Punch feeding stroke synchronous adaptive control system

Abstract

The application discloses a punch feeding stroke synchronous adaptive control system and relates to the technical field of punch feeding. The system comprises the following steps: collecting the crank angle, servo position speed, tension and to-position signal, aligning the synchronous observation package according to the unified timestamp, calculating the command to-position and actual to-position difference to obtain the synchronization error, reconstructing the command trajectory and actual trajectory and differentiating to obtain the phase offset, length deviation, acceleration and deceleration residual error and angle jitter; constructing the abnormality discrimination index set based on the tension fluctuation, micro-slip indication and pulse integrity, checking the output error source label and reliability according to the error decomposition quantity consistency, generating the compensation update control condition according to the label and the reliability, updating the phase compensation quantity, length compensation quantity and control gain and reserving the rollback benchmark; performing the boundary constraint on the compensation increment, freezing and rolling back in the abnormal state, adjusting the prohibited feeding interval according to the calculation window margin and generating the self-checking record; and the system can inhibit the compensation drift, improve the safety and stability of the narrow window high-speed punch.

Description

Technical Field

[0001] This invention relates to the field of stamping feeding technology, specifically to a synchronous adaptive control system for the feeding stroke of a stamping press. Background Technology

[0002] In a stamping production line for coiled material, the feeding mechanism typically employs a servo feeder or a roller feeder. Using the press crankshaft angle as the primary time base, the feeding start, acceleration / deceleration, position holding, and prohibited feeding intervals are synchronously bound to the press stroke window, thus forming a synchronous control scheme for the press feeding stroke. To adapt to the following errors caused by different materials, different mold conditions, and variations in the number of strokes, existing schemes often introduce adaptive control or online compensation mechanisms. For example, they update the phase offset, feeding length, position compensation amount, or control gain online to reduce synchronization errors and improve stability under high-speed cycles.

[0003] However, in actual operation, the error sources faced by the punch press feeding stroke synchronization system have multiple superposition and time-varying characteristics. The errors may come from insufficient servo axis following ability, mechanical backlash and elastic deformation, as well as from material strip slight slippage, tension fluctuation, changes in material guiding resistance, changes in lubrication state, thermal drift, and jitter, pulse loss or short-term abnormality of crankshaft angle acquisition link. Among the above errors, some are stable deviations that can be compensated and offset in the long term, while others are uncontrollable disturbances or occasional abnormalities. Moreover, different error sources may appear to be similar in terms of positioning error or phase error.

[0004] In existing technologies, adaptive compensation mechanisms often use synchronization errors or position residuals as direct learning signals, interpreting error changes as changes in controllable deviations by default, and continuously accumulating compensation accordingly. However, in the absence of error source attribution, compensation update gating, and compensation freeze or rollback mechanisms under abnormal conditions, adaptive compensation may still be misled and learn incorrectly when the error is triggered by uncontrollable disturbances or sensor link anomalies. This leads to the compensation amount drifting over time, resulting in a chronic instability phenomenon where the more compensation is applied, the more skewed the result becomes. This instability is not easily exposed in low-speed jogging or short-term production stages, but in high-speed continuous stamping and narrow-window conditions, it gradually pushes the feeding action towards the window boundary, ultimately leading to frequent false stops, increased risk of missed stops, die rubbing and misaligned punching, guide pin breakage, and die damage. It also increases the difficulty of on-site adjustment and fault location, reducing production line stability.

[0005] Therefore, a synchronous adaptive control system for press feed stroke is needed, which can effectively distinguish and attribute the source of synchronization error under the condition of multiple disturbances and signal anomalies. On this basis, it can perform closed-loop management of the update conditions, boundary constraints and recovery strategies of adaptive compensation to avoid the adaptive compensation being misled by anomalies and causing cumulative drift, thereby improving the long-term stability and safety of stroke synchronization control. Summary of the Invention

[0006] The purpose of this invention is to provide a synchronous adaptive control system for the feed stroke of a punch press, so as to overcome the shortcomings of the prior art.

[0007] To achieve the above objectives, the present invention provides the following technical solution: The punch press feed stroke synchronization adaptive control system includes: The observation construction module is used to acquire the crankshaft angle, servo position speed, tension, and positioning signal of the punch press, and generate a synchronous observation package by aligning it with a unified timestamp; the synchronization error is calculated from the difference between the command positioning and the actual positioning. The reconstruction decomposition module reconstructs the command trajectory and the actual trajectory from the synchronous observation package, and obtains the error decomposition quantity by differentiating the command trajectory and the actual trajectory. The anomaly detection module calculates anomaly detection index set based on synchronous observation packages; performs consistency verification between the anomaly detection index set and error decomposition quantity, and outputs error source labels and confidence levels. The error source labels are used to distinguish between offsettable deviations, uncontrollable disturbances and acquisition anomalies. The update control module generates compensation update control conditions based on the error source label and confidence level; the confidence level is compared with a preset threshold. If the comparison result meets the compensation update control conditions and the error source label is a cancelable deviation, the compensation amount is updated; otherwise, the compensation amount remains unchanged. The constraint safety module is used to apply boundary constraints to the compensation increment generated by the update. When the error source label is an uncontrollable disturbance or acquisition anomaly, the update is frozen and rolled back to the rollback baseline. The rolled-back parameters are used to generate the feeding command for the next stroke, the window margin is calculated and the prohibited feeding range is adjusted accordingly, and a self-verification record is generated.

[0008] Furthermore, methods for calculating the synchronization error from the difference between the command arrival and the actual arrival include: Based on the crankshaft angle and preset feeding instruction generation rules in the synchronous observation package, calculate the command arrival reference quantity corresponding to each unified timestamp. The command arrival reference quantity includes at least the target position value of the command arrival and the target time or target crankshaft angle of the command arrival. The actual arrival time is determined based on the arrival status edge marker in the synchronous observation packet, and the servo position is obtained at the unified timestamp corresponding to the actual arrival time as the actual arrival position; when the arrival status is not triggered, the actual arrival time is taken as the time when the servo position reaches the target position corresponding to the command arrival reference quantity. The command position is obtained by taking the command position reference value at the actual position time, and the difference between the command position and the actual position is calculated as the synchronization error.

[0009] Furthermore, methods for reconstructing command trajectories and actual trajectories based on synchronous observation packets include: Read the unified timestamp sequence and crankshaft angle sequence from the synchronous observation package, and read the preset feeding instruction generation rule corresponding to the crankshaft angle; The command position and command velocity values ​​are calculated point by point according to the unified timestamp sequence to obtain the command position sequence and command velocity sequence. The command position sequence and command velocity sequence are then combined into the command trajectory. The servo position sequence and servo velocity sequence in the synchronous observation package are read and arranged according to the unified timestamp sequence to obtain the actual position sequence and actual velocity sequence. The actual position sequence and actual velocity sequence are then combined into the actual trajectory.

[0010] Furthermore, methods for obtaining the error decomposition quantity by differentiating the command trajectory from the actual trajectory include: Error decomposition includes phase offset, length deviation, acceleration / deceleration residuals, and angle jitter; Calculate position and velocity residuals using a unified timestamp; enumerate time translation amounts within a preset translation search range, take the time translation amount corresponding to the minimum sum of squares of position residuals, and convert it into phase offset according to the sampling interval; calculate the length deviation by averaging the position residuals during the position holding time interval; calculate the acceleration and deceleration residuals by averaging the velocity residuals during acceleration and deceleration phases, and obtain the acceleration / deceleration residuals by calculating the peak value or root mean square value of the position residuals; calculate the angle increment and adjacent differences from the crankshaft angle, using the middle value of the angle increment sliding window as the reference increment, and take the angle jitter as the combination of the deviation of the angle increment from the reference increment and the absolute value of the increment difference.

[0011] Furthermore, the position residual is the command position value minus the actual position value under the same unified timestamp; the velocity residual is the command velocity value minus the actual velocity value under the same unified timestamp.

[0012] Furthermore, methods for calculating anomaly discrimination index sets based on synchronous observation packets include: Read the unified timestamp sequence, tension sequence, servo position sequence, and crankshaft angle sequence from the synchronous observation package, and set the width of the sliding window bound to the unified timestamp sequence; within the sliding window, obtain the tension fluctuation index from the difference between the maximum and minimum tension values; Within the sliding window, the position increment sequence and tension increment sequence are calculated from the servo position sequence and tension sequence, respectively, and the micro-slip indication is calculated from the position increment sequence and tension increment sequence; Within a sliding window, the angle increment sequence is calculated from the crankshaft angle sequence. The reference increment is taken as the median of the angle increment sequence. The number of samples where the angle increment is equal to zero, the number of samples where the absolute value of the difference between the angle increment and the reference increment exceeds the preset deviation threshold, and the number of samples where the signs of adjacent terms of the angle increment are reversed are counted. The pulse integrity indicator is then normalized according to the window length. Tension fluctuation index, microslippage indicator, and pulse integrity indicator are written into the synchronous observation package according to a unified timestamp sequence to form an anomaly discrimination index set.

[0013] Furthermore, methods for performing consistency checks on the anomaly detection index set and error decomposition quantities, and outputting error source labels and confidence levels, include: Read the anomaly discrimination index set and error decomposition quantity from the synchronous observation package; generate the acquisition anomaly discrimination condition set, uncontrollable disturbance discrimination condition set, and offset deviation discrimination condition set respectively; count each item in each discrimination condition set to form the corresponding condition count value, compare it with the corresponding count threshold, and output the error source label; calculate the consistency score for each discrimination condition set, and take the maximum consistency score as the confidence level.

[0014] Furthermore, the method for generating compensation update control conditions based on error source labels and confidence levels includes: The confidence level is compared with a preset threshold to obtain the confidence level comparison result; within the most recent preset window number, the number of times the error source label is equal to the number of times the offset deviation occurs is counted, and the number of occurrences is compared with a preset consistency number threshold to obtain the label consistency comparison result; the confidence level comparison result and the label consistency comparison result are merged to generate compensation update control conditions, which include update allow and update prohibit.

[0015] Furthermore, the method and steps for updating the compensation amount include: Read the compensation update control conditions and error source labels corresponding to the unified timestamp in the synchronous observation package; when the compensation update control conditions are update allowed and the error source labels are offset deviations, read the phase compensation amount, length compensation amount and control gain, and save the parameters before the update as the rollback reference. The phase compensation increment is obtained based on the phase offset according to the preset phase mapping rule, the length compensation increment is obtained based on the length deviation according to the preset length mapping rule, and the control gain increment is obtained based on the acceleration / deceleration residual according to the preset gain mapping rule. The corresponding update increments of the phase compensation, length compensation and control gain are superimposed to obtain the update parameters. When the compensation update control condition is update prohibited, the phase compensation, length compensation and control gain remain unchanged.

[0016] Furthermore, methods for calculating window margin and adjusting the prohibited feeding range accordingly include: Read the starting crankshaft angle and ending crankshaft angle of the window corresponding to the boundary of the punch press stroke window, and read the next stroke command trajectory to obtain the feeding start crankshaft angle and feeding end crankshaft angle; The window margin is the minimum of the absolute values ​​of the difference between the starting crankshaft angle of the window and the starting crankshaft angle of the feeding start, and the absolute values ​​of the difference between the ending crankshaft angle of the window and the ending crankshaft angle of the feeding finish. The window margin is compared with the preset margin threshold. If the window margin is not greater than the preset margin threshold, the starting crankshaft angle of the prohibited feeding interval is moved by a preset adjustment step size in the direction of the starting crankshaft angle of the window, and the ending crankshaft angle of the prohibited feeding interval is moved by a preset adjustment step size in the direction of the ending crankshaft angle of the window. When the window margin is greater than the preset margin threshold, the starting crankshaft angle of the prohibited feeding interval will be moved away from the direction of the starting crankshaft angle of the window, and the preset adjustment step size will be moved. The ending crankshaft angle of the prohibited feeding interval will be moved away from the direction of the ending crankshaft angle of the window, and the preset adjustment step size will be moved.

[0017] The technical effects and advantages of the punch press feed stroke synchronous adaptive control system provided by this invention are as follows: Synchronization observation packets are generated based on unified timestamp alignment, and the synchronization error is obtained from the difference between command arrival and actual arrival. At the same time, the command trajectory and actual trajectory are reconstructed, and the position residual and velocity residual are calculated. The overall deviation is then decomposed into phase offset, length deviation, acceleration / deceleration residual, and angle jitter, so that the synchronization error is expanded from a single residual quantity to a set of locatable components. The effect of this is that the same arrival error can be decomposed into time offset, length offset, dynamic tracking difference, and sampling fluctuation difference, reducing the numerical overlap of different error sources, improving the repeatability of error assessment, and providing clear calculation input for subsequent compensation update control.

[0018] Anomaly discrimination index set is calculated based on synchronous observation packages. The anomaly discrimination index set includes tension fluctuation index, micro-slip indication, and pulse integrity indication. The anomaly discrimination index set and error decomposition quantity are subjected to consistency verification to output error source labels and confidence levels. Then, compensation update control conditions are generated based on error source labels and confidence levels. The compensation quantity is updated only when the error source label is a cancelable deviation and the confidence level meets the update requirements compared with the preset threshold. The learning signal of adaptive compensation is transformed from simple residual driving to update judgment with source constraints. This avoids uncontrollable disturbances or acquisition anomalies being misinterpreted as controllable deviation changes, which would drive the continuous accumulation of phase compensation quantity, length compensation quantity, and control gain. It also suppresses the drift of compensation quantity over time and reduces the risk of accidental stop, missed stop, die rubbing, and misaligned punching caused by the feeding action gradually approaching the boundary of the punching stroke window.

[0019] Boundary constraints are applied to the compensation increment generated by the update. When the error source label is an uncontrollable disturbance or an abnormal acquisition, the update is frozen and rolled back to the rollback baseline. The rolled-back parameters are then used to generate the feeding command for the next stroke. The window margin is calculated and the prohibited feeding range is adjusted accordingly. At the same time, a self-verification record is generated. The change of the compensation amount is limited to a controllable range, and the parameters before the update are quickly restored in abnormal conditions, so that the generation of the next stroke command is not affected by the parameter changes driven by the abnormality. The allowable offset margin of the feeding action relative to the window boundary is quantified by the window margin and the prohibited feeding range is adjusted to keep the feeding sequence and stroke window safe margin. The self-verification record collects the error source label, confidence level, frozen update mark, rollback baseline index, boundary constraint parameters, window margin and prohibited feeding range and outputs the verification results, shortening the on-site verification and fault location time and improving operational traceability and maintenance efficiency. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0021] Figure 1 This is a schematic diagram of the punch press feed stroke synchronous adaptive control system of the present invention. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] It should be noted that when a component is said to be "fixed to" another component, it can be directly attached to the other component or there may be an intervening component. When a component is said to be "connected to" another component, it can be directly connected to the other component or there may be an intervening component. Example

[0024] Please see Figure 1 As shown, this embodiment provides a synchronous adaptive control system for the feed stroke of a punch press, including an observation and construction module, a reconstruction and decomposition module, an anomaly detection module, an update control module, and a constraint safety module. The modules are connected by wired and / or wireless means.

[0025] The observation construction module is used to acquire crankshaft angle, servo position speed, tension, and positioning signals, and generate a synchronous observation package by aligning them with a unified timestamp; the synchronization error is calculated from the difference between the command positioning and the actual positioning.

[0026] In this embodiment, the method steps for obtaining crankshaft angle, servo position speed, tension, and positioning signal include: The system reads the output of the angle acquisition device rigidly associated with the punch press crankshaft to form a crankshaft angle sequence, and adds a sampling time to each angle sample value; it acquires the feedback from the servo drive device to obtain the servo position sequence and servo speed sequence, and adds a sampling time to each position and speed sample value; it acquires the output of the tension acquisition device installed on the material belt path to form a tension sequence, and adds a sampling time to each tension sample value; it acquires the position trigger signal associated with the position detector to form a position signal sequence, and records the trigger time and trigger state corresponding to each position trigger; the purpose is to obtain the original observations describing the correspondence between the feeding action and the punch press stroke using a unified sampling time as an index, providing a data source for subsequent generation of synchronous observation packages aligned with a unified timestamp.

[0027] In this embodiment, the method for generating synchronization observation packets by aligning with a unified timestamp includes the following steps: A unified timestamp sequence is determined based on the sampling times of the crankshaft angle sequence. This unified timestamp sequence is a set of times that increase at fixed sampling intervals. The servo position sequence, servo speed sequence, and tension sequence are resampled on the unified timestamp sequence. Resampling uses adjacent time-interpolation or a preservation rule to ensure that each unified timestamp corresponds to a unique servo position, servo speed, and tension value. The trigger times in the position signal sequence are mapped to the unified timestamp sequence to obtain the position state corresponding to each unified timestamp, and the edge markers of the position trigger are retained. The crankshaft angle sequence is synchronously resampled on the unified timestamp sequence, ensuring that each unified timestamp corresponds to a unique crankshaft angle value. The unified timestamp sequence, along with the aligned crankshaft angle, servo position, servo speed, tension, and position state, are combined by field to generate a synchronous observation package. Missing sample markers and abnormal jump markers are recorded in the synchronous observation package. The purpose is to convert multi-source signals into an observation set that can be directly calculated and compared under the same time reference, avoiding error aliasing caused by asynchronous sampling and providing consistent data input for synchronization error calculation.

[0028] In one implementation of this embodiment, the unified timestamp sequence is generated by a monotonically increasing clock inside the controller and a time mapping relationship is established with the crankshaft angle sampling event. When resampling the servo position, servo speed, and tension signals at the unified timestamp, interpolation or holding is used to obtain the values ​​when the interval between adjacent samples is within the preset maximum interval. When the interval between adjacent samples exceeds the preset maximum interval, the corresponding field is marked as a missing sample and written as a missing mark. At the same time, the missing samples are prohibited from participating in the phase offset estimation and micro-slip indication calculation to avoid distortion of the error decomposition caused by the missing samples.

[0029] In this embodiment, the method for calculating the synchronization error based on the difference between the command arrival and the actual arrival includes the following steps: Based on the crankshaft angle and preset feeding command generation rules in the synchronous observation package, the command arrival reference quantity is calculated for each unified timestamp. The command arrival reference quantity includes at least the target position value of the command arrival and the target time or target crankshaft angle of the command arrival. The actual arrival time is determined based on the arrival status edge marker in the synchronous observation package, and the servo position is obtained as the actual arrival position at the unified timestamp corresponding to the actual arrival time. When the arrival status is not triggered, the time when the servo position reaches the target position corresponding to the command arrival reference quantity is used as the actual arrival time. The command arrival reference quantity is taken at the actual arrival time to obtain the command arrival position, and the difference between the command arrival position and the actual arrival position is calculated as the synchronization error. The synchronization error, along with the corresponding unified timestamp, crankshaft angle, servo position speed, tension, and arrival status, is written into the error field of the synchronous observation package to form a traceable synchronization error. The purpose is to quantify and output the deviation of the feeding action relative to the press stroke in the form of a verifiable difference, and to bind and store the synchronization error with the observation conditions that generated the error to support the calculation input for subsequent error source differentiation and compensation update control. A specific example is as follows: In one implementation of this embodiment, when the positioning state is not triggered, the actual positioning time is defined as the moment when the servo position first enters the tolerance band of the target position corresponding to the command positioning reference quantity and does not leave the tolerance band within a continuously preset holding time; when it leaves the tolerance band within the preset holding time, it continues to search backward until the holding condition is met, so as to ensure that the actual positioning time is unique even when there is overshoot, oscillation or rebound.

[0030] The unified timestamp sampling interval is 1 millisecond, and the example is a single arrival determination in the same pulse; Calculate the command arrival reference value: At a unified timestamp of 18 milliseconds, the crankshaft angle recorded by the synchronous observation package is 175 degrees. The preset feeding command generation rule stipulates that when the crankshaft angle reaches 175 degrees, the feeding command enters the arrival holding section, and its command arrival target position value is 120.0 mm, the command arrival target crankshaft angle is 175 degrees, and based on this, the command arrival reference value is the target position of 120.0 mm and the target crankshaft angle of 175 degrees.

[0031] Determine the actual arrival time and actual arrival position: The arrival status of the synchronous observation packet shows an edge marker changing from 0 to 1 at a unified timestamp of 19 milliseconds, so 19 milliseconds is determined as the actual arrival time; the servo position is read as 119.6 mm at the unified timestamp corresponding to the actual arrival time, which is taken as the actual arrival position; where, the arrival status is a binary state quantity, the arrival status is 0 when the arrival detection signal is not triggered, and the arrival status is 1 when the arrival detection signal is triggered, and the arrival status edge marker is the marker corresponding to the moment when the arrival status changes from 0 to 1, which is used to determine the actual arrival time.

[0032] Calculate the synchronization error and write it into the synchronization observation package: At the actual arrival time of 19 milliseconds, the command arrival reference value still corresponds to a command arrival position of 120.0 mm; calculate the difference between the command arrival position of 120.0 mm and the actual arrival position of 119.6 mm, and obtain a synchronization error of 0.4 mm. Write this synchronization error, along with the corresponding unified timestamp of 19 milliseconds, crankshaft angle of 176 degrees, servo position of 119.6 mm, servo speed of 0.0 mm / s, tension of 45 N, and arrival status 1, into the error record item of the synchronization observation package to form a traceable synchronization error record.

[0033] Example of the same rule when the arrival status is not triggered: If the arrival status edge marker does not appear in the synchronization observation packet of this stroke, then retrieve the moment when the servo position first reaches the target position of 120.0 mm in the unified timestamp sequence. For example, if the servo position is 120.0 mm at the unified timestamp of 20 milliseconds, then take 20 milliseconds as the actual arrival time, and take the servo position of 120.0 mm at that time as the actual arrival position; then take the command arrival position of 120.0 mm at that time, calculate the difference between the two to get the synchronization error of 0.0 mm, and write it into the synchronization observation packet according to the above writing rule.

[0034] The reconstruction decomposition module reconstructs the command trajectory and the actual trajectory based on the synchronous observation package. It obtains the error decomposition quantity by differentiating the command trajectory and the actual trajectory. The error decomposition quantity includes phase offset, length deviation, acceleration / deceleration residual and angle jitter.

[0035] In this embodiment, the method steps for reconstructing the command trajectory and the actual trajectory based on the synchronous observation packet include: Read the unified timestamp sequence and crankshaft angle sequence from the synchronous observation package, and read the preset feeding instruction generation rule corresponding to the crankshaft angle; The command position and command velocity values ​​are calculated point by point according to the unified timestamp sequence to obtain the command position sequence and command velocity sequence that change over time. The command position sequence and command velocity sequence are then combined to form the command trajectory. Read the servo position sequence and servo velocity sequence from the synchronous observation package, arrange them according to the unified timestamp sequence to obtain the actual position sequence and actual velocity sequence, and combine the actual position sequence and actual velocity sequence into the actual trajectory; When the servo speed sequence is missing, the actual speed sequence is calculated and completed by dividing the servo position difference at adjacent unified timestamps by the sampling interval, and the completed result is written back to the synchronous observation package. The purpose is to transform the crankshaft angle information and preset feed command generation rules in the synchronous observation package into a directly comparable command trajectory, and to organize the servo feedback quantities in the synchronous observation package into a directly comparable actual trajectory, providing input for subsequent error decomposition by differentiating the command trajectory and the actual trajectory.

[0036] To facilitate understanding of the process of reconstructing the command trajectory and the actual trajectory, the following example is provided: The unified timestamp sequence in the synchronous observation package is selected as 17 milliseconds, 18 milliseconds, and 19 milliseconds; the crankshaft angle sequence is 174 degrees, 175 degrees, and 176 degrees respectively; the preset feeding instruction generation rule stipulates that when the crankshaft angle is 174 degrees, it is in the deceleration stage before reaching the position; when the crankshaft angle reaches 175 degrees, the command to reach the target position value is 120.0 mm and enter the position holding stage.

[0037] Based on the preset feeding command generation rules, the following values ​​were calculated: at 17 milliseconds, the command position was 119.6 mm and the command speed was 400 mm / s; at 18 milliseconds, the command position was 120.0 mm and the command speed was 0 mm / s; at 19 milliseconds, the command position was 120.0 mm and the command speed was 0 mm / s. The command position sequence and command speed sequence were combined to obtain the command trajectory.

[0038] The servo position sequence within the synchronous observation package is 119.3 mm, 119.5 mm, and 119.6 mm at 17 ms, 18 ms, and 19 ms, respectively; the servo velocity sequence is 380 mm / s at 17 ms, missing at 18 ms, and 0 mm / s at 19 ms; the actual position sequence and the actual velocity sequence are arranged and combined according to a unified timestamp sequence to obtain the actual trajectory.

[0039] For the missing servo velocity at 18 milliseconds, the complete value is calculated based on the difference between the servo positions at adjacent unified timestamps and divided by the sampling interval. The completed servo velocity is equal to (119.5 mm minus 119.3 mm) divided by 0.001 seconds, which gives 200 mm per second. The completed result is written back to the servo velocity field corresponding to 18 milliseconds in the synchronous observation packet.

[0040] In this embodiment, the method steps for obtaining the error decomposition quantity by differentiating the command trajectory from the actual trajectory include: The position residual sequence is calculated point by point according to a unified timestamp. The position residual is the command position value minus the actual position value at the same unified timestamp. The velocity residual sequence is also calculated point by point according to a unified timestamp. The velocity residual is the command velocity value minus the actual velocity value at the same unified timestamp. The time shift is defined as the number of sampling points that shift the command position sequence forward or backward along the unified timestamp direction. Within a preset shift search range, the number of sampling points is enumerated. The shifted command position sequence is subtracted from the actual position sequence point by point, and the sum of squared residuals is calculated. The number of sampling points corresponding to the smallest sum of squared residuals is taken as the time shift. The time shift is multiplied by the sampling interval to obtain the time offset. This time offset is converted into a phase offset. The phase offset is the difference between the crankshaft angle corresponding to the actual arrival time and the crankshaft angle corresponding to the command arrival time after shift.

[0041] In one improved embodiment, the phase offset estimation is preferentially based on the similarity between the command velocity sequence and the actual velocity sequence. The phase offset is determined by the time shift with the smallest alignment error to reduce the influence of the length deviation on the time shift. After obtaining the phase offset, the position residual is averaged over the position holding time interval to obtain the length deviation, thereby forming a separate calculation order of timing first and amplitude second, reducing misjudgment caused by the coupling of phase offset and length deviation.

[0042] Based on the alignment result after phase offset correction, select the time interval for the command position to enter the position and maintain it, and calculate the average of the position residuals within this interval to obtain the length deviation. After completing the phase offset correction and length deviation deduction, the time intervals in which the command speed is in the acceleration and deceleration phases are selected respectively. The mean value of the velocity residuals in the corresponding intervals is calculated and the peak value or root mean square value of the position residuals are calculated to obtain the acceleration and deceleration residuals. An angle increment sequence is calculated based on the crankshaft angle sequence using adjacent unified timestamps. The angle increment is taken as the difference between the crankshaft angles at adjacent times. Then, an increment difference sequence is calculated on the angle increment sequence, with the increment difference taken as the difference between adjacent angle increments. The median of the angle increment within the sliding window is used as the reference increment. The angle jitter is taken as the combination of the deviation of the angle increment from the reference increment and the absolute value of the increment difference. The angle jitter and the abrupt change in phase offset are marked on a unified timestamp. The purpose is to decompose the overall deviation between the command trajectory and the actual trajectory into four types of error decomposition quantities: phase offset, length deviation, acceleration / deceleration residual, and angle jitter. This enables subsequent steps to perform differentiated processing and compensation control for different error decomposition quantities.

[0043] To facilitate the explanation of the calculation process for the above error decomposition, we continue to use the 17-19 ms samples from the same impulse in the aforementioned synchronous observation package, and supplement them with 15-ms and 16-ms samples. A set of numerical calculation examples is given, with a unified timestamp sampling interval of 1 ms, as follows: The synchronous observation package recorded the following values ​​between 15 and 19 milliseconds: 15 milliseconds: crankshaft angle 172.0 degrees, command position 118.8 mm, command speed 200 mm / s, actual position 118.6 mm, actual speed 180 mm / s; 16 milliseconds: crankshaft angle 173.2 degrees, command position 119.4 mm, command speed 600 mm / s, actual position 119.1 mm, actual speed 560 mm / s; 17 milliseconds: crankshaft angle 174.0 degrees, command position 119.6 mm, command speed 40 mm / s. 0 mm / s, actual position 119.3 mm, actual speed 380 mm / s; 18 ms, crankshaft angle 175.0 degrees, command position 120.0 mm, command speed 0 mm / s, actual position 119.5 mm (actual speed field missing), padded to 200 mm / s by the difference between adjacent unified timestamp servo positions and divided by the sampling interval; 19 ms, crankshaft angle 176.0 degrees, command position 120.0 mm, command speed 0 mm / s, actual position 119.6 mm, actual speed 0 mm / s.

[0044] Calculate the position residual sequence and velocity residual sequence point by point. The position residual equals the command position minus the actual position at the same unified timestamp, and the velocity residual equals the command velocity minus the actual velocity at the same unified timestamp. The point-by-point calculation method according to the unified timestamp is as follows: 15 millimeters: command position 118.8 mm, actual position 118.6 mm, position residual 118.8 minus 118.6 equals 0.2 mm; command velocity 200 mm / s, actual velocity 180 mm / s, velocity residual 200 minus 180 equals 20 mm / s; 16 millimeters: command position 119.4 mm, actual position 119.1 mm, position residual 119.4 minus 119.1 equals 0.3 mm; command velocity 600 mm / s, actual velocity 560 mm / s, velocity residual 600 minus 560 equals 40 mm / s; 17 millimeters... 18 milliseconds: Command position 119.6 mm, actual position 119.3 mm, position residual 119.6 minus 119.3 equals 0.3 mm; command speed 400 mm / s, actual speed 380 mm / s, speed residual 400 minus 380 equals 20 mm / s; 18 milliseconds: Command position 120.0 mm, actual position 119.5 mm, position residual 120.0 minus 119.5 equals 0.5 mm; command speed 0 mm / s, actual speed is 200 mm / s based on the completion result, speed residual 0 minus 200 equals -200 mm / s; 19 milliseconds: Command position 120.0 mm, actual position 119.6 mm, position residual 120.0 minus 119.6 equals 0.4 mm; command speed 0 mm / s, actual speed 0 mm / s, speed residual 0 minus 0 equals 0 mm / s.

[0045] Among them, the unified timestamp is The sampling interval is denoted as The command location is recorded as The actual location is recorded as The command speed is recorded as The actual speed is recorded as The crankshaft angle is denoted as The positional residual is denoted as The velocity residual is denoted as The calculation formula is as follows:

[0046]

[0047] Enumerate the time shifts and take the one with the smallest sum of squared residuals; the time shift is defined as the number of sampling points to shift the command position sequence forward or backward along the unified timestamp direction. For example, the search range is shifted forward by 1 point, no shift, and shifted backward by 1 point; let the time shift be... ,in This indicates that the command position sequence is shifted forward by one sampling point. This indicates that the command position sequence is not shifted. This indicates that the command position sequence is shifted forward by one sampling point; the sum of squared residuals after the shift is denoted as... The calculation formula is as follows:

[0048] in, Time shift amount The corresponding set of overlapping unified timestamps.

[0049] In this example, the sum of squared residuals without translation is:

[0050] The sum of squared residuals when the command position sequence is shifted forward by one sampling point is:

[0051] The sum of squared residuals when the command position sequence is shifted one sampling point to the right is:

[0052] The time shift measures the number of sampling points corresponding to the minimum sum of squared residuals, calculated using the following formula:

[0053] Candidate time shift In the middle, calculate the corresponding sum of squared residuals respectively. And select to make The time shift corresponding to the minimum value is taken as the optimal time shift. .

[0054] less than and ,get .

[0055] Therefore, the command position sequence is shifted back by one sampling point. The time offset is denoted as... The calculation formula is as follows:

[0056] The time offset was converted into a phase offset and recorded; the actual arrival time in this stroke was 19 milliseconds, corresponding to a crankshaft angle of 176.0 degrees; the target crankshaft angle given by the command arrival reference was 175.0 degrees; the phase offset was recorded as... The calculation formula is as follows:

[0057] in, The actual arrival time. This refers to the crankshaft angle corresponding to the actual moment of arrival. The target crankshaft angle is given as the reference value for the command's positioning; substituting the values ​​in this example, we get:

[0058] Will As a phase offset, and recorded corresponding to 19 milliseconds, and recorded corresponding to 19 milliseconds.

[0059] Calculate the length deviation based on the alignment result after phase offset correction; the command position enters the hold-in time interval of 18 to 19 milliseconds; let the hold-in time interval be... The length deviation is denoted as The calculation formula is as follows:

[0060] in, To ensure the number of consistent timestamps within the given time interval is maintained, substituting the values ​​from this example yields:

[0061] 0.45 mm was taken as the length deviation and recorded in relation to the corresponding number of strokes.

[0062] Calculate the acceleration / deceleration residuals on the alignment result after phase offset correction; in the example, the command speed increases from 15 ms to 16 ms, selected as the acceleration segment, and the command speed decreases from 17 ms to 18 ms, selected as the deceleration segment; let the acceleration segment be... The deceleration phase is The mean value of the velocity residual during the acceleration phase is denoted as The mean value of the speed residual during the deceleration phase is denoted as The calculation formulas are as follows:

[0063]

[0064] Substituting the values ​​from this example, we get:

[0065]

[0066] Let the root mean square of the position residual during the acceleration segment be denoted as . The root mean square of the position residual of the deceleration section is denoted as The calculation formulas are as follows:

[0067]

[0068] Substituting the values ​​from this example, we get:

[0069]

[0070] The mean velocity residual of the acceleration phase (30 mm / s), the root mean square position residual of the acceleration phase (0.255 mm), the mean velocity residual of the deceleration phase (−90 mm / s), and the root mean square position residual of the deceleration phase (0.412 mm) are recorded as acceleration and deceleration residuals.

[0071] Calculate and mark the angle jitter; the angle increments of adjacent unified timestamps are 1.2 degrees for 15 ms to 16 ms, 0.8 degrees for 16 ms to 17 ms, 1.0 degrees for 17 ms to 18 ms, and 1.0 degrees for 18 ms to 19 ms. Let the angle increment be denoted as... The calculation formula is as follows:

[0072] Let the median of the angle increments within the sliding window be the baseline increment, denoted as . The angle increment difference is denoted as The calculation formula is as follows:

[0073]

[0074]

[0075] This represents the median operation, used to sort the angle increments within the sliding window by numerical value and take the angle increment in the middle position after sorting as the base increment; when the number of angle increments is even, the average of the two middle angle increments after sorting is taken as the base increment.

[0076] Angular jitter is recorded as The calculation formula is as follows:

[0077] Between 16 and 17 milliseconds, the baseline increment is 1.0 degree, the angle increment is 0.8 degrees, and the previous angle increment is 1.2 degrees. Substituting these values ​​into the formula, we get:

[0078] Between 18 and 19 milliseconds, the baseline increment is 1.0 degree, the angle increment is 1.0 degree, and the previous angle increment is 1.0 degree. Substituting these values ​​into the formula, we get:

[0079] Angle jitter and phase offset are recorded together at the corresponding unified timestamp in the synchronous observation package for direct retrieval in subsequent steps.

[0080] The anomaly detection module calculates anomaly detection index set based on the synchronous observation package. The anomaly detection index set includes tension fluctuation index, micro-slippage indicator, and pulse integrity indicator. The anomaly detection index set and error decomposition quantity are subjected to consistency verification, and the error source label and confidence level are output. The error source label is used to distinguish between offsettable deviations, uncontrollable disturbances and acquisition anomalies.

[0081] The method for calculating the anomaly discrimination index set based on the synchronous observation package includes the following steps: Read the unified timestamp sequence, tension sequence, servo position sequence, servo speed sequence, and crankshaft angle sequence from the synchronous observation package, and set the sliding window width bound to the unified timestamp sequence; Within each sliding window, the tension fluctuation index is calculated for the tension sequence. The tension fluctuation index is the difference between the maximum and minimum tension values. Simultaneously, the mean of the absolute values ​​of adjacent differences in the tension sequence is calculated as the tension fluctuation supplement, which is used to characterize the tension fluctuation amplitude and the tension change rate. Within the same sliding window, the position increment sequence is calculated for the servo position sequence, and the tension increment sequence is calculated for the tension sequence. The sign agreement rate and amplitude ratio deviation of the position increment and tension increment are also calculated. The sign agreement rate is the proportion of samples with the same sign within the window, and the amplitude ratio deviation is the deviation of the ratio of the absolute values ​​of the position increment and tension increment relative to the median value within the window. The two are combined according to a preset weight to form a micro-slip indication quantity, which is used to characterize the inconsistency between the servo displacement change and the strip force change. In one implementation of this embodiment, the micro-slip indicator is composed of the correlation characteristics between the position increment and the tension increment. The correlation characteristics include at least the sign consistency rate and the normalized amplitude consistency. The normalized amplitude consistency is obtained by normalizing the absolute value of the tension increment after adding a preset minimum scale factor, so as to avoid ratio singularity and noise amplification when the tension increment is close to zero. When the tension signal is missing or in an invalid range, the micro-slip indicator is output as invalid and an invalid mark is written. At the same time, the micro-slip indicator is prohibited from participating in the calculation of the uncontrollable disturbance discrimination condition.

[0082] Within the same sliding window, calculate the angle increment sequence for the crankshaft angle sequence, and use the median of the angle increment sequence as the reference increment. Count the number of samples where the angle increment is equal to zero, the angle increment deviates too much from the reference increment, and the angle increment sign is reversed. Normalize the result according to the window length to obtain the pulse integrity indicator, which is used to characterize the missing and jump signs of the crankshaft angle acquisition sequence. Within the same sliding window, calculate the angle increment sequence for the crankshaft angle sequence, and use the median of the angle increment sequence as the reference increment; count the number of samples where the angle increment is equal to zero, the number of samples where the absolute value of the difference between the angle increment and the reference increment exceeds a preset deviation threshold, and the number of samples where the signs of adjacent terms of the angle increment are reversed; normalize the number of each sample according to the window length to obtain the pulse integrity indicator.

[0083] Tension fluctuation index, micro-slippage indicator, and pulse integrity indicator are written into the synchronous observation package according to a unified timestamp sequence to form an anomaly discrimination index set. The purpose is to convert the multi-source signals in the synchronous observation package into a computable, alignable, and comparable anomaly discrimination index set, providing quantitative input for consistency verification.

[0084] To facilitate understanding of the calculation process of the above anomaly detection index set, we will now use the same synchronous observation packet sample from the previous event and provide a numerical example for the sliding window corresponding to a unified timestamp of 19 milliseconds. The sliding window width is 3 sampling points, and the sampling interval is 1 millisecond, as shown in the example below: The synchronous observation package recorded the following values ​​between 17 and 19 milliseconds: 17 milliseconds: tension 44 N, servo position 119.3 mm, servo speed 380 mm / s, crankshaft angle 174.0 degrees; 18 milliseconds: tension 50 N, servo position 119.5 mm, servo speed 200 mm / s, crankshaft angle 175.0 degrees; 19 milliseconds: tension 45 N, servo position 119.6 mm, servo speed 0 mm / s, crankshaft angle 176.0 degrees.

[0085] The tension fluctuation index is calculated as follows: within the sliding window, the maximum tension is 50 N and the minimum tension is 44 N. The tension fluctuation index is calculated by subtracting the minimum value from the maximum value, resulting in 6 N. The absolute value of the tension difference between adjacent values ​​is 6 N for 17 ms to 18 ms and 5 N for 18 ms to 19 ms. The mean value is (6 + 5) divided by 2, resulting in 5.5 N, which is used as the tension fluctuation supplement. The tension fluctuation index of 6 N and the tension fluctuation supplement of 5.5 N are written into the corresponding field of 19 ms in the synchronous observation package.

[0086] For the micro-slip indication calculation, within the sliding window, the servo position increment is 0.2 mm for 17-18 ms and 0.1 mm for 18-19 ms; the tension increment is 6 N for 17-18 ms and -5 N for 18-19 ms; the first segment consists of samples where the position increment and tension increment have the same sign, and the sign agreement rate is 1 divided by 2, which equals 0.5; the ratio of the absolute values ​​of the position increment and tension increment is approximately 0.0333 for 0.2 divided by 6 and 0.02 for 0.1 divided by 5, with the median ratio being approximately equal to the median of (0.0333 and 0.02). At 0.0267, the amplitude ratio deviation is taken as the average absolute deviation of the two ratios relative to the median ratio, resulting in (absolute value 0.0333 minus 0.0267 plus absolute value 0.02 minus 0.0267) divided by 2, which is approximately equal to 0.0067. The preset weights are 0.7 and 0.3. The micro-slip indicator is taken as 0.7 multiplied by (1 minus sign consistency rate) plus 0.3 multiplied by (amplitude ratio deviation divided by median ratio), resulting in 0.7 multiplied by 0.5 plus 0.3 multiplied by 0.25, which equals 0.425. The micro-slip indicator of 0.425 is written into the corresponding field of 19 milliseconds in the synchronous observation package.

[0087] The pulse integrity indicator is calculated as follows: within the sliding window, the crankshaft angle increment is 1.0 degree for increments of 17-18 ms and 1.0 degree for increments of 18-19 ms, and the reference increment is the median angle increment of 1.0 degree. The number of samples where the angle increment is zero is 0; the number of samples where the absolute value of the difference between the angle increment and the reference increment exceeds a preset deviation threshold of 0.3 degrees is 0; the number of samples where the signs of adjacent angle increment terms are reversed is 0. Normalized to a window length of 3, the pulse integrity indicator is (0 + 0 + 0) divided by 3, resulting in 0. The pulse integrity indicator 0 is then written into the corresponding field in the synchronous observation package at 19 ms.

[0088] Anomaly detection index set is written into the synchronous observation package, which includes the tension fluctuation index, micro-slippage indicator, and pulse integrity indicator corresponding to 19 milliseconds.

[0089] In this embodiment, the method steps for performing consistency verification between the anomaly discrimination index set and the error decomposition quantity, and outputting error source labels and confidence levels, include: Read the set of anomaly discrimination indicators corresponding to the unified timestamp sequence in the synchronous observation package, as well as the error decomposition quantity corresponding to the unified timestamp sequence.

[0090] A set of acquisition anomaly discrimination conditions is generated, which includes: the pulse integrity indicator exceeding the preset pulse threshold; the angle jitter exceeding the preset jitter threshold; and the phase offset difference component exceeding the preset phase difference threshold. The phase offset difference component is the absolute value of the difference between the phase offsets of adjacent sliding windows. The acquisition anomaly discrimination conditions are counted item by item to obtain the acquisition anomaly condition count value, and the acquisition anomaly condition count value is compared with the acquisition anomaly count threshold. When the comparison result is not less than the threshold, the error source label is output as acquisition anomaly.

[0091] A set of uncontrollable disturbance discrimination conditions is generated, which includes: the tension fluctuation index exceeding the preset tension threshold or the micro-slippage indicator exceeding the preset micro-slippage threshold; the pulse integrity indicator not exceeding the preset pulse threshold; and the acceleration / deceleration residual increment exceeding the preset residual increment threshold. The acceleration / deceleration residual increment is the absolute value of the difference between the acceleration / deceleration residuals output by adjacent sliding windows. The uncontrollable disturbance condition count value is obtained by counting each item in the set of uncontrollable disturbance discrimination conditions, and the uncontrollable disturbance condition count value is compared with the uncontrollable disturbance count threshold. When the comparison result is not less than the threshold, the error source label is output as uncontrollable disturbance.

[0092] A set of conditions for offsetting deviations is generated, which includes: the tension fluctuation index is within the preset tension threshold; the micro-slip indicator is within the preset micro-slip threshold; the pulse integrity indicator is within the preset pulse threshold; the phase offset difference component is within the preset phase difference threshold; and the length deviation difference component is within the preset length difference threshold. The length deviation difference component is the absolute value of the difference between the length deviations output by adjacent sliding windows. The offset deviation condition count is obtained by counting each item in the offset deviation condition set, and the offset deviation condition count is compared with the offset deviation count threshold. When the comparison result is not less than the threshold, the error source label is output as offset deviation.

[0093] The reliability is calculated and written into the synchronous observation package. Consistency scores are calculated for the set of acquisition anomaly discrimination conditions, the set of uncontrollable disturbance discrimination conditions, and the set of offsettable deviation discrimination conditions, respectively. The consistency score is the sum of the number of condition items in each set whose comparison results are above or below the threshold, according to a preset weight, and normalized according to the number of items in the set. The maximum consistency score is taken as the reliability, and the reliability and the error source label are written into the synchronous observation package according to a unified timestamp sequence. The purpose is to use the anomaly discrimination index set to constrain the interpretation of the error decomposition quantity, map the error performance corresponding to the synchronization error to the three types of error source labels: offsettable deviation, uncontrollable disturbance, and acquisition anomaly, and provide verifiable output basis with reliability.

[0094] In one implementation of this embodiment, when both the set of acquisition anomaly discrimination conditions and the set of uncontrollable disturbance discrimination conditions meet the output conditions, the acquisition anomaly label is output first; when both the set of uncontrollable disturbance discrimination conditions and the set of offsettable deviation discrimination conditions meet the output conditions, the uncontrollable disturbance label is output first; and conflict markers and corresponding consistency scores are written into the synchronous observation package for subsequent verification.

[0095] To facilitate understanding of the error source labeling and confidence level output process, the following example is provided: The anomaly detection index set corresponding to 19 milliseconds within the sliding window is as follows: pulse integrity indicator 0, tension fluctuation index 6 N, and micro-slippage indicator 0.425. These values ​​are from the aforementioned anomaly detection index set calculation example. The error decomposition within the 19-millisecond sliding window is as follows: phase offset 1.0 degree, angle jitter 0 degree, and acceleration / deceleration residual increment 30 mm / s. The phase offset of 1.0 degree comes from the aforementioned error decomposition example. The acceleration / deceleration residual increment is the absolute value of the difference between the acceleration / deceleration residuals output by adjacent sliding windows. For example, the acceleration / deceleration residual of the previous sliding window is 60 mm / s, and the acceleration / deceleration residual of this sliding window is 30 mm / s, resulting in an acceleration / deceleration residual increment of 30 mm / s. The phase offset difference component is the absolute value of the difference between the phase offsets output by adjacent sliding windows. For example, the phase offset of the previous sliding window is 1.1 degree, and the phase offset of this sliding window is 1.0 degree, resulting in a phase offset difference component of 0.1 degree. The length deviation difference component is, for example, the length deviation of the previous sliding window is 0.50 mm, and the length deviation of this sliding window is 0.45 mm, resulting in a length deviation difference component of 0.05 mm.

[0096] The acquisition anomaly detection condition set is compared and counted. The preset pulse threshold is 0.3, and the pulse integrity indicator 0 is compared with the preset pulse threshold and the result is within the threshold. The preset jitter threshold is 0.5 degrees, and the angle jitter 0 degrees is compared with the preset jitter threshold and the result is within the threshold. The preset phase difference threshold is 0.5 degrees, and the phase offset difference component 0.1 degrees is compared with the preset phase difference threshold and the result is within the threshold. The acquisition anomaly detection condition set is counted item by item, and the acquisition anomaly condition count value is 0. The acquisition anomaly count threshold is 2, and the acquisition anomaly condition count value 0 is compared with the acquisition anomaly count threshold 2 and the result is less than. Therefore, this sliding window does not output the error source label corresponding to the acquisition anomaly.

[0097] The set of uncontrollable disturbance discrimination conditions is compared and counted. The preset tension threshold is 5 N, and the tension fluctuation index of 6 N is found to exceed the preset tension threshold. The preset micro-slip threshold is 0.6, and the micro-slip indication of 0.425 is found to be within the preset micro-slip threshold. In this example, the combination of the tension fluctuation index exceeding the threshold or the micro-slip indication exceeding the threshold is considered to exceed the threshold. The preset pulse threshold is still 0.3, and the pulse integrity indication of 0 is found to be within the preset pulse threshold. The preset residual increment threshold is 20 mm / s, and the acceleration / deceleration residual increment of 30 mm / s is found to exceed the preset residual increment threshold. The set of uncontrollable disturbance discrimination conditions is counted item by item, and the uncontrollable disturbance condition count value is 3. The uncontrollable disturbance count threshold is 2, and the uncontrollable disturbance condition count value 3 is found to be not less than the uncontrollable disturbance count threshold 2. Therefore, the error source label of this sliding window output is uncontrollable disturbance.

[0098] The set of conditions for offsetting deviations is compared and counted. The preset tension threshold is 5 N. The tension fluctuation index of 6 N is compared with the preset tension threshold and the result is that it exceeds the threshold. The condition corresponding to the tension fluctuation index not exceeding the threshold is "not passed" in this example. The preset micro-slip threshold is 0.6. The micro-slip indication of 0.425 is compared with the preset micro-slip threshold and the result is that it does not exceed the threshold. The preset pulse threshold is 0.3. The pulse integrity indication of 0 is compared with the preset pulse threshold and the result is that it does not exceed the threshold. The preset phase difference threshold is 0.5 degrees. The phase offset difference component of 0.1 degrees is compared with the preset phase difference threshold and the result is that it does not exceed the threshold. The preset length difference threshold is 0.2 mm. The length deviation difference component of 0.05 mm is compared with the preset length difference threshold and the result is that it does not exceed the threshold. The set of conditions for offsetting deviations is counted item by item, and the count value of offsetting deviations is 4. The count value of offsetting deviations is 4. The count value of offsetting deviations is 4. The result of comparing the count value of offsetting deviations 4 with the count value of offsetting deviations is not less than 4. Therefore, the label output condition also appears in the direction of offsetting deviations in this example.

[0099] The credibility calculation and label selection are written. The consistency score is normalized according to the number of set items. The consistency score for collection anomalies is 0 divided by 3, which equals 0; the consistency score for uncontrollable disturbances is 3 divided by 3, which equals 1; the consistency score for offsettable deviations is 4 divided by 5, which equals 0.8; the maximum consistency score is 1, corresponding to a credibility of 1, and the corresponding error source label is selected as uncontrollable disturbance.

[0100] The update control module generates compensation update control conditions based on the error source label and confidence level. The confidence level is compared with a preset threshold. When the comparison result meets the compensation update control conditions and the error source label is a cancelable deviation, the compensation amount is updated and the parameters before the update are saved as the rollback reference. Otherwise, the compensation amount remains unchanged. The compensation amount includes phase compensation amount, length compensation amount and control gain.

[0101] In one implementation of this embodiment, the rollback reference is the parameter snapshot that most recently met the update permission and did not have an abnormal acquisition tag appear in subsequent consecutive preset windows; when a freeze update event occurs, the controller restores the phase compensation amount, length compensation amount and control gain to the rollback reference, and writes the freeze update flag and rollback reference index into the self-verification record to avoid abnormally driven parameter drift from propagating in subsequent pulses.

[0102] In this embodiment, the method steps for generating compensation update control conditions based on error source labels and confidence levels include: Read the error source label and confidence level corresponding to the unified timestamp from the synchronous observation package, and synchronously read the phase offset, length deviation and acceleration / deceleration residual corresponding to the same unified timestamp; The credibility is compared with a preset threshold to obtain a credibility comparison result, which is used to indicate whether the credibility is passed or not. Perform consistency counting on the error source labels, count the number of times the error source label is equal to the offset deviation within the most recent preset window number, and compare the number of occurrences with the preset consistency count threshold to obtain the label consistency comparison result; The confidence comparison results are merged with the label consistency comparison results to generate compensation update control conditions. The compensation update control conditions include at least update allow and update prohibit, and are associated with a unified timestamp and written into the synchronous observation package. The reason is that the error source label and confidence are transformed into an executable update judgment basis, so that the compensation amount update is only triggered under the interpretation path of offsetting deviation, and the compensation amount is avoided from being driven to change under uncontrollable disturbances or abnormal acquisition conditions.

[0103] To illustrate the process of generating compensation update control conditions, a specific example is provided below: When the unified timestamp is 19 milliseconds, the confidence level of the synchronous observation packet is 1, and the error source label is uncontrollable disturbance. The result comes from the aforementioned consistency verification example. The preset threshold is 0.8, and the confidence level is passed when compared with the preset threshold. The most recent preset window number is 3 sliding windows, corresponding to unified timestamps of 17 milliseconds, 18 milliseconds, and 19 milliseconds respectively. The error source labels for the three are cancelable deviation, uncontrollable disturbance, and uncontrollable disturbance respectively. The 17 millisecond and 18 millisecond labels are output records of the same impulse in adjacent sliding windows.

[0104] The credibility is compared with the preset threshold to obtain the credibility comparison result. The credibility of 1 corresponding to 19 milliseconds is compared with the preset threshold of 0.8, and the credibility comparison result is passed.

[0105] The label consistency comparison result is obtained by counting the consistency of the error source label. Within the most recent 3 sliding windows, the number of times the error source label is equal to the offsetting deviation is counted. 17 milliseconds is an offsetting deviation and is counted as 1 time. 18 milliseconds and 19 milliseconds are uncontrollable disturbances and are counted as 0 times. The total number of occurrences is 1. The preset consistency number threshold is 2. The occurrence 1 is compared with the preset consistency number threshold 2. The label consistency comparison result is not passed.

[0106] The merged comparison results generate compensation update control conditions and are written into the synchronous observation package. If the confidence comparison results pass and the label consistency comparison results fail, the compensation update control conditions are output as update prohibited and associated with a unified timestamp of 19 milliseconds and written into the synchronous observation package. This is used in subsequent steps to directly read and control whether the compensation amount is updated.

[0107] In this embodiment, the method steps for updating the compensation amount include: Read the compensation update control conditions, error source labels and confidence comparison results corresponding to the unified timestamp in the synchronous observation package; When the compensation update control condition indicates that the update is allowed and the error source label is a cancelable deviation, read the current phase compensation amount, length compensation amount and control gain, save the parameters before the update as the rollback reference, and store the rollback reference with a unified timestamp. The phase compensation update increment is calculated. The phase compensation update increment is obtained by converting the phase offset through a preset phase mapping rule. The phase mapping rule includes converting the phase offset into a time offset according to the sampling interval and performing a proportional conversion according to a preset compensation coefficient. Calculate the length compensation update increment, which is obtained by converting the length deviation through a preset length mapping rule. The length mapping rule includes scaling the length deviation proportionally and keeping the update direction consistent with the sign of the length deviation. Calculate the control gain update increment, which is obtained by converting the acceleration / deceleration residuals through a preset gain mapping rule. The gain mapping rule includes mapping the acceleration / deceleration residuals into piecewise gain increments and applying a limit to the gain increments. The phase compensation, length compensation, and control gain are each added to their corresponding update increments to obtain update parameters, which are then written into the synchronous observation package. When the compensation update control condition indicates that update is prohibited, the phase compensation, length compensation, and control gain are kept unchanged and written into the synchronous observation package. The purpose is to trigger the update of the compensation amount with error source label and confidence constraint, and to provide a rollback reference when the update is triggered, so that the changes in the phase compensation, length compensation, and control gain have traceable and recoverable attributes.

[0108] To facilitate understanding of the execution process of compensation amount update and rollback baseline saving, the following example is provided: At a unified timestamp of 19 milliseconds, the compensation update control condition recorded in the synchronous observation package is "update prohibited," the error source label is "uncontrollable disturbance," and the confidence comparison result is "passed." This result comes from the aforementioned example of compensation update control conditions. Therefore, at 19 milliseconds, the phase compensation, length compensation, and control gain remain unchanged and are written into the synchronous observation package. To demonstrate the update-allowed scenario, at a unified timestamp of 18 milliseconds, the compensation update control condition recorded in the synchronous observation package is "update allowed," the error source label is "cancellable deviation," and the confidence comparison result is "passed." At 18 milliseconds, the error decomposition measures a phase offset of 1.0 degree, a length deviation of 0.45 mm, and an acceleration / deceleration residual of 30 mm / s. The phase offset and length deviation are taken from the aforementioned example of error decomposition measures, and the acceleration / deceleration residual is taken from the aforementioned example of acceleration / deceleration residual calculation.

[0109] Read the compensation update control conditions, error source labels, and confidence comparison results. At 18 milliseconds, read that the compensation update control conditions are allowed, the error source labels are offset deviations, and the confidence comparison results are passed. Use these as the inputs to trigger this update.

[0110] Save the parameters before the update as the rollback baseline and store them together. At 18 milliseconds, read the current phase compensation amount as 0.20 milliseconds, the length compensation amount as 0.10 millimeters, and the control gain as 1.50. Save the above parameters before the update as the rollback baseline and store them together with a unified timestamp of 18 milliseconds for easy rollback calls in the future.

[0111] The phase compensation update increment is calculated according to the preset phase mapping rule. The preset phase mapping rule takes the phase offset as the time offset and multiplies it by the preset compensation coefficient. When the phase offset of 1.0 degree is converted into the time offset, the crankshaft angular velocity of the current stroke is used as the conversion basis. For example, the crankshaft angular velocity is taken as 1 degree per millisecond, then the time offset is 1.0 millisecond. The preset compensation coefficient is 0.1, and the phase compensation update increment is 0.1 multiplied by 1.0 millisecond, which gives 0.10 milliseconds.

[0112] The length compensation update increment is calculated according to the preset length mapping rule. The preset length mapping rule takes the length deviation ratio and scales it to keep it consistent with the length deviation sign. A length deviation of 0.45 mm is positive. The preset scaling factor is 0.2. The length compensation update increment is 0.2 multiplied by 0.45 mm, which is 0.09 mm. The update direction is positive.

[0113] The control gain update increment is calculated according to the preset gain mapping rules, and a limiting is applied. The preset gain mapping rules adopt a piecewise mapping: the gain increment is 0 when the acceleration / deceleration residual is no greater than 20 mm / s; the gain increment is 0.01 multiplied by (acceleration / deceleration residual minus 20 mm / s) when the acceleration / deceleration residual is greater than 20 mm / s but no greater than 60 mm / s; and the gain increment is 0.40 when the acceleration / deceleration residual is greater than 60 mm / s. In the example, the acceleration / deceleration residual is 30 mm / s, and the gain increment is 0.01 multiplied by 10, resulting in 0.10. The gain increment limiting range is from -0.20 to 0.20. 0.10 falls within the limiting range, so the limiting result is still 0.10.

[0114] The updated compensation values ​​are written to the synchronous observation package. The phase compensation value of 0.20 ms is added to the phase compensation value update increment of 0.10 ms to obtain the updated phase compensation value of 0.30 ms; the length compensation value of 0.10 mm is added to the length compensation value update increment of 0.09 mm to obtain the updated length compensation value of 0.19 mm; the control gain of 1.50 is added to the control gain update increment of 0.10 to obtain the updated control gain of 1.60; the updated phase compensation value, updated length compensation value, and updated control gain are written to the corresponding field in the synchronous observation package at 18 ms; for records where the compensation update control condition at 19 ms is update prohibited, the phase compensation value of 0.30 ms, the length compensation value of 0.19 mm, and the control gain of 1.60 are kept unchanged, and the unchanged results are written to the corresponding field in the synchronous observation package at 19 ms, consistent with the aforementioned update prohibited logic.

[0115] The constraint safety module applies boundary constraints to the compensation increment generated by the update; when the error source label is uncontrollable disturbance or abnormal acquisition, the update is frozen and rolled back to the rollback baseline; the rolled-back parameters are used to generate the feeding command for the next stroke, the window margin is calculated and the prohibited feeding range is adjusted accordingly, and a self-verification record is generated; the window margin is the allowable offset margin of the feeding action relative to the boundary of the punch press stroke window.

[0116] In this embodiment, the method steps for applying boundary constraints to the compensation increment generated by the update include: Read the phase compensation amount, length compensation amount, control gain and their update parameters corresponding to the same unified timestamp in the synchronous observation package, calculate the phase compensation increment, length compensation increment and gain compensation increment, and take the update parameter minus the parameter before update for the compensation increment; Each compensation increment is compared with the preset upper limit and the preset lower limit. Compensation increments exceeding the upper limit are truncated to the upper limit, and compensation increments below the lower limit are truncated to the lower limit. The boundary constraint parameters are obtained by adding the compensation increment obtained from the truncation process to the parameters before the update, and then writing the boundary constraint parameters into the synchronous observation package. The purpose is to limit the change of compensation amount to a controllable range and reduce the risk of cumulative offset of compensation amount.

[0117] In this embodiment, the method steps for freezing updates and rolling back to the rollback baseline include: Read the error source label from the synchronous observation package; when the error source label is an uncontrollable disturbance or acquisition anomaly, read the rollback reference stored with a unified timestamp, replace the phase compensation amount, length compensation amount, and control gain with the corresponding parameters of the rollback reference, and write the freeze update flag into the synchronous observation package; use the parameters corresponding to the rollback reference as the generation parameters for the next stroke feeding command, read the crankshaft angle sequence and the preset feeding command generation rules, and generate command position values ​​and command speed values ​​point by point according to the crankshaft angle, where the phase compensation amount is used to correct the target time or target crankshaft angle of the command arrival, the length compensation amount is used to correct the target position value of the command arrival, and the control gain is used to correct the speed change rate of the command acceleration and deceleration segment, and write the generation results into the synchronous observation package; the purpose is to prevent uncontrollable disturbances or acquisition anomalies from driving the accumulation of compensation amount, while ensuring that the next stroke command generation uses rollback parameters.

[0118] In this embodiment, the method for calculating the window margin and adjusting the prohibited feeding range accordingly includes the following steps: Read the starting crankshaft angle and ending crankshaft angle of the punch press stroke window corresponding to the window boundary, and read the next stroke command trajectory to obtain the feeding start crankshaft angle and feeding end crankshaft angle.

[0119] The window margin is calculated by taking the minimum of the absolute values ​​of the difference between the starting crankshaft angle of the window and the starting crankshaft angle of the feed, and the absolute values ​​of the difference between the ending crankshaft angle of the window and the ending crankshaft angle of the feed. The window margin is used to characterize the allowable offset margin of the feed action relative to the boundary of the punch press stroke window.

[0120] The window margin is compared with a preset margin threshold. If the window margin is not greater than the preset margin threshold, an expansion adjustment is performed on the prohibited feeding interval. The expansion adjustment includes moving the starting crankshaft angle of the prohibited feeding interval towards the starting crankshaft angle of the window by a preset adjustment step, and moving the ending crankshaft angle of the prohibited feeding interval towards the ending crankshaft angle of the window by a preset adjustment step. If the window margin is greater than the preset margin threshold, a convergence adjustment is performed on the prohibited feeding interval. The convergence adjustment includes moving the starting crankshaft angle of the prohibited feeding interval away from the starting crankshaft angle of the window by a preset adjustment step, and moving the ending crankshaft angle of the prohibited feeding interval away from the ending crankshaft angle of the window by a preset adjustment step. The adjusted prohibited feeding interval is written into the synchronous observation package. The purpose is to drive the angle boundary adjustment of the prohibited feeding interval with the window margin comparison result, so that the feeding action retains a preset allowable offset margin within the stroke window, thereby reducing the risk of mold rubbing and accidental stop.

[0121] In one implementation of this embodiment, the adjustment of the prohibited feeding range is constrained by a preset maximum adjustment step size and a preset boundary range in each stroke, and the adjusted prohibited feeding range does not overlap with the range corresponding to the feeding start crankshaft angle and the feeding end crankshaft angle; when the adjustment will cause overlap, the prohibited feeding range of the previous stroke remains unchanged and a constraint trigger mark is written.

[0122] To illustrate the process of calculating the window margin and adjusting the prohibited feeding range, the following example is provided: The punch press stroke window boundary is pre-calibrated with a starting crankshaft angle of 160 degrees and an ending crankshaft angle of 200 degrees. The next stroke command trajectory reads the starting crankshaft angle as 168 degrees and the ending crankshaft angle as 192 degrees. The command trajectory is obtained by correcting the parameters after the aforementioned rollback. The preset margin threshold is 8 degrees, and the preset adjustment step size is 2 degrees. Before adjustment, the prohibited feeding range is a starting crankshaft angle of 162 degrees and an ending crankshaft angle of 198 degrees.

[0123] Calculate the window margin. The absolute value of the difference between the starting crankshaft angle of the window and the starting crankshaft angle of the feeding is 160 minus 168, which equals 8 degrees. The absolute value of the difference between the ending crankshaft angle of the window and the ending crankshaft angle of the feeding is 200 minus 192, which also equals 8 degrees. The window margin is the minimum of the absolute values ​​of the differences between the above two values, which equals 8 degrees. This means that the allowable offset margin of the feeding action relative to the window boundary is 8 degrees.

[0124] The window margin is compared with the preset margin threshold, and the prohibited feeding range is adjusted. The window margin of 8 degrees is compared with the preset margin threshold of 8 degrees. Since the window margin is not greater than the preset margin threshold, the prohibited feeding range is expanded. During the expansion adjustment, the starting crankshaft angle of the prohibited feeding range is moved by a preset adjustment step of 2 degrees towards the starting crankshaft angle of 160 degrees in the window direction, and the starting crankshaft angle is adjusted from 162 degrees to 160 degrees. The ending crankshaft angle of the prohibited feeding range is moved by a preset adjustment step of 2 degrees towards the ending crankshaft angle of 200 degrees in the window direction, and the ending crankshaft angle is adjusted from 198 degrees to 200 degrees. The adjusted starting crankshaft angle of 160 degrees and ending crankshaft angle of 200 degrees of the prohibited feeding range are written into the synchronous observation package for the next stroke control to read.

[0125] In this embodiment, the method steps for generating a self-verification record include: The synchronous observation package includes error source labels, confidence levels, freeze update flags, rollback baseline indexes, boundary constraint parameters, window margins, and prohibited feeding intervals. Perform a comparison and verification between the boundary constraint parameters and the preset parameter boundaries; perform an interval verification between the crankshaft angle at the start of feeding and the crankshaft angle at the end of feeding and the crankshaft angle at the beginning and end of the window; and output the verification results. The aforementioned aggregated fields and verification results are associated with a unified timestamp and written into the self-verification record. The purpose is to provide traceable evidence for freeze updates, rollback to the rollback baseline, window margin calculation, and adjustment of prohibited feeding ranges, supporting on-site verification and fault location.

[0126] To illustrate how the self-verification record is generated, a specific example is provided below: The unified timestamp is 19 milliseconds. The corresponding fields of the synchronous observation package are: error source label "uncontrollable disturbance", confidence level 1, freeze update mark "frozen", rollback baseline index "18 milliseconds", boundary constraint parameters "phase compensation 0.20 milliseconds", "length compensation 0.10 mm", "control gain 1.50", "window margin 8 degrees", and "prohibited feeding interval" "starting crankshaft angle 160 degrees" and "ending crankshaft angle 200 degrees". The window margin and prohibited feeding interval are from the aforementioned window margin example, and the rollback baseline index and boundary constraint parameters are from the aforementioned rollback baseline example.

[0127] The aggregated fields include error source labels, confidence level, freeze update flag, rollback baseline index, boundary constraint parameters, window margin, and prohibited feeding interval, all grouped together with a unified timestamp of 19 milliseconds into a set of fields to be written.

[0128] The comparison and verification results are output. The preset parameter boundaries are set as follows: phase compensation from 0.00 ms to 1.00 ms, length compensation from -1.00 mm to 1.00 mm, and control gain from 0.50 to 2.00. If the boundary constraint parameters fall within the corresponding boundary range, the comparison and verification result is recorded as passed. The starting crankshaft angle of the window is 160 degrees, the ending crankshaft angle of the window is 200 degrees, the starting crankshaft angle of the feeding is 168 degrees, and the ending crankshaft angle of the feeding is 192 degrees. If the starting crankshaft angle of the feeding and the ending crankshaft angle of the feeding are within the limits of the starting crankshaft angle of the window and the ending crankshaft angle of the window, the interval verification result is recorded as passed. The two verification results are combined into a verification result field.

[0129] Write a self-verification record. Associate the set of fields to be written with the verification result field with a unified timestamp of 19 milliseconds and write them into the self-verification record. The record content includes timestamp, error source label, confidence level, freeze update flag, rollback baseline index, boundary constraint parameters, window margin, prohibited feeding range and verification result, which are used for subsequent on-site verification and fault location.

[0130] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A synchronous adaptive control system for the feed stroke of a punch press, characterized in that, include: The observation construction module is used to acquire the crankshaft angle, servo position speed, tension, and positioning signal of the punch press, and generate a synchronous observation package by aligning it with a unified timestamp. The synchronization error is calculated from the difference between the command being received and the actual command being received. The reconstruction decomposition module reconstructs the command trajectory and the actual trajectory from the synchronous observation package, and obtains the error decomposition quantity by differentiating the command trajectory and the actual trajectory. The anomaly detection module calculates anomaly detection index set based on synchronous observation packets; Perform consistency verification between the anomaly discrimination index set and the error decomposition quantity, and output the error source label and confidence level. The error source label is used to distinguish between offsettable deviations, uncontrollable disturbances and acquisition anomalies. The update control module generates compensation update control conditions based on the error source label and confidence level; the confidence level is compared with a preset threshold. If the comparison result meets the compensation update control conditions and the error source label is a cancelable deviation, the compensation amount is updated; otherwise, the compensation amount remains unchanged. The constraint safety module is used to apply boundary constraints to the compensation increment generated by the update. When the error source label is an uncontrollable disturbance or acquisition anomaly, the update is frozen and rolled back to the rollback baseline. The rolled-back parameters are used to generate the feeding command for the next stroke, the window margin is calculated and the prohibited feeding range is adjusted accordingly, and a self-verification record is generated.

2. The press feed stroke synchronous adaptive control system according to claim 1, characterized in that, Methods for calculating synchronization error from the difference between command arrival and actual arrival include: Based on the crankshaft angle and preset feeding instruction generation rules in the synchronous observation package, calculate the command arrival reference quantity corresponding to each unified timestamp. The command arrival reference quantity includes at least the target position value of the command arrival and the target time or target crankshaft angle of the command arrival. The actual arrival time is determined based on the arrival status edge marker in the synchronous observation packet, and the servo position is obtained at the unified timestamp corresponding to the actual arrival time as the actual arrival position; when the arrival status is not triggered, the actual arrival time is taken as the time when the servo position reaches the target position corresponding to the command arrival reference quantity. The command position is obtained by taking the command position reference value at the actual position time, and the difference between the command position and the actual position is calculated as the synchronization error.

3. The press feed stroke synchronous adaptive control system according to claim 1, characterized in that, Methods for reconstructing command trajectories and actual trajectories based on synchronous observation packets include: Read the unified timestamp sequence and crankshaft angle sequence from the synchronous observation package, and read the preset feeding instruction generation rule corresponding to the crankshaft angle; The command position and command velocity values ​​are calculated point by point according to the unified timestamp sequence to obtain the command position sequence and command velocity sequence. The command position sequence and command velocity sequence are then combined into the command trajectory. The servo position sequence and servo velocity sequence in the synchronous observation package are read and arranged according to the unified timestamp sequence to obtain the actual position sequence and actual velocity sequence. The actual position sequence and actual velocity sequence are then combined into the actual trajectory.

4. The press feed stroke synchronous adaptive control system according to claim 3, characterized in that, Methods for obtaining the error decomposition by differentiating the command trajectory from the actual trajectory include: Error decomposition includes phase offset, length deviation, acceleration / deceleration residuals, and angle jitter; Calculate position and velocity residuals using a unified timestamp; enumerate time translation amounts within a preset translation search range, take the time translation amount corresponding to the minimum sum of squares of position residuals, and convert it into phase offset according to the sampling interval; calculate the length deviation by averaging the position residuals during the position holding time interval; calculate the acceleration and deceleration residuals by averaging the velocity residuals during acceleration and deceleration phases, and obtain the acceleration / deceleration residuals by calculating the peak value or root mean square value of the position residuals; calculate the angle increment and adjacent differences from the crankshaft angle, using the middle value of the angle increment sliding window as the reference increment, and take the angle jitter as the combination of the deviation of the angle increment from the reference increment and the absolute value of the increment difference.

5. The press feed stroke synchronous adaptive control system according to claim 4, characterized in that, The position residual is the command position value minus the actual position value under the same unified timestamp; the velocity residual is the command velocity value minus the actual velocity value under the same unified timestamp.

6. The press feed stroke synchronous adaptive control system according to claim 4, characterized in that, Methods for calculating anomaly discrimination index sets based on synchronous observation packets include: Read the unified timestamp sequence, tension sequence, servo position sequence, and crankshaft angle sequence from the synchronous observation package, and set the width of the sliding window bound to the unified timestamp sequence; within the sliding window, obtain the tension fluctuation index from the difference between the maximum and minimum tension values; Within the sliding window, the position increment sequence and tension increment sequence are calculated from the servo position sequence and tension sequence, respectively, and the micro-slip indication is calculated from the position increment sequence and tension increment sequence; Within a sliding window, the angle increment sequence is calculated from the crankshaft angle sequence. The reference increment is taken as the median of the angle increment sequence. The number of samples where the angle increment is equal to zero, the number of samples where the absolute value of the difference between the angle increment and the reference increment exceeds the preset deviation threshold, and the number of samples where the signs of adjacent terms of the angle increment are reversed are counted. The pulse integrity indicator is then normalized according to the window length. Tension fluctuation index, microslippage indicator, and pulse integrity indicator are written into the synchronous observation package according to a unified timestamp sequence to form an anomaly discrimination index set.

7. The press feed stroke synchronous adaptive control system according to claim 6, characterized in that, Methods for performing consistency checks on the anomaly detection index set and error decomposition quantities, and outputting error source labels and confidence levels, include: Read the anomaly discrimination index set and error decomposition quantity from the synchronous observation package; generate the acquisition anomaly discrimination condition set, uncontrollable disturbance discrimination condition set, and offset deviation discrimination condition set respectively; count each item in each discrimination condition set to form the corresponding condition count value, compare it with the corresponding count threshold, and output the error source label; calculate the consistency score for each discrimination condition set, and take the maximum consistency score as the confidence level.

8. The press feed stroke synchronous adaptive control system according to claim 7, characterized in that, Methods for generating compensation update control conditions based on error source labels and confidence levels include: The confidence level is compared with a preset threshold to obtain the confidence level comparison result; within the most recent preset window number, the number of times the error source label is equal to the number of times the offset deviation occurs is counted, and the number of occurrences is compared with a preset consistency number threshold to obtain the label consistency comparison result; the confidence level comparison result and the label consistency comparison result are merged to generate compensation update control conditions, which include update allow and update prohibit.

9. The press feed stroke synchronous adaptive control system according to claim 8, characterized in that, The steps for updating the compensation amount include: Read the compensation update control conditions and error source labels corresponding to the unified timestamp in the synchronous observation package; when the compensation update control conditions are update allowed and the error source labels are offset deviations, read the phase compensation amount, length compensation amount and control gain, and save the parameters before the update as the rollback reference. The phase compensation increment is obtained based on the phase offset according to the preset phase mapping rule, the length compensation increment is obtained based on the length deviation according to the preset length mapping rule, and the control gain increment is obtained based on the acceleration / deceleration residual according to the preset gain mapping rule. The corresponding update increments of the phase compensation, length compensation and control gain are superimposed to obtain the update parameters. When the compensation update control condition is update prohibited, the phase compensation, length compensation and control gain remain unchanged.

10. The press feed stroke synchronous adaptive control system according to claim 9, characterized in that, Methods for calculating window margin and adjusting the prohibited feeding range accordingly include: Read the starting crankshaft angle and ending crankshaft angle of the punch press stroke window corresponding to the window boundary, and read the next stroke command trajectory to obtain the feeding start crankshaft angle and feeding end crankshaft angle; The window margin is the minimum of the absolute values ​​of the difference between the starting crankshaft angle of the window and the starting crankshaft angle of the feeding start, and the absolute values ​​of the difference between the ending crankshaft angle of the window and the ending crankshaft angle of the feeding finish. The window margin is compared with the preset margin threshold. If the window margin is not greater than the preset margin threshold, the starting crankshaft angle of the prohibited feeding interval is moved by a preset adjustment step size in the direction of the starting crankshaft angle of the window, and the ending crankshaft angle of the prohibited feeding interval is moved by a preset adjustment step size in the direction of the ending crankshaft angle of the window. When the window margin is greater than the preset margin threshold, the starting crankshaft angle of the prohibited feeding interval will be moved away from the direction of the starting crankshaft angle of the window, and the preset adjustment step size will be moved. The ending crankshaft angle of the prohibited feeding interval will be moved away from the direction of the ending crankshaft angle of the window, and the preset adjustment step size will be moved.

Images