Intelligent control method for force and speed increasing of large-tonnage high-speed precision punch

Abstract

The application discloses a large-tonnage high-speed precise punch press force and speed increasing intelligent control method, and particularly relates to the technical field of metal plate stamping equipment control, which comprises a pre-pressing part arranged in a waste area of a high-quality aluminum alloy plate corresponding to a die, wherein the height difference between the pre-pressing part and the lower end of a main forming part is a pre-pressing stroke; when the pre-pressing part is pressed into the waste area and the main forming part does not contact the product area, an edge computing node collects the displacement of a sliding block and the hydraulic pressure, and the load displacement sequence is formed by multiplying the effective area of the force increasing cylinder by the hydraulic pressure; the load displacement sequence is interpolated according to a uniform displacement step, and a multiscale sequence is extracted according to the integral power of two, the ratio of the stamping force difference and the displacement difference of adjacent points is calculated to form a slope sequence, the square residual sum in a section is taken as a section cost, and the square median of the adjacent slope difference is taken as a section penalty; the load response of the same high-quality aluminum alloy plate waste area before the pre-pressing part is pressed in the product area is obtained, and the speed increasing exit position and the force increasing pressure supply path are determined by the edge computing node.

Description

Technical Field

[0001] This invention relates to the field of control technology for metal sheet stamping equipment, and more specifically, to an intelligent control method for increasing the force and speed of a large-tonnage, high-speed, precision stamping press. Background Technology

[0002] When processing high-quality aluminum alloy sheets, existing high-speed precision punching machines with large tonnage often pre-set the force and speed increase curve of the hydraulic servo valve based on the material grade, nominal plate thickness and previous punching data, and then correct the control parameters by combining the pressure value and slide displacement after the plate is pressed. Taking continuous stamping of automotive body panels as an example, the same roll of material will still have differences in thickness, hardening degree and lubricant film distribution along the length direction. The single stamping cycle is relatively short, and the force-increasing and speed-increasing mechanism must complete the switching before the product enters the main plastic deformation area. During the production process, it is not possible to stop the machine to cut off the material of the current stamping for inspection, nor can a load be applied in advance in the product's effective area. Therefore, the existing control can only obtain material response after the effective area of ​​the product has been subjected to the impact pressure. The parameter correction is later than the switching point of the force and speed increase. In actual production, the peak position of the first impact pressure will drift, the pressure increment corresponding to the same slider displacement will be inconsistent, and further cause local indentation, thinning fluctuation and springback size dispersion. The technical problem to be solved by this application is: how to obtain the actual deformation response of the same high-quality aluminum alloy sheet in the current stroke before the effective area of ​​the product is subjected to pressure, and to determine the force increase and speed control parameters accordingly. Summary of the Invention

[0003] To overcome the aforementioned deficiencies of the prior art, embodiments of the present invention provide an intelligent control method for increasing force and speed in high-tonnage high-speed precision punching presses. This method obtains the load response of the same high-quality aluminum alloy sheet scrap area before the pre-pressing section is pressurized in the product area, and the edge computing node determines the speed-up exit position and the force-increasing pressure supply path accordingly, thereby solving the problems mentioned in the background art.

[0004] To achieve the above objectives, the present invention provides the following technical solution: an intelligent control method for increasing the force and speed of a large-tonnage, high-speed, precision punching press, comprising: S1. A pre-pressing section is set in the scrap area of ​​high-quality aluminum alloy sheet corresponding to the mold. The height difference between the pre-pressing section and the lower end of the main forming section is the pre-pressing stroke. When the pre-pressing section is pressed into the scrap area and the main forming section does not contact the product area, the edge calculation node collects the slider displacement and hydraulic pressure. The hydraulic pressure is multiplied by the effective area of ​​the booster cylinder to form a load displacement sequence. S2. Interpolate the load displacement sequence according to the uniform displacement step size and extract the multi-scale sequence according to the integer power of two. Calculate the ratio of the impact force difference to the displacement difference between adjacent points to form a slope sequence. Use the sum of squared residuals within the segment as the segment cost and the median of the squared difference between adjacent slopes as the segment penalty. S3. The multi-scale PELT change point detection algorithm is used to recursively calculate the cumulative cost. The first term of the cumulative cost is taken as the predecessor point and the change point is read in reverse. The intervals with the highest residual decrease are retained from coarse to fine. The midpoint of the last interval is taken to form the segment boundary. S4. Fit the load slope according to the segment boundary and convert the main forming load range according to the pressure area ratio. Use the branch limit interval dynamic programming algorithm to generate the pressure supply value and switching position, and connect the interval where the pressure supply increment does not exceed the pressure rise of the force-increasing cavity. S5. Cut off the branches of the completed path with the time lower bound of the remaining displacement and the work lower bound of the remaining supply pressure to obtain the force-increasing and speed-up control path; when the slider reaches the switching position, cut off the oil supply to the speed-up chamber and supply oil to the force-increasing chamber according to the supply pressure value, and supply pressure according to the pressure difference of the next main forming load range until the forming end point returns to reset.

[0005] In a preferred embodiment, S1 includes: S11. Before the first pressing section contacts the waste area, control the hydraulic servo valve to maintain the oil supply to the speed-up chamber, record the slider displacement and hydraulic pressure according to the sampling cycle, and record the first sampling point when the hydraulic pressure increases for two consecutive sampling cycles and the increase is not less than the resolution of the pressure sensor as the first pressing start point, and take the sampling point before the slider displacement reaches the first pressing stroke relative to the first pressing start point as the first pressing end point. The pre-pressing section refers to the local pressing structure set at the lower end of the mold and corresponding to the waste area of ​​the high-quality aluminum alloy sheet. When the pre-pressing section moves down with the slider, it contacts the waste area before the main forming section and forms pre-pressing deformation within the range that does not penetrate the high-quality aluminum alloy sheet and does not extend to the product area. A hydraulic servo valve is an electro-hydraulic control valve that is driven by electronic control commands output from an edge computing node and is used to change the oil inlet and return passages and flow area of ​​the speed-increasing and force-increasing chambers. The speed-increasing chamber refers to the hydraulic working chamber in the hydraulic speed-increasing mechanism that receives pressurized oil before the main forming part contacts the product area and drives the slider to move rapidly downward through the piston movement; Slider displacement refers to the distance the slider moves along the stamping direction, measured by the displacement sensor with the starting position of the slider's downward movement as the zero point; hydraulic pressure refers to the oil pressure measured by the pressure sensor in the oil inlet passage of the booster cylinder. Pressure sensor resolution refers to the unit pressure difference between two adjacent pressure readings that the pressure sensor can distinguish after calibration. An increase in hydraulic pressure reaching one unit pressure difference is recorded as an effective pressure increase. The first stroke refers to the distance the slide moves along the stamping direction from when the first pressing part begins to contact the scrap area to when the main forming part begins to contact the product area. The value of the first stroke is equal to the height difference along the stamping direction between the lower end of the first pressing part and the lower end of the main forming part. S12. For the slider displacement and hydraulic pressure before the first pressing point, the no-load pressure line of the slider displacement is obtained by least squares fitting. Substitute each slider displacement from the first pressing point to the first pressing end point into the no-load pressure line to obtain the corresponding no-load pressure. Subtract the corresponding no-load pressure from the hydraulic pressure to obtain the plate pressure. S13. Multiply the plate pressure at each sampling point by the effective area of ​​the booster cylinder to obtain the plate load, and then write the plate load in ascending order of slider displacement to form the current stroke load displacement sequence. The effective area of ​​the booster cylinder refers to the net projected area formed by the piston bearing surface on the side of the booster cylinder connected to the booster chamber along the axial direction of the booster cylinder. It is used to convert the hydraulic pressure in the booster chamber into the load output by the booster cylinder along the pressing direction.

[0006] In a preferred embodiment, S2 includes: S21. Using the first displacement value of the load displacement sequence as the interpolation starting point, subtract the first displacement value from the last displacement value and divide by the number of sampling points minus one to obtain a uniform displacement step size. Distribute the corresponding impact force difference between adjacent slider displacements according to the distance ratio to form an equal displacement load sequence. The displacement value refers to the distance the slider moves along the stamping direction, measured by the displacement sensor with the starting position of the slider's downward movement as the zero point and according to the sampling period; the first displacement value is the displacement value corresponding to the first sampling point in the load displacement sequence, and the last displacement value is the displacement value corresponding to the last sampling point in the load displacement sequence. The punching force difference refers to the difference between the punching force of the second punching force and the punching force of the first punching force corresponding to the displacement of two adjacent sliders. In the linear interpolation process, the punching force difference is distributed according to the ratio of the distance of the interpolated displacement from the previous slider displacement to the difference between the adjacent slider displacements, which is used to determine the punching force corresponding to the interpolated displacement. S22. For the equal displacement load sequence, the initial extraction interval is two to the power of zero, and the power is increased successively. The impact force is read according to the first point index plus an integer multiple of the extraction interval. When the last point index cannot be divided by the extraction interval, the last point is added to form the load sequence of each scale. Then, the slope sequence of each scale is formed by dividing the difference between adjacent impact forces by the corresponding displacement difference. S23. Using the continuous slope between any two indices in the slope sequence of each scale as candidate segments, subtract the mean slope from each slope in the candidate segments and then sum the squares to obtain the segment cost. Arrange the squares of the differences between adjacent slopes in ascending order of value, take the middle term for odd-numbered terms and half the sum of the two middle terms for even-numbered terms to obtain the segment penalty for the corresponding scale.

[0007] In a preferred embodiment, S3 includes: S31. For the slope sequence of each scale, divide the square of the impact force resolution by the square of the corresponding displacement difference, and then add the result obtained by multiplying the square of the impact force difference by the square of the displacement sensor resolution and dividing by the fourth power of the corresponding displacement difference to obtain the error of each slope. Then, divide the sum of the slope multiplied by the reciprocal of the error by the sum of the reciprocals of the error to obtain the weighted slope of the candidate segment. Finally, divide the square of each slope deviating from the weighted slope by the corresponding error and accumulate them to form the error constraint segment cost. The impact pressure resolution refers to the product of the difference between adjacent hydraulic pressure readings that the pressure sensor can distinguish after calibration and the effective area of ​​the booster cylinder. It is used to represent the unit difference that adjacent impact pressure readings in the load displacement sequence can be identified. Displacement difference refers to the value obtained by subtracting the slider displacement corresponding to the previous sampling point from the slider displacement corresponding to the subsequent sampling point in the load sequence of each scale. Displacement difference is used to calculate the slope and slope error between two corresponding sampling points. Displacement sensor resolution refers to the unit distance between adjacent slider displacement readings that the displacement sensor can distinguish after calibration, and is used to represent the position quantization error introduced during the slider displacement acquisition process; S32. Using the multi-scale PELT change point detection algorithm, with the cumulative cost of the first position of the sequence as the negative segmentation penalty, calculate the cumulative cost of each predecessor point, the sum of the error constraint segmentation cost from the predecessor point to the endpoint and the segmentation penalty for each endpoint. Read the first predecessor point in ascending order of the calculation results and write it into the predecessor chain. At the same time, delete the predecessor point whose cumulative cost of the predecessor point and the sum of the error constraint segmentation cost from the predecessor point to the current endpoint reach the current cumulative cost. Continue until the last endpoint is recursively calculated and then the change points of each scale are formed in reverse along the predecessor chain. The cumulative cost refers to the first value in the sorting result obtained by adding the error constraint segment cost and segment penalty of each candidate segment when forming all candidate segments from the first position of the slope sequence to the current end position. The cumulative cost of the first position of the sequence is the negative segment penalty, which is used to offset the segment penalty that is repeatedly included in the first candidate segment. The predecessor point is the slope index located before the current endpoint and serving as the starting boundary of the current candidate segment. The cumulative cost corresponding to the predecessor point is added to the error constraint segment cost and segment penalty from the predecessor point to the current endpoint to determine the cumulative cost of the current endpoint and the connection position of the predecessor chain.

[0008] In a preferred embodiment, S3 further includes: S33. Using each change point as the segmentation position, calculate the error constraint segmentation cost of the unsegmented interval minus the difference between the error constraint segmentation cost and the segmentation penalty of the two intervals after segmentation. Move the segmentation position from the change point to both sides one by one, convert the continuous position with non-negative difference into a change interval, and then take the coarse-scale change interval as the current intersection according to the scale from coarse to fine. Only connect the fine-scale change intervals that intersect with the current intersection and whose interval width is shortened, and update the current intersection with the intersection to form a cross-scale interval path. The segmentation interval refers to the continuous slope index range between adjacent change points or between the sequence boundary and adjacent change points, including the current change point; the current change point divides the segmentation interval into a left interval and a right interval, which are used to calculate the change in error constraint segmentation cost before and after segmentation; From coarse to fine means processing the multi-scale sequence in descending order of sampling interval. Scales with longer sampling intervals contain fewer sampling points and correspond to coarse scales, while scales with shorter sampling intervals contain more sampling points and correspond to fine scales.

[0009] In a preferred embodiment, S3 further includes: S34. Accumulate the reduction amount of each intersection and the loss amount of the segmentation difference in the cross-scale interval path in the directions from coarse to fine and from fine to coarse, respectively. Read the first path in ascending order of the accumulation results. When there is an intersection between two first paths, take the intersection as the boundary interval. When there is no intersection between two first paths, take the last level interval of the accumulation result in the preceding path as the boundary interval. Take the midpoint of the boundary interval to form the segmented boundary for the exit of the hydraulic speed-increasing chamber and the switching of the pressure supply of the force-increasing chamber. The intersection reduction refers to the difference in interval width between two adjacent intersection updates across scale interval paths. The interval width is obtained by subtracting the lower end displacement from the upper end displacement of the interval, and is used to represent the degree of shrinkage of the fine-scale change interval on the current intersection. The segmentation difference loss refers to the absolute value of the difference between the segmentation differences of two adjacent change intervals in a cross-scale interval path. The segmentation difference is the value obtained by subtracting the error constraint segmentation cost of the two intervals after segmentation and the segmentation penalty term from the error constraint segmentation cost before segmentation. It is used to indicate the degree of change in the segmentation criteria for the same change position between adjacent scales.

[0010] In a preferred embodiment, S4 includes: S41. For the load and error within each adjacent segment boundary, using the reciprocal of the error as the weight, solve the segment load slope when the partial derivatives of the weighted residual sum of squares with respect to the slope and intercept are both zero. Then, divide the last term of the absolute value of the load deviating from the fitted line in ascending order by the corresponding displacement span to obtain the slope correction amount. After adding and subtracting the slope correction amount from the segment load slope, convert it according to the ratio of the pressure area of ​​the main forming part to the pressure area of ​​the first pressing part to form the main forming load range. The partial derivative of the intercept refers to the rate of change of the weighted sum of squared residuals with respect to the intercept of the fitted line when the slope of the piecewise fitted line remains constant. It is used to determine the translation position of the fitted line along the load direction. The process of finding the piecewise load slope when the partial derivatives of the weighted residual sum of squares with respect to both the slope and the intercept are zero is as follows: multiply each displacement value in the segment by the slope to be found and add it to the intercept to be found to obtain the fitted load. Subtract the fitted load from the measured load to obtain the load residual. Multiply the square of the load residual by the reciprocal of the corresponding error and sum them to form the weighted residual sum of squares. Calculate the partial derivatives of the slope and the intercept to be found separately and set both partial derivatives to zero. Solve the two simultaneous linear equations to obtain the piecewise load slope and intercept. Load deviation from the fitted line refers to the load residual obtained by subtracting the load calculated by the piecewise fitted line from the measured load corresponding to the same displacement position. Its absolute value represents the load distance between the measured load at that displacement position and the piecewise fitted result. The pressure-bearing area of ​​the main forming part refers to the net projected area of ​​the contact surface of the main forming part applying load to the high-quality aluminum alloy sheet along the stamping direction after contacting the product area; the pressure-bearing area of ​​the pre-pressing part refers to the net projected area of ​​the contact surface of the pre-pressing part applying load to the high-quality aluminum alloy sheet along the stamping direction when it is pressed into the scrap area. S42. Divide the upper boundary of each main forming load interval by the effective area of ​​the booster cylinder, and round up according to the hydraulic servo valve pressure resolution to obtain the initial supply pressure value. Increase the initial supply pressure value according to the hydraulic servo valve pressure resolution to the pump rated pressure to form candidate supply pressure values. Then, subtract the slider movement during the hydraulic servo valve response from the interval start point to obtain the switching position corresponding to each candidate supply pressure value. The hydraulic servo valve pressure resolution refers to the unit pressure difference between two adjacent executable pressure commands after the hydraulic servo valve completes the pressure command calibration. It is used to discretely generate candidate supply pressure values ​​for the boosting chamber. The rated pressure of a pump refers to the working pressure that a hydraulic pump is allowed to continuously output under the conditions of rated speed, rated oil temperature and continuous operation. The candidate supply pressure value shall not exceed the rated pressure of the pump. The slider movement refers to the difference between the slider displacement when the hydraulic servo valve receives the pressure switching command and the slider displacement when the pressure in the boosting chamber increases for the first two consecutive sampling cycles. It is used to convert the hydraulic response time into the switching position compensation distance.

[0011] In a preferred embodiment, S4 includes: S43. Using the branch-bound interval dynamic programming algorithm, the control state is composed of interval index, candidate supply pressure value and oil supply state of the speed-up chamber. The state transition edge is established only when the difference between the next candidate supply pressure value and the previous candidate supply pressure value does not exceed the pressure increase obtained by dividing the pump flow rate by the interval passage time and dividing by the volume of the speed-up chamber and the bulk modulus of the oil, the next switching position is not earlier than the previous switching position, and the oil supply state of the speed-up chamber changes from oil supply to oil stop only once. The oil supply status of the speed-up chamber refers to the oil supply status recorded when the hydraulic servo valve controls the pressure oil to enter the speed-up chamber, or the oil stop status recorded when the hydraulic servo valve cuts off the oil inlet passage of the speed-up chamber; the oil supply status of the speed-up chamber is only allowed to change from the oil supply status to the oil stop status once in the same stroke. Pumping flow rate refers to the volume of oil pumped per unit time to the inlet passage of the booster chamber via the hydraulic servo valve. The pumping flow rate is measured by the inlet passage flow sensor within the corresponding main forming load range. The boosting chamber refers to the hydraulic working chamber in the hydraulic boosting and speed-increasing mechanism that is connected to the piston bearing surface of the boosting cylinder. After the pressurized oil enters the boosting chamber, it pushes the boosting cylinder to output load along the pressing direction. The bulk modulus of oil is the value obtained by dividing the increase in oil pressure by the corresponding relative volume compression of the oil. The bulk modulus of oil is calibrated according to the current oil temperature and gas content of the hydraulic system and is used to convert the oil volume entering the booster chamber into the pressure increase. S44. The path cost is formed by multiplying the candidate supply pressure value by the oil volume of the boosting chamber and then adding the rated power of the main motor by the interval passage time. The lower bound of the remaining cost is formed by multiplying the starting supply pressure value of the remaining main forming load interval by the corresponding oil volume and then adding the rated power of the main motor by the remaining bit and removing the result obtained by the rated speed of the speed-increasing cylinder. The branches whose sum of the current path cost and the lower bound of the remaining cost is not earlier than the completed path cost are deleted. When the control state to be expanded is empty, the candidate supply pressure value and switching position are read in reverse along the predecessor chain of the completed path to form the boosting speed control path. The rated power of the main motor refers to the mechanical power output under rated voltage, rated speed and continuous operation conditions as indicated on the main motor nameplate. It is used to calculate the work done by the main drive during the slide block passing through the corresponding main forming load range. The rated speed of the speed-increasing cylinder refers to the speed at which the speed-increasing cylinder drives the piston to move along the stamping direction under rated oil supply flow and rated load conditions; when there is a transmission ratio between the speed-increasing cylinder and the slider, it is converted into the moving speed of the slider along the stamping direction according to the transmission ratio.

[0012] In a preferred embodiment, S5 includes: S51. Read each pressure value in reverse from the forming end point to the current main forming load interval. Subtract the achievable pressure increase segment by segment according to the passage time, pump flow rate, boosting chamber volume and oil volume modulus of adjacent main forming load intervals, and retain the pressure interval that can continuously reach the corresponding pressure value to form the remaining pressure domain. The forming endpoint refers to the slider displacement position when the main forming part completes the specified forming stroke of the current stroke. The forming endpoint is calculated by subtracting the thickness of the sheet metal after forming and the mold limit clearance from the mold closing position. The achievable pressure increase refers to the pressure increase calculated based on the pump flow rate, the volume of oil input, the volume of the boosting chamber, and the bulk modulus of the oil within the time interval of a main forming load range. It is used to limit the continuous pressure increment that can be completed between adjacent main forming load ranges. S52. For each branch entering the remaining pressure supply domain, the time lower bound is obtained by removing the remaining position at the rated speed of the speed-increasing cylinder. The work lower bound is obtained by multiplying the lower boundary of the pressure supply of each remaining main forming load interval by the corresponding oil inlet volume and then summing them. When the sum of the used time and the time lower bound is after the completed path passage time, or when the two are the same and the sum of the used work and the work lower bound is not earlier than the completed path work, the corresponding branch is deleted, and a force-increasing speed control path is formed along the precursor chain of the retained branch.

[0013] In a preferred embodiment, S5 includes: S53. When the slider reaches the switching position in the force-increasing and speed-up control path, cut off the oil supply to the speed-up chamber and supply oil to the force-increasing chamber according to the corresponding supply pressure value. Divide the pressure difference obtained by subtracting the measured pressure from the target punching pressure of the corresponding main forming load range by the effective area of ​​the force-increasing cylinder to obtain the pressure correction amount. Then add the pressure correction amount to the supply pressure value of the next main forming load range to form the correction supply pressure value. S54. When the corrected pressure supply value is within the remaining pressure supply domain of the next main forming load range, continue to supply oil according to the corrected pressure supply value. When the corrected pressure supply value is outside the remaining pressure supply domain, re-execute the remaining branch search with the current slider displacement, the current force-increasing cavity pressure and the speed-increasing cavity oil stop state until the slider reaches the forming end point, then control the force-increasing cavity to depressurize and drive the slider to return to reset.

[0014] The technical effects and advantages of this invention are as follows: 1. This solution involves pressing the pre-pressed section into the same sheet metal waste area before the product area is pressurized, thereby obtaining the deformation response of the current punch in advance, so that the force and speed increase parameters can be determined before the main forming. 2. By synchronously acquiring the slider displacement and hydraulic pressure and subtracting the no-load pressure, the interference of inertia, friction and hydraulic fluctuations on load identification can be reduced, and the measurement consistency can be improved. 3. Performing equal displacement interpolation and multi-scale PELT change point detection on the load displacement sequence can enable cross-verification between different sampling scales and reduce load segment position drift. 4. Solving the main forming pressure supply path under the constraints of hydraulic boosting capacity and switching timing can improve the matching relationship between the speed-up exit position and the force-boosting pressure value; 5. Correcting the load range with the measured punching force and then re-searching when the correction value exceeds the remaining supply pressure range can relatively suppress the accumulation of load deviation in continuous punching. Attached Figure Description

[0015] Figure 1This is a flowchart of the intelligent control process for increasing force and speed of the high-tonnage, high-speed precision punch press of the present invention. Detailed Implementation

[0016] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0017] Refer to the instruction manual appendix Figure 1 The present invention provides an intelligent control method for increasing the force and speed of a high-tonnage, high-speed precision punch press, comprising: S1. A pre-pressing section is set in the scrap area of ​​high-quality aluminum alloy sheet corresponding to the mold. The height difference between the pre-pressing section and the lower end of the main forming section is the pre-pressing stroke. When the pre-pressing section is pressed into the scrap area and the main forming section does not contact the product area, the edge calculation node collects the slider displacement and hydraulic pressure. The hydraulic pressure is multiplied by the effective area of ​​the booster cylinder to form a load displacement sequence. This embodiment utilizes the pre-pressing section to press the same high-quality aluminum alloy sheet into the scrap area before the main forming section contacts the product area, so that the load change of the sheet in the current punch is reflected in the pressure of the force-enhancing chamber in advance. The speed-increasing cylinder and the force-increasing cylinder are connected to the slider. After receiving pressurized oil, the speed-increasing chamber drives the slider to move downward. The force-increasing chamber is kept filled with liquid during the initial pressurization period. A pressure sensor is set in the oil inlet passage of the force-increasing cylinder. The displacement sensor, pressure sensor, and hydraulic servo valve are all connected to the edge computing node and use the same hardware clock to generate sampling numbers. The pre-pressing section is located in the waste area to be removed later. The plastic deformation boundary formed by the pre-pressing section does not intersect with the product area boundary. After mold assembly, the slider displacements of the pre-pressing section and the main forming section when they contact the calibration sheet of the same thickness are measured respectively. The difference between the two slider displacements is written as the pre-pressing stroke. The specific execution process is as follows. After the mold calibration is completed, the edge computing node writes the oil supply command for the speed-up chamber to the hydraulic servo valve and keeps the valve opening unchanged during the pre-pressure contact recognition period; The sampling period is determined according to the upper limit of the slider's downward speed, so that the slider's movement distance within a single sampling period does not exceed one-thirty-second of the initial pressure stroke; at the end of each sampling period, the displacement sensor and the pressure sensor output the slider displacement and hydraulic pressure, respectively, and the edge computing node writes the two readings into the initial pressure sampling table according to the sampling sequence number. During continuous sampling, S11 is executed: when the hydraulic pressure at the second sampling point minus the hydraulic pressure at the first sampling point reaches the resolution of the pressure sensor, and the hydraulic pressure at the third sampling point minus the hydraulic pressure at the second sampling point reaches the resolution of the pressure sensor, the second sampling point is written as the first pressure starting point. When a change in the hydraulic servo valve opening occurs between three sampling points, the continuously added records are cleared, and the calculation is recalculated from the first sampling point after the valve opening is restored to a fixed value. The edge computing node adds the slider displacement corresponding to the first pressure start point to the first pressure stroke to obtain the first pressure termination displacement; when the slider displacement of the sampling point is equal to the first pressure termination displacement, the sampling point is written as the first pressure end point; when two adjacent sampling points cross the first pressure termination displacement, the previous sampling point is written as the first pressure end point, and the hydraulic pressure corresponding to the first pressure termination displacement is linearly inserted between the two points. If there are fewer than seventeen valid sampling points from the start point to the end point of the first press, the slider displacement changes in the opposite direction, or any sampling sequence number lacks slider displacement or hydraulic pressure, the current stroke is written with a first press sampling failure flag, and the control record of the previous qualified stroke for the same mold and the same sheet material specification is read. After the initial pressure point is written, the edge computing node reads the slider displacement and hydraulic pressure during the period when the hydraulic servo valve opening remains unchanged and before the initial pressure point, in order to separate the slider inertia, sealing friction and inherent pressure of the hydraulic circuit. The no-load pressure calculation is completed in S12: when the same slider displacement corresponds to multiple hydraulic pressures, the hydraulic pressure with the later sampling order is retained; when the later slider displacement is less than the previous slider displacement, the reverse displacement position and subsequent records are truncated. Let the number of sampling points be n, the slider displacement be x, and the hydraulic pressure be p. The fitting denominator is n multiplied by the sum of the squares of the slider displacements minus the square of the sum of the slider displacements. When the fitting denominator is non-zero and there are more than three different slider displacements, the slope of the no-load pressure is obtained by multiplying n by the sum of the products of the slider displacement and the hydraulic pressure minus the product of the slider displacement and the hydraulic pressure, and then dividing by the fitting denominator. The intercept of the no-load pressure is obtained by subtracting the product of the no-load pressure slope and the sum of the slider displacements from the sum of the hydraulic pressures and then dividing by n. Each slider displacement from the initial pressure point to the initial pressure point is multiplied by the no-load pressure slope and the no-load pressure intercept is added to obtain the corresponding no-load pressure. The hydraulic pressure under the same sampling number is subtracted from the no-load pressure to obtain the plate pressure; when the plate pressure is less than zero, it is written as zero. When the fitting denominator is zero, there are fewer than three different slider displacements, or the sampling sequence is interrupted, the edge calculation node writes an empty fitting failure flag and stops generating the plate pressure for the current stroke. After the plate pressure is formed, the edge computing node reads the net projected area of ​​the piston bearing surface on the side where the booster cylinder and booster chamber are connected, and writes the net projected area as the effective area of ​​the booster cylinder. Load conversion is performed in S13: the plate pressure at each sampling point is multiplied by the effective area of ​​the booster cylinder to obtain the plate load; when the booster cylinder and the slider are connected by a fixed transmission ratio, the plate load is multiplied by the fixed transmission ratio to obtain the load of the first pressing part acting on the scrap area along the stamping direction. When the same slider displacement corresponds to multiple plate loads, retain the plate load with the later sampling order; when the slider displacement changes in the opposite direction, truncate the reverse position and the subsequent records. Edge computing nodes write the slider displacement and plate load in ascending order of slider displacement, and simultaneously write the first displacement value, the last displacement value and the number of sampling points to form the current stroke load displacement sequence; If the load displacement sequence has fewer than seventeen different slider displacements, the final displacement value is not greater than the first displacement value, or the plate load lacks a corresponding slider displacement, the current punch load displacement sequence will not be sent to subsequent processing. Taking the processing of a 2.0 mm thick high-quality aluminum alloy sheet of the sixth series as an example, the pre-pressing part is arranged in the scrap area to be removed by subsequent punching. The mold calibration shows that the pre-pressing stroke is 1.6 mm, the upper limit of the slider downward speed is 0.8 m / s, the sampling period is set to 0.05 ms, and a single sampling period corresponds to a slider movement distance of 0.04 mm. During the period when the opening of the hydraulic servo valve remains unchanged, the edge computing node determines the starting point of the pressure based on two consecutive effective pressure increments, and determines the ending displacement of the pressure by adding 1.6 mm to the displacement of the starting point of the pressure. Then, the unloaded pressure slope and unloaded pressure intercept are obtained by using the sampled data before the starting point of the pressure. The unloaded pressure is deducted point by point, and the plate pressure is multiplied by the effective area of ​​the booster cylinder to form the current stroke load displacement sequence with strictly increasing slider displacement. The load variation in the current stamping waste area is thus converted into a uniform load record before the main forming part contacts the product area, and is used for subsequent equal displacement interpolation and multi-scale slope calculation.

[0018] S2. Interpolate the load displacement sequence according to the uniform displacement step size and extract the multi-scale sequence according to the integer power of two. Calculate the ratio of the impact force difference to the displacement difference between adjacent points to form a slope sequence. Use the sum of squared residuals within the segment as the segment cost and the median of the squared difference between adjacent slopes as the segment penalty. After the load displacement sequence enters S2, the edge computing node first eliminates the displacement interval difference caused by the fixed sampling period and inconsistent slider movement distance, and then generates load expressions with different extraction intervals from the same displacement sequence for subsequent change points to be read together. The load displacement sequence involved in the processing is arranged along the downward direction of the slider, where the displacement value is the moving distance measured by the displacement sensor with the starting position of the slider's downward movement as the zero point. The first displacement value and the last displacement value are taken from the first and last terms of the sequence, respectively. The sequence retains at least three different displacement values, with the last displacement value being greater than the first displacement value, and adjacent displacement values ​​strictly increasing. The edge computing node completes the end data writing after the current stroke's pre-pressing endpoint is formed, and performs interpolation position update using the sequence prefix already written during the pre-pressing period, so that the calculation result of S2 is formed before the main forming pressure supply path is solved; The displacement reference of the load displacement sequence is unified by S21; assuming that the load displacement sequence contains n sampling points, the edge calculation node subtracts the first displacement value from the last displacement value, and then divides by n minus one to obtain the unified displacement step size; then, starting from the first displacement value, the unified displacement step size is accumulated sequentially to generate n interpolated displacements, the first interpolated displacement is directly taken from the first displacement value, and the last interpolated displacement is directly taken from the last displacement value. For the interpolated displacement located between two adjacent original displacements, first find adjacent sampling points that satisfy the condition that the previous original displacement is not greater than the interpolated displacement and the interpolated displacement is not greater than the subsequent original displacement. Then, subtract the previous impact force from the subsequent impact force to obtain the impact force difference. Divide the distance obtained by subtracting the previous original displacement from the interpolated displacement by the difference between the two original displacements, and multiply the resulting ratio by the impact force difference and add it to the previous impact force to obtain the interpolated impact force. If the interpolated displacement is the same as the original displacement, directly read the original impact force. When the difference between adjacent original displacements is zero, retain the impulse force that is sampled later and rearrange the load displacement sequence; if there are still fewer than three different displacement values ​​after rearrangement, or if the last displacement value is not greater than the first displacement value, write an interpolation failure flag and stop the scale extraction of the current impulse; write all interpolated displacements and interpolated impulse forces in pairs according to the interpolation index to form an equal displacement load sequence. After completing the displacement unification, S22 establishes load sequences for each scale from the same equal displacement load sequence; the edge calculation nodes make the extraction interval take the zero power, first power, and second power of two in sequence, and read the interpolation punch according to the first point index plus an integer multiple of the extraction interval; when the last point index cannot be divided by the extraction interval, the last point of the equal displacement load sequence is added to the end of the current scale. After each extraction, the number of load points at the current scale is counted, and a scale with no less than four load points is retained; if fewer than four load points are generated in the next extraction interval, the exponential increment is terminated. The slope between adjacent scale load points is obtained by subtracting the previous impact force from the subsequent impact force, and then dividing by the subsequent displacement value minus the previous displacement value; when the end point is formed by the supplementation method, the slope still uses the actual displacement difference between the end point and the previous extraction point, and is not replaced by the standard extraction interval of the current scale. The edge computing node writes the scale number, extraction interval, original sequence index, displacement value and impact force into the scale load table, and writes the slope formed by each adjacent point into the corresponding scale slope sequence in index order, so that the change position of different scales can be uniformly converted into slider displacement in the later stage. After the slope sequences at each scale are formed, S23 is used to establish the candidate segmentation cost and segmentation penalty required for multi-scale PELT change point detection; Edge computing nodes form candidate segments based on the continuous slope between any two segmentation indices, and each candidate segment contains at least two slopes; index combinations with fewer than two slopes are not written into the candidate segment table. For each valid candidate segment, first add up all the slopes in the segment and divide by the number of slopes to get the average slope. Then calculate the square of the difference between each slope and the average slope and sum them up to get the segment cost. The segmented penalty is determined by the square of the difference between adjacent slopes in the current scale slope sequence: all squared differences are arranged in ascending order of value. When the number of squared differences is odd, the middle term is read. When the number of squared differences is even, half of the sum of the two middle terms is read. When the obtained value is zero, the slope resolution is obtained by dividing the impact force resolution by the actual displacement difference between adjacent load points in the current scale, and the square of the slope resolution is taken as the segmented penalty. The segment cost, segment penalty, candidate segment start point and candidate segment end point are written into the current scale cost table respectively. The segment cost is used for initial residual comparison and subsequent segment difference calculation. S31 introduces the slope error on the same candidate segment and generates error constraint segment cost separately. The two cost fields are saved separately. Taking a load displacement sequence containing 33 different displacement values ​​as an example, the edge computing nodes first generate 33 equally spaced interpolation points, and then form four scales according to interpolation step sizes of one, two, four and eight; when the extraction interval is increased to sixteen, only three load points are retained, so the extraction is terminated before the scale is generated. The slope is calculated using the actual difference between adjacent displacements at each scale, and candidate segments containing only one slope are excluded; when the median of the square of the difference between adjacent slopes is zero, the square of the slope resolution at the current scale is directly used as the substitute. The resulting slope sequences, segmented costs, and segmented penalties at each scale have a unified index source and can be directly read by the subsequent multi-scale PELT change point detection process.

[0019] S3. The multi-scale PELT change point detection algorithm is used to recursively calculate the cumulative cost. The first term of the cumulative cost is taken as the predecessor point and the change point is read in reverse. The intervals with the highest residual decrease are retained from coarse to fine. The midpoint of the last interval is taken to form the segment boundary. After the slope sequence at each scale is formed, the edge calculation nodes first convert the pressure measurement error and displacement measurement error to the slope calculation caliber, and then complete the recursion of change points, expansion of change intervals, and cross-scale position verification. All scales use the scale number, extraction interval, load point index and slider displacement written in S22. Change points uniformly represent the division position between adjacent slopes, and change intervals and segment boundaries are uniformly converted into slider displacements. To reduce repetitive calculations within the current stroke, edge computing nodes store the sum of the inverse of the error, the sum of the product of the slope and the inverse of the error, and the sum of the product of the square of the slope and the inverse of the error, according to the scale. The data required for candidate segments are obtained through prefix differences. The pressure and displacement resolution capability is first passed to each slope; the edge computing node reads the impact pressure difference, displacement difference, impact pressure resolution and displacement sensor resolution of adjacent load points at the same scale, and when executing S31, the value obtained by dividing the square of the impact pressure resolution by the square of the displacement difference is added to the value obtained by multiplying the square of the impact pressure difference by the square of the displacement sensor resolution and then dividing by the fourth power of the displacement difference to obtain the error of the current slope. Candidate segments contain at least two continuous slopes. The reciprocal of the error of each slope is used as the corresponding weight. The weighted slope is obtained by dividing the sum of the products of the slope and the weight by the sum of the weights. The square of the deviation of each slope from the weighted slope is then divided by the corresponding error and accumulated to form the error-constrained segment cost. The segmented penalty term generated by S23 is divided by the median of the current scale error and then written to the current scale, so that the segmented penalty term and the error constraint segmented cost adopt the same dimensionless caliber; when the candidate segment has less than two slopes or the weight sum is zero, the corresponding segment record is not generated and the scale calculation failure flag is written. After the cost caliber is unified, the change points are recursively calculated according to the slope index. Assuming that the current scale contains m slopes, a zero boundary is set before the first slope and a final boundary is set after the last slope. The change point is located between two adjacent slopes. The edge computing node writes the cumulative cost of the zero boundary as a negative segmented penalty and adds the zero boundary to the predecessor candidate set. For any endpoint, when executing S32, only the predecessor point that contains at least two slopes between it and the current endpoint is read. The cumulative cost of the predecessor point, the sum of the error constraint segment cost and the segment penalty from the predecessor point to the current endpoint are calculated respectively, and the results are sorted in ascending order. If the results are the same, the predecessor point with the earlier index is read. The first result is written as the current endpoint cumulative cost, and the first predecessor point is written into the predecessor chain; when the sum of the predecessor point cumulative cost and the error constraint segment cost from the predecessor point to the current endpoint reaches the current endpoint cumulative cost, the corresponding predecessor point is deleted from the candidate set. After the final boundary is recursively calculated, the edge computing nodes read backward along the predecessor chain to the zero boundary, and then form the current scale change points in ascending index order; if no internal change points are formed, the current scale is recorded as a scale with no change. Each change point is then expanded into a continuous range that allows movement; the slope of the previous change point is the local starting point of the edge computing node, and the slope of the next change point is the local ending point. The first change point takes the zero boundary as the previous boundary, and the last change point takes the last boundary as the last boundary. When executing S33 around the current point of change, first calculate the error constraint segmentation cost of the unsegmented local interval, then subtract the left and right costs and a segmentation penalty after segmentation to obtain the original segmentation difference; then move the segmentation position left and right bit by bit and recalculate the segmentation difference. When the next position results in less than two slopes on either side, a segmentation difference of less than zero, or the local boundary is reached, the expansion in the corresponding direction stops, and the continuous index between the left and right stop positions is converted into the slider displacement range, forming a change interval. For scales with long extraction intervals, cross-scale connections are first entered. Each coarse-scale variation interval intersects with all variation intervals of the next scale. When an intersection exists and the intersection width is shortened, a connection is established for each fine-scale variation interval that meets the conditions, and the current path is updated with the new intersection. When there is no variation interval that meets the conditions at a certain scale, the current path terminates at that scale. Two-way verification is used to eliminate position offset caused by a single traversal direction; edge computing nodes establish cross-scale interval paths according to the extraction interval from long to short and from short to long respectively, the difference in interval width before and after each intersection update is recorded as the intersection reduction amount, and the absolute value of the difference in the segmentation difference between adjacent changed intervals is recorded as the segmentation difference loss amount; After the path calculation enters S34, the intersection reduction amount is divided by the width of the initial change interval of the path to obtain the intersection reduction ratio, the segmentation difference loss amount is divided by the sum of the absolute values ​​of the two adjacent segmentation differences to obtain the segmentation loss ratio, and the intersection reduction ratio and the segmentation loss ratio of each time are accumulated to form the path cost. When the corresponding denominator is zero, the correlation ratio is written as zero; when there is a path covering all scales, only the complete path is compared; when there is no complete path, only the path with the most covered scales is compared. When the final intersections of the first terms of the paths in two directions intersect, the intersection is taken as the boundary interval; when the two paths do not intersect, the final intersection of the paths with the path cost preceding the first path is taken as the boundary interval. The midpoint of the boundary interval is written as the segment boundary, and then sorted in ascending order of slider displacement; for those with the same midpoint, the item with the path cost first is retained; when there are less than two slopes between adjacent segment boundaries, the segment boundary with the path cost first is retained. Taking the four-scale slope sequence formed by thirty-three equal displacement load points as an example, the edge computing nodes first obtain the slope error based on the impact force resolution and displacement sensor resolution, and then calculate the candidate segment cost based on the prefix quantity. After the change points are recursively calculated, each change point is expanded into a slider displacement interval. The coarse-scale change interval and the fine-scale change interval are intersected step by step, and the final intersection is checked from the two traversal directions. Repeated boundaries are merged, and boundaries with less than two slopes in adjacent intervals are deleted. The final segmented boundaries are written into S4 for reading when fitting the main forming load interval and calculating the hydraulic switching position.

[0020] S4. Fit the load slope according to the segment boundary and convert the main forming load range according to the pressure area ratio. Use the branch limit interval dynamic programming algorithm to generate the pressure supply value and switching position, and connect the interval where the pressure supply increment does not exceed the pressure rise of the force-increasing cavity. After the segment boundary enters S4, the edge computing node first converts the load change in the waste area into the load range that the product area needs to bear in each displacement range, and then combines the discrete pressure supply capability of the hydraulic servo valve, the pressure boosting capability of the boosting chamber, and the slider running sequence to search for the pressure supply path. The main forming load range uniformly includes the range start point, range end point, load lower boundary, and load upper boundary; each control state in the pressure supply path uniformly includes the range index, candidate pressure supply value, and speed-up chamber oil supply state. The hydraulic pump is driven by an independent pump motor, and the main motor is used to drive the main transmission mechanism of the punch press. The hydraulic input power and the main transmission time are calculated separately and then written into the same path cost; the specific execution process is as follows; The segmented load fitting first processes the dimensional conversion between the slope error and the load residual. For each load point within a segment, the edge calculation nodes multiply the adjacent slope error by the square of the corresponding displacement difference to obtain the load error. The first point of the segment reads the load error on the right, the last point of the segment reads the load error on the left, and the remaining load points take half of the sum of the load errors on the left and right sides. Load fitting is completed in S41. The reciprocal of the load error is used as the weight. Each displacement value is multiplied by the slope to be determined and the intercept to be determined to obtain the fitted load. The square of the residual obtained by subtracting the fitted load from the measured load is multiplied by the corresponding weight and then accumulated. The partial derivatives of the accumulated results with respect to the slope and intercept are set to zero at the edge calculation nodes. The two linear equations are combined to obtain the piecewise load slope and intercept. When a segment contains fewer than two distinct displacement values ​​or the denominator of the simultaneous equations is zero, the current segment is merged with the next segment; when the current segment is at the end, it is merged with the previous segment. Then, the absolute values ​​of each load residual are arranged in ascending order, and the last term is divided by the span obtained by subtracting the starting displacement of the segment from the end displacement of the segment to obtain the slope correction amount. The pressure area of ​​the main forming part and the pressure area of ​​the first pressing part are both calculated based on the net projected area of ​​the contact surface in the stamping direction, and are limited to the displacement range in which the main forming part and the first pressing part are subjected to normal pressing loads in the stamping direction; the edge calculation node adds or subtracts the slope correction amount of the segmented load slope, and then multiplies it by the ratio of the two pressure areas to obtain the lower boundary and upper boundary of the main forming load slope. The starting load of the first main forming load interval is taken as the fitted starting load of the first pressing part multiplied by the ratio of the two pressing areas. The starting load of subsequent main forming load intervals is taken as the ending load of the previous main forming load interval. Each slope boundary is multiplied by the displacement increment of the current displacement relative to the starting point of the interval, and then added to the starting load of the interval to form the lower boundary and upper boundary of the load. When the lower boundary of the load is less than zero, it is written as zero. After obtaining the main forming load range, the hydraulic pressure level is generated according to the pressure command that the hydraulic servo valve can actually execute; The pressure supply discretization is performed within S42. The edge computing node divides the upper boundary of the load of each main forming load interval by the effective area of ​​the booster cylinder, then divides it by the pressure resolution of the hydraulic servo valve, reads the first integer not less than the obtained quotient, and multiplies the integer by the pressure resolution of the hydraulic servo valve to obtain the initial pressure supply value. The pump rated pressure, hydraulic servo valve allowable pressure, booster cylinder allowable pressure and oil inlet pipeline allowable pressure are arranged in ascending order of value, with the first item written as the upper limit of supply pressure; when the initial supply pressure value exceeds the upper limit of supply pressure, the current main forming load range is written with the supply pressure unreachable flag. When the initial supply pressure value does not exceed the supply pressure limit, the pressure is increased incrementally from the initial supply pressure value according to the pressure resolution of the hydraulic servo valve until the supply pressure limit is reached, thus forming a candidate supply pressure value. The slider movement during the hydraulic servo valve response is read from the installation calibration table. The installation calibration table uses the supply pressure value before switching, the candidate supply pressure value after switching, and the slider speed during switching as indexes to record the slider movement distance during the two consecutive sampling cycles from when the hydraulic servo valve receives the command to when the pressure in the boosting chamber increases. When the index value is between two calibrated slider speeds, it is interpolated according to the distance ratio. When the index value exceeds the range of the calibration table, the corresponding candidate supply pressure value is deleted. The switching position is obtained by subtracting the corresponding slider movement from the starting point of the interval. If the switching position is earlier than the current slider displacement, the corresponding candidate supply pressure value is also deleted. After the candidate pressure value and switching position are written into the status table, the edge computing node begins to establish the control state transition between adjacent main forming load intervals; In S43, the branch-bound interval dynamic programming algorithm uses the first main forming load interval, the pressure level after the current boosting chamber pressure is rounded up according to the hydraulic servo valve pressure resolution, and the oil supply state of the speed-up chamber to form the initial control state. The subsequent candidate supply pressure value must not be lower than the previous candidate supply pressure value, the subsequent switching position must not be earlier than the previous switching position, and the oil supply status of the speed-increasing chamber is only allowed to switch from oil supply to oil stop once, and oil supply will not be restored after oil stop. The achievable pressure boost of adjacent main forming load intervals is calculated from the net compressed oil quantity: first, multiply the pumping flow rate measured by the inlet flow sensor by the interval passage time to obtain the inlet oil volume, then subtract the sweeping volume obtained by multiplying the effective area of ​​the booster cylinder by the equivalent movement of the booster piston, and subtract the leakage volume obtained by multiplying the machine leakage flow rate by the interval passage time to obtain the net compressed oil quantity; when the net compressed oil quantity is not greater than zero, the achievable pressure boost is written as zero; The starting and ending volumes of the booster chamber are calculated based on the position of the booster piston. The sum of the two volumes divided by two yields the volume of the booster chamber in the interval. The bulk modulus of the oil is read from the calibration table using bidirectional linear interpolation based on the current oil temperature and gas content. The net compressed oil volume is divided by the volume of the booster chamber in the interval and then multiplied by the bulk modulus of the oil to obtain the achievable pressure boost. When the difference between the previous and subsequent candidate supply pressure values ​​does not exceed the achievable pressure boost, a state transition edge is established between the two control states; when the oil temperature or oil gas content exceeds the calibration range, the boosting chamber volume is zero, or the net compressed oil quantity is missing, no corresponding state transition edge is established. Before the path search begins, the edge computing nodes first attempt to connect the complete path according to the first item of the candidate pressure supply value in the ascending order within each main forming load interval; when the connection is interrupted, the candidate pressure supply value at the interruption position is incremented sequentially and backtracked until the first completed path reaching the forming endpoint is formed. In S44, the path calculation multiplies the candidate supply pressure value by the corresponding oil inlet volume of the booster chamber to obtain the hydraulic input power, and then multiplies the rated power of the main motor by the interval transit time to obtain the main drive time cost. The two are added together and accumulated segment by segment to form the path cost. The rated power of the main motor is only used to unify the energy dimension of the time item and is not written into the actual power consumption record of the equipment. For branches that have not yet extended to the forming endpoint, the edge calculation nodes are accumulated by multiplying the initial supply pressure value of each remaining main forming load interval by the base oil inlet volume, and then adding the rated power of the main motor multiplied by the remaining bit removal to obtain the rated speed of the speed-increasing cylinder, forming the lower bound of the remaining cost; the base oil inlet volume is the sum of the piston sweep volume and leakage volume of the corresponding interval; When the sum of the current path cost and the lower bound of the remaining cost reaches the cost of the completed path, delete the current branch; when the path costs are the same, read the paths with shorter interval traversal time and fewer supply pressure switching times in sequence. When the extended control state is empty and the completed path set is not empty, the candidate pressure supply value and switching position are read in reverse along the predecessor chain of the first item in ascending order of path cost to form a force-increasing and speed-up control path for S5 to continue to verify; when the completed path set is empty, the pressure supply unreachable flag is written, the main forming pressure supply command is stopped, and the speed-up chamber is controlled to stop oiling and the slider is controlled to return. If the edge computing node has not completed all searches before the slider reaches the first switching position, it reads the first completed path and enters S5; if no completed path has been formed, it performs a power supply unreachable handling procedure. After completing the four processing steps, in the actual scenario: when the pressure area of ​​the first pressing part is 20 square centimeters and the pressure area of ​​the main forming part is 800 square centimeters, the pressure area ratio is written as 40; after the load slope and slope correction of a certain segment are converted, the corresponding main forming load range is formed by combining the segment starting load and displacement increment. The hydraulic servo valve has a pressure resolution of 0.1 MPa. When the upper boundary of the load is divided by the effective area of ​​the booster cylinder to obtain 12.03 MPa, the initial supply pressure value is written as 12.1 MPa, and then candidate supply pressure values ​​are generated by incrementing by 0.1 MPa. The edge computing node calculates the achievable pressure boost in adjacent intervals based on the current oil temperature, oil gas content, variable volume of boosting chamber, and net compressed oil quantity. It only connects to control states that can complete the pressure boost on time and whose switching position sequence is valid. Then, it uses the lower bound of the completed path cost and the remaining cost to delete branches that cannot obtain the previous result, and finally forms a boosting speed control path that includes segmented pressure supply values ​​and switching positions.

[0021] S5. Cut off the branches of the completed path with the time lower bound of the remaining displacement and the work lower bound of the remaining supply pressure to obtain the force and speed control path; when the slider reaches the switching position, cut off the oil supply to the speed-increasing chamber and supply oil to the force-increasing chamber according to the supply pressure value, and supply pressure according to the pressure difference of the next main forming load range, and return to the forming end point for reset. After the main forming load range and candidate supply pressure values ​​are determined, the edge calculation nodes, based on the hydraulic system's ability to continuously increase pressure from the current slider position to the forming endpoint, reversely delete pressure levels that cannot reach the endpoint supply pressure requirements, and then screen the force and speed increase control path by time and hydraulic work. The starting position of the slider's downward movement is set as the zero point of displacement, and the downward direction is the direction of displacement increase. The mold closing position in the state without sheet metal is converted to the same displacement coordinate. The forming end point is obtained by subtracting the thickness of the sheet metal after forming and the mold limiting gap from the mold closing position. The main forming load range, candidate supply pressure value, switching position, range passage time and state transition edge all use the records generated by S4 to avoid the remaining path search changing the previous pressure level and the range boundary diameter. The main forming load interval where the forming endpoint is located is first used as the starting point of reverse recursion; the edge calculation node reads the candidate supply pressure value within this interval that is not lower than the initial supply pressure value and does not exceed the supply pressure upper limit, and writes the first candidate supply pressure value and the last candidate supply pressure value as the lower boundary and upper boundary of the final stage remaining supply pressure domain, respectively; When retreating step by step towards the current main forming load range, S51 uses the achievable pressure increase calculated in S43 from the net compressed oil quantity, the range boosting cavity volume, and the oil bulk modulus. The achievable pressure increase of the current range is subtracted from the lower boundary of the next remaining pressure supply range, and the latter term of the result and the initial pressure supply value of the current range is written as the current lower pressure limit; the former term of the upper boundary of the next remaining pressure supply range and the upper limit of the current range is written as the current upper pressure limit. The current lower pressure limit and the current upper pressure limit are then rounded inward according to the pressure resolution of the hydraulic servo valve to form the remaining pressure range of the current interval. If the current lower pressure limit exceeds the current upper pressure limit, or if there is no pressure level that the hydraulic servo valve can execute after rounding, delete the reverse branch; if there is a pressure level, write the current remaining pressure domain into the corresponding interval and continue to push forward. When the pressure in the booster chamber is limited to not decreasing along the downward direction of the slider, the upper limit of the current supply pressure does not exceed the upper boundary of the next remaining supply pressure domain. When re-searching from the current slider position, the lower boundary of the current remaining supply pressure domain also needs to be compared with the pressure level after the current booster chamber pressure is rounded up according to the pressure resolution, and the result with the value read is the result of the second step. Branches entering the remaining pressure supply domain are selected according to time priority and work order; the rated speed of the speed increase cylinder is converted into the slider speed after being converted by the transmission ratio, and is also used as the upper limit of the slider's downward speed; the distance from the current slider displacement to the forming end point is divided by this upper limit of speed to obtain the lower limit of the time used by S52; The lower limit of hydraulic power is calculated item by item according to the remaining main forming load interval: for each interval, the lower boundary of the remaining pressure supply domain is read, and the volume of the booster piston sweeping is added to the leakage volume obtained by multiplying the installed leakage flow rate by the interval passage time to obtain the basic oil inlet volume. Then, the lower boundary of the remaining pressure supply domain is multiplied by the basic oil inlet volume, and the sum is accumulated to form the lower limit of power. The time used is accumulated from the moment the main forming part contacts the product area, and the work done is accumulated by multiplying the actual supply pressure value of the already executed interval by the actual oil inlet volume. The first completed path obtained in S44 is used to provide an initial feasible path; when comparing branches in S52, branches whose sum of used time and time lower bound exceeds the time taken by the current completed path are deleted; when the two are equal, branches whose sum of used work and work lower bound reaches the work done by the current completed path are deleted. After forming a new complete path, first compare the path travel time, then compare the work done on the path if the travel time is the same, and update the current completed path with the complete path that is ranked first. After all extended branches have been processed, the edge computing node reads the pressure value and switching position of each interval in reverse along the predecessor chain of the currently completed path to form a power boosting and speed control path; if a complete path is not formed, a remaining pressure unreachable flag is written. After the slider enters the switching position in the boosting and speed control path, the hydraulic servo valve cuts off the oil inlet passage of the speed-increasing chamber and keeps the speed-increasing chamber in the oil-stop follow-up oil return state, while supplying oil to the boosting chamber according to the current interval supply pressure value. Within a complete sampling period before the end of the current main forming load range, the edge calculation node reads the upper boundary of the load at the current slider displacement as the target punching force; the measured punching force is obtained by subtracting the calibrated back pressure on the return oil side from the oil inlet pressure of the boosting chamber, multiplying it by the effective area of ​​the boosting cylinder, and then multiplying it by the fixed transmission ratio from the boosting cylinder to the slider. S53 subtracts the measured punching pressure from the target punching pressure to form the punching pressure difference, and then divides it by the effective area of ​​the booster cylinder and the fixed transmission ratio to obtain the pressure correction amount; the pressure correction amount is added to the original supply pressure value of the next main forming load range, and then rounded up according to the pressure resolution of the hydraulic servo valve to form the corrected supply pressure value. When the corrected supply pressure value is lower than the current booster chamber pressure, it is written as the current booster chamber pressure; when it exceeds the supply pressure limit, it is directly recorded as outside the supply pressure range. The corrected supply pressure value for the next main forming load range is no longer generated for the last main forming load range. After the corrected pressure value is formed, the edge computing nodes check the corrected pressure value using an interval judgment method that includes the upper and lower boundaries of the remaining pressure domain; When the corrected pressure value is within the remaining pressure range of the next main forming load interval, S54 replaces the original pressure value of the next main forming load interval with this pressure level, and rewrites the next switching position according to the previous switching position, the hydraulic servo valve response movement amount and the starting point of the next main forming load interval. When the corrected supply pressure value is lower than the lower boundary of the remaining supply pressure domain or higher than the upper boundary of the remaining supply pressure domain, the index of the main forming load interval where the current slider displacement is located, the pressure level obtained by rounding up the current force-increasing cavity pressure, and the oil stop state of the speed-increasing cavity together constitute a new initial control state. The used time and the used work remain at their original cumulative values, and the edge computing node re-executes the reverse recursion of the remaining supply pressure domain and the remaining branch search. The deadline for re-searching is the moment when the slider reaches the starting point of the next unexecuted main forming load interval. During the search, branches with earlier lower time bound values ​​are prioritized for expansion. When a complete path is formed before the deadline, the unexecuted supply pressure value and switching position are replaced. If a complete path is not formed by the deadline, the hydraulic servo valve maintains the current pressure of the boosting chamber, the main motor performs deceleration braking, and after the slider stops descending, the oil inlet passage of the boosting chamber is cut off and the oil return passage is opened. After the pressure of the boosting chamber drops to the return drive load divided by the effective area of ​​the boosting cylinder to obtain the return allowable pressure, the slider is driven to return. When the slider reaches the forming endpoint normally, the edge calculation node maintains the pressure supply to the booster cavity according to the holding time in the current stamping process record; when the holding time is written as zero, it directly enters the depressurization stage. After the pressure holding period ends, the hydraulic servo valve closes the oil inlet passage of the boosting chamber and opens the oil return passage step by step according to the oil return flow area. After the pressure of the boosting chamber drops to the return stroke allowable pressure, the return stroke drive is started. When the slider returns to the starting position of the downward movement, the oil return passage is closed and the initial oil supply state of the boosting chamber is restored. The actual passage time, actual oil inlet volume, measured stroke pressure in each interval, pressure correction amount, and re-search results of the current stroke are written into the control record by stroke number, so that the hydraulic response movement and leakage flow can be read in the next stroke. In conjunction with continuous stamping conditions, the candidate pressure range for the final stage main forming load interval is 18.4 MPa to 20 MPa. When the achievable pressure increase of the previous interval is 1.2 MPa, the edge calculation node subtracts 1.2 MPa from the lower boundary of the final stage pressure domain and compares it with the initial pressure value of the previous interval to form the lower limit of the pressure of the previous interval. The pressure domain is discretized by pressure resolution and then backtracked to the current slider position step by step. The path selection first compares the time required to reach the forming endpoint, and then compares the hydraulic work done; when the corrected supply pressure value generated during the execution is within the remaining supply pressure domain, it is directly written into the next main forming load range; when it exceeds the remaining supply pressure domain, it is searched again from the current force-enhancing chamber pressure. Therefore, the force-increasing and speed-up control path is generated and executed under the joint constraints of hydraulic pressure boosting capacity, slider running time, and forming load requirements.

[0022] exist Figure 1 It should be noted that this diagram illustrates the overall execution flow of the intelligent control method for increasing force and speed in large-tonnage, high-speed, precision punching machines. After the current stroke begins, the pre-pressing section first performs pre-pressing in the high-quality aluminum alloy sheet scrap area. The edge computing node synchronously collects the slider displacement and hydraulic pressure to form a load displacement sequence. Then, the load displacement sequence is subjected to iso-displacement interpolation and multi-scale extraction to obtain a multi-scale slope sequence. Then, the load change range and segment boundaries are determined by multi-scale PELT change point detection. The segment boundaries are used to convert the main forming load range, and the supply pressure value and switching position are solved by dynamic programming of the branch limit range to form a force-increasing and speed-up control path. After the slider reaches the switching position, the hydraulic servo valve cuts off the oil supply to the speed-up chamber and supplies oil to the force-increasing chamber. At the same time, the corrected supply pressure value is calculated based on the target punching pressure and the measured punching pressure. When the corrected supply pressure value is within the remaining supply pressure domain of the next main forming load range, the oil supply continues according to the corrected supply pressure value. When the corrected supply pressure value is outside the remaining supply pressure domain, the remaining control path is searched again based on the current slider displacement, force-increasing chamber pressure, and speed-up chamber oil stop status. After the slider reaches the forming end point, it completes depressurization, return stroke, and reset. In the diagram, rounded rectangles represent data processing or control execution processes, diamonds represent the judgment of whether the corrected supply pressure value is within the remaining supply pressure domain, ellipses represent the start and end of the current stroke, branch lines marked yes or no represent the subsequent flow corresponding to the judgment result, solid lines with arrows represent the execution order and data transmission direction between each processing process, and solid lines returning to execute the power boosting and speed increasing control path represent the continued execution relationship after the remaining control path is re-searched.

[0023] Overall Execution and Implementation: This scheme first presses the scrap area of ​​the same high-quality aluminum alloy sheet into the mold before the main forming part contacts the product area. The edge computing node synchronously collects the slider displacement and hydraulic pressure. After deducting the no-load pressure, the load displacement sequence of the current stroke is obtained. The load displacement sequence is then transformed into a slope sequence after equal displacement interpolation and multi-scale extraction. The multi-scale PELT change point detection algorithm identifies the load change position of the sheet and forms the segment boundary. The edge computing node then converts the scrap area load of each segment into the main forming load interval. The branch boundary interval dynamic programming algorithm, combined with the hydraulic boosting capacity, slider passage time, and pressure supply work, determines the speed-up exit position and segment pressure supply value. When the slider reaches the switching position, the oil supply to the speed-up chamber stops and the oil supply to the force-increasing chamber stops. Subsequently, the pressure supply value of the next load interval is corrected according to the difference between the target punching pressure and the measured punching pressure until the punching is completed and the return is reset. During the continuous stamping process of automotive body panels, the thickness, hardening degree, and lubrication state of the same roll of high-quality aluminum alloy sheet will change at different locations. After each feeding, the first pressing section forms a local pressing in the scrap area to be cut off. The edge calculation node obtains the actual load response of the current sheet position based on this and completes the load segmentation and pressure supply path calculation before the main forming section contacts the product area. If the load at the current sheet position increases ahead of schedule, the system will stop the oil supply to the speed-increasing chamber and increase the pressure supply to the force-increasing chamber accordingly. If the load increase is delayed, the system will switch and adjust the subsequent pressure supply later. Thus, each stroke completes the force and speed matching based on the actual state of the current sheet, avoiding the inability of fixed control parameters to adapt to local differences in the roll material.

[0024] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for intelligent control of force and speed increase in a large-tonnage, high-speed precision punching press, characterized in that: include: S1. A pre-pressing section is set in the scrap area of ​​high-quality aluminum alloy sheet corresponding to the mold. The height difference between the pre-pressing section and the lower end of the main forming section is the pre-pressing stroke. When the pre-pressing section is pressed into the scrap area and the main forming section does not contact the product area, the edge calculation node collects the slider displacement and hydraulic pressure. The hydraulic pressure is multiplied by the effective area of ​​the booster cylinder to form a load displacement sequence. S2. Interpolate the load displacement sequence according to the uniform displacement step size and extract the multi-scale sequence according to the integer power of two. Calculate the ratio of the impact force difference to the displacement difference between adjacent points to form a slope sequence. Use the sum of squared residuals within the segment as the segment cost and the median of the squared difference between adjacent slopes as the segment penalty. S3. The multi-scale PELT change point detection algorithm is used to recursively calculate the cumulative cost. The first term of the cumulative cost is taken as the predecessor point and the change point is read in reverse. The intervals with the highest residual decrease are retained from coarse to fine. The midpoint of the last interval is taken to form the segment boundary. S4. Fit the load slope according to the segment boundary and convert the main forming load range according to the pressure area ratio. Use the branch limit interval dynamic programming algorithm to generate the pressure supply value and switching position, and connect the interval where the pressure supply increment does not exceed the pressure rise of the force-increasing cavity. S5. Cut off the branches of the completed path with the time lower bound of the remaining displacement and the work lower bound of the remaining supply pressure to obtain the force-increasing and speed-up control path; when the slider reaches the switching position, cut off the oil supply to the speed-up chamber and supply oil to the force-increasing chamber according to the supply pressure value, and supply pressure according to the pressure difference of the next main forming load range until the forming end point returns to reset.

2. The intelligent control method for increasing force and speed of a large-tonnage high-speed precision punch press according to claim 1, characterized in that: S1 includes: S11. Before the first pressing section contacts the waste area, control the hydraulic servo valve to maintain the oil supply to the speed-up chamber, record the slider displacement and hydraulic pressure according to the sampling cycle, and record the first sampling point when the hydraulic pressure increases for two consecutive sampling cycles and the increase is not less than the resolution of the pressure sensor as the first pressing start point, and take the sampling point before the slider displacement reaches the first pressing stroke relative to the first pressing start point as the first pressing end point. S12. For the slider displacement and hydraulic pressure before the first pressing point, the no-load pressure line of the slider displacement is obtained by least squares fitting. Substitute each slider displacement from the first pressing point to the first pressing end point into the no-load pressure line to obtain the corresponding no-load pressure. Subtract the corresponding no-load pressure from the hydraulic pressure to obtain the plate pressure. S13. Multiply the plate pressure at each sampling point by the effective area of ​​the booster cylinder to obtain the plate load, and then write the plate load in ascending order of slider displacement to form the load displacement sequence of the current stroke.

3. The intelligent control method for increasing force and speed of a large-tonnage high-speed precision punch press according to claim 2, characterized in that: S2 includes: S21. Using the first displacement value of the load displacement sequence as the interpolation starting point, subtract the first displacement value from the last displacement value and divide by the number of sampling points minus one to obtain a uniform displacement step size. Distribute the corresponding impact force difference between adjacent slider displacements according to the distance ratio to form an equal displacement load sequence. S22. For the equal displacement load sequence, the initial extraction interval is two to the power of zero, and the power is increased successively. The impact force is read according to the first point index plus an integer multiple of the extraction interval. When the last point index cannot be divided by the extraction interval, the last point is added to form the load sequence of each scale. Then, the slope sequence of each scale is formed by dividing the difference between adjacent impact forces by the corresponding displacement difference. S23. Using the continuous slope between any two indices in the slope sequence of each scale as candidate segments, subtract the mean slope from each slope in the candidate segments and then sum the squares to obtain the segment cost. Arrange the squares of the differences between adjacent slopes in ascending order of value, take the middle term for odd-numbered terms and half the sum of the two middle terms for even-numbered terms to obtain the segment penalty for the corresponding scale.

4. The intelligent control method for increasing force and speed of a large-tonnage high-speed precision punch press according to claim 3, characterized in that: S3 includes: S31. For the slope sequence of each scale, divide the square of the impact force resolution by the square of the corresponding displacement difference, and then add the result obtained by multiplying the square of the impact force difference by the square of the displacement sensor resolution and dividing by the fourth power of the corresponding displacement difference to obtain the error of each slope. Then, divide the sum of the slope multiplied by the reciprocal of the error by the sum of the reciprocals of the error to obtain the weighted slope of the candidate segment. Finally, divide the square of each slope deviating from the weighted slope by the corresponding error and accumulate them to form the error constraint segment cost. S32. Using the multi-scale PELT change point detection algorithm, with the cumulative cost of the first position of the sequence as the negative segmentation penalty, calculate the cumulative cost of each predecessor point, the sum of the error constraint segmentation cost from the predecessor point to the endpoint and the segmentation penalty for each endpoint. Read the first predecessor point in ascending order of the calculation results and write it into the predecessor chain. At the same time, delete the predecessor point whose cumulative cost of the predecessor point and the sum of the error constraint segmentation cost from the predecessor point to the current endpoint reach the current cumulative cost. Continue until the last endpoint is recursively calculated and the change points of each scale are formed in reverse along the predecessor chain.

5. The intelligent control method for increasing force and speed of a large-tonnage high-speed precision punch press according to claim 4, characterized in that: S3 also includes: S33. Using each change point as the segmentation position, calculate the error constraint segmentation cost of the unsegmented interval minus the difference between the error constraint segmentation cost and the segmentation penalty of the two intervals after segmentation. Move the segmentation position from the change point to both sides one position at a time, converting the continuous positions where the difference is non-negative into change intervals. Then, take the coarse-scale change intervals as the current intersection according to the scale from coarse to fine. Only connect the fine-scale change intervals that intersect with the current intersection and whose interval width is shortened. Update the current intersection with the intersection to form a cross-scale interval path.

6. The intelligent control method for increasing force and speed of a large-tonnage high-speed precision punch press according to claim 5, characterized in that: S3 also includes: S34. Accumulate the reduction amount of each intersection and the segmentation difference loss in the cross-scale interval path in the directions from coarse to fine and from fine to coarse, respectively. Read the first path in ascending order of the accumulation results. When there is an intersection between two first paths, take the intersection as the boundary interval. When there is no intersection between two first paths, take the last level interval of the accumulated result in the preceding path as the boundary interval, and take the midpoint of the boundary interval to form the segment boundary.

7. The intelligent control method for increasing force and speed of a large-tonnage high-speed precision punch press according to claim 6, characterized in that: S4 includes: S41. For the load and error within each adjacent segment boundary, using the reciprocal of the error as the weight, solve the segment load slope when the partial derivatives of the weighted residual sum of squares with respect to the slope and intercept are both zero. Then, divide the last term of the absolute value of the load deviating from the fitted line in ascending order by the corresponding displacement span to obtain the slope correction amount. After adding and subtracting the slope correction amount from the segment load slope, convert it according to the ratio of the pressure area of ​​the main forming part to the pressure area of ​​the first pressing part to form the main forming load range. S42. Divide the upper boundary of each main forming load interval by the effective area of ​​the booster cylinder, and round up according to the hydraulic servo valve pressure resolution to obtain the initial supply pressure value. Increase the initial supply pressure value according to the hydraulic servo valve pressure resolution to the pump rated pressure to form candidate supply pressure values. Then, subtract the slider movement during the hydraulic servo valve response from the interval start point to obtain the switching position corresponding to each candidate supply pressure value.

8. The intelligent control method for increasing force and speed of a large-tonnage high-speed precision punch press according to claim 7, characterized in that: S4 includes: S43. Using the branch-bound interval dynamic programming algorithm, the control state is composed of interval index, candidate supply pressure value and oil supply state of the speed-up chamber. The state transition edge is established only when the difference between the next candidate supply pressure value and the previous candidate supply pressure value does not exceed the pressure increase obtained by dividing the pump flow rate by the interval passage time and dividing by the volume of the speed-up chamber and the bulk modulus of the oil, the next switching position is not earlier than the previous switching position, and the oil supply state of the speed-up chamber changes from oil supply to oil stop only once. S44. The path cost is formed by multiplying the candidate supply pressure value by the oil inlet volume of the boosting chamber and then adding the rated power of the main motor by the interval passage time. The lower bound of the remaining cost is formed by multiplying the starting supply pressure value of the remaining main forming load interval by the corresponding oil inlet volume and then adding the rated power of the main motor by the remaining bit and removing the result obtained by the rated speed of the speed-increasing cylinder. The branches whose sum of the current path cost and the lower bound of the remaining cost is not earlier than the completed path cost are deleted. When the control state to be expanded is empty, the candidate supply pressure value and switching position are read in reverse along the predecessor chain of the completed path to form the boosting speed control path.

9. The intelligent control method for increasing force and speed of a large-tonnage high-speed precision punch press according to claim 8, characterized in that: S5 includes: S51. Read each pressure value in reverse from the forming end point to the current main forming load interval. Subtract the achievable pressure increase segment by segment according to the passage time, pump flow rate, boosting chamber volume and oil volume modulus of adjacent main forming load intervals, and retain the pressure interval that can continuously reach the corresponding pressure value to form the remaining pressure domain. S52. For each branch entering the remaining pressure supply domain, the time lower bound is obtained by removing the remaining position at the rated speed of the speed-increasing cylinder. The work lower bound is obtained by multiplying the lower boundary of the pressure supply of each remaining main forming load interval by the corresponding oil inlet volume and then summing them. When the sum of the used time and the time lower bound is after the completed path passage time, or when the two are the same and the sum of the used work and the work lower bound is not earlier than the completed path work, the corresponding branch is deleted, and a force-increasing speed control path is formed along the precursor chain of the retained branch.

10. The intelligent control method for increasing force and speed of a large-tonnage high-speed precision punch press according to claim 9, characterized in that: S5 includes: S53. When the slider reaches the switching position in the force-increasing and speed-up control path, cut off the oil supply to the speed-up chamber and supply oil to the force-increasing chamber according to the corresponding supply pressure value. Divide the pressure difference obtained by subtracting the measured pressure from the target punching pressure of the corresponding main forming load range by the effective area of ​​the force-increasing cylinder to obtain the pressure correction amount. Then add the pressure correction amount to the supply pressure value of the next main forming load range to form the correction supply pressure value. S54. When the corrected pressure supply value is within the remaining pressure supply domain of the next main forming load range, continue to supply oil according to the corrected pressure supply value. When the corrected pressure supply value is outside the remaining pressure supply domain, re-execute the remaining branch search with the current slider displacement, the current force-increasing cavity pressure and the speed-increasing cavity oil stop state until the slider reaches the forming end point, then control the force-increasing cavity to depressurize and drive the slider to return to reset.

Images