First bending position deviation correction method and system for servo pipe bending machine

Abstract

The invention relates to the technical field of numerical control compensation, in particular to a first bending position deviation correction method and system for a servo pipe bender, in the first bending position deviation correction method and system for the servo pipe bender, fuzzy logic judgment is carried out on deviation direction consistency, a collaborative deviation behavior between feeding and clamping is converted into a trend quantity, parameter items influencing first bending error changes are accurately extracted, and the deviation correction accuracy is improved. Linkage combination of three types of control section instructions is realized, a time difference change trend is learned and extracted through a long-short-term memory network, a control time period with overlapped structure characteristics is identified, stable interval priority ordering and cross-section buffering of an error high-incidence area are realized in instruction section structure rearrangement, and the accuracy of instruction section structure rearrangement is improved. Multiple structural variables such as error trend, trigger delay and control index compression ratio are introduced into a control regulation system in participation item selection, so that the response timeliness of a key decision in a deviation correction path and the instruction chain stability under error interference are remarkably improved, and the transmission amplification risk of errors in a forming path is effectively controlled.

Description

Technical Field

[0001] This invention relates to the field of CNC compensation technology, and in particular to a method and system for correcting the first bend position deviation of a servo pipe bending machine. Background Technology

[0002] The field of CNC compensation technology mainly involves real-time or preset compensation for positioning errors caused by clamping errors, initial zero point offset, mechanical backlash, thermal deformation and initial calibration errors during the execution process under the control of CNC system, so as to ensure the accurate execution of the target machining path and actions. The key lies in correcting the machining path, displacement command and coordinate data through software and control commands to achieve closed-loop control, accuracy repair and dynamic correction.

[0003] The method for correcting the first bend position deviation of a servo pipe bending machine aims to adjust the position by using error detection and CNC system compensation logic. By correcting the deviation of the first bend position, the overall pipe bending accuracy is improved, error accumulation is prevented, product consistency and yield are enhanced, and the pipe forming process meets the set process parameters, bending angle, plane deviation and end coordinates. It also meets the application requirements of high-precision continuous pipe bending structures.

[0004] Existing technologies rely on pre-set compensation for a single source of error, mainly focusing on pre-correction of control commands and static replacement of coordinate paths. They lack the expression of multi-factor cross-characteristics, which makes it impossible for the system to identify the dynamic error structure caused by the coordinated offset between clamping, feeding, servo and other units in a timely manner. Since the lack of trend judgment and time-series overlap segment screening mechanism, the system can usually only retrospectively analyze the cause of deviation after the first bend is completed, resulting in insufficient real-time adjustment capability. Moreover, the errors formed in batch forming are often continuously superimposed, making it difficult to establish a fast-response control feedback channel, resulting in the accumulation of endpoint coordinate deviation and angle distortion, which limits the batch stability of high-precision continuous bending pipes. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of the prior art and to propose a method and system for correcting the first bend position deviation of a servo pipe bending machine.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: a method for correcting the first bend position deviation of a servo pipe bending machine, comprising the following steps:

[0007] Step 1: Based on the displacement error of the feeding mechanism, the lag start behavior of the clamping mechanism, the response deviation of the servo corner unit, and the initial drift of the first forming angle, extract the feeding point difference, clamping delay sequence, and identify the servo delay section to generate the first bend start control behavior feature set.

[0008] Step 2: Based on the first bend start control behavior feature set, compare the feeding fluctuation point with the clamping delay value, use fuzzy logic to determine the consistency of the offset direction, identify the error curve fluctuation interval, select the feeding step distance, clamping advance amount and angle compression ratio, and generate an error trend control parameter group.

[0009] Step 3: Based on the error trend control parameter group, obtain the feeding path interpolation and clamping trigger time parameters, adjust the servo node compression segment, match and combine the three types of index parameters, and generate a linkage control instruction set;

[0010] Step 4: Based on the aforementioned linkage control instruction set, a long short-term memory network is used to identify the time difference trend structure, calculate the feeding path and clamping time difference, determine the angle curvature difference, rearrange the order of instruction segments, adjust the combination of feeding instructions, clamping instructions and cornering instructions, and generate the first curve execution path adjustment model set.

[0011] Step 5: Based on the first bend execution path adjustment model set, set the feeding start point, align the clamping task identifier, synchronize the servo initial sequence, write the feeding, clamping and servo execution segment labels to the sequence register, and generate the first bend control execution signal group.

[0012] As a further aspect of the present invention, the specific steps for generating the first curve initiation control behavior feature set are as follows:

[0013] Based on the displacement error of the feeding mechanism, the lag start behavior of the clamping mechanism, the response deviation of the servo rotation unit, and the initial drift of the first forming angle, the three-axis coordinate set between the feeding start point and the actual path offset point is extracted. The difference sequence between the coordinate set and the standard trajectory is calculated in batches, and the clamping control command trigger point is extracted to obtain the time sequence segment between the signal response points and obtain the feeding clamping timing offset data set.

[0014] Based on the feeding clamping timing offset data set, the rotation change value of the starting angle segment in the servo angle control sequence is extracted, and the speed peak difference within the continuous time slice segment is compared. The overlap range of the angle drift interval between the clamping start segment and the initial servo rotation segment is identified, and the combination pairs of position points and time points within the intersection segment are screened to establish the first bend start control behavior feature set.

[0015] As a further aspect of the present invention, the specific steps for generating the error trend control parameter group are as follows:

[0016] Based on the first bend start control behavior feature set, extract the peak position points in the fluctuation segment of the feeding control sequence, pair the corresponding delay points in the clamping control segment according to the time axis, extract the positive and negative signs of the displacement direction difference after each pairing, and use fuzzy logic to judge the trend consistency between the feeding direction and the clamping direction, mark the start and end index positions of the matching segment, and generate a feeding clamping direction consistency data group.

[0017] Based on the consistent data set of the feeding clamping direction, the interval distance between continuous peak points in the error fluctuation segment is extracted, the error peak index interval is scanned, the continuous fluctuation range that matches the time segment in the clamping control segment is selected, the degree of overlap of the difference change trend of the two segments is calculated, and the cross structure segment segment is obtained to establish the error fluctuation structure feature group.

[0018] Based on the error fluctuation structure feature group, the step number corresponding to the distance from the start point to the end point of the descent section in the feeding control section is extracted, the start and end index sections of the forward trigger point within the clamping control time period are obtained, the ratio of the duration of the compression section to the total angle change within the servo angle change section is calculated, and the error trend control parameter group is generated.

[0019] As a further aspect of the present invention, the fuzzy logic is based on the formula:

[0020]

[0021] Where: μ′ represents the membership degree value of the consistency between the feeding and clamping directions after fuzzy logic judgment, x represents the displacement difference between the feeding and clamping directions at the same time point, Δt represents the time interval between the peak point of the feeding fluctuation and the clamping delay response point, ω represents the weighting coefficient of the difference in the rate of change between the feeding and clamping directions, θ represents the time offset between the feeding trigger point and the clamping response point, k represents the adjustment factor used to control the degree of change in the slope of the membership function curve, c represents the central reference point for judging the consistency of trends, and e represents the base constant of the natural logarithm.

[0022] As a further aspect of the present invention, the specific steps for generating the linkage control instruction set are as follows:

[0023] Based on the error trend control parameter group, extract the coordinate values ​​of the start and end nodes of the feeding path, calculate the distance difference between nodes, establish a complete trajectory coordinate set by inserting fixed interval points, extract the peak value of the response delay time in the clamping trigger segment and reverse locate the start point in advance through the reference time point, filter the valley segment of continuous deceleration in the servo speed change curve and calculate the time range after compression, and establish a path trigger compression control dataset.

[0024] Based on the path-triggered compression control dataset, the interpolated point numbers of each segment in the feeding trajectory are extracted and summarized into a feeding index set in sequence. The early start time index number set in the clamping trigger section is obtained, and the continuous node number set in the servo compression section is obtained. The timing indexes of the three sets are aligned and an instruction package group is formed to generate a linkage control instruction set.

[0025] As a further aspect of the present invention, the specific steps for generating the first curve execution path adjustment model set are as follows:

[0026] Based on the aforementioned linkage control instruction set, the time values ​​of the starting and ending points in the feeding control segment are extracted, and the time value of the trigger point in the clamping control segment is subtracted. The control point numbers in the corresponding time difference segment are compared and the error offset point is located. The long short-term memory network is used to learn the trend of the time difference sequence, identify the overlapping time period structure in the repetitive pattern, filter out the index point group of the overlapping segment in the time series, and obtain the feeding clamping time difference index group.

[0027] Based on the feeding clamping time difference index group, the angle change value of adjacent nodes in the servo corner segment is extracted and the angle change rate is calculated. Continuous node segments with angle change rates exceeding the average change value are selected and the time index is recorded. After comparing the intersection of the servo segment change interval and the feeding clamping index segment, the overlapping segment is extracted, and the angle path deviation segment data group is generated.

[0028] Based on the angle path deviation segment data group, the start and end indices of overlapping segments in the feeding control segment are extracted and sorted by time. The task instruction index positions of the clamping trigger segment and the instruction index of the servo compression segment are rearranged. The three instruction sets are rearranged as a whole in the control chain and a combined structure is generated to generate the first curve execution path adjustment model set.

[0029] As a further aspect of the present invention, the long short-term memory network is configured according to the formula:

[0030]

[0031] in: q represents the overlap trend value of the predicted output at time p. p δ represents the time difference input value between the start and end points within the p-th feeding control cycle. p η represents the trigger delay time of the p-th clamping response signal relative to the corresponding feeding signal. p ψ represents the error drift magnitude generated by the p-th control point in the previous control cycle. p denoted as the instruction trigger frequency coefficient per unit time in the p-th feeding control segment, V represents the fully connected weight matrix between the network input layer and the prediction output layer, u represents the bias vector used to shift the output range of the activation function, and σ represents the Sigmoid activation function, used to compress the prediction result into a probability output value between 0 and 1.

[0032] As a further aspect of the present invention, the first-bend execution path adjustment model set refers to a set of reconstructed control models for the execution of the first-bend action, formed by sorting, combining, and structurally mapping the key action commands in the control chain based on the time series overlap relationship, error change characteristics, and angle compression characteristics of the three control segments: feeding, clamping, and servo. Specifically, it includes the sorted and adjusted start and end index pairs in the feeding control segment, the repositioned trigger position number in the clamping task command, the continuous execution node index of the compression section in the servo control segment, and the time-series coordination structure of the three within a unified control cycle. The model set is constructed based on the principles of error trend optimization, control response matching, and angle compression adaptation, forming a set of dynamic control path combination templates that can be called by the controller.

[0033] As a further aspect of the present invention, the step of generating the first curve control execution signal group is as follows:

[0034] Based on the first curve execution path adjustment model set, extract the starting coordinates of the feeding control section and compare them with the process reference point and determine the displacement difference, set the starting index of the feeding module, extract the marked position in the clamping task, align it to the feeding time reference point, synchronize the sequence number corresponding to the first angle change point of the servo section and set it as the starting index, and establish the first curve control starting matching set.

[0035] Based on the initial matching set of the first bend control, the feeding start point index is extracted and input into the first address of the sequence register, and the clamping alignment identifier is written into the second address area. The servo initial number is input into the third address segment to complete the continuous registration of the three control values ​​and generate the first bend control execution signal group.

[0036] A first-bend position deviation correction system for a servo pipe bending machine, wherein the first-bend position deviation correction system for a servo pipe bending machine is used to execute the aforementioned first-bend position deviation correction method for a servo pipe bending machine, the system comprising:

[0037] Error feature extraction module: Based on the displacement error of the feeding mechanism, the lag start behavior of the clamping mechanism, the response deviation of the servo rotation unit and the initial drift of the first forming angle, extract the three-axis coordinate difference between the feeding nodes, the identification of the clamping trigger signal delay time slice and the servo execution angle segment response lag interval, and generate the first bend start control behavior feature set.

[0038] Trend rule modeling module: Based on the first bend start control behavior feature set, extract the peak amplitude of the feeding fluctuation peak segment and pair it with the clamping response delay segment. Combine the direction judgment result and use fuzzy logic to make consistency judgment on the trend interval. Statistically calculate the error peak interval and the compression segment change trend to establish an error trend control parameter group.

[0039] Path combination calculation module: Based on the error trend control parameter group, it constructs continuous coordinate points in the feeding path interpolation segment, extracts the index of the clamping advance time point, evaluates the amplitude of the compression angle segment within the servo response segment, and combines and matches the feeding, clamping and servo index numbers to generate a linkage control instruction set.

[0040] Structural rearrangement identification module: Based on the linkage control instruction set, a long short-term memory network is used to identify repeating structural segments in the feeding clamping time difference sequence, extract stable trend time windows and establish a mapping relationship with the corresponding index of control instruction nodes, jointly judge the matching relationship between the feeding angle change rate and the compression amplitude, reconstruct the control instruction sequence, and generate the first curve execution path adjustment model set;

[0041] Signal sequence generation module: Based on the first curve execution path adjustment model set, register the starting instruction number of the feeding execution segment, mark the return position of the clamping task identifier, synchronously integrate the control nodes of the servo initial angle segment, and write the three types of execution labels into the controller sequence register in address order to generate the first curve control execution signal group.

[0042] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0043] 1. In this invention, by performing fuzzy logic judgment on the consistency of the offset direction, the coordinated offset behavior between feeding and clamping is transformed into a trend quantity. Combined with the local fluctuation interval of the error curve, the parameter items affecting the change of the first bend error are accurately extracted, and the linkage combination of three types of control segment commands is realized.

[0044] 2. In this invention, the trend of time difference change is learned and extracted by long short-term memory network, and control time periods with overlapping structural characteristics are identified. In the rearrangement of instruction segment structure, the stable interval is prioritized and the cross-segment buffer of error high-incidence area is realized, so that the first curve control output is smoother and the control rhythm is more robust.

[0045] 3. In this invention, by introducing multiple structural variables such as error trend, trigger delay, and control index compression ratio into the control and regulation system through the selection of participating items, the responsiveness of key decisions in the deviation correction path and the stability of the instruction chain under error interference are significantly improved, and the risk of error propagation and amplification in the forming path is effectively controlled. Attached Figure Description

[0046] Figure 1 This is a schematic diagram of the workflow of the present invention;

[0047] Figure 2 This is a system flowchart of the present invention. Detailed Implementation

[0048] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0049] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0050] Example

[0051] Please see Figure 1 This invention provides a technical solution: a method for correcting the first bend position deviation of a servo pipe bending machine, comprising the following steps:

[0052] Step 1: Based on the displacement error of the feeding mechanism, the lag start behavior of the clamping mechanism, the response deviation of the servo corner unit, and the initial drift of the first forming angle, extract the feeding point difference, clamping delay sequence, and identify the servo delay section to generate the first bend start control behavior feature set.

[0053] Step 2: Based on the feature set of the initial control behavior of the first bend, compare the feeding fluctuation point with the clamping delay value, use fuzzy logic to judge the consistency of the offset direction, identify the fluctuation interval of the error curve, select the feeding step distance, clamping advance amount and angle compression ratio, and generate a group of error trend control parameters.

[0054] Step 3: Based on the error trend control parameter group, obtain the feeding path interpolation and clamping trigger time parameters, adjust the servo node compression segment, match and combine the three types of index parameters, and generate a linkage control instruction set;

[0055] Step 4: Based on the linkage control instruction set, a long short-term memory network is used to identify the time difference trend structure, calculate the feeding path and clamping time difference, determine the angle curvature difference, rearrange the order of instruction segments, adjust the combination of feeding instructions, clamping instructions and cornering instructions, and generate the first curve execution path adjustment model set.

[0056] Step 5: Based on the first curve execution path adjustment model set, set the feeding start point, align the clamping task identifier, synchronize the servo initial sequence, write the feeding, clamping and servo execution segment labels to the sequence register, and generate the first curve control execution signal group.

[0057] The specific steps for generating the feature set of the initial control behavior of the first curve are as follows:

[0058] Based on the displacement error of the feeding mechanism, the lag start behavior of the clamping mechanism, the response deviation of the servo rotation unit, and the initial drift of the first forming angle, the three-axis coordinate set between the feeding start point and the actual path offset point is extracted. The difference sequence between the coordinate set and the standard trajectory is calculated in batches, and the clamping control command trigger point is extracted to obtain the time sequence segment between the signal response points and obtain the feeding clamping timing offset data set.

[0059] Based on the feeding clamping timing offset data set, the rotation change value of the starting angle segment in the servo angle control sequence is extracted, and the speed peak difference within the continuous time slice segment is compared. The overlap range of the angle drift interval between the clamping start segment and the initial servo rotation segment is identified, and the combination pairs of position points and time points within the cross segment are screened to establish the first bend start control behavior feature set.

[0060] Based on the displacement error of the feeding mechanism, the lag start behavior of the clamping mechanism, the response deviation of the servo rotation unit, and the initial drift of the first forming angle, the three-axis coordinate group between the feeding start point and the actual path offset point is compared point by point with the same index point in the reference path coordinate set. The point-to-point difference in the X, Y, and Z directions is calculated respectively. The coordinate difference threshold is set to 0.2 mm. The registration rule is called to output the error data of each coordinate point according to the linear sequence number and complete the offset record. The clamping control command trigger point is calibrated using the signal edge extraction algorithm. The trigger monitoring window width is set to 5 milliseconds. The first abrupt change point of the rising edge signal located at the center of the monitoring window is extracted as the response start point. The time difference interval set between the trigger and the response is established by the continuous recording method of the time axis, and the feeding clamping timing offset data group is generated.

[0061] Based on the feeding clamping timing offset data set, the start and end angle values ​​of the continuous execution segment in the servo angle control sequence are extracted. Angle difference statistics are performed by dividing the segment into 0.5-degree segments. The width of the continuous time slice window is set to 3 control cycles. The angle change in each window is scanned and the position number of the segment with the fastest speed increase is recorded. At the same time, the overlapping segment filtering method is applied. Based on the clamping response start point, the time points of each servo angle change record are extracted from the window segments extending 15 milliseconds before and after. Then, the angle start and end numbers of the previous continuous compression angle segment are extracted from the servo control command sequence. Two time intervals are established according to the timestamp. The overlapping segment judgment logic is called to compare the intervals and filter out the position point number pairs corresponding to the overlapping time slices to establish the first curve start control behavior feature set.

[0062] The specific steps for generating the error trend control parameter group are as follows:

[0063] Based on the feature set of the initial control behavior of the first bend, the peak position points in the fluctuation segment of the feeding control sequence are extracted, and the corresponding delay points in the clamping control segment are paired according to the time axis. The positive and negative signs of the displacement direction difference after each pairing are extracted, and fuzzy logic is used to determine the trend consistency between the feeding direction and the clamping direction. The start and end index positions of the matching segments are marked, and the feeding and clamping direction consistency data group is generated.

[0064] Based on the consistent data set of feeding clamping direction, the interval distance between continuous peak points in the error fluctuation segment is extracted, the error peak index interval is scanned, the continuous fluctuation range that matches the time segment in the clamping control segment is selected, the degree of overlap of the difference change trend of the two segments is calculated, and the cross structure segment segment is obtained to establish the error fluctuation structure feature group.

[0065] Based on the error fluctuation structure feature group, the step number corresponding to the distance from the start point to the end point of the descent section in the feeding control section is extracted, the start and end index sections of the forward trigger point within the clamping control time period are obtained, the ratio of the duration of the compression section to the total angle change within the servo angle change section is calculated, and the error trend control parameter group is generated.

[0066] Based on the feature set of the initial control behavior of the first bend, the peak position points of each fluctuation segment in the feeding control sequence are extracted. The delay points in the corresponding clamping control segment are paired according to the time axis. The pairing index interval is set to not exceed 5 cycles. The feeding position and clamping trigger displacement records in the data are read. The direction change values ​​of the first axis and the second axis in the direction coordinate sequence are read according to the time alignment method. The positive and negative signs of the difference are judged respectively. The number of results with the same sign and the index position are counted. The fuzzy logic judgment method is adopted, and the fuzzy membership level of positive and negative trends is set to three levels. Low, medium and high level membership weights are assigned to each group of positive and negative combinations. The trend consistency score sequence is accumulated through the membership values. The index segments with continuous scores higher than 0.7 are selected as trend segment boundary points. The start and end index positions of the matching segments are marked to generate feeding clamping direction consistency data groups.

[0067] Based on the consistent data set of feeding clamping direction, the index position interval between continuous peak points in the error fluctuation segment is extracted. The continuous peak segment identification window size is set to 3 peak units. The corresponding time period index in the clamping control sequence is extracted in sequence. Through the index interval center point alignment logic, the sliding window moves the continuous fluctuation segment. The sliding step size is set to 1 index unit. The continuous fluctuation range matching the time segment in the clamping control segment is filtered according to the index overlap ratio. The cross segment index mapping list is constructed through the corresponding positions of the first and last points of the segment. The set of structural connection points of the control sequence in the overlapping area of ​​clamping and feeding is established. The effective cross segments are filtered and the position point number and time slice number combination pair are output in sequence to obtain the cross structure segment segment and establish the error fluctuation structure feature group.

[0068] Based on the error fluctuation structure feature group, the starting point to the ending point of the descending section in the feeding control segment is extracted. After sorting by timestamp in ascending order, the cumulative sum of the distances between adjacent points is extracted. The step number is generated using the cumulative distance value corresponding to the index number range. The index range of the forward movement trigger point within the clamping control time period is obtained. Time alignment correction is completed within 2 index units by setting the start and end point offset compensation value. The continuous descending interval in the servo angle change segment is filtered, and the total change in angle value and the duration of the corresponding time period are recorded. The proportional parameters of the two variables in the compression zone are calculated and included in the segment matching table. Then, a three-item index fusion relationship is established with the feeding step number and the clamping forward movement segment to generate an error trend control parameter group.

[0069] Fuzzy logic is based on the formula:

[0070]

[0071] Where: μ′ represents the membership degree value of the consistency between the feeding and clamping directions after fuzzy logic judgment, x represents the displacement difference between the feeding and clamping directions at the same time point, Δt represents the time interval between the peak point of the feeding fluctuation and the clamping delay response point, ω represents the weighting coefficient of the difference in the rate of change between the feeding and clamping directions, θ represents the time offset between the feeding trigger point and the clamping response point, k represents the adjustment factor used to control the degree of change of the slope of the membership function curve, c represents the central reference point for judging the consistency of trends, and e represents the base constant of the natural logarithm;

[0072] Execution process: First, extract the instantaneous displacement change of the peak value of each fluctuation segment in the feeding control sequence. Combined with the delayed response point of the clamping mechanism on the corresponding time axis, calculate the displacement direction difference at the pairing point as the input variable x. Obtain the time interval between the peak point of each feeding group and the clamping delay point as the parameter Δt. Simultaneously extract the difference in the rate of change of the feeding and clamping directions in the nearest segments and calculate the normalized weighted result as the weight coefficient ω. Extract the phase offset length between the feeding start time and the clamping response trigger time as the input parameter θ. Set the curve slope control factor k and the center reference point c. Substitute the above parameters into the improved formula μ′ to calculate the trend consistency membership value of each pairing within the control cycle, form the time series trend judgment curve, and mark the start and end of continuous segments with membership values ​​greater than the threshold to generate feeding and clamping direction consistency data groups.

[0073] The specific steps for generating a linkage control instruction set are as follows:

[0074] Based on the error trend control parameter group, the coordinate values ​​of the start and end nodes of the feeding path are extracted, the distance difference between nodes is calculated, a complete trajectory coordinate set is established by inserting fixed interval points, the peak value of the response delay time in the clamping trigger segment is extracted, and the start point is located in advance by reverse positioning through the reference time point. The valley segment of continuous deceleration in the servo speed change curve is screened and the time range after compression is calculated to establish a path trigger compression control dataset.

[0075] Based on the path-triggered compression control dataset, the interpolated point numbers of each segment in the feeding trajectory are extracted and summarized into a feeding index set in order. The early start time index number set in the clamping trigger section is obtained, and the continuous node number set in the servo compression section is obtained. The timing indexes of the three sets are aligned and formed into an instruction package group to generate a linkage control instruction set.

[0076] Based on the error trend control parameter group, the X, Y, and Z coordinate values ​​of the start and end nodes of the feeding path are extracted. The distance is executed for each node in the segment, and the intermediate points are inserted at intervals of 0.5 mm. The numbers are recorded in the insertion order and a coordinate index sequence is generated. The complete trajectory coordinate set is summarized. The delay time sequence is scanned in the clamping trigger segment. 1.5 times the standard deviation is used as the peak value judgment boundary. After locating the maximum value index, five time steps are traced back. The corresponding control point number is recorded as the early start point. The minimum length of the continuous falling section in the servo speed change sequence is set to five time slices. The minimum value point is marked and the left and right boundaries are expanded by traversing the curve. The duration of the compression section is generated according to the start and end time points. The path trigger compression control dataset is established.

[0077] Based on the path-triggered compression control dataset, a cumulative mapping is established according to the interpolated feeding trajectory in numerical order. A number point is set with a step size of 0.5 mm, the number is incremented and the sequence of each segment is marked. The output is organized into a feeding index set. The early response time point index is extracted from the clamping trigger segment, arranged in ascending order and limited to a maximum interval of two cycles to form an early start time index set. The angle sequence is obtained according to the servo compression segment. It is determined whether the difference between any two consecutive nodes is less than 0.8 degrees. The consecutive nodes that meet the condition are marked as a segment and the start and end numbers are recorded to construct a continuous node number set. The segments in the three sets are matched according to the time index. The number difference is determined to be no more than ±2 as the alignment condition. All index rows are merged in alignment order and the structural instructions are packaged to generate a linkage control instruction set.

[0078] The specific steps for generating the first curve execution path adjustment model set are as follows:

[0079] Based on the linkage control instruction set, the time values ​​of the starting point and the ending point in the feeding control section are extracted, and the time value of the trigger point in the clamping control section is subtracted. The control point numbers in the corresponding time difference section are compared and the error offset point is located. The long short-term memory network is used to learn the trend of the time difference sequence, identify the overlapping time period structure in the repetitive pattern, filter out the index point group of the overlapping section in the time series, and obtain the feeding clamping time difference index group.

[0080] Based on the feeding clamping time difference index group, the angle change value of adjacent nodes in the servo corner segment is extracted and the angle change rate is calculated. Continuous node segments with angle change rates exceeding the average change value are filtered and the time index is recorded. After comparing the intersection of the servo segment change interval and the feeding clamping index segment, the overlapping segment is extracted and the angle path deviation segment data group is generated.

[0081] Based on the angle path deviation segment data group, the start and end indices of overlapping segments in the feeding control segment are extracted and sorted by time. The task instruction index positions of the clamping trigger segment and the instruction index of the servo compression segment are rearranged. The three instruction sets are rearranged as a whole in the control chain and a combined structure is generated to generate the first curve execution path adjustment model set.

[0082] Based on the linkage control instruction set, the time values ​​of the starting and ending points in the feeding control section are extracted. The time format is set to integer millisecond precision timestamp. The system time record corresponding to each instruction point in the number interval is read and the time value corresponding to the trigger point in the clamping control section is subtracted to generate a set of time difference sequences. Each time difference data item is labeled with the control point and numbered and an index mapping table is established. The number points with an absolute difference value exceeding the preset offset threshold of 1.2 milliseconds are found and marked as abnormal index points. A long short-term memory network is used to construct a training data window with a length of 6 steps and a sliding step size of 1. The continuous segments of the current time difference sequence are read as input tensors. The input tensor dimension is set to 6×1 and the number of hidden layer nodes is set to 16. The state return structure is set to unidirectional and non-overlapping. The weight matrix is ​​initialized using a normal distribution. The sequential time step prediction task is executed and a probability value sequence is generated. The overlap probability threshold is set to 0.65. The corresponding index point segments are filtered as repeating structure segments. The numbers in the middle interval are extracted as index point groups to obtain the feeding clamping time difference index group.

[0083] Based on the feeding clamping time difference index group, the angle change values ​​of adjacent nodes are extracted from the servo angle control segment. The sampling interval is set to 2 milliseconds. The angle control value sequence is read in the order of index number. The difference is obtained point by point to generate the angle change amount sequence. The average baseline is established in the order of sequence number. The floating baseline is set to the average value of a 3-point sliding window. Points with angle change values ​​greater than the average baseline value are selected. Segments that exceed the average baseline value for a continuous number of times are counted. The start and end index numbers of the segments are recorded to obtain continuous angle change segments. The change interval of the servo segment is compared with the start and end number intervals in the feeding clamping time difference index group. A time range comparison matrix is ​​established for the two segments. If the number intersection is not empty, it is identified as an overlapping segment. The numbers of the intersection segments are extracted to form an index set and generate the angle path deviation segment data group.

[0084] Based on the angle path deviation segment data group, the start and end indices of the corresponding overlapping segments in the feeding control segment are extracted. The index time tags are sorted in ascending order according to the system time record, and the sorted index pairs are output. The task instruction list is obtained from the clamping trigger segment. The response execution number of each task is read according to the task number. A mapping index group between task instructions and control numbers is established. The instruction number position of the task belonging to the time overlapping segment is relocated and the calling order is rewritten. The instruction sequence of continuous angle segments is obtained from the servo compression segment. The instruction index contained in the intersection segment number is extracted. The instruction number table is rearranged according to the time sequence of the overlapping segment. The indexes of the feeding segment, clamping segment and servo segment are integrated and rearranged into three segments. The numbers are marked and the numbering standard is unified. The data is packaged into a unified data structure and output to generate the first curve execution path adjustment model set.

[0085] Long Short-Term Memory (LSTM) networks are based on the formula:

[0086]

[0087] in: q represents the overlap trend value of the predicted output at time p. p δ represents the time difference input value between the start and end points within the p-th feeding control cycle. p η represents the trigger delay time of the p-th clamping response signal relative to the corresponding feeding signal. p ψ represents the error drift magnitude generated by the p-th control point in the previous control cycle. p denoted as the instruction trigger frequency coefficient per unit time in the p-th feeding control segment, V represents the fully connected weight matrix between the network input layer and the prediction output layer, u represents the bias vector used to shift the output range of the activation function, and σ represents the Sigmoid activation function, used to compress the prediction result into a probability output value between 0 and 1.

[0088] Execution process: First, the time interval between the start and end points of the feeding control cycle is extracted as the sequence input parameter q.p Simultaneously, the response delay value of the clamping trigger signal relative to the feeding control point is extracted as δ. p And obtain the error drift amplitude of the corresponding point in the previous control cycle as η. p Then, the command trigger frequency per unit time in the feeding control section is converted into a normalized frequency coefficient ψ. p The four inputs are combined to form a vector, which is then fed into the prediction model. This vector, along with the weight matrix V and the bias vector u, is processed by the activation function σ to output the predicted trend probability value for the current time step. Determine whether the position belongs to the stable overlapping section of feeding and clamping.

[0089] The first-bend execution path adjustment model set refers to a set of reconstructed control models for the execution of the first-bend action, formed by sorting, combining, and structurally mapping the key action commands in the control chain based on the time series overlap relationship, error change characteristics, and angle compression characteristics of the three control segments: feeding, clamping, and servo. Specifically, it includes the sorted and adjusted start and end index pairs in the feeding control segment, the repositioned trigger position numbers in the clamping task commands, the continuous execution node indexes of the compression section in the servo control segment, and the time-series coordination structure of the three within a unified control cycle. The model set is constructed based on the principles of error trend optimization, control response matching, and angle compression adaptation, forming a set of dynamic control path combination templates that can be called by the controller.

[0090] The steps for generating the first curve control execution signal group are as follows:

[0091] Based on the first curve execution path adjustment model set, extract the starting coordinates of the feeding control section and compare them with the process reference point and judge the displacement difference. Set the starting index of the feeding module, extract the marked position in the clamping task, align it to the feeding time reference point, synchronize the sequence number corresponding to the first angle change point of the servo section and set it as the starting index, and establish the first curve control starting matching set.

[0092] Based on the initial matching set of the first bend control, the feeding start point index is extracted and input into the first address of the sequence register, and the clamping alignment identifier is written into the second address area. The servo initial number is input into the third address segment to complete the continuous registration of the three control values ​​and generate the first bend control execution signal group.

[0093] Based on the first curve execution path adjustment model set, the starting coordinates of the feeding control segment are read, and the reading dimensions are set as X, Y, and Z axes. The position reading function is called and the coordinates of the first point are recorded according to the number. The process reference point is set as the center point of the preset mold. The three-axis values ​​of the center point are read, and the displacement difference vector is generated by subtracting them according to the axes. The position threshold setting function is used to set the maximum difference to ±1 mm. It is determined whether it is within the offset tolerance. If it is, the current feeding start point index is confirmed as a valid index point. The identification position in the clamping task is extracted, the task number is compared, and the control execution time point is read and converted into a timestamp. Then, the feeding time reference point is aligned, and the difference between the two is calculated to not exceed the preset range of 2 milliseconds. The task number is marked and confirmed as the clamping reference identifier. The control node number of the first angle change in the servo segment is obtained. The index is extracted according to the number sequence and set as the starting command index value. A three-field matching record table is established with the previous two items to generate the first curve control starting matching set.

[0094] Based on the initial matching set of the first curve control, the index number of the feeding start point is extracted as the address content input to the sequence register. The first address position is set to 0000H, and the number is written into the first address content area of ​​the register in decimal format. The number of the clamping task identifier is extracted, converted into 16-bit address content, and written into the second address area at offset address 0001H in the sequence register. The servo initial number is extracted, and the matching index in the angle control sequence is converted into the register input value. The address encoding conversion module is used to write it into the address segment 0002H. The continuous write operation command MOVW3 segment is executed to fill the contents of the three address areas in sequence and register the data verification flag. The mapping of the register is completed, and the contents of the control register segment are exported as binary control signal format data to generate the first curve control execution signal group.

[0095] Please see Figure 2 A system for correcting the first bend position deviation of a servo pipe bending machine, comprising:

[0096] Error feature extraction module: Based on the displacement error of the feeding mechanism, the lag start behavior of the clamping mechanism, the response deviation of the servo rotation unit and the initial drift of the first forming angle, extract the three-axis coordinate difference between the feeding nodes, the identification of the clamping trigger signal delay time slice and the servo execution angle segment response lag interval, and generate the first bend start control behavior feature set.

[0097] Trend rule modeling module: Based on the feature set of the initial control behavior of the first bend, extract the peak amplitude of the peak section of the feeding fluctuation and pair it with the clamping response delay section. Combine the direction judgment result and use fuzzy logic to make consistency judgment on the trend interval. Statistically calculate the error peak interval and the change trend of the compression section, and establish error trend control parameter group.

[0098] Path combination calculation module: Based on the error trend control parameter group, it constructs continuous coordinate points in the feeding path interpolation segment, extracts the index of the clamping advance time point, evaluates the amplitude of the compression angle segment within the servo response segment, and combines and matches the feeding, clamping and servo index numbers to generate a set of linkage control instructions.

[0099] Structural rearrangement identification module: Based on the linkage control instruction set, it uses a long short-term memory network to identify repetitive structural segments in the feeding clamping time difference sequence, extracts stable trend time windows and establishes a mapping relationship with the corresponding index of control instruction nodes, jointly judges the matching relationship between the feeding angle change rate and the compression amplitude, reconstructs the control instruction sequence, and generates the first curve execution path adjustment model set.

[0100] Signal sequence generation module: Based on the first curve execution path adjustment model set, register the starting instruction number of the feeding execution segment, mark the return position of the clamping task identifier, synchronously integrate the control node of the servo initial angle segment, and write the three types of execution labels into the controller sequence register in address order to generate the first curve control execution signal group.

[0101] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A method for correcting the initial bend position deviation of a servo pipe bending machine, characterized in that, Includes the following steps: Step 1: Based on the displacement error of the feeding mechanism, the lag start behavior of the clamping mechanism, the response deviation of the servo corner unit, and the initial drift of the first forming angle, extract the feeding point difference, clamping delay sequence, and identify the servo delay section to generate the first bend start control behavior feature set. Step 2: Based on the first bend start control behavior feature set, compare the feeding fluctuation point with the clamping delay value, use fuzzy logic to determine the consistency of the offset direction, identify the error curve fluctuation interval, select the feeding step distance, clamping advance amount and angle compression ratio, and generate an error trend control parameter group. Step 3: Based on the error trend control parameter group, obtain the feeding path interpolation and clamping trigger time parameters, adjust the servo node compression segment, match and combine the three types of index parameters, and generate a linkage control instruction set; Step 4: Based on the aforementioned linkage control instruction set, a long short-term memory network is used to identify the time difference trend structure, calculate the feeding path and clamping time difference, determine the angle curvature difference, rearrange the order of instruction segments, adjust the combination of feeding instructions, clamping instructions and cornering instructions, and generate the first curve execution path adjustment model set. Step 5: Based on the first bend execution path adjustment model set, set the feeding start point, align the clamping task identifier, synchronize the servo initial sequence, write the feeding, clamping and servo execution segment labels to the sequence register, and generate the first bend control execution signal group.

2. The method for correcting the first bend position deviation of a servo pipe bending machine according to claim 1, characterized in that, The specific steps for generating the first curve initiation control behavior feature set are as follows: Based on the displacement error of the feeding mechanism, the lag start behavior of the clamping mechanism, the response deviation of the servo rotation unit, and the initial drift of the first forming angle, the three-axis coordinate set between the feeding start point and the actual path offset point is extracted. The difference sequence between the coordinate set and the standard trajectory is calculated in batches, and the clamping control command trigger point is extracted to obtain the time sequence segment between the signal response points and obtain the feeding clamping timing offset data set. Based on the feeding clamping timing offset data set, the rotation change value of the starting angle segment in the servo angle control sequence is extracted, and the speed peak difference within the continuous time slice segment is compared. The overlap range of the angle drift interval between the clamping start segment and the initial servo rotation segment is identified, and the combination pairs of position points and time points within the intersection segment are screened to establish the first bend start control behavior feature set.

3. The method for correcting the first bend position deviation of a servo pipe bending machine according to claim 1, characterized in that, The specific steps for generating the error trend control parameter group are as follows: Based on the first bend start control behavior feature set, extract the peak position points in the fluctuation segment of the feeding control sequence, pair the corresponding delay points in the clamping control segment according to the time axis, extract the positive and negative signs of the displacement direction difference after each pairing, and use fuzzy logic to judge the trend consistency between the feeding direction and the clamping direction, mark the start and end index positions of the matching segment, and generate a feeding clamping direction consistency data group. Based on the consistent data set of the feeding clamping direction, the interval distance between continuous peak points in the error fluctuation segment is extracted, the error peak index interval is scanned, the continuous fluctuation range that matches the time segment in the clamping control segment is selected, the degree of overlap of the difference change trend of the two segments is calculated, and the cross structure segment segment is obtained to establish the error fluctuation structure feature group. Based on the error fluctuation structure feature group, the step number corresponding to the distance from the start point to the end point of the descent section in the feeding control section is extracted, the start and end index sections of the forward trigger point within the clamping control time period are obtained, the ratio of the duration of the compression section to the total angle change within the servo angle change section is calculated, and the error trend control parameter group is generated.

4. The method for correcting the first bend position deviation of a servo pipe bending machine according to claim 3, characterized in that, The fuzzy logic is based on the formula: Where: μ′ represents the membership degree value of the consistency between the feeding and clamping directions after fuzzy logic judgment, x represents the displacement difference between the feeding and clamping directions at the same time point, Δt represents the time interval between the peak point of the feeding fluctuation and the clamping delay response point, ω represents the weighting coefficient of the difference in the rate of change between the feeding and clamping directions, θ represents the time offset between the feeding trigger point and the clamping response point, k represents the adjustment factor used to control the degree of change in the slope of the membership function curve, c represents the central reference point for judging the consistency of trends, and e represents the base constant of the natural logarithm.

5. The method for correcting the first bend position deviation of a servo pipe bending machine according to claim 1, characterized in that, The specific steps for generating the linkage control instruction set are as follows: Based on the error trend control parameter group, extract the coordinate values ​​of the start and end nodes of the feeding path, calculate the distance difference between nodes, establish a complete trajectory coordinate set by inserting fixed interval points, extract the peak value of the response delay time in the clamping trigger segment and reverse locate the start point in advance through the reference time point, filter the valley segment of continuous deceleration in the servo speed change curve and calculate the time range after compression, and establish a path trigger compression control dataset. Based on the path-triggered compression control dataset, the interpolated point numbers of each segment in the feeding trajectory are extracted and summarized into a feeding index set in sequence. The early start time index number set in the clamping trigger section is obtained, and the continuous node number set in the servo compression section is obtained. The timing indexes of the three sets are aligned and an instruction package group is formed to generate a linkage control instruction set.

6. The method for correcting the first bend position deviation of a servo pipe bending machine according to claim 1, characterized in that, The specific steps for generating the first curve execution path adjustment model set are as follows: Based on the aforementioned linkage control instruction set, the time values ​​of the starting and ending points in the feeding control segment are extracted, and the time value of the trigger point in the clamping control segment is subtracted. The control point numbers in the corresponding time difference segment are compared and the error offset point is located. The long short-term memory network is used to learn the trend of the time difference sequence, identify the overlapping time period structure in the repetitive pattern, filter out the index point group of the overlapping segment in the time series, and obtain the feeding clamping time difference index group. Based on the feeding clamping time difference index group, the angle change value of adjacent nodes in the servo corner segment is extracted and the angle change rate is calculated. Continuous node segments with angle change rates exceeding the average change value are selected and the time index is recorded. After comparing the intersection of the servo segment change interval and the feeding clamping index segment, the overlapping segment is extracted, and the angle path deviation segment data group is generated. Based on the angle path deviation segment data group, the start and end indices of overlapping segments in the feeding control segment are extracted and sorted by time. The task instruction index positions of the clamping trigger segment and the instruction index of the servo compression segment are rearranged. The three instruction sets are rearranged as a whole in the control chain and a combined structure is generated to generate the first curve execution path adjustment model set.

7. The method for correcting the first bend position deviation of a servo pipe bending machine according to claim 1, characterized in that, The Long Short-Term Memory (LSTM) network is configured according to the formula: in: q represents the overlap trend value of the predicted output at time p. p δ represents the time difference input value between the start and end points within the p-th feeding control cycle. p η represents the trigger delay time of the p-th clamping response signal relative to the corresponding feeding signal. p ψ represents the error drift magnitude generated by the p-th control point in the previous control cycle. p denoted as the instruction trigger frequency coefficient per unit time in the p-th feeding control segment, V represents the fully connected weight matrix between the network input layer and the prediction output layer, u represents the bias vector used to shift the output range of the activation function, and σ represents the Sigmoid activation function, used to compress the prediction result into a probability output value between 0 and 1.

8. The method for correcting the first bend position deviation of a servo pipe bending machine according to claim 1, characterized in that, The first-bend execution path adjustment model set refers to a set of reconstructed control models for the execution of the first-bend action, formed by sorting, combining, and structurally mapping the key action commands in the control chain based on the time series overlap relationship, error change characteristics, and angle compression characteristics of the three control segments: feeding, clamping, and servo. Specifically, it includes the sorted and adjusted start and end index pairs in the feeding control segment, the repositioned trigger position number in the clamping task command, the continuous execution node index of the compression section in the servo control segment, and the time-series coordination structure of the three within a unified control cycle. The model set is constructed based on the principles of error trend optimization, control response matching, and angle compression adaptation, forming a set of dynamic control path combination templates that can be called by the controller.

9. The method for correcting the first bend position deviation of a servo pipe bending machine according to claim 1, characterized in that, The steps for generating the first curve control execution signal group are as follows: Based on the first curve execution path adjustment model set, extract the starting coordinates of the feeding control section and compare them with the process reference point and determine the displacement difference, set the starting index of the feeding module, extract the marked position in the clamping task, align it to the feeding time reference point, synchronize the sequence number corresponding to the first angle change point of the servo section and set it as the starting index, and establish the first curve control starting matching set. Based on the initial matching set of the first bend control, the feeding start point index is extracted and input into the first address of the sequence register, and the clamping alignment identifier is written into the second address area. The servo initial number is input into the third address segment to complete the continuous registration of the three control values ​​and generate the first bend control execution signal group.

10. A system for correcting the first bend position deviation of a servo pipe bending machine, characterized in that, The method for correcting the first bend position deviation for a servo pipe bending machine according to any one of claims 1-9, wherein the system comprises: Error feature extraction module: Based on the displacement error of the feeding mechanism, the lag start behavior of the clamping mechanism, the response deviation of the servo rotation unit and the initial drift of the first forming angle, extract the three-axis coordinate difference between the feeding nodes, the identification of the clamping trigger signal delay time slice and the servo execution angle segment response lag interval, and generate the first bend start control behavior feature set. Trend rule modeling module: Based on the first bend start control behavior feature set, extract the peak amplitude of the feeding fluctuation peak segment and pair it with the clamping response delay segment. Combine the direction judgment result and use fuzzy logic to make consistency judgment on the trend interval. Statistically calculate the error peak interval and the compression segment change trend to establish an error trend control parameter group. Path combination calculation module: Based on the error trend control parameter group, it constructs continuous coordinate points in the feeding path interpolation segment, extracts the index of the clamping advance time point, evaluates the amplitude of the compression angle segment within the servo response segment, and combines and matches the feeding, clamping and servo index numbers to generate a linkage control instruction set. Structural rearrangement identification module: Based on the linkage control instruction set, a long short-term memory network is used to identify repeating structural segments in the feeding clamping time difference sequence, extract stable trend time windows and establish a mapping relationship with the corresponding index of control instruction nodes, jointly judge the matching relationship between the feeding angle change rate and the compression amplitude, reconstruct the control instruction sequence, and generate the first curve execution path adjustment model set; Signal sequence generation module: Based on the first curve execution path adjustment model set, register the starting instruction number of the feeding execution segment, mark the return position of the clamping task identifier, synchronously integrate the control nodes of the servo initial angle segment, and write the three types of execution labels into the controller sequence register in address order to generate the first curve control execution signal group.

Images