tension feedforward compensation amount.The application solves the problem that detection and control are disconnected in traditional methods, realizes real-time sensing and feedback control integration of a contact state, improves the authenticity and integrity of control information, significantly enhances the prediction ability and self-adaptive ability to working condition changes, avoids disturbance from entering a closed-loop control link, improves the stability and smoothness of pressure control, breaks through the problem that traditional fixed PID parameters or experience compensation tables cannot adapt to complex working conditions, effectively solves the problem of control mismatch of traditional fixed parameters, effectively identifies and corrects local partial load and edge collapse, and makes the control object improve from'single-point pressure stability' to 'overall contact uniformity control'.

Description

Technical Field

[0001] This invention relates to the field of stainless steel winding equipment control, and more specifically to a method for controlling the winding pressure of stainless steel winding drums based on online indentation detection. Background Technology

[0002] Stainless steel winding bobbins are widely used in the winding manufacturing process of precision steel wire, metal strip, alloy wire, and high-performance wire. During the winding process, the winding pressure applied by the pressure roller directly determines the degree of interlayer compaction, winding density, and finished product dimensional accuracy. If the winding pressure is too high, it can easily lead to plastic crushing, localized stress concentration, or even surface damage to the wire; if the winding pressure is too low, it will cause loose winding layers, interlayer slippage, and a decrease in the stability of the winding structure. Therefore, the winding pressure needs to be adjusted in real time according to the number of winding layers, material condition, and equipment operating status.

[0003] Current winding pressure control methods primarily rely on closed-loop regulation based on hydraulic pressure, roller displacement, or tension feedback. The controlled object is typically the hydraulic system itself, rather than the actual contact state between the roller and the winding drum. As the number of winding layers increases, the overall stiffness of the winding drum, contact area, and material deformation characteristics all dynamically change. Controlling solely with fixed pressure or fixed PID parameters fails to reflect the actual stress at the contact interface in a timely manner, easily leading to discrepancies between the pressure reaching the target and the actual compaction effect. Therefore, how to utilize online indentation detection results to reflect the winding contact state in real time, establish a dynamic correlation between indentation characteristics and friction state, contact stiffness, and pressure control, and achieve predictive adjustment, adaptive optimization, and closed-loop control of the contact state for winding pressure, while also considering mechanical disturbance compensation and local off-center load correction, has become a key technical problem urgently needing to be solved in the field of intelligent control for stainless steel winding equipment.

[0004] Existing methods for controlling the winding pressure of stainless steel bobbins cannot achieve real-time sensing and feedback control of the contact state, which reduces the authenticity and completeness of control information, has poor predictive and adaptive capabilities for changes in working conditions, and reduces the stability and smoothness of pressure control. Furthermore, traditional fixed PID parameters or empirical compensation tables cannot adapt to complex working conditions and cannot solve the mismatch problem of traditional fixed parameter control. Therefore, we propose a method for controlling the winding pressure of stainless steel bobbins based on online indentation detection. Summary of the Invention

[0005] The purpose of this invention is to address the deficiencies in the existing technology and provide a method for controlling the winding pressure of stainless steel winding drums based on online indentation detection.

[0006] This invention proposes a method for controlling the winding pressure of stainless steel winding drums based on online indentation detection. The technical solution adopted to solve the technical problem is as follows: Ⅰ. Establish an online detection environment and collect indentation feature data simultaneously: Install the stainless steel winding drum to be processed onto the rotating spindle, control the pressure roller to apply the set winding pressure to the winding drum, enter the normal winding state, and set up a high-speed laser displacement sensor, focal displacement detection probe, rotary encoder, pressure sensor and speed sensor, and collect multi-source sensor data simultaneously to construct an indentation feature dataset of the entire winding process; II. Identify the dynamic friction coefficient and generate tension feedforward compensation: After the indentation feature dataset is constructed, the unloading window is entered when the pressure roller periodically leaves the surface of the winding drum. At the same time, the displacement curve of the entire indentation recovery process is continuously collected. The indentation rebound rate and attenuation slope in the initial stage of unloading are extracted. Based on the information, the dynamic friction coefficient of the winding contact surface is identified in real time. Then, the tension feedforward compensation is generated according to the friction change trend, and the target pressure and tension of the next winding layer are pre-adjusted. III. Separate the spatial harmonics of the indentation through a digital phase-locked loop, and compensate for the mechanical eccentricity disturbance: resample the corresponding data in the indentation feature dataset in the spatial domain according to the rotation angle of the winding drum, analyze and extract the periodic harmonic components corresponding to the mechanical eccentricity and ellipticity, generate the corresponding virtual compensation signal, and superimpose it on the pressure control command. IV. Online identification of contact stiffness and reconstruction of pressure control parameters: Based on the compensated indentation detection results, instantaneous contact stiffness is obtained, and the equivalent contact stiffness that changes continuously during the winding process is identified online. Based on the rate of change of winding speed, the gain parameters and output limit of the pressure controller are reconstructed in real time. V. Constructing a three-dimensional topology model of indentation and identifying abnormal contact states: Based on the indentation feature dataset, a three-dimensional stress topology point cloud of the winding contact area is generated, a three-dimensional stress topology model of the winding contact area is reconstructed, and different abnormal states are identified by combining various statistical features. At the same time, a multi-segment pressure adjustment strategy is generated to dynamically correct the attitude of the pressure roller. VI. Predicting Wear Trends Based on a Three-Dimensional Stress Topology Model: Based on a continuously constructed three-dimensional stress topology model, the contact arc length, edge curvature, and local wear evolution trends are analyzed, and early warnings are given for the wear status of the pressure roller. VII. Optimize the closed-loop control of winding pressure based on the acquired data: Based on the acquired dynamic friction coefficient, disturbance compensation results, online identification of contact stiffness and contact topology, jointly optimize the target pressure, target tension and multi-segment pressure distribution for the next winding cycle.

[0007] It should be further explained that the high-speed laser displacement sensor mentioned in step I is arranged near the pressure roller, and at least three sets of confocal displacement detection probes are evenly arranged along the axial direction of the winding drum to simultaneously collect the indentation depth, indentation width, edge protrusion height and indentation contour. At the same time, a rotary encoder is used to collect the real-time rotation angle of the winding drum, a pressure sensor is used to collect the actual pressure of the hydraulic cylinder, and a speed sensor is used to collect the winding speed and acceleration.

[0008] As a further aspect of the present invention, the specific steps for simultaneously acquiring multi-source sensor data and constructing an indentation feature dataset for the entire winding process in step I are as follows: S1.1: After the winding equipment is started, the winding drum is brought into a stable rotation state according to the set process speed. At the same time, the pressure roller is controlled to slowly apply the initial winding pressure so that the pressure roller and the winding drum form a stable contact area. After the equipment is running stably, various sensing devices are started. Then, the central controller reads the parameters of each sensing device and performs zero-point calibration on each detection device. After the calibration is completed, the initial stability of the corresponding sensing channel of each sensing device is detected. If the stability is less than the preset threshold, it means that the winding equipment has entered a stable working state, and then enters the multi-source data synchronous acquisition stage. S1.2: When the synchronous acquisition phase begins, the central controller sends a synchronous trigger signal to all sensing devices according to a unified sampling period. After that, various sensing devices begin to collect multi-source operating status data. After each synchronous sampling is completed by each sensing device, a complete set of synchronous sampling records is formed and stored continuously in the sampling order. S1.3: The central controller selects the synchronous clock of the winding equipment as a common time reference, then maps the multi-source operating status data in each synchronous sampling record to a unified sampling time, and calculates the time offset between each sensing channel and the common time reference. Subsequently, time compensation is performed on the corresponding operating status data according to the time offset, and the data of different sensing devices are uniformly mapped to the same sampling time. S1.4: After time synchronization is completed, the circumferential coordinates of the winding drum are established based on the real-time rotation angle of the winding drum collected by the rotary encoder, the axial coordinates of the winding drum are established based on the installation position of the confocal displacement detection probe, and the actual contact position of the wire on the surface of the winding drum is obtained by combining the winding speed collected by the speed sensor, and the actual contact position is used as the winding space coordinates. S1.5: Each time-aligned synchronous sampling record is mapped to its corresponding winding space coordinates. After the spatial mapping is completed, the central controller fuses all synchronous sampling records according to the winding order to form a unified feature vector. Then, all feature vectors are stored continuously according to the winding time order to form an indentation feature dataset of the entire winding process.

[0009] Furthermore, it should be further explained that the parameters of the sensing device described in S1.1 specifically include the device number, sampling frequency, communication status, and zero-point drift parameters; The zero-point calibration described in S1.1 is as follows: for high-speed laser displacement sensors, the reference distance measured when the pressure roller is not in contact with the winding drum is taken as the zero point of displacement; for pressure sensors, the hydraulic system in no-load state is taken as the zero point of pressure; for rotary encoders, the initial installation angle of the winding drum is used to establish a zero-angle reference coordinate; for confocal displacement detection probes, the measurement reference is corrected using a standard calibration block.

[0010] As a further aspect of the present invention, the specific steps of step II, which involve real-time identification of the dynamic friction coefficient of the winding contact surface, generating tension feedforward compensation based on the friction change trend, and pre-adjusting the target pressure and tension of the next winding layer, are as follows: S2.1: After the indentation feature dataset is constructed, the central controller continuously monitors the pressure change of the pressure roller. When it detects that the pressure roller has entered the periodic lifting stage from the loading state and the hydraulic pressure drops below the preset unloading threshold, it determines that the pressure roller has entered an unloading window and immediately switches the high-speed laser displacement sensor to the high-frequency sampling mode. The starting time of the pressure roller unloading is taken as the new time zero point, and the recovery displacement of the indentation surface within the unloading window is continuously collected to establish a recovery displacement sequence. S2.2: Based on the established recovery displacement sequence, calculate the recovery velocity between adjacent sampling points in time order to obtain the instantaneous changes in the indentation rebound process. Then, based on the instantaneous changes in the indentation rebound process, calculate the degree of decay of the rebound rate along the time direction and construct a rebound decay index. S2.3: The indentation recovery process is taken as an elastic contact unloading process, and the maximum rebound rate during the indentation recovery process is selected. Then, the indentation recovery response parameters are statistically analyzed based on the rebound attenuation index and the maximum rebound rate. Subsequently, the recovery response parameters are converted into real-time dynamic friction coefficients based on the Hertz dynamic unloading mapping relationship obtained from historical experiments. S2.4: Continuously save the dynamic friction coefficients obtained in real time from multiple unloading windows, and form a friction history sequence in chronological order. Based on the dynamic friction coefficients of two adjacent unloading windows, calculate the friction change rate between two adjacent unloading windows, and cumulatively analyze multiple consecutive friction change rates to obtain the corresponding friction growth trend index. S2.5: Based on the friction growth trend index and the current dynamic friction coefficient, obtain the compensation coefficient corresponding to the friction trend, and then calculate the tension feedforward compensation amount required for the next winding cycle based on the compensation coefficient. Then, based on the obtained tension feedforward compensation amount, correct the target tension in advance before the start of the next winding cycle.

[0011] It should be further explained that the specific steps of obtaining the Hertz dynamic unloading mapping relationship based on historical experiments as described in S2.3 are as follows: P1.1: Before online indentation detection, a stainless steel winding drum sample and corresponding wire of the same material as those used in actual production are selected as calibration objects. The structure, contact area, and surface treatment of the pressure roller are kept consistent with the actual equipment to construct a standard test environment for calibrating the Hertz dynamic unloading mapping relationship. Then, the pressure roller is gradually loaded according to multiple preset pressure levels. When the set loading pressure is reached, the pressure roller is controlled to enter the unloading process, the indentation recovery displacement sequence is obtained, and the initial loading pressure, material surface condition, and environmental parameters corresponding to each unloading process are recorded to construct the corresponding calibration sample. P1.2: After the calibration samples are constructed, the recovery speed between adjacent sampling points in each indentation recovery displacement sequence is calculated in chronological order. Then, the maximum recovery speed is selected as the elastic response strength index in the initial stage of unloading. Curve fitting is then performed on the initial recovery stage to obtain the recovery attenuation coefficient and the corresponding recovery stability bias. P1.3: Based on the maximum rebound speed, recovery attenuation coefficient and recovery stability bias, a comprehensive recovery response feature is generated by weighted fusion. A mapping relationship between the dynamic friction coefficient and the comprehensive recovery response feature is established to generate the Hertz unloading mapping model. The friction coefficient obtained from the actual friction test is then used as the target value. The Hertz unloading mapping model is then used to predict the predicted friction coefficient for multiple subsequent time periods. P1.4: Statistically predict the error between the friction coefficient and the target value, and then optimize the mapping parameters of the Hertz unloading mapping model using the gradient descent method based on the obtained error value until the error value no longer changes, and then stop the optimization. P1.5: After the mapping parameters are optimized, the dynamic unloading response parameters corresponding to different pressures, surface conditions and contact conditions are archived and a corresponding calibration database is established. At the same time, during the actual winding process, when a new indentation unloading recovery curve is collected, the corresponding recovery feature parameters are extracted, and the corresponding dynamic friction coefficient is obtained by using the calibrated mapping relationship in the calibration database.

[0012] As a further aspect of the present invention, the specific steps of step III—analyzing and extracting the periodic harmonic components corresponding to mechanical eccentricity and ellipticity, generating the corresponding virtual compensation signal, and superimposing it onto the pressure control command—are as follows: S3.1: Extract continuously collected indentation depth data and real-time rotation angle of the winding drum from the indentation feature dataset. Establish a unified spatial angle grid based on the real-time rotation angle of the winding drum. Divide a complete rotation cycle into multiple groups of equal angle sampling nodes. Then map the indentation depth to the corresponding spatial position based on the original sampling angle. After completing the spatial domain resampling, generate a spatial indentation sequence that is uniformly arranged according to the rotation angle. S3.2: Establish a digital phase-locked loop reference phase, and compare it with the spatial indentation sequence point by point to obtain the phase error between the current sampling point and the reference signal. Then, continuously adjust the internal oscillator of the digital phase-locked loop according to the phase error to keep the reference phase synchronized with the indentation period. After phase locking is completed, the synchronous projection values ​​corresponding to the first-order fundamental frequency component and the second-order harmonic frequency component are statistically analyzed. S3.3: Based on the synchronous projection values ​​corresponding to the first-order fundamental frequency component and the second-order harmonic component, the corresponding harmonic amplitude and phase are statistically analyzed. Then, a disturbance expression is established based on the amplitude and phase of each harmonic. After the expression is established, the pressure disturbance generated at each rotation angle position of the winding drum is predicted, and the spatial distribution of the corresponding mechanical periodic error is constructed to complete the construction of the periodic disturbance model. S3.4: Based on the periodic disturbance model, an adaptive notch filter is established, and the amplitude and phase of the first and second harmonics output by the digital phase-locked loop are tracked in real time through the notch filter. Then, the difference between the current indentation depth and the pressure disturbance predicted by the current periodic disturbance model is calculated, and the difference is used as the current filter residual. Subsequently, the center parameters of the notch filter are updated in real time based on the filter residual. At the same time, the notch filter is continuously iterated and updated to gradually eliminate random noise and non-periodic disturbances. S3.5: Rotate the separated harmonic phase as a whole by 180°, and at the same time set the compensation gain according to the disturbance amplitude to generate a compensation signal. Then, the central controller reads the target pressure command output by the current pressure controller in real time, and superimposes the compensation signal with the target pressure command according to the current rotation angle. At the same time, it outputs the final control command and pre-adjusts the target pressure and tension of the next winding layer according to the corrected control command.

[0013] As a further aspect of the present invention, the specific steps for real-time reconstruction of the gain parameters and output limiting of the pressure controller in step IV are as follows: S4.1: After the mechanical periodic disturbance compensation is completed, the indentation depth and corresponding pressure value are read at two consecutive sampling times according to a fixed sampling period. Then, the difference between the indentation depth at adjacent times is calculated to obtain the change in indentation depth corresponding to this round of sampling. At the same time, the difference between the pressure values ​​at adjacent times is calculated to obtain the corresponding change in pressure. Based on the change in indentation depth and the change in pressure obtained from multiple consecutive sampling periods, an indentation change sequence and a pressure change sequence are constructed. S4.2: Based on the indentation change sequence and the pressure change sequence, the instantaneous contact stiffness at each sampling moment is statistically analyzed. Then, multiple consecutive instantaneous contact stiffnesses are combined into an online identification input sequence. After the sequence is established, the winding speed at each sampling moment is extracted from the indentation feature dataset. Then, based on the winding speed at adjacent sampling moments, the speed change rate at each sampling moment is obtained. Finally, the current preset forgetting factor is dynamically updated based on the speed change rate at each sampling moment. S4.3: Based on the instantaneous contact stiffness corresponding to two consecutive sampling times, construct the corresponding sampling input vector, then use the sampling input vector and the current forgetting factor to establish a recursive gain matrix, and then recursively update the current stiffness identification parameters based on the established recursive gain matrix. S4.4: After completing the recursive update, calculate the real-time equivalent contact stiffness based on the current stiffness identification parameters, then adjust the proportional gain based on the real-time equivalent contact stiffness, and update the current integral gain based on the adjusted proportional gain. Then, based on the basic output limit and the real-time equivalent contact stiffness, adjust the controller output limit synchronously, and automatically reconstruct the parameters based on the subsequent increase in the number of winding layers.

[0014] As a further aspect of the present invention, the specific steps of step V, which involves generating a three-dimensional stress topology point cloud of the winding contact region based on the indentation feature dataset and reconstructing a three-dimensional stress topology model of the winding contact region, are as follows: S5.1: After the pressure controller parameters are adjusted, the geometric features of each confocal detection probe in the indentation feature dataset are extracted synchronously according to a unified sampling period. Then, the axial coordinates are established based on the installation position of each confocal detection probe, the circumferential coordinates are established based on the rotary encoder, and the radial height coordinates are established using the indentation depth. Each measuring point is then associated with its corresponding spatial position. S5.2: Map the collected geometric features to the corresponding spatial coordinates and establish spatial points with attributes. After completing the mapping of all measurement points, obtain the discrete spatial sampling points of the entire winding contact area, search the neighborhood sampling points around each spatial sampling point, and then establish a weighted interpolation model based on the distance. Through this weighted interpolation model, calculate the topological height corresponding to any spatial location. Then, combine the topological heights of all spatial locations to form a three-dimensional stress topological point cloud. S5.3: After the reconstruction is completed, calculate the average topological height of the current space based on all the topological heights within the current spatial location. Then, based on the topological height of each spatial location and the average topological height, obtain the three-dimensional topological skewness and the kurtosis of the point cloud height distribution. After that, based on each spatial point in the current three-dimensional stress topological point cloud and its corresponding neighboring spatial points, obtain the local curvature index of each spatial point. S5.4: Based on the obtained topological skewness, point cloud height distribution kurtosis and local curvature index, a comprehensive morphological evaluation value is established. Based on the comprehensive morphological evaluation value, the entire three-dimensional stress topological point cloud is continuously analyzed to identify the various morphological features existing in the contact area and form a complete three-dimensional topological analysis result.

[0015] It should be further explained that the morphological features described in S5.4 specifically include: force displacement, local stress concentration, edge collapse, or non-uniform contact, etc.

[0016] As a further aspect of the present invention, the specific steps for generating a multi-segment pressure adjustment strategy and dynamically correcting the pressure roller posture in step V are as follows: S6.1: Scan the three-dimensional stress topology point cloud region by region, divide the entire winding contact area into multiple continuous analysis units along the winding cylinder axis, calculate the difference in indentation height distribution on the left and right sides of each analysis unit to obtain the corresponding asymmetry index, then analyze the change in topological edge height to obtain the degree of edge collapse, and then calculate the degree of local stress concentration based on the comprehensive morphology evaluation value. S6.2: Construct an off-center load judgment value by combining the asymmetry index, the degree of edge collapse, and the local stress concentration factor. When the off-center load judgment value corresponding to any analysis unit exceeds the preset threshold, it is determined that the analysis unit has an off-center load. Arrange all the analysis units with off-center loads in the axial installation order, and calculate the overall off-center load center position according to the degree of off-center load of each analysis unit. Then, calculate the difference between the rightmost position and the leftmost position of the off-center load area to obtain the off-center load influence range. S6.3: After completing the off-center load positioning, the entire pressure roller contact area is divided into multiple pressure control sections according to the axial length of the winding drum. Each section corresponds to a control area. Pressure distribution weights are established based on the distance between each section and the off-center load center. At the same time, a pressure distribution matrix for the entire pressure roller contact area is constructed based on the pressure weights of each section. S6.4: Calculate the required pressure adjustment amount for each control section according to the weight of each section, form a pressure adjustment vector from all the pressure adjustment amounts of all sections, and establish a corresponding multi-section pressure adjustment strategy. Then, according to the multi-section pressure adjustment strategy, allocate the off-center load correction requirement to the left and right sides of the pressure roller, and obtain the corresponding pressure correction amount for the left and right sides. S6.5: Based on the obtained pressure correction quality, adjust the pressure on both sides of the pressure roller. After the adjustment is completed, call each confocal detection probe to collect new indentation topology information in real time, and recalculate the attitude correction error of each analysis unit. When the attitude correction residual error converges to the preset allowable range, it is considered that the pressure roller attitude has been adjusted to the optimal state.

[0017] As a further aspect of the present invention, the pressure distribution matrix described in S6.3 is used to reflect the degree of influence of different control sections on the off-center load area, providing a weighting basis for the calculation of pressure regulation strategy.

[0018] As a further aspect of the present invention, the specific steps of step VI, which involves analyzing the contact arc length, edge curvature, and local wear evolution trend based on a continuously constructed three-dimensional stress topology model, and providing early warning of the wear state of the pressure roller, are as follows: S7.1: After completing the construction of the three-dimensional stress topology point cloud of the winding contact area, new three-dimensional topology models are continuously acquired at fixed time intervals. The dynamic changes of the contact area between the pressure roller and the winding drum during the winding process are collected in real time as the pressure roller wears, the wire accumulates, and the pressure increases. At the same time, the three-dimensional topology point cloud at the current moment is read. The adjacent three-dimensional stress topology models obtained by continuous sampling are spatially registered to maintain the correspondence of topology points at the same position. The topology spatial matching error is established. When the matching error meets the preset requirements, the three-dimensional topology models corresponding to multiple time nodes are combined into a continuous three-dimensional topology evolution model in chronological order. S7.2: After the continuous three-dimensional topological evolution model is established, based on the preset contact height threshold, topological points higher than the contact height threshold are selected from the three-dimensional topological model, and an effective contact area is established based on the selected topological points. Then, a spatial contact curve is fitted based on the boundary points of the effective contact area, and the spatial distance between adjacent boundary points is calculated. After that, the effective contact arc length at the current moment is obtained by accumulating the spatial distances. At the same time, the change value of the effective contact arc length at different time nodes is calculated to obtain the trend of contact area expansion or contraction. S7.3: Based on the currently obtained effective contact arc length, extract each edge point of the effective contact area and calculate the directional change in the neighborhood of the edge point to obtain the local curvature of each edge point. Then, calculate the average curvature of the contact edge area in the current effective contact area and compare it with the average curvature of the previous time interval to obtain the corresponding curvature change rate. At the same time, monitor the curvature change rate in real time. If the curvature change rate continues to increase, it indicates that there is abnormal wear or pressure concentration in the contact edge area. S7.4: Statistically calculate the change in height of effective topological points at the same spatial location in the effective contact area in the continuous three-dimensional topological evolution model. Then, accumulate the changes in height of each effective topological point in the effective contact area, and calculate the average wear change in the effective contact area in the current time interval based on the accumulated value. Then, compare it with the average wear change in the previous time interval to obtain the wear evolution rate of the effective contact area in real time. S7.5: Integrate the effective contact arc length, curvature change rate, and wear evolution speed within the current time window into a wear feature vector. Then, based on the historical wear feature vector, calculate the current wear accumulation state. By performing continuous wear accumulation state statistics, obtain the predicted wear state at different future times. Then, compare the predicted wear state with the preset maximum allowable wear state threshold and obtain the corresponding wear risk assessment value. S7.6: If the risk assessment value is less than the preset lower limit threshold of wear risk, the pressure roller is determined to be in normal wear condition; if the preset lower limit threshold of wear risk is less than or equal to the risk assessment value, the pressure roller is determined to enter the wear trend warning state; if the risk assessment value is greater than or equal to the preset upper limit threshold of wear risk, the pressure roller needs to be replaced in advance.

[0019] As a further aspect of the present invention, the specific steps of jointly optimizing the target pressure, target tension, and multi-segment pressure distribution for the next winding cycle in step VII are as follows: S8.1: Collect the dynamic friction coefficient, tension feedforward compensation, real-time equivalent contact stiffness, and effective contact area morphology information of the current moment. Normalize the collected state parameters and then combine them to generate the control state vector corresponding to the winding cycle. S8.2: Based on the dynamic friction coefficient of the current winding cycle and the preset reference friction coefficient under standard working conditions, obtain the friction correction factor and corresponding friction state deviation for the corresponding winding cycle, monitor the change of dynamic friction coefficient for different winding cycles in real time, obtain the sliding resistance between the pressure roller and the wire in real time, and dynamically generate the corresponding pressure demand value according to the friction correction factor. S8.3: After the friction effect correction is completed, the difference between the disturbance amplitude after compensation based on the tension feedforward compensation amount in the current winding cycle and the original disturbance amplitude is calculated to obtain the residual error after mechanical disturbance compensation in the previous winding cycle. Then, the mechanical disturbance pressure correction amount in the current winding cycle is calculated based on the residual error, and the obtained pressure demand value is mechanically disturbed based on the obtained mechanical disturbance pressure correction amount. S8.4: After the mechanical disturbance correction is completed, calculate the difference between the real-time equivalent contact stiffness currently identified online and the preset reference stiffness. Then, based on the difference, calculate the stiffness deviation value within the corresponding winding cycle. Then, use the corrected pressure demand value and the corresponding stiffness deviation value to calculate the corresponding stiffness compensation pressure amount. Based on the stiffness compensation pressure amount, perform stiffness correction on the corrected pressure demand value. S8.5: Count the number of topological points in the effective contact area and the average topological height of the area. Then, based on the height of each topological point and the average topological height of the area, obtain the pressure distribution deviation of the effective contact area, and generate pressure adjustment weights according to the deviations of different areas. Then, calculate the product between each pressure adjustment weight and the corresponding stiffness-corrected pressure demand value to generate the target pressure of different pressure sections. S8.6: After completing the pressure space allocation, the deviation between the current actual applied pressure and the expected pressure of the next winding cycle is statistically analyzed to obtain the corresponding pressure deviation. Then, the wire tension in the current winding cycle is continuously collected, and the average tension in the current winding cycle is statistically analyzed. Then, based on the wire tension and average tension at different times in the current winding cycle, the standard deviation of tension fluctuation in the current winding cycle is statistically analyzed, and the corresponding tension stability is obtained based on the statistical standard deviation of tension fluctuation. S8.7: Based on the pressure deviation, pressure distribution deviation, friction state deviation, and tension stability within the current winding cycle, establish a corresponding comprehensive control evaluation function. Statistically calculate the corresponding comprehensive control evaluation value through the comprehensive control evaluation function. Minimize the comprehensive control evaluation value as the objective. Iteratively adjust the target pressure and target tension for the next winding cycle until the changes in target pressure and target tension converge to a preset range. Simultaneously, generate control parameters for the next winding cycle, including target pressure, target tension, and target pressure for different pressure zones.

[0020] It should be further explained that the dynamic friction coefficient mentioned in S8.1 is used to reflect the change in sliding resistance at the contact interface between the pressure roller and the winding drum; the tension feedforward compensation amount is used to describe the influence of factors such as current mechanical eccentricity and bearing runout on pressure stability; the real-time equivalent contact stiffness is used to reflect the structural bearing capacity after the change in winding layer thickness; and the contact area morphology parameters include the contact uniformity, the degree of local off-center loading, and the effective contact range.

[0021] The beneficial effects of this invention are: This invention establishes an online indentation detection system, simultaneously acquiring multi-source operating status data and constructing an indentation feature dataset for the entire winding process. It utilizes the periodic unloading window of the pressure roller to collect the indentation recovery process, extracting rebound attenuation characteristics, identifying the dynamic friction coefficient, and generating tension feedforward compensation based on friction change trends. The indentation depth is spatially resampled according to the winding drum rotation angle. A mechanical periodic disturbance model is constructed using a digital phase-locked loop, spatial domain Fourier analysis, and adaptive notch filtering. Pressure and tension pre-compensation is achieved through reverse harmonic injection. An instantaneous contact stiffness model is established based on the changes in indentation depth and pressure. An online equivalent contact stiffness is identified using a recursive least squares algorithm with a time-varying forgetting factor, and the proportional gain, integral gain, and output limiting parameters of the pressure controller are dynamically reconstructed. Finally, multiple sets of confocal detection probes are integrated. Data reconstruction of the three-dimensional stress topology point cloud of the winding contact area is used to calculate morphological features such as skewness, kurtosis, and local curvature. Off-center loading areas are identified, and a multi-segment pressure distribution matrix is ​​established. Zoned pressure fine-tuning is implemented on the auxiliary hydraulic cylinders on both sides of the pressure roller, solving the problem of disconnect between detection and control in traditional methods. This achieves real-time perception and feedback control of the contact state, improving the authenticity and completeness of control information, significantly enhancing the predictive and adaptive capabilities for changes in operating conditions, preventing disturbances from entering the closed-loop control link, and improving the stability and smoothness of pressure control. It overcomes the problem that traditional fixed PID parameters or empirical compensation tables cannot adapt to complex operating conditions, effectively solving the mismatch problem of traditional fixed parameter control, effectively identifying and correcting local off-center loading and edge collapse, and upgrading the control object from "single-point pressure stability" to "overall contact uniformity control." Attached Figure Description

[0022] The present invention will now be further described with reference to the accompanying drawings.

[0023] Figure 1 This is a framework diagram of a stainless steel winding drum winding pressure control method based on online indentation detection; Figure 2 This is a flowchart of the dynamic friction coefficient identification and tension feedforward compensation process in the stainless steel winding drum winding pressure control method based on online indentation detection. Figure 3 This is a flowchart of the off-center load identification and multi-segment pressure adjustment process in the stainless steel winding drum winding pressure control method based on online indentation detection. Detailed Implementation

[0024] 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.

[0025] Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0026] See Figure 1-3 This invention provides a method for controlling the winding pressure of a stainless steel winding drum based on online indentation detection. The method includes the following steps: Establish an online detection environment and collect indentation feature data simultaneously: Install the stainless steel winding drum to be processed onto the rotating spindle, control the pressure roller to apply the set winding pressure to the winding drum, enter the normal winding state, deploy a high-speed laser displacement sensor, a focal displacement detection probe, a rotary encoder, a pressure sensor and a speed sensor, and collect multi-source sensor data simultaneously to construct an indentation feature dataset of the entire winding process.

[0027] Specifically, after the winding equipment starts, the winding drum enters a stable rotational state according to the set process speed. Simultaneously, the pressure roller is slowly applied with initial winding pressure to create a stable contact area between the pressure roller and the winding drum. After the equipment stabilizes, various sensors are activated. The central controller then reads the parameters of each sensor and performs zero-point calibration on each sensor. After calibration, the initial stability of the corresponding sensing channel of each sensor is checked. If the stability is less than a preset threshold, it indicates that the winding equipment has entered a stable working state, and then enters the multi-source data synchronous acquisition stage. At the start of the synchronous acquisition stage, the central controller sends a synchronous trigger signal to all sensors according to a unified sampling period. Afterwards, various sensors begin to collect multi-source operating status data, and after each synchronous sampling, a complete synchronous sampling record is formed. These synchronous sampling records are stored continuously according to the sampling order. The central controller selects the winding equipment's synchronous clock as a common time reference. The multi-source operating status data in each synchronous sampling record is mapped to a unified sampling time, and the time offset between each sensing channel and the common time reference is calculated. Then, time compensation is performed on the corresponding operating status data according to the time offset, and the data of different sensing devices are uniformly mapped to the same sampling time. After time synchronization is completed, the circumferential coordinate of the winding drum is established according to the real-time rotation angle of the winding drum collected by the rotary encoder, and the axial coordinate of the winding drum is established according to the installation position of the confocal displacement detection probe. Then, the actual contact position of the wire on the surface of the winding drum is obtained by combining the winding speed collected by the speed sensor, and the obtained actual contact position is used as the winding spatial coordinate. Each synchronous sampling record after time alignment is mapped to the corresponding winding spatial coordinate. After the spatial mapping is completed, the central controller fuses all synchronous sampling records according to the winding sequence to form a unified feature vector. Then, all feature vectors are stored continuously according to the winding time sequence to form an indentation feature dataset of the entire winding process.

[0028] It should be further explained that the high-speed laser displacement sensor is arranged near the pressure roller, and at least three sets of confocal displacement detection probes are evenly arranged along the axis of the winding drum to simultaneously collect the indentation depth, indentation width, edge protrusion height and indentation contour. At the same time, the rotary encoder is used to collect the real-time rotation angle of the winding drum, the pressure sensor is used to collect the actual pressure of the hydraulic cylinder, and the speed sensor is used to collect the winding speed and acceleration. The specific parameters of the sensing device include the device number, sampling frequency, communication status, and zero-point drift parameters; Zero-point calibration is performed as follows: for high-speed laser displacement sensors, the reference distance measured when the pressure roller is not in contact with the winding drum is taken as the zero point of displacement; for pressure sensors, the hydraulic system in no-load condition is taken as the zero point of pressure; for rotary encoders, the initial installation angle of the winding drum is used to establish a zero-angle reference coordinate; for confocal displacement detection probes, the measurement reference is corrected using a standard calibration block.

[0029] Identify the dynamic friction coefficient and generate tension feedforward compensation: After the indentation feature dataset is constructed, the unloading window is entered when the pressure roller periodically leaves the surface of the winding drum. At the same time, the displacement curve of the entire indentation recovery process is continuously collected. The springback rate and attenuation slope of the indentation in the initial stage of unloading are extracted. Based on the information, the dynamic friction coefficient of the winding contact surface is identified in real time. Then, the tension feedforward compensation is generated according to the friction change trend, and the target pressure and tension of the next winding layer are pre-adjusted.

[0030] For details, please refer to Figure 2It can be seen that after the indentation feature dataset is constructed, the central controller continuously monitors the pressure change of the pressure roller. When it detects that the pressure roller has entered the periodic lifting stage from the loading state and the hydraulic pressure drops below the preset unloading threshold, it determines that the pressure roller has entered an unloading window and immediately switches the high-speed laser displacement sensor to high-frequency sampling mode. Taking the start time of the pressure roller unloading as the new time zero point, it continuously collects the recovery displacement of the indentation surface within the unloading window and establishes a recovery displacement sequence. Based on the established recovery displacement sequence, the recovery speed between adjacent sampling points is calculated in chronological order to obtain the instantaneous changes in the indentation rebound process. Then, based on the instantaneous changes in the indentation rebound process, the degree of attenuation of the rebound rate along the time direction is statistically analyzed, and a rebound attenuation index is constructed. The indentation recovery process is regarded as an elastic contact unloading process, and the most elastic contact unloading process is selected. The maximum rebound rate is used, and then the indentation recovery response parameters are statistically analyzed based on the rebound attenuation index and the maximum rebound rate. Subsequently, based on the Hertz dynamic unloading mapping relationship obtained from historical experiments, the recovery response parameters are converted into real-time dynamic friction coefficients. The dynamic friction coefficients obtained in real time from multiple unloading windows are continuously saved and arranged into a friction history sequence in chronological order. Based on the dynamic friction coefficients of two adjacent unloading windows, the friction change rate between two adjacent unloading windows is statistically analyzed. Multiple consecutive friction change rates are cumulatively analyzed to obtain the corresponding friction growth trend index. Based on the friction growth trend index and the current dynamic friction coefficient, the compensation coefficient corresponding to the friction trend is obtained. Then, based on the compensation coefficient, the tension feedforward compensation amount required for the next winding cycle is calculated. Finally, based on the obtained tension feedforward compensation amount, the target tension is corrected in advance before the start of the next winding cycle.

[0031] Furthermore, it should be noted that if the friction growth trend index continues to increase, it indicates that the lubrication condition of the winding contact surface is deteriorating, and the system enters the tension feedforward compensation calculation stage.

[0032] The spatial harmonics of the indentation are separated by a digital phase-locked loop, while the mechanical eccentricity disturbance is compensated: the corresponding data in the indentation feature dataset are spatially resampled according to the rotation angle of the winding drum, and the periodic harmonic components corresponding to the mechanical eccentricity and ellipticity are analyzed and extracted, and then the corresponding virtual compensation signal is generated and superimposed on the pressure control command.

[0033] For details, please refer to Figure 2It can be seen that continuously acquired indentation depth data and real-time rotation angle of the winding drum are extracted from the indentation feature dataset. A unified spatial angle grid is established based on the real-time rotation angle of the winding drum, and a complete rotation cycle is divided into multiple groups of equal-angle sampling nodes. Then, the indentation depth is mapped to the corresponding spatial position according to the original sampling angle. After spatial domain resampling, a spatial indentation sequence uniformly arranged according to the rotation angle is generated. A digital phase-locked loop (PLL) reference phase is established and compared with the spatial indentation sequence point by point to obtain the phase error between the current sampling point and the reference signal. Then, the internal oscillator of the PLL is continuously adjusted according to the phase error to keep the reference phase synchronized with the indentation cycle. After phase locking is completed, the synchronous projection values ​​corresponding to the first-order fundamental frequency component and the second-order harmonic component are statistically analyzed. Based on the obtained synchronous projection values ​​corresponding to the first-order fundamental frequency component and the second-order harmonic component, the corresponding harmonic amplitude and phase are statistically analyzed. Then, a perturbation expression is established based on the amplitude and phase of each harmonic. After the establishment is completed, the perturbation expression is used to predict... The pressure disturbance generated at each rotation angle of the winding drum is analyzed, and the spatial distribution of the corresponding mechanical periodic error is constructed to complete the periodic disturbance model. Based on the periodic disturbance model, an adaptive notch filter is established, and the amplitude and phase of the first and second harmonics output by the digital phase-locked loop are tracked in real time through this notch filter. The difference between the current indentation depth and the pressure disturbance predicted by the current periodic disturbance model is calculated, and the difference is used as the current filter residual. Subsequently, the center parameters of the notch filter are updated in real time according to the filter residual. At the same time, the notch filter is continuously iterated and updated to gradually eliminate random noise and non-periodic disturbances. The separated harmonic phases are rotated by 180° as a whole. Simultaneously, the compensation gain is set according to the disturbance amplitude to generate a compensation signal. Then, the central controller reads the target pressure command output by the current pressure controller in real time, and synchronously superimposes the compensation signal with the target pressure command according to the current rotation angle. At the same time, the final control command is output, and the target pressure and tension of the next winding layer are pre-adjusted according to the corrected control command.

[0034] Online identification of contact stiffness and reconstruction of pressure control parameters: Based on the compensated indentation detection results, instantaneous contact stiffness is obtained, and the equivalent contact stiffness that changes continuously during the winding process is identified online. Based on the rate of change of winding speed, the gain parameters and output limit of the pressure controller are reconstructed in real time.

[0035] Specifically, after mechanical periodic disturbance compensation is completed, the indentation depth and corresponding pressure value are read at two consecutive sampling times according to a fixed sampling period. Then, the difference between the indentation depth at adjacent times is calculated to obtain the change in indentation depth corresponding to this round of sampling. At the same time, the difference between the pressure values ​​at adjacent times is calculated to obtain the corresponding change in pressure. Based on the changes in indentation depth and pressure obtained from multiple consecutive sampling periods, an indentation change sequence and a pressure change sequence are constructed. Based on the indentation change sequence and the pressure change sequence, the instantaneous contact stiffness at each sampling time is calculated. Then, multiple consecutive instantaneous contact stiffnesses are combined into an online identification input sequence. After the sequence is established, the winding velocity at each sampling time is extracted from the indentation feature dataset. Then, based on the winding velocity at adjacent sampling times... The system obtains the velocity change rate at each sampling moment, and then dynamically updates the current preset forgetting factor based on the velocity change rate at each sampling moment. Based on the instantaneous contact stiffness corresponding to two consecutive sampling moments, it constructs the corresponding sampling input vector, and then uses the sampling input vector and the current forgetting factor to establish a recursive gain matrix. Then, based on the established recursive gain matrix, it recursively updates the current stiffness identification parameters. After the recursive update is completed, it calculates the real-time equivalent contact stiffness based on the current stiffness identification parameters, then adjusts the proportional gain according to the real-time equivalent contact stiffness, and updates the current integral gain based on the adjusted proportional gain. Then, based on the basic output limit and the real-time equivalent contact stiffness, it synchronously adjusts the controller output limit, and automatically reconstructs the parameters based on the subsequent increase in the number of winding layers.

[0036] It should be further explained that when the winding enters the acceleration phase, the rate of change of speed increases and the forgetting factor automatically decreases, making the model pay more attention to the latest sampled data; while when the winding enters the stable phase, the rate of change of speed tends to decrease and the forgetting factor gradually increases, improving the model's noise resistance and providing dynamic weights for subsequent online identification.

[0037] Construct a three-dimensional topology model of the indentation and identify abnormal contact states: Based on the indentation feature dataset, generate a three-dimensional stress topology point cloud of the winding contact area, reconstruct the three-dimensional stress topology model of the winding contact area, and identify different abnormal states by combining various statistical features. At the same time, generate a multi-segment pressure adjustment strategy to dynamically correct the attitude of the pressure roller.

[0038] Specifically, after the pressure controller parameters are adjusted, the geometric features of each confocal detection probe in the indentation feature dataset are synchronously extracted according to a unified sampling period. Then, axial coordinates are established based on the installation position of each confocal detection probe, circumferential coordinates are established based on the rotary encoder, and radial height coordinates are established using the indentation depth. Each measuring point is then mapped to its corresponding spatial position, and the collected geometric features are mapped to their corresponding spatial coordinates. Attributed spatial points are established. After mapping all measuring points, discrete spatial sampling points of the entire winding contact area are obtained. Neighborhood sampling points around each spatial sampling point are searched, and a weighted interpolation model is established based on distance. Using this weighted interpolation model, the topological height corresponding to any spatial position is statistically calculated. Then... The topological heights of all spatial locations are combined to form a three-dimensional stress topological point cloud. After reconstruction, the average topological height of the current space is calculated based on all the topological heights within the current spatial location. Then, based on the topological heights of each spatial location and the average topological height, the three-dimensional topological skewness and the kurtosis of the point cloud height distribution are obtained. Subsequently, the local curvature index of each spatial point in the current three-dimensional stress topological point cloud and its corresponding neighboring spatial points are obtained. Based on the obtained topological skewness, point cloud height distribution kurtosis, and local curvature index, a comprehensive morphological evaluation value is established. Based on the comprehensive morphological evaluation value, the entire three-dimensional stress topological point cloud is continuously analyzed to identify the various morphological features existing in the contact area and form a complete three-dimensional topological analysis result.

[0039] For details, please refer to Figure 3It is known that by scanning the three-dimensional stress topology point cloud region by region, the entire winding contact area is divided into multiple continuous analysis units along the winding drum axis. The difference in the height distribution of the indentation on the left and right sides of each analysis unit is calculated to obtain the corresponding asymmetry index. Then, the change in the height of the topological edge is analyzed to obtain the degree of edge collapse. Based on the comprehensive morphology evaluation value, the degree of local stress concentration is calculated. The off-center load judgment value is constructed by combining the asymmetry index, the degree of edge collapse, and the local stress concentration coefficient. When the off-center load judgment value corresponding to any analysis unit exceeds the preset threshold, it is determined that the analysis unit has an off-center load. All analysis units with off-center loads are arranged in the axial installation order, and the position of the overall off-center load center of gravity is calculated based on the degree of off-center load of each analysis unit. Then, the difference between the rightmost position and the leftmost position of the off-center load area is calculated to obtain the off-center load influence range. After completing the off-center load positioning, the entire winding drum is divided into multiple units according to the axial length of the winding drum. The contact area of ​​the pressure roller is divided into multiple pressure control sections, each corresponding to a control area. Pressure distribution weights are established based on the distance between each section and the off-center load center. Simultaneously, based on the pressure weights of each section, a pressure distribution matrix for the entire contact area of ​​the pressure roller is constructed. The required pressure adjustment amount for each control section is calculated according to the weights of each section. All section pressure adjustment amounts are combined into a pressure adjustment vector to establish a corresponding multi-segment pressure adjustment strategy. Then, according to the multi-segment pressure adjustment strategy, the off-center load correction requirement is allocated to the left and right sides of the pressure roller, and the corresponding pressure correction amount for the left and right sides is obtained. Based on the obtained pressure correction amount, the pressure on both sides of the pressure roller is adjusted. After the adjustment is completed, each confocal detection probe is called to collect new indentation topology information in real time, and the attitude correction error of each analysis unit is recalculated. When the attitude correction residual error converges to the preset allowable range, the pressure roller attitude is considered to have been adjusted to the optimal state.

[0040] Furthermore, as a further explanation of the present invention, the morphological features specifically include: force displacement, local stress concentration, edge collapse, or non-uniform contact, etc. The pressure distribution matrix is ​​used to reflect the degree of influence of different control sections on the off-center load area, providing a weighting basis for the calculation of pressure regulation strategies.

[0041] Based on a three-dimensional stress topology model, wear trends are predicted: the contact arc length, edge curvature and local wear evolution trends are analyzed based on a continuously constructed three-dimensional stress topology model, and early warning of the wear status of the pressure roller is given.

[0042] Specifically, after constructing the 3D stress topology point cloud of the winding contact area, new 3D topology models are continuously acquired at fixed time intervals. The dynamic changes in the contact area between the pressure roller and the winding drum during the winding process are also monitored in real time, considering roller wear, wire accumulation, and pressure. Simultaneously, the current 3D topology point cloud is read, and adjacent 3D stress topology models obtained through continuous sampling are spatially registered to maintain the correspondence of topology points at the same location. A topology spatial matching error is established. Once the matching error meets preset requirements, the 3D topology models corresponding to multiple time nodes are arranged in chronological order to form a continuous 3D topology evolution model. After the continuous 3D topology evolution model is established, it is then used according to preset connection parameters. A contact height threshold is used to select topological points from the 3D topological model that exceed the threshold. An effective contact region is then established based on these selected points. A spatial contact curve is fitted to the boundary points of the effective contact region, and the spatial distances between adjacent boundary points are calculated. These spatial distances are then accumulated to obtain the effective contact arc length at the current moment. Simultaneously, the changes in the effective contact arc length at different time points are statistically analyzed to obtain the expansion or contraction trend of the contact region. Based on the currently obtained effective contact arc length, each edge point of the effective contact region is extracted, and the directional changes within the neighborhood of each edge point are calculated to obtain the local curvature of each edge point. Finally, the average curvature of the contact edge region within the current effective contact region is calculated. Then, the average curvature of the previous time interval is compared to obtain the corresponding rate of curvature change. Simultaneously, the rate of curvature change is monitored in real time. If the rate of curvature change continues to increase, it indicates abnormal wear or pressure concentration in the contact edge area. The change in height of effective topological points at the same spatial location in the effective contact area is statistically analyzed in the continuous three-dimensional topological evolution model. Then, the height changes of each effective topological point in the effective contact area are accumulated, and based on the accumulated value, the average wear change of the effective contact area in the current time interval is calculated. This is then compared with the average wear change of the previous time interval to obtain the wear evolution speed of the effective contact area in real time. The effective contact arc length and curvature within the current time window are then analyzed. The rate of change of wear rate and the wear evolution speed are integrated into a wear feature vector. Then, based on the historical wear feature vector, the current wear accumulation state is statistically analyzed. By performing statistical analysis of each consecutive wear accumulation state, the predicted wear state at different future times is obtained. The predicted wear state is then compared with the preset maximum allowable wear state threshold, and the corresponding wear risk assessment value is obtained. If the risk assessment value is less than the preset lower limit threshold for wear risk, the pressure roller is determined to be in a normal wear state. If the preset lower limit threshold for wear risk is less than or equal to the risk assessment value, the pressure roller is determined to enter a wear trend warning state. If the risk assessment value is greater than or equal to the preset upper limit threshold for wear risk, the pressure roller needs to be replaced in advance.

[0043] Based on the acquired data, the closed-loop control of winding pressure is optimized: based on the acquired dynamic friction coefficient, disturbance compensation results, online identification of contact stiffness and contact topology, the target pressure, target tension and multi-segment pressure distribution for the next winding cycle are jointly optimized.

[0044] Specifically, the system collects various state parameters, including the dynamic friction coefficient, tension feedforward compensation, real-time equivalent contact stiffness, and effective contact area morphology, and normalizes these parameters. Then, it combines these parameters to generate a control state vector for the corresponding winding cycle. Based on the dynamic friction coefficient of the current winding cycle and a preset reference friction coefficient under standard operating conditions, it obtains the friction correction factor and corresponding friction state deviation for the corresponding winding cycle. The system monitors the changes in the dynamic friction coefficient across different winding cycles in real time to obtain the sliding resistance between the pressure roller and the wire. Based on the friction correction factor, it dynamically generates the corresponding pressure requirement value. After the friction effect correction is completed, the system... The difference between the disturbance amplitude after compensation based on tension feedforward compensation and the original disturbance amplitude in the current winding cycle is calculated to obtain the residual error after mechanical disturbance compensation in the previous winding cycle. Then, based on this residual error, the mechanical disturbance pressure correction amount for the current winding cycle is calculated. The obtained pressure demand value is then mechanically corrected based on this mechanical disturbance pressure correction amount. After the mechanical disturbance correction is completed, the difference between the currently identified real-time equivalent contact stiffness and the preset reference stiffness is calculated. Then, based on this difference, the stiffness deviation value within the corresponding winding cycle is calculated. Finally, the corresponding stiffness compensation pressure amount is calculated using the corrected pressure demand value and the corresponding stiffness deviation value. Finally, the corrected pressure is adjusted based on this stiffness compensation pressure amount. The demand value is stiffened, and the number of topological points in the effective contact area and the average topological height of the area are counted. Then, based on the height of each topological point and the average topological height of the area, the pressure distribution deviation of the effective contact area is obtained, and pressure adjustment weights are generated according to the deviations of different areas. Then, the product between each pressure adjustment weight and the corresponding stiffness-corrected pressure demand value is calculated to generate the target pressure of different pressure segments. After completing the pressure spatial allocation, the deviation between the current actual applied pressure and the expected pressure of the next winding cycle is counted to obtain the corresponding pressure deviation. Then, the wire tension is continuously collected within the current winding cycle, and the average tension within the current winding cycle is counted. Finally, based on different times within the current winding cycle... The wire tension and average tension are calculated, and the standard deviation of tension fluctuation within the current winding cycle is statistically analyzed. Based on the statistical standard deviation of tension fluctuation, the corresponding tension stability is obtained. A corresponding comprehensive control evaluation function is established based on the pressure deviation, pressure distribution deviation, friction state deviation, and tension stability within the current winding cycle. The corresponding comprehensive control evaluation value is statistically analyzed through the comprehensive control evaluation function. The target pressure and target tension for the next winding cycle are repeatedly iteratively adjusted with the goal of minimizing the comprehensive control evaluation value. The adjustment stops when the change values ​​of target pressure and target tension converge to the preset range. At the same time, the control parameters for the next winding cycle are generated, including target pressure, target tension, and target pressure for different pressure sections.

[0045] It should be further explained that after the reconstruction of the three-dimensional topological point cloud of the contact area of ​​the winding is completed, the topological model is serialized based on the continuous time window, and the three-dimensional stress topology at adjacent time points is aligned to form a contact morphology evolution data stream with temporal continuity. Then, by calculating the continuous set of points above the effective contact threshold of the pressure roller in the topological surface, its spatial projection curve is fitted, and the actual contact arc length change is obtained accordingly.

[0046] Subsequently, the edge curvature of the contact area is calculated locally, and the edge curvature gradient is extracted by analyzing the rate of change of the normal vector at the topological boundary points. When the edge curvature shows a continuous increase or uneven distribution, it is determined that the local contact stress begins to concentrate towards the edge, indicating a potential risk of accelerated wear. Then, the topological height field is processed by time difference to construct a local wear evolution index, which is used to characterize the rate of change of material compaction or surface wear per unit time.

[0047] Based on the aforementioned multidimensional characteristics, the system constructs a sliding wear accumulation prediction model. It uses the contact arc length change rate, edge curvature change rate, and local wear evolution index as input variables to estimate the cumulative wear state of the pressure roller surface through a time-recursive approach. When the predicted value exceeds a set threshold, a wear warning mechanism is triggered in advance, enabling early identification of pressure roller failure trends.

[0048] At the control optimization level, a multi-source fusion control state vector is constructed by integrating dynamic friction coefficient, mechanical cycle disturbance compensation results, online identification of contact stiffness, and current contact topology. Based on this state vector, the target pressure and target tension for the next winding cycle are jointly optimized to match changes in the current system contact conditions in advance. Simultaneously, based on the local off-center loading and uneven topographic distribution in the three-dimensional topology, the multi-segment pressure distribution matrix is ​​dynamically adjusted to achieve coordinated optimization of pressure on both sides of the pressure roller and in each axial segment.

[0049] Ultimately, a closed-loop mechanism of "topology perception - wear prediction - state fusion - control reconstruction" is formed, enabling the winding to complete parameter adjustments before the wear of the pressure roller affects product quality, and to achieve adaptive optimal control of pressure and tension in each winding cycle.

[0050] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the claims of this invention should still fall within the patent coverage of this invention.

Claims

1. A method for controlling the winding pressure of a stainless steel winding drum based on online indentation detection, characterized in that, Includes the following steps: Ⅰ. Establish an online detection environment and collect indentation feature data simultaneously: Install the stainless steel winding drum to be processed onto the rotating spindle, control the pressure roller to apply the set winding pressure to the winding drum, enter the normal winding state, and set up a high-speed laser displacement sensor, focal displacement detection probe, rotary encoder, pressure sensor and speed sensor, and collect multi-source sensor data simultaneously to construct an indentation feature dataset of the entire winding process; II. Identify the dynamic friction coefficient and generate tension feedforward compensation: After the indentation feature dataset is constructed, the unloading window is entered when the pressure roller periodically leaves the surface of the winding drum. At the same time, the displacement curve of the entire indentation recovery process is continuously collected. The indentation rebound rate and attenuation slope in the initial stage of unloading are extracted. Based on the information, the dynamic friction coefficient of the winding contact surface is identified in real time. Then, the tension feedforward compensation is generated according to the friction change trend, and the target pressure and tension of the next winding layer are pre-adjusted. III. Separate the spatial harmonics of the indentation through a digital phase-locked loop, and compensate for the mechanical eccentricity disturbance: resample the corresponding data in the indentation feature dataset in the spatial domain according to the rotation angle of the winding drum, analyze and extract the periodic harmonic components corresponding to the mechanical eccentricity and ellipticity, generate the corresponding virtual compensation signal, and superimpose it on the pressure control command. IV. Online identification of contact stiffness and reconstruction of pressure control parameters: Based on the compensated indentation detection results, instantaneous contact stiffness is obtained, and the equivalent contact stiffness that changes continuously during the winding process is identified online. Based on the rate of change of winding speed, the gain parameters and output limit of the pressure controller are reconstructed in real time. V. Constructing a three-dimensional topology model of indentation and identifying abnormal contact states: Based on the indentation feature dataset, a three-dimensional stress topology point cloud of the winding contact area is generated, a three-dimensional stress topology model of the winding contact area is reconstructed, and different abnormal states are identified by combining various statistical features. At the same time, a multi-segment pressure adjustment strategy is generated to dynamically correct the pressure roller posture. VI. Predicting wear trends based on a three-dimensional stress topology model: Based on a continuously constructed three-dimensional stress topology model, the contact arc length, edge curvature, and local wear evolution trends are analyzed, and early warnings are given for the wear status of the pressure roller. VII. Based on the acquired data, optimize the closed-loop control of winding pressure: Based on the acquired dynamic friction coefficient, disturbance compensation results, online identification of contact stiffness and contact topology, jointly optimize the target pressure, target tension and multi-segment pressure distribution for the next winding cycle; The specific steps of step II, which involve real-time identification of the dynamic friction coefficient of the winding contact surface, generating tension feedforward compensation based on the friction change trend, and pre-adjusting the target pressure and tension of the next winding layer, are as follows: S2.1: After the indentation feature dataset is constructed, the central controller continuously monitors the pressure change of the pressure roller. When it detects that the pressure roller has entered the periodic lifting stage from the loading state and the hydraulic pressure drops below the preset unloading threshold, it determines that the pressure roller has entered an unloading window and immediately switches the high-speed laser displacement sensor to the high-frequency sampling mode. The starting time of the pressure roller unloading is taken as the new time zero point, and the recovery displacement of the indentation surface within the unloading window is continuously collected to establish a recovery displacement sequence. S2.2: Based on the established recovery displacement sequence, calculate the recovery velocity between adjacent sampling points in time order to obtain the instantaneous changes in the indentation rebound process. Then, based on the instantaneous changes in the indentation rebound process, calculate the degree of decay of the rebound rate along the time direction and construct a rebound decay index. S2.3: The indentation recovery process is taken as an elastic contact unloading process, and the maximum rebound rate during the indentation recovery process is selected. Then, the indentation recovery response parameters are statistically analyzed based on the rebound attenuation index and the maximum rebound rate. Subsequently, the recovery response parameters are converted into real-time dynamic friction coefficients based on the Hertz dynamic unloading mapping relationship obtained from historical experiments. S2.4: Continuously save the dynamic friction coefficients obtained in real time from multiple unloading windows, and form a friction history sequence in chronological order. Based on the dynamic friction coefficients of two adjacent unloading windows, calculate the friction change rate between two adjacent unloading windows, and cumulatively analyze multiple consecutive friction change rates to obtain the corresponding friction growth trend index. S2.5: Based on the friction growth trend index and the current dynamic friction coefficient, obtain the compensation coefficient corresponding to the friction trend, and then calculate the tension feedforward compensation amount required for the next winding cycle based on the compensation coefficient. Then, based on the obtained tension feedforward compensation amount, correct the target tension in advance before the start of the next winding cycle.

2. The method for controlling the winding pressure of a stainless steel winding drum based on online indentation detection according to claim 1, characterized in that, The high-speed laser displacement sensor described in step I is arranged near the pressure roller, and at least three sets of confocal displacement detection probes are evenly arranged along the axial direction of the winding drum to simultaneously collect the indentation depth, indentation width, edge protrusion height, and indentation contour. At the same time, a rotary encoder is used to collect the real-time rotation angle of the winding drum, a pressure sensor is used to collect the actual pressure of the hydraulic cylinder, and a speed sensor is used to collect the winding speed and acceleration.

3. The method for controlling the winding pressure of a stainless steel winding drum based on online indentation detection according to claim 2, characterized in that, The specific steps for synchronously acquiring multi-source sensor data and constructing an indentation feature dataset for the entire winding process, as described in Step I, are as follows: S1.1: After the winding equipment is started, the winding drum is brought into a stable rotation state according to the set process speed. At the same time, the pressure roller is controlled to slowly apply the initial winding pressure so that the pressure roller and the winding drum form a stable contact area. After the equipment is running stably, various sensing devices are started. Then, the central controller reads the parameters of each sensing device and performs zero-point calibration on each detection device. After the calibration is completed, the initial stability of the corresponding sensing channel of each sensing device is detected. If the stability is less than the preset threshold, it means that the winding equipment has entered a stable working state, and then enters the multi-source data synchronous acquisition stage. S1.2: When the synchronous acquisition phase begins, the central controller sends a synchronous trigger signal to all sensing devices according to a unified sampling period. After that, various sensing devices begin to collect multi-source operating status data. After each synchronous sampling is completed by each sensing device, a complete set of synchronous sampling records is formed and stored continuously in the sampling order. S1.3: The central controller selects the synchronous clock of the winding equipment as a common time reference, then maps the multi-source operating status data in each synchronous sampling record to a unified sampling time, and calculates the time offset between each sensing channel and the common time reference. Subsequently, time compensation is performed on the corresponding operating status data according to the time offset, and the data of different sensing devices are uniformly mapped to the same sampling time. S1.4: After time synchronization is completed, the circumferential coordinates of the winding drum are established based on the real-time rotation angle of the winding drum collected by the rotary encoder, the axial coordinates of the winding drum are established based on the installation position of the confocal displacement detection probe, and the actual contact position of the wire on the surface of the winding drum is obtained by combining the winding speed collected by the speed sensor, and the actual contact position is used as the winding space coordinates. S1.5: Each time-aligned synchronous sampling record is mapped to its corresponding winding space coordinates. After the spatial mapping is completed, the central controller fuses all synchronous sampling records according to the winding order to form a unified feature vector. Then, all feature vectors are stored continuously according to the winding time order to form an indentation feature dataset of the entire winding process.

4. The method for controlling the winding pressure of a stainless steel winding drum based on online indentation detection according to claim 3, characterized in that, The specific steps of step III, which involve analyzing and extracting the periodic harmonic components corresponding to mechanical eccentricity and ellipticity, generating the corresponding virtual compensation signal, and superimposing it onto the pressure control command, are as follows: S3.1: Extract continuously collected indentation depth data and real-time rotation angle of the winding drum from the indentation feature dataset. Establish a unified spatial angle grid based on the real-time rotation angle of the winding drum. Divide a complete rotation cycle into multiple groups of equal angle sampling nodes. Then map the indentation depth to the corresponding spatial position based on the original sampling angle. After completing the spatial domain resampling, generate a spatial indentation sequence that is uniformly arranged according to the rotation angle. S3.2: Establish a digital phase-locked loop reference phase, and compare it with the spatial indentation sequence point by point to obtain the phase error between the current sampling point and the reference signal. Then, continuously adjust the internal oscillator of the digital phase-locked loop according to the phase error to keep the reference phase synchronized with the indentation period. After phase locking is completed, the synchronous projection values ​​corresponding to the first-order fundamental frequency component and the second-order harmonic frequency component are statistically analyzed. S3.3: Based on the synchronous projection values ​​corresponding to the first-order fundamental frequency component and the second-order harmonic component, the corresponding harmonic amplitude and phase are statistically analyzed. Then, a disturbance expression is established based on the amplitude and phase of each harmonic. After the expression is established, the pressure disturbance generated at each rotation angle position of the winding drum is predicted, and the spatial distribution of the corresponding mechanical periodic error is constructed to complete the construction of the periodic disturbance model. S3.4: Based on the periodic disturbance model, an adaptive notch filter is established, and the amplitude and phase of the first and second harmonics output by the digital phase-locked loop are tracked in real time through the notch filter. Then, the difference between the current indentation depth and the pressure disturbance predicted by the current periodic disturbance model is calculated, and the difference is used as the current filter residual. Subsequently, the center parameters of the notch filter are updated in real time based on the filter residual. At the same time, the notch filter is continuously iterated and updated to gradually eliminate random noise and non-periodic disturbances. S3.5: Rotate the separated harmonic phase as a whole by 180°, and at the same time set the compensation gain according to the disturbance amplitude to generate a compensation signal. Then, the central controller reads the target pressure command output by the current pressure controller in real time, and superimposes the compensation signal with the target pressure command according to the current rotation angle. At the same time, it outputs the final control command and pre-adjusts the target pressure and tension of the next winding layer according to the corrected control command.

5. The method for controlling the winding pressure of a stainless steel winding drum based on online indentation detection according to claim 4, characterized in that, The specific steps for real-time reconfiguration of the gain parameters and output limiting of the pressure controller described in step IV are as follows: S4.1: After the mechanical periodic disturbance compensation is completed, the indentation depth and corresponding pressure value are read at two consecutive sampling times according to a fixed sampling period. Then, the difference between the indentation depth at adjacent times is calculated to obtain the change in indentation depth corresponding to this round of sampling. At the same time, the difference between the pressure values ​​at adjacent times is calculated to obtain the corresponding change in pressure. Based on the change in indentation depth and the change in pressure obtained from multiple consecutive sampling periods, an indentation change sequence and a pressure change sequence are constructed. S4.2: Based on the indentation change sequence and the pressure change sequence, the instantaneous contact stiffness at each sampling moment is statistically analyzed. Then, multiple consecutive instantaneous contact stiffnesses are combined into an online identification input sequence. After the sequence is established, the winding speed at each sampling moment is extracted from the indentation feature dataset. Then, based on the winding speed at adjacent sampling moments, the speed change rate at each sampling moment is obtained. Finally, the current preset forgetting factor is dynamically updated based on the speed change rate at each sampling moment. S4.3: Based on the instantaneous contact stiffness corresponding to two consecutive sampling times, construct the corresponding sampling input vector, then use the sampling input vector and the current forgetting factor to establish a recursive gain matrix, and then recursively update the current stiffness identification parameters based on the established recursive gain matrix. S4.4: After completing the recursive update, calculate the real-time equivalent contact stiffness based on the current stiffness identification parameters, then adjust the proportional gain based on the real-time equivalent contact stiffness, and update the current integral gain based on the adjusted proportional gain. Then, based on the basic output limit and the real-time equivalent contact stiffness, adjust the controller output limit synchronously, and automatically reconstruct the parameters based on the subsequent increase in the number of winding layers.

6. The method for controlling the winding pressure of a stainless steel winding drum based on online indentation detection according to claim 3, characterized in that, The specific steps for generating a three-dimensional stress topology point cloud of the winding contact area based on the indentation feature dataset and reconstructing the three-dimensional stress topology model of the winding contact area in step V are as follows: S5.1: After the pressure controller parameters are adjusted, the geometric features of each confocal detection probe in the indentation feature dataset are extracted synchronously according to a unified sampling period. Then, the axial coordinates are established based on the installation position of each confocal detection probe, the circumferential coordinates are established based on the rotary encoder, and the radial height coordinates are established using the indentation depth. Each measuring point is then associated with its corresponding spatial position. S5.2: Map the collected geometric features to the corresponding spatial coordinates and establish spatial points with attributes. After completing the mapping of all measurement points, obtain the discrete spatial sampling points of the entire winding contact area, search the neighborhood sampling points around each spatial sampling point, and then establish a weighted interpolation model based on the distance. Through this weighted interpolation model, calculate the topological height corresponding to any spatial location. Then, combine the topological heights of all spatial locations to form a three-dimensional stress topological point cloud. S5.3: After the reconstruction is completed, calculate the average topological height of the current space based on all the topological heights within the current spatial location. Then, based on the topological height of each spatial location and the average topological height, obtain the three-dimensional topological skewness and the kurtosis of the point cloud height distribution. After that, based on each spatial point in the current three-dimensional stress topological point cloud and its corresponding neighboring spatial points, obtain the local curvature index of each spatial point. S5.4: Based on the obtained topological skewness, point cloud height distribution kurtosis and local curvature index, a comprehensive morphological evaluation value is established. Based on the comprehensive morphological evaluation value, the entire three-dimensional stress topological point cloud is continuously analyzed to identify the various morphological features existing in the contact area and form a complete three-dimensional topological analysis result.

7. The method for controlling the winding pressure of a stainless steel winding drum based on online indentation detection according to claim 4, characterized in that, The specific steps for generating the multi-segment pressure regulation strategy and dynamically correcting the pressure roller posture described in step V are as follows: S6.1: Scan the three-dimensional stress topology point cloud region by region, divide the entire winding contact area into multiple continuous analysis units along the winding cylinder axis, calculate the difference in indentation height distribution on the left and right sides of each analysis unit to obtain the corresponding asymmetry index, then analyze the change in topological edge height to obtain the degree of edge collapse, and then calculate the degree of local stress concentration based on the comprehensive morphology evaluation value. S6.2: Construct an off-center load judgment value by combining the asymmetry index, the degree of edge collapse, and the local stress concentration factor. When the off-center load judgment value corresponding to any analysis unit exceeds the preset threshold, it is determined that the analysis unit has an off-center load. Arrange all the analysis units with off-center loads in the axial installation order, and calculate the overall off-center load center position according to the degree of off-center load of each analysis unit. Then, calculate the difference between the rightmost position and the leftmost position of the off-center load area to obtain the off-center load influence range. S6.3: After completing the off-center load positioning, the entire pressure roller contact area is divided into multiple pressure control sections according to the axial length of the winding drum. Each section corresponds to a control area. Pressure distribution weights are established based on the distance between each section and the off-center load center. At the same time, a pressure distribution matrix for the entire pressure roller contact area is constructed based on the pressure weights of each section. S6.4: Calculate the required pressure adjustment amount for each control section according to the weight of each section, form a pressure adjustment vector from all the pressure adjustment amounts of all sections, and establish a corresponding multi-section pressure adjustment strategy. Then, according to the multi-section pressure adjustment strategy, allocate the off-center load correction requirement to the left and right sides of the pressure roller, and obtain the corresponding pressure correction amount for the left and right sides. S6.5: Based on the obtained pressure correction quality, adjust the pressure on both sides of the pressure roller. After the adjustment is completed, call each confocal detection probe to collect new indentation topology information in real time, and recalculate the attitude correction error of each analysis unit. When the attitude correction residual error converges to the preset allowable range, it is considered that the pressure roller attitude has been adjusted to the optimal state.

8. The method for controlling the winding pressure of a stainless steel winding drum based on online indentation detection according to claim 7, characterized in that, Step VI, which involves analyzing the contact arc length, edge curvature, and local wear evolution trend based on a continuously constructed three-dimensional stress topology model, and providing early warning of the wear state of the pressure roller, follows these specific steps: S7.1: After completing the construction of the three-dimensional stress topology point cloud of the winding contact area, new three-dimensional topology models are continuously acquired at fixed time intervals. The dynamic changes of the contact area between the pressure roller and the winding drum during the winding process are collected in real time as the pressure roller wears, the wire accumulates, and the pressure increases. At the same time, the three-dimensional topology point cloud at the current moment is read. The adjacent three-dimensional stress topology models obtained by continuous sampling are spatially registered to maintain the correspondence of topology points at the same position. The topology spatial matching error is established. When the matching error meets the preset requirements, the three-dimensional topology models corresponding to multiple time nodes are combined into a continuous three-dimensional topology evolution model in chronological order. S7.2: After the continuous three-dimensional topological evolution model is established, based on the preset contact height threshold, topological points higher than the contact height threshold are selected from the three-dimensional topological model, and an effective contact area is established based on the selected topological points. Then, a spatial contact curve is fitted based on the boundary points of the effective contact area, and the spatial distance between adjacent boundary points is calculated. After that, the effective contact arc length at the current moment is obtained by accumulating the spatial distances. At the same time, the change value of the effective contact arc length at different time nodes is calculated to obtain the trend of contact area expansion or contraction. S7.3: Based on the currently obtained effective contact arc length, extract each edge point of the effective contact area and calculate the directional change in the neighborhood of the edge point to obtain the local curvature of each edge point. Then, calculate the average curvature of the contact edge area in the current effective contact area and compare it with the average curvature of the previous time interval to obtain the corresponding curvature change rate. At the same time, monitor the curvature change rate in real time. If the curvature change rate continues to increase, it indicates that there is abnormal wear or pressure concentration in the contact edge area. S7.4: Statistically calculate the change in height of effective topological points at the same spatial location in the effective contact area in the continuous three-dimensional topological evolution model. Then, accumulate the changes in height of each effective topological point in the effective contact area, and calculate the average wear change in the effective contact area in the current time interval based on the accumulated value. Then, compare it with the average wear change in the previous time interval to obtain the wear evolution rate of the effective contact area in real time. S7.5: Integrate the effective contact arc length, curvature change rate, and wear evolution speed within the current time window into a wear feature vector. Then, based on the historical wear feature vector, calculate the current wear accumulation state. By performing continuous wear accumulation state statistics, obtain the predicted wear state at different future times. Then, compare the predicted wear state with the preset maximum allowable wear state threshold and obtain the corresponding wear risk assessment value. S7.6: If the risk assessment value is less than the preset lower limit threshold for wear risk, the pressure roller is determined to be in a normal wear state. If the preset lower limit of wear risk is less than or equal to the risk assessment value and the preset upper limit of wear risk, then the wear trend warning state is entered. If the risk assessment value is greater than or equal to the preset upper limit of wear risk, then the pressure roller needs to be replaced in advance.

9. The method for controlling the winding pressure of a stainless steel winding drum based on online indentation detection according to claim 8, characterized in that, The specific steps for jointly optimizing the target pressure, target tension, and multi-segment pressure distribution for the next winding cycle, as described in step VII, are as follows: S8.1: Collect the dynamic friction coefficient, tension feedforward compensation, real-time equivalent contact stiffness, and effective contact area morphology information of the current moment. Normalize the collected state parameters and then combine them to generate the control state vector corresponding to the winding cycle. S8.2: Based on the dynamic friction coefficient of the current winding cycle and the preset reference friction coefficient under standard working conditions, obtain the friction correction factor and corresponding friction state deviation for the corresponding winding cycle, monitor the change of dynamic friction coefficient for different winding cycles in real time, obtain the sliding resistance between the pressure roller and the wire in real time, and dynamically generate the corresponding pressure demand value according to the friction correction factor. S8.3: After the friction effect correction is completed, the difference between the disturbance amplitude after compensation based on the tension feedforward compensation amount in the current winding cycle and the original disturbance amplitude is calculated to obtain the residual error after mechanical disturbance compensation in the previous winding cycle. Then, the mechanical disturbance pressure correction amount in the current winding cycle is calculated based on the residual error, and the obtained pressure demand value is mechanically disturbed based on the obtained mechanical disturbance pressure correction amount. S8.4: After the mechanical disturbance correction is completed, calculate the difference between the real-time equivalent contact stiffness currently identified online and the preset reference stiffness. Then, based on the difference, calculate the stiffness deviation value within the corresponding winding cycle. Then, use the corrected pressure demand value and the corresponding stiffness deviation value to calculate the corresponding stiffness compensation pressure amount. Based on the stiffness compensation pressure amount, perform stiffness correction on the corrected pressure demand value. S8.5: Count the number of topological points in the effective contact area and the average topological height of the area. Then, based on the height of each topological point and the average topological height of the area, obtain the pressure distribution deviation of the effective contact area, and generate pressure adjustment weights according to the deviations of different areas. Then, calculate the product between each pressure adjustment weight and the corresponding stiffness-corrected pressure demand value to generate the target pressure of different pressure sections. S8.6: After completing the pressure space allocation, the deviation between the current actual applied pressure and the expected pressure of the next winding cycle is statistically analyzed to obtain the corresponding pressure deviation. Then, the wire tension in the current winding cycle is continuously collected, and the average tension in the current winding cycle is statistically analyzed. Then, based on the wire tension and average tension at different times in the current winding cycle, the standard deviation of tension fluctuation in the current winding cycle is statistically analyzed, and the corresponding tension stability is obtained based on the statistical standard deviation of tension fluctuation. S8.7: Based on the pressure deviation, pressure distribution deviation, friction state deviation, and tension stability within the current winding cycle, establish a corresponding comprehensive control evaluation function. Statistically calculate the corresponding comprehensive control evaluation value through the comprehensive control evaluation function. Minimize the comprehensive control evaluation value as the objective. Iteratively adjust the target pressure and target tension for the next winding cycle until the changes in target pressure and target tension converge to a preset range. Simultaneously, generate control parameters for the next winding cycle, including target pressure, target tension, and target pressure for different pressure zones.

Images