Intelligent stamping device control method and system with adaptive adjustment function

Abstract

The present application relates to the technical field of stamping, and particularly relates to a control method and system of intelligent stamping equipment with self-adaptive adjusting function, which comprises the following steps: acquiring the spatial pose of a stamping workpiece and extracting the anisotropic characteristics of surface microtexture to generate multi-modal feeding characteristics; constructing an acoustic-force anisotropic coupling fingerprint in real time during stamping forming; predicting the critical driving force of necking propagation in a transient forming limit stress space based on the coupling fingerprint; triggering a nonlinear compliant switching strategy when a high-order buckling mode precursor of wrinkling is identified; generating a coordinated compensation instruction with the transient forming limit boundary as a constraint to guide the dangerous point strain state to a safe domain along a high-tolerance deformation corridor; constructing a damage tolerance model after forming and recursively feeding back to the multi-modal feeding characteristics and the switching strategy. Through micro-mechanical perception and multi-modal information fusion, the present application improves the instability early warning and timeliness, defect avoidance capability and process robustness of stamping of difficult-to-form materials.

Description

Technical Field

[0001] This invention relates to the field of stamping technology, and in particular to a control method and system for intelligent stamping equipment with adaptive adjustment function. Background Technology

[0002] Stamping is a core process for the efficient processing of sheet metal, and its level of intelligent control directly affects the forming quality and yield. In existing technologies, intelligent stamping equipment mostly adopts a technical approach that combines vision-guided positioning systems with force-position hybrid servo control. By monitoring the deviation of stamping force and displacement in real time during the forming process, and using preset thresholds or simple PID adjustments to compensate for press movements, it can address forming defects. However, conventional force-position feedback regulation exhibits significant limitations when dealing with difficult-to-form materials such as high-strength steel and lightweight alloys. On the one hand, it can only respond to macroscopic force and displacement deviations, failing to perceive deeper forming state information such as microscopic dislocation movement, texture evolution, and anisotropic flow during plastic deformation, resulting in a significant lag in the identification of early instability features such as wrinkling and necking. On the other hand, traditional control strategies often employ rigid switching or linear decay methods when switching execution modes, making it difficult to balance rapid response to instability and stability during the forming process, frequently leading to excessive thinning or even cracking of the sheet metal. Furthermore, the inspection data after part forming is usually only used for batch quality statistics, lacking a closed-loop feedback mechanism to trace back from finished product defects and iteratively optimize the feeding strategy and instability criteria, thus limiting the self-evolution capability of the control system. In summary, existing technologies are deficient in three key aspects: the information dimension of forming state perception, the compliance of instability response, and the adaptive iteration across batches. There is an urgent need for an intelligent stamping control method and system that can integrate multimodal information, have the ability to predict micro-damage, and achieve cross-batch self-learning. Summary of the Invention

[0003] This invention overcomes the shortcomings of the prior art and provides an intelligent stamping equipment control method and system with adaptive adjustment function.

[0004] To achieve the above objectives, the technical solution adopted by the present invention is as follows: The first aspect of this invention discloses a control method for an intelligent stamping equipment with adaptive adjustment function, comprising the following steps: Step 1: Obtain the spatial pose of the stamped workpiece and extract the anisotropic features of the surface micro-texture to generate a multimodal feeding feature that includes macro-geometric positioning data and micro-texture orientation distribution tensor. Step 2: During the stamping process, the servo torque signal and the acoustic emission spectrum envelope signal extracted by phase lock with the punch movement are collected in real time to construct an acoustic-mechanical anisotropic coupling fingerprint. Step 3: Based on the aforementioned acoustic-mechanical anisotropic coupling fingerprint, predict the critical driving force for necking propagation in the transient forming limit stress space considering the strain path. Step 4: When the precursor of the high-order buckling mode of wrinkling is identified, a nonlinear compliant switching strategy is triggered to switch the servo actuator from the position-force composite master control mode to the stiffness guided control mode. Step 5: In the stiffness-guided control mode, using the real-time updated transient forming limit boundary as a constraint, a coordinated compensation command is generated to guide the strain state of the danger point back to the safe domain along the preset high-tolerance deformation corridor trajectory. Step 6: After the forming is completed, a damage tolerance model is constructed based on the trajectory deviation data of the high tolerance deformation corridor, and the damage tolerance model is recursively fed back to the multimodal feeding feature and nonlinear compliant switching strategy to update the initial weights of the microtexture orientation distribution tensor and the identification threshold of the high-order buckling mode.

[0005] The second aspect of the present invention discloses an intelligent stamping equipment control system with adaptive adjustment function. The system includes a memory and a processor. The memory stores an intelligent stamping equipment control method program. When the intelligent stamping equipment control method program is executed by the processor, the steps of the intelligent stamping equipment control method described in any one of the present invention are implemented.

[0006] This invention addresses the technical deficiencies in the prior art and offers the following advantages: It introduces material anisotropy characteristics into the material feeding stage and quantifies them, integrating them into the control benchmark, thus basing forming decisions on an understanding of the material's intrinsic properties; during forming, it uses acoustic-mechanical coupling fingerprints instead of single force parameters for instability prediction, enabling early detection of early signs of necking propagation and high-order buckling, improving the timeliness of wrinkling and cracking warnings; employing a nonlinear compliant switching strategy and stiffness-guided control mode, it achieves a smooth transition rather than rigid switching after instability identification, effectively avoiding secondary forming defects caused by transient unloading; through high-tolerance deformation corridor planning and coordinated compensation spectrum, it compliantly guides the strain state at critical points back to the safe domain, balancing forming defect avoidance with maintaining plate thickness uniformity; it constructs a damage tolerance model using acoustic emission b-value attenuation recursion and provides feedback across batches, forming a self-evolving closed loop from finished product quality back to initial feeding and instability criteria, allowing the control strategy to continuously converge and optimize with each production batch, effectively improving the process robustness and yield of intelligent stamping for difficult-to-form materials. Attached Figure Description

[0007] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other embodiments can be obtained from these drawings without creative effort.

[0008] Figure 1 This is a flowchart of the control method of the present invention; Figure 2 This is a diagram of the control system architecture of the present invention. Detailed Implementation

[0009] To better understand the above-mentioned objectives, features, and advantages of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in these embodiments can be combined with each other.

[0010] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and therefore the scope of protection of the invention is not limited to the specific embodiments disclosed below.

[0011] like Figure 1 As shown, the first aspect of this invention discloses a control method for an intelligent stamping equipment with adaptive adjustment function, comprising the following steps: Step 1: Obtain the spatial pose of the stamped workpiece and extract the anisotropic features of the surface micro-texture to generate a multimodal feeding feature that includes macro-geometric positioning data and micro-texture orientation distribution tensor. Step 2: During the stamping process, the servo torque signal and the acoustic emission spectrum envelope signal extracted by phase lock with the punch movement are collected in real time to construct an acoustic-mechanical anisotropic coupling fingerprint. Step 3: Based on the aforementioned acoustic-mechanical anisotropic coupling fingerprint, predict the critical driving force for necking propagation in the transient forming limit stress space considering the strain path. Step 4: When the precursor of the high-order buckling mode of wrinkling is identified, a nonlinear compliant switching strategy is triggered to switch the servo actuator from the position-force composite master control mode to the stiffness guided control mode. Step 5: In the stiffness-guided control mode, using the real-time updated transient forming limit boundary as a constraint, a coordinated compensation command is generated to guide the strain state of the danger point back to the safe domain along the preset high-tolerance deformation corridor trajectory. Step 6: After the forming is completed, a damage tolerance model is constructed based on the trajectory deviation data of the high tolerance deformation corridor, and the damage tolerance model is recursively fed back to the multimodal feeding feature and nonlinear compliant switching strategy to update the initial weights of the microtexture orientation distribution tensor and the identification threshold of the high-order buckling mode.

[0012] Step one specifically includes: The multi-angle sequence lighting unit is activated to project polarized structured light onto the stamping workpiece located at the pre-calibrated station from multiple azimuth angles, and simultaneously triggers the multi-view vision system to acquire polarization state image sequences at different azimuth angles. The spatial pose of the workpiece is calculated from the polarization state image sequences to generate macroscopic geometric positioning data. The polarization state image sequence is analyzed pixel by pixel to calculate the surface normal azimuth angle and polarization phase angle of each pixel, and a normal-polarization fusion gradient field reflecting the micro-undulation characteristics of the workpiece surface is constructed. Extract the spatial distribution pattern of gradient direction in the normal-polarization fused gradient field, and use the reference coordinate system established by the macroscopic geometric positioning data as a reference to segment and identify microtexture primitives with consistent orientation, and assign each microtexture primitive a microtexture orientation vector describing its spatial orientation. All microtexture orientation vectors are statistically mapped and tensorized in the reference coordinate system to construct a microtexture orientation distribution tensor with the macroscopic geometric coordinate system as the reference. The macroscopic geometric positioning data and the microtexture orientation distribution tensor are encapsulated into a multimodal feeding feature package, and the optimal placement angle of the stamping workpiece is determined based on the matching relationship between the principal direction component of the microtexture orientation distribution tensor and the force direction of the stamping die.

[0013] Specifically, in the stamping and loading process, the workpiece is first transferred to a pre-calibrated station. At this point, a multi-angle sequential lighting unit positioned above the station is activated sequentially. Each lighting unit contains several polarized light sources arranged along different azimuth angles, projecting polarized structured light onto the workpiece surface in a time-division manner. A synchronously controlled multi-view vision system acquires corresponding polarization-state image sequences under illumination conditions at each azimuth angle, recording the light intensity response of each point on the workpiece surface under the current polarized illumination. Since the cameras in the multi-view vision system have known baseline geometric relationships, the workpiece's pose information in space can be directly calculated from the polarization-state image sequences from different viewpoints using the principle of multi-view parallax, thereby generating macroscopic geometric positioning data.

[0014] Because the scattering and reflection of polarized structured light at the micro-texture undulations of a workpiece surface are anisotropic, the polarization state of the outgoing light after surface modulation by incident light from different orientations will carry information reflecting the surface's micro-geometry. The analysis process involves fitting the light intensity variation curve of each pixel under illumination from multiple azimuth angles, extracting the surface normal azimuth angle and polarization phase angle of that point. The former characterizes the spatial orientation of the micro-region, while the latter reflects the steepness of the height undulation of the micro-region. Fusing these two in the pixel coordinate system forms a normal-polarization fused gradient field containing the micro-morphological gradient information of each pixel. For example, for aluminum alloy sheets with obvious rolling textures, they exhibit different polarization response characteristics in the rolling direction and perpendicular to the rolling direction. This difference will be reflected by the changes in the dominant direction of the normal azimuth angle and the amplitude of the polarization phase angle in the corresponding region of the gradient field.

[0015] Based on this, the spatial distribution pattern of gradient directions is extracted from the normal-polarization fused gradient field. Specifically, the gradient directions of each pixel in the gradient field are determined by neighborhood consistency. Pixels with significantly consistent and spatially adjacent gradient directions are grouped into the same micro-texture primitive. Each micro-texture primitive represents a segment of uniformly oriented micro-texture units on the workpiece surface. Each micro-texture primitive is assigned a micro-texture orientation vector describing its overall spatial orientation, determined by the statistical principal direction of the normal azimuth angles of all pixels within that primitive. The segmentation and labeling of all micro-texture primitives are based on the reference coordinate system established by the macroscopic geometric positioning data, ensuring that the position and orientation of each primitive are expressed in a unified physical coordinate system.

[0016] All microtexture orientation vectors are statistically mapped and tensorized in a reference coordinate system to construct a microtexture orientation distribution tensor. In practice, each microtexture orientation vector can be considered as a unit direction vector. A dextral operation is performed on all orientation vectors, and a weighted sum is calculated. The weights can be set based on the area or internal gradient consistency of each microtexture primitive. This yields a symmetric second-order tensor. The eigenvalues ​​of the tensor reflect the concentration of texture orientation, while the eigenvectors indicate the dominant orientation direction. This tensor uses a macroscopic geometric coordinate system as a reference, thus its orientation information directly corresponds to the force coordinate system of the mold. For example, through this construction process, a second-order tensor with clear physical meaning can be obtained, and the eigenvector corresponding to its largest eigenvalue represents the dominant orientation direction of the microtexture on the workpiece surface.

[0017] Finally, the aforementioned macroscopic geometric positioning data and the microtexture orientation distribution tensor are encapsulated into a multimodal loading feature package, serving as prior information for subsequent forming control. When determining the loading pose, the processor optimizes the solution based on the matching relationship between the principal direction component of this tensor and the preset force direction of the stamping die, calculating the placement posture that minimizes the angle between the principal direction of the workpiece's microtexture and the forming force direction of the die. This determines the optimal placement angle for the stamped workpiece. Through this angle adjustment, the anisotropic directional characteristics of the material can be incorporated into the reference frame of the control system in the early stages of forming.

[0018] Step two specifically includes: Using the punch movement cycle trigger sequence generated when the punch moves to the preset phase-locked window as the time reference, the servo torque sensor and acoustic emission sensor are synchronously triggered to acquire the servo torque time-domain sequence and acoustic emission waveform stream locked with the punching cycle, respectively. The acoustic emission waveform stream is segmented with the punch movement beat trigger sequence as the frame boundary. Time-frequency transformation is performed on each frame to extract the acoustic emission signal envelope that crosses the forming frequency band, thus forming a phase-locked acoustic emission spectrum envelope signal that reflects the spectral evolution characteristics of the material's micro-deformation. The servo torque time-domain sequence is subjected to sliding window filtering to separate the dynamic torque fluctuation component associated with the fluctuation of the material's plastic flow resistance. The phase-locked acoustic emission spectrum envelope signal and the dynamic torque fluctuation component are dynamically time-normalized and aligned using the punch motion beat trigger sequence. The aligned dynamic torque fluctuation components and the phase-locked acoustic emission spectrum envelope signal are correlated and mapped in the time domain and frequency domain to construct an acoustic-force time-varying coupling matrix with time as the index and torque fluctuation amplitude and acoustic emission radio frequency band energy distribution as the two axes. The acoustic-force time-varying coupling matrix is ​​defined as an acoustic-force anisotropic coupling fingerprint.

[0019] Specifically, at the start of forming, the control system uses the position of the punch reaching the preset phase-locked window as the trigger condition. The preset phase-locked window refers to a specific stroke interval before and after the punch contacts the workpiece during its downward movement. Whenever the punch reaches the start or end point of this window, the control system automatically generates a trigger pulse. All pulses together constitute the punch movement cycle trigger sequence. This sequence serves as a unified time reference, synchronously triggering the servo torque sensor and acoustic emission sensor to enter the acquisition state, thereby obtaining the servo torque time-domain sequence and acoustic emission waveform stream that are strictly locked to the stamping cycle, respectively.

[0020] For the acquired acoustic emission waveform stream, the control system segments it using the trigger sequence of the punch movement as the frame boundary, defining the acoustic emission waveform between two adjacent trigger pulses as an analysis frame. For each frame, a generalized S-transform is used to perform a time-frequency transformation to obtain the time-frequency distribution spectrum of the acoustic emission signal within the frame, from which multiple characteristic frequency bands spanning the forming frequency band can be identified. Based on this, the amplitude within each characteristic frequency band is enveloped along the frequency axis to obtain an acoustic emission signal envelope line that evolves with the frame sequence. The envelope line reflects the evolution characteristics of the micro-deformation spectrum of the material during continuous stamping. For example, when the dislocation density inside the material increases or local necking begins to initiate, the acoustic emission energy in a specific frequency band will show significant migration or accumulation, and this change will be directly reflected in the spectral envelope signal of the corresponding frame.

[0021] Synchronously, the control system performs sliding window filtering on the servo torque time-domain sequence. A low-pass filter window function with a suitable cutoff frequency is used to separate the trend component reflecting the steady-state deformation resistance of the material from the dynamic component reflecting the instantaneous fluctuations in plastic flow resistance in the servo torque signal. The separated dynamic torque fluctuation component is essentially contributed by the instantaneous resistance changes caused by the evolution of the material's internal microstructure (such as dislocation pile-up, local yielding, etc.). To further achieve time-domain alignment of the two heterogeneous signals, the precise timestamp of each pulse in the punch movement trigger sequence is used to perform dynamic time warping alignment on the phase-locked acoustic emission spectrum envelope signal and the dynamic torque fluctuation component, ensuring a precise correspondence between the mechanical fluctuation information and acoustic micro-features within the same forming cycle on the time axis.

[0022] After alignment, the aligned dynamic torque fluctuation components and the phase-locked loop acoustic emission spectral envelope signal are correlated and mapped in the time and frequency domains. A three-dimensional acoustic-force time-varying coupling matrix is ​​constructed with time as the first index axis, torque fluctuation amplitude as the second axis, and acoustic emission radio frequency band energy distribution as the third axis. The matrix records the coupling strength distribution between the servo torque fluctuation and the acoustic emission energy at each frequency band at each time step. Its time evolution trajectory characterizes the synergistic relationship between the macroscopic mechanical response and microscopic deformation mechanism of the material during the forming process. Since the anisotropy of the material leads to differences in the acoustic-force response modes under different loading directions, the coupling matrix naturally carries characteristic information reflecting anisotropic flow, and is therefore defined as an acoustic-force anisotropic coupling fingerprint.

[0023] Step three specifically includes: The directional characteristics of the acoustic emission main frequency energy migration trajectory and dynamic torque fluctuation component are decoupled from the acoustic-mechanical anisotropic coupling fingerprint, and the main axis direction of the energy distribution in the acoustic-mechanical time-varying coupling matrix is ​​identified as the acoustic-mechanical main response direction of the material anisotropic flow under the current strain path. The acoustic-force principal response direction is mapped to the transient forming limit stress space that takes into account the strain path, so as to construct an equivalent stress space distance field containing the current stress state point and the transient forming limit boundary. The transient forming limit stress space is updated in real time with the strain path ergodicity characterized by the acoustic emission principal frequency energy migration trajectory as a parameter. In the equivalent stress space distance field, the nearest neighbor failure point between the current stress state point and the transient forming limit boundary is searched along the acoustic-force principal response direction. Starting from the nearest neighbor failure point, a necking propagation mode vector pointing to the principal strain increment direction of the current stress state point is defined. The projection components of the necking propagation mode vector on each principal stress axis in the transient forming limit stress space are obtained. Combined with the energy proportion of the secondary axis orthogonal to the principal acoustic-force response direction in the acoustic-force time-varying coupling matrix, the anisotropic critical driving force that drives the necking to propagate along the anisotropic direction of the material is calculated. The anisotropic critical driving force is determined as the critical driving force for necking propagation, and the strain state corresponding to the nearest neighbor failure point is used as the instability critical criterion and output together.

[0024] It should be noted that the transient forming limit stress space refers to a multi-dimensional stress state space spanned by coordinate axes composed of stress tensor components and boundaries defined by the stress states of the sheet metal reaching local necking under different strain paths. Unlike the traditional equivalent strain forming limit diagram, this space uses stress components as descriptive parameters, thus exhibiting better adaptability to non-proportional loading paths. Furthermore, its boundaries are not static initial forming limit lines, but are adjusted in real-time based on the degree of strain path traversal reflected by the acoustic emission dominant frequency energy migration trajectory during forming; hence, it is called the "transient" forming limit stress space.

[0025] Specifically, feature decoupling is first performed from the acoustic-mechanical anisotropic coupling fingerprint, i.e., the acoustic-mechanical time-varying coupling matrix. The decoupling process involves two parallel extraction paths: first, tracking the movement of the energy weighted center frequency of each frequency band along the time axis of the acoustic-mechanical time-varying coupling matrix to obtain a trajectory of acoustic emission dominant frequency energy migration that evolves with the forming process. This trajectory reflects the transition of the material's micro-deformation dominant mechanism from dislocation slip to micropore aggregation; second, performing principal component analysis on the dynamic torque fluctuation components in the acoustic-mechanical time-varying coupling matrix to extract the direction of their maximum rate of change at each time point, which is used as the directional feature of the dynamic torque fluctuation components. Then, the principal axis direction of energy distribution in the acoustic-mechanical time-varying coupling matrix is ​​identified, i.e., the direction along which the coupling matrix energy is most concentrated and decays the slowest. This direction physically corresponds to the orientation where the material is most prone to plastic flow under the current strain path, and is thus defined as the acoustic-mechanical principal response direction.

[0026] Then, the boundary of the transient forming limit stress space is updated in real time using the current progression rate of the acoustic emission dominant frequency energy migration trajectory as the strain path ergodicity parameter, thus ensuring that the boundary conditions are always associated with the actual forming history. Based on the position of the acoustic-force principal response direction and the current stress state in the stress space, an equivalent stress spatial distance field is constructed, which gives the equivalent stress distance distribution in each direction from the current stress state point to the transient forming limit boundary. Next, in the equivalent stress spatial distance field, a search is performed along the acoustic-force principal response direction to determine the nearest neighbor failure point between the current stress state point and the transient forming limit boundary, i.e., the critical position where the stress state first touches the boundary in that principal response direction. Starting from this nearest neighbor failure point and pointing towards the principal strain increment direction of the current stress state point, a necking propagation mode vector is defined from the failure boundary to the danger zone. This vector describes the spatial path of the necking band expanding along the anisotropic orientation of the material once local necking occurs.

[0027] Next, the projection components of the necking propagation mode vector on each principal stress axis in the transient forming limit stress space are obtained. These projection components reflect the distribution of the necking propagation driving force on different stress components. Simultaneously, the secondary axis energy percentage orthogonal to the principal acoustic-mechanical response direction is extracted from the acoustic-mechanical time-varying coupling matrix. The secondary axis energy percentage characterizes the material's ability to resist deformation in non-dominant flow directions; a lower value indicates more significant flow anisotropy and a more concentrated driving force propagating along the principal response direction. The projection components and the secondary axis energy percentage are correlated and calculated to obtain the anisotropic critical driving force driving the necking propagation along the material's anisotropic direction. Finally, this anisotropic critical driving force is determined as the critical driving force for necking propagation, and the strain state corresponding to the nearest neighbor failure point is used as the instability critical criterion. Both are output to subsequent steps.

[0028] Step four specifically includes: The high-frequency components of the phase-locked acoustic emission spectrum envelope signal are extracted from the acoustic-mechanical anisotropic coupling fingerprint. The time evolution curve of the energy ratio of each frequency band is monitored. When the energy ratio of the high-order buckling mode characteristic frequency band representing the stress wave reflection in the thickness direction of the plate is alternately reversed, it is determined to be a precursor of the high-order buckling mode of wrinkling. At the instant the precursor of the high-order buckling mode of wrinkling is identified, the end stiffness value of the current position-force composite master control mode of the servo actuator is recorded. Starting from the end stiffness value, a nonlinear decaying compliant switching transition trajectory is generated based on the growth rate of the proportion of secondary axis energy orthogonal to the main acoustic-force response direction in the acoustic-force time-varying coupling matrix. Based on the compliant switching transition trajectory, the position loop gain of the servo actuator is synchronously modulated to generate a position loop stiffness attenuation envelope that is synchronous with the nonlinear attenuation, so that the end stiffness gradually decreases along the compliant switching transition trajectory. During the end stiffness decay process, the dynamic response spectrum of the anti-buckling force sensor is collected synchronously. Based on the degree of coupling between the dynamic response spectrum and the characteristic frequency band of the higher buckling mode, the decay rate of the position loop stiffness decay envelope is adjusted in real time to form a closed-loop compliant switching control. When the end stiffness decays to a preset stiffness guidance threshold, the servo actuator disengages from the position-force composite master control mode. Based on the final stiffness of the compliant switching transition trajectory and the current reaction force value of the anti-top force sensor, a stiffness guidance vector field is constructed with the virtual elastic constraint of the workpiece as the target. The servo actuator then switches to a stiffness guidance control mode with the stiffness guidance vector field as the control reference.

[0029] It should be noted that the "high-frequency component" refers to the spectral region in the phase-locked acoustic emission spectrum envelope signal located above the material's inherent forming frequency band. This region is highly sensitive to stress wave reflection and local buckling deformation in the thickness direction of the sheet material. The "higher-order buckling mode characteristic frequency band" is a specific frequency range in this high-frequency component corresponding to the sheet material excited under higher-order buckling modes. Its physical essence is a standing wave vibration response formed by repeated stress wave reflection in the thickness direction of the sheet material before wrinkling.

[0030] In specific implementation, the high-frequency components of the phase-locked acoustic emission spectrum envelope signal are extracted from the acoustic-mechanical anisotropic coupling fingerprint, and the energy proportion of each sub-band within this high-frequency band is calculated to establish a time evolution curve of the energy proportion of each band with the forming process. The energy proportion change pattern of the high-order buckling mode characteristic band in this curve is continuously monitored. Under normal forming conditions, the energy proportion of the high-order buckling mode characteristic band is relatively stable; when wrinkling precursors begin to appear locally in the sheet metal, the stress wave reflection in the thickness direction will cause the energy dominance of this characteristic band to alternate between the high-order mode and the adjacent low-order mode, which is manifested as an alternating reversal of the energy proportion of this band and the energy proportion of the adjacent band on the time evolution curve. For example, when the energy proportion of the characteristic band corresponding to the second buckling mode alternately surpasses the energy proportion of its adjacent first or third order band in more than two consecutive stamping cycles, the control system determines it as a high-order buckling mode precursor to wrinkling. The preset threshold for the number of alternating reversals of energy percentage can be set to two consecutive stamping cycles. This value balances recognition sensitivity and avoidance of false triggers.

[0031] Upon identifying the precursor to the higher-order buckling mode, the end-effector stiffness value exhibited by the current servo actuator under the position-force composite control mode is immediately recorded. The end-effector stiffness under the position-force composite control mode refers to the ratio of the force response increment to the position response increment of the servo actuator under this composite control law, determined jointly by the current control parameters and structural stiffness. Starting from this end-effector stiffness value, the control system generates an online nonlinear decaying compliant switching transition trajectory based on the growth rate of the secondary axis energy proportion orthogonal to the acoustic-force principal response direction in the aforementioned acoustic-force time-varying coupling matrix. The decay rate of this transition trajectory is positively correlated with the growth rate of the secondary axis energy proportion; that is, the more significant the anisotropy and the more urgent the buckling tendency, the faster the stiffness decays, thus directly integrating the degree of instability urgency into the switching dynamic process.

[0032] Next, the position loop gain of the servo actuator is synchronously modulated based on the compliant switching transition trajectory. In the position-force composite master control mode, the position loop gain directly determines the end effector stiffness level. By reducing the position loop gain along the compliant switching transition trajectory, a corresponding position loop stiffness attenuation envelope with the same nonlinear decay is generated, causing the end effector stiffness to gradually decrease along the compliant switching transition trajectory. This achieves a compliant transition from high stiffness to low stiffness, avoiding workpiece springback or increased vibration caused by instantaneous unloading.

[0033] It should also be noted that this embodiment does not perform stiffness decay in an open-loop manner, but rather introduces a counter-force sensor to form a closed-loop regulation. During the end-effector stiffness decay process, the control system synchronously acquires the dynamic response spectrum of the counter-force sensor and determines in real time the degree of coupling between this spectrum and the characteristic frequency band of the higher-order buckling mode. The degree of coupling refers to the correlation between the energy amplitude of the counter-force dynamic response spectrum at the characteristic frequency band of the higher-order buckling mode and the energy amplitude of the acoustic emission signal at the same frequency band. If the degree of coupling is high, it indicates that the excitation of the buckling mode is intensifying due to stiffness decay. At this time, the control system reduces the decay rate of the position loop stiffness decay envelope to slow down the pace of stiffness decrease and allow the modal excitation to gradually converge. Conversely, if the degree of coupling is maintained at a low level, the decay rate is allowed to remain or be appropriately increased, thereby forming a compliant switching control based on the buckling mode coupling state, so that the switching process always takes into account the dual constraints of instability prevention and forming continuity.

[0034] When the end-effector stiffness decays to a preset stiffness guidance threshold along the compliant transition trajectory, the control system causes the servo actuator to officially disengage from the position-force composite master control mode. The stiffness guidance threshold can be set to 1 / 3 to 1 / 5 of the initial end-effector stiffness value. This range ensures sufficient flexibility to adapt to workpiece deformation requirements while retaining basic force transmission capability. At the end of the mode switch, based on the final stiffness of the compliant transition trajectory and the current reaction force value of the anti-jacking force sensor, the range of virtual elastic constraints that the workpiece can withstand at the current position is determined, and a stiffness guidance vector field is constructed with the virtual elastic constraints of the workpiece as the target. This vector field defines the permissible displacement-force mapping relationship in each direction in the form of a spatial vector, and the servo actuator then switches to a stiffness guidance control mode with this stiffness guidance vector field as the control reference.

[0035] Step five specifically includes: Using the stiffness-guided vector field as the force control reference, the dangerous strain state corresponding to the nearest neighbor failure point is mapped to the real-time updated transient forming limit boundary to determine the current instability margin position of the dangerous point. Among them, "dangerous strain state corresponding to the nearest neighbor failure point" refers to the strain tensor state corresponding to the nearest neighbor failure point position determined in the equivalent stress space distance field in step three. This state is above the transient forming limit boundary, so once touched by the strain evolution trajectory of the current dangerous point, it can induce irreversible development of local necking or wrinkling. "Dangerous point" refers to the current stress state point, which is located in the region where the instability margin inside the transient forming limit boundary has been significantly reduced.

[0036] Using the transient forming limit boundary as a constraint, and combining the acoustic-mechanical principal response direction with the current energy proportion of the high-order buckling mode characteristic frequency band, a high-tolerance deformation corridor is generated in the equivalent stress spatial distance field, extending from the danger point to the safe region and along the direction of decreasing energy proportion of the high-order buckling mode characteristic frequency band. The high-tolerance deformation corridor is discretized into a series of intermediate target strain state points. Based on the direction of the stiffness guiding vector field and the secondary axis energy ratio weight in the acoustic-force time-varying coupling matrix, each intermediate target strain state point is transformed into a strain deviation compensation vector containing the increment of the blanking force, the increment of the anti-top force, and the punch attitude correction. Construct a coordinated compensation spectrum with the strain deviation compensation vector as input and the blank holder force adjustment channel and the anti-jacking force adjustment channel as execution objects, so that each channel performs synchronous coordinated output according to the stiffness allocation weight corresponding to the compliant switching transition trajectory; During the guidance process along the high-tolerance deformation corridor, the alternating reversal of the energy proportion of the high-order buckling mode characteristic frequency band in the phase-locked acoustic emission spectrum envelope signal is continuously monitored to see if it disappears. When the alternating reversal state is detected to have disappeared and the acoustic emission main frequency energy migration trajectory returns to the steady-state frequency band, it is determined that the danger point has fallen back to the safe domain.

[0037] In specific implementation, the stiffness guiding vector field constructed in step four is used as the force control benchmark to map the aforementioned dangerous strain state onto the real-time updated transient forming limit boundary. During the mapping process, along the current acoustic-force principal response direction of the dangerous point, the corresponding projection position of the dangerous strain state on the transient forming limit boundary is determined in the stress space. This determines the current instability margin position of the dangerous point, i.e., the stress space distance between the current stress state point and the aforementioned mapped projection position. Using the transient forming limit boundary as an insurmountable constraint boundary, and combining the current energy proportion of the acoustic-force principal response direction and the characteristic frequency band of the higher-order buckling modes, a high-tolerance deformation corridor is generated in the equivalent stress space distance field constructed in step three. The specific method for generating this corridor is as follows: starting from the danger point and ending at a predetermined safe zone within the transient forming limit boundary, the spatial distribution gradient of the energy proportion of the characteristic frequency band of the higher buckling modes is calculated in the equivalent stress spatial distance field. This decreasing energy proportion gradient direction is used as the guiding direction for path searching, ensuring that the path extends towards a lower energy proportion at each step. This guarantees that the strain state moving along the corridor always stays away from the stress conditions that trigger the higher buckling modes. For example, if the energy proportion of the characteristic frequency band of the higher buckling modes in a certain local area is about an order of magnitude higher than that of its neighboring areas, then the decreasing energy proportion gradient direction in that area will point in a direction that avoids the high energy proportion region. The high-tolerance deformation corridor is planned along this direction. The corridor generated in this way is essentially an optimal strain regression path in the equivalent stress spatial distance field that simultaneously satisfies the transient forming limit boundary constraint and the buckling mode energy proportion decreasing constraint.

[0038] After obtaining the high-tolerance deformation corridor, the corridor is discretized into a series of adjacent and equally spaced intermediate target strain state points in the stress space. For each intermediate target strain state point, based on the vector direction and amplitude of the stiffness guiding vector field at that point, and the anisotropy weight represented by the secondary axis energy proportion in the acoustic-force time-varying coupling matrix, the strain state point is transformed into a strain deviation compensation vector. This compensation vector specifically includes three executable dimensions: the increment of the blank holder force, the increment of the anti-jacking force, and the correction of the punch attitude. Subsequently, a coordinated compensation spectrum is constructed, using the strain deviation compensation vector as the input command and the blank holder force adjustment channel and the anti-jacking force adjustment channel as parallel execution objects. The coordinated compensation spectrum refers to the command generation mechanism that maps the components in the strain deviation compensation vector to each execution channel according to a certain weight distribution relationship. In this embodiment, the allocation weight obtained by each channel is determined according to the stiffness value corresponding to the compliant switching transition trajectory at the current moment as described in step four. That is, the blank holder force channel undertakes the compensation share inversely proportional to the current stiffness value, and the anti-jacking force channel undertakes the remaining share. This ensures that when the end stiffness is low and the servo actuator is already in a compliant state, the compensation task is completed more by adjusting the blank holder force, avoiding the application of excessive local concentrated loads to the workpiece.

[0039] During the gradual guidance along the high-tolerance deformation corridor, the alternating reversal state of the energy proportion of the characteristic frequency band of the high-order buckling mode in the phase-locked acoustic emission spectrum envelope signal is continuously monitored. When it is detected that this alternating reversal state has disappeared, that is, there is no longer an alternating transcendence of the energy proportion between the characteristic frequency band and the adjacent frequency band, and the energy migration trajectory of the acoustic emission main frequency has returned to the steady-state frequency band interval corresponding to the stable plastic flow, the control system determines that the strain state of the danger point has successfully fallen back to the safe region along the high-tolerance deformation corridor, and at this time the instability risk of this forming has been effectively eliminated.

[0040] Step six specifically includes: The trajectory deviation data of the high tolerance deformation corridor is obtained, the residual offset between the actual fall trajectory of the danger point under the action of each adjustment channel of the coordinated compensation spectrum and the high tolerance deformation corridor is extracted, and the steady-state regression rate of the acoustic emission main frequency energy migration trajectory in the entire forming cycle is simultaneously traced back. Based on the residual offset and the steady-state regression rate of the acoustic emission main frequency energy migration trajectory, an acoustic emission b-value attenuation recursive sequence is established to correlate the accumulation of microscopic damage and the deviation of macroscopic strain in the material. Based on the projection weight of the necking propagation mode vector on each stress principal axis, the acoustic emission b-value attenuation recursive sequence is mapped to a damage tolerance model characterizing the anisotropic damage tolerance of the plate. The damage sensitivity of each principal direction component of the microtexture orientation distribution tensor along the damage tolerance model is calculated, and a microtexture weight correction factor is generated to adjust the initial contribution weight of each microtexture orientation vector in the statistical mapping. Meanwhile, the energy proportion offset corresponding to the b-value attenuation recursion process of the high-order buckling mode feature frequency band is extracted from the damage tolerance model, and the energy proportion offset is converted into the high-order buckling mode recognition threshold update amount. The microtexture weight correction factor is fed back to the multimodal feeding feature to update the initial weight of the microtexture orientation distribution tensor; and the high-order buckling mode recognition threshold update is fed back to the nonlinear compliant switching strategy to update the recognition threshold of the high-order buckling mode, thus completing cross-batch adaptive iteration. When the fluctuation range of the damage tolerance model across multiple consecutive batches is lower than the preset threshold convergence criterion, the control strategy is determined to have converged.

[0041] Specifically, the actual fall trajectory of the danger point under the action of each adjustment channel of the coordinated compensation spectrum constructed in step five is extracted. The actual fall trajectory is obtained by back-calculating the measured execution amount of the blank holder force adjustment channel and the anti-top force adjustment channel and the measured value of the punch attitude to the stress space. The actual fall trajectory is compared point by point with the high tolerance deformation corridor planned in step five, and the spatial deviation distance between the two at corresponding positions in the stress space is calculated and defined as the residual offset. At the same time, the acoustic emission main frequency energy migration trajectory recorded throughout the entire forming cycle is traced back, and the rate at which the trajectory returns from the unsteady region to the steady-state frequency band during the fall of the danger point is calculated and defined as the steady-state return rate. Then, a recursive sequence of acoustic emission b-value attenuation related to the accumulation of microscopic damage and macroscopic strain deviation of the associated material is established. It should be noted that the "acoustic emission b-value" mentioned in this embodiment refers to the power exponent of acoustic emission amplitude-frequency distribution, which is derived from the Gutenberg-Richard relation in seismology. It is a characteristic parameter used to characterize the scale distribution of micro-damage events within a material. A higher b-value indicates a larger proportion of low-energy micro-damage events, suggesting the material is in a uniform damage stage; a lower b-value indicates an increased proportion of high-energy micro-damage events, suggesting the material has entered a damage localization stage. The "attenuation recursive sequence" refers to the recursive expression of the attenuation evolution of the b-value during the forming process. The b-value at each moment is recursively updated by the residual offset amplitude and the statistical results of the acoustic emission event amplitude distribution at the current moment. In specific implementation, the control system uses the projection components of the residual offset on each stress principal axis as a measure of macroscopic strain deviation, and the power exponent fitting result of the acoustic emission event amplitude-frequency distribution corresponding to each stress principal axis as a measure of micro-damage in that direction. The reciprocal of the steady-state regression rate is used as the time reference for the recursive step size, thereby establishing a b-value attenuation recursive sequence encompassing each stress principal axis direction along the forming time axis.

[0042] Subsequently, the control system, based on the projection weights of the necking propagation mode vectors on each stress principal axis, weights and fuses the b-value attenuation curves in each principal axis direction of the acoustic emission b-value attenuation recursive sequence, mapping them into a damage tolerance model that comprehensively characterizes the anisotropic damage tolerance of the plate material. This model, based on the b-value distribution in anisotropic coordinates, can reflect the differences in damage tolerance of the material in different directions. Next, the damage sensitivity of each principal direction component along the microtexture orientation distribution tensor is calculated from the damage tolerance model. Specifically, the b-value attenuation rate in each stress principal axis direction of the damage tolerance model is correlated and compared with the projection directions of each principal direction component of the microtexture orientation distribution tensor generated in step one on the corresponding stress principal axis to determine the correspondence between the texture principal direction and the damage-sensitive direction; along each texture principal direction, the normalized value of the b-value attenuation rate in that direction is taken as the damage sensitivity in that direction. Based on the damage sensitivity of each main texture direction, a micro-texture weight correction factor is generated to adjust the initial contribution weight of each micro-texture orientation vector in the statistical mapping described in step one. This correction factor is negatively correlated with the damage sensitivity, that is, the texture orientation weight in the damage sensitive direction is reduced, so as to guide the material anisotropy direction and the forming force direction to achieve a better match during the feeding stage.

[0043] Simultaneously, the control system extracts the energy proportion offset of the high-order buckling mode characteristic frequency band during the b-value decay recursion process from the damage tolerance model. This energy proportion offset refers to the cumulative shift of the energy proportion of the high-order buckling mode characteristic frequency band relative to its initial value at each time point in the b-value decay recursion sequence. This energy proportion offset is converted into an update amount for the high-order buckling mode identification threshold. A positive correlation can be established between the two; that is, a larger energy proportion offset indicates that the identification of high-order buckling modes in the current batch is either too conservative or too aggressive, and the identification threshold is adjusted accordingly.

[0044] After completing the above calculations, the microtexture weight correction factor is fed back to the multimodal feeding feature in step one to update the initial weight of the microtexture orientation distribution tensor, so that the feeding angle optimization of the next batch is based on the understanding of the damage characteristics of the current batch; at the same time, the update amount of the high-order buckling mode recognition threshold is fed back to the nonlinear compliant switching strategy described in step four to update the recognition threshold of the high-order buckling mode, so that the mode switching decision of the next batch is more in line with the actual wrinkling characteristics of the material under the current process conditions.

[0045] The aforementioned feedback process iterates continuously with each batch. When the fluctuation range of the b-value distribution in the damage tolerance model across multiple consecutive batches is lower than the preset threshold convergence criterion, it indicates that the control system has fully adapted to the damage evolution characteristics under the current material and process conditions, and the control strategy is determined to have converged. The threshold convergence criterion can be set as the relative change in the b-value attenuation rate along each principal stress axis in the damage tolerance model across three consecutive batches being less than a preset proportion, such as 5%. This value can be configured based on the tolerance of the actual process for stability and the number of iteration cycles.

[0046] In a preferred embodiment of the present invention, it further includes: Using the independent components of the microtexture orientation distribution tensor as nodes and the workpiece surface established by the macroscopic geometric positioning data as the manifold basis, the direction of the plastic strain increment borne by each node at the discrete target point of the high tolerance deformation corridor is mapped to the anisotropic rheological traction vector on the manifold basis, thus constructing a plastic strain energy transport network composed of all nodes and their adjacency relationships. The proportion of secondary axis energy orthogonal to the necking propagation mode vector in the acoustic-mechanical time-varying coupling matrix is ​​extracted and used as the strain energy diffusion fugacity of each side in the plastic strain energy transport network. Each node is constrained and calibrated according to the transient forming limit boundary to form a non-equilibrium steady-state transport field containing node potential and edge fugacity. In the non-equilibrium steady-state transport field, the current working node of the edge pressure force adjustment channel and the anti-top force adjustment channel is taken as the source point, and the first target node in the safe domain is taken as the trap point. Particle transport is simulated along the negative direction of the edge fugacity gradient, the frequency of each edge being crossed is recorded, and a strain energy transport streamline density distribution map is generated. From the strain energy transport streamline density distribution diagram, the main path that runs through the source point and the sink point and has the maximum and minimum streamline density is identified. The node sequence connected by the main path is defined as the plastic flexibility ridge. The plastic flexibility ridge is the main channel for coordinated and compensated force energy transmission. Along the plastic flexibility ridge, the anisotropic rheological traction vector at each node and the strain energy diffusion fugacity of the corresponding edge are vector synthesized to obtain the compensation deflection angle at each node. The compensation deflection angle is then decomposed into the coordination compensation increment coefficient of the blank holder force adjustment channel and the anti-jacking force adjustment channel to form the minimum energy distribution path of the coordination compensation spectrum of blank holder force and anti-jacking force.

[0047] It should be noted that the manifold substrate refers to a continuous surface model on which a differential geometric structure is defined, using the three-dimensional curved surface of the workpiece established by macroscopic geometric positioning data as the geometric carrier; each point on the surface model inherits each independent component of the microtexture orientation distribution tensor, thereby endowing the manifold substrate with anisotropic material properties.

[0048] In implementation, each independent component of the microtexture orientation distribution tensor is used as a node, and the workpiece surface established by the macroscopic geometric positioning data is used as the manifold basis to construct a node network on the workpiece surface. For each node, the discrete intermediate target strain state point of the high-tolerance deformation corridor at the corresponding position of the node is extracted to obtain the direction of the plastic strain increment experienced by the node in the current forming stage. The direction of the plastic strain increment is mapped to an anisotropic rheological traction vector acting on the node by projecting the local tangent plane at the node on the manifold basis and the tensor action, which characterizes the directional traction effect of the material microtexture on the plastic flow at the node. All nodes and their spatial adjacency relationships on the manifold basis are connected as edges, each node carries the rheological traction vector, and each edge represents the strain energy transfer channel between nodes, thereby constructing a complete plastic strain energy transport network.

[0049] Then, the secondary axis energy percentage, orthogonal to the necking propagation mode vector, is extracted from the acoustic-mechanical time-varying coupling matrix. This secondary axis energy percentage reflects the material's tendency to dissipate energy in the non-dominant flow direction and is assigned to each edge of the plastic strain energy transport network as the strain energy diffusion fugacity of each edge. Simultaneously, each node is constrained and calibrated according to the transient forming limit boundary, classifying nodes into three categories: safe nodes, critical nodes, and dangerous nodes. This forms a non-equilibrium steady-state transport field containing node potential and edge fugacity. The non-equilibrium steady-state transport field describes the potential energy and resistance distribution of plastic strain energy directionally transported from the dangerous region to the safe region under the current forming state.

[0050] In the non-equilibrium steady-state transport field, the particle transport process is simulated along the negative direction of the edge fugacity gradient (i.e., the direction in which strain energy dissipation resistance decreases) using the current active nodes of the edge pressure force adjustment channel and the counter-pressure force adjustment channel as the source point and the first target node at the end of the high-tolerance deformation corridor within the safety domain as the sink point. During the simulation, the frequency of particle traversal of each edge in the transport network is recorded, generating a strain energy transport streamline density distribution map. This map reflects the statistical tendency of preferential transport of plastic strain energy along different paths. Based on this, the main path with the maximum and minimum streamline density, which runs through the source and sink points, is identified from the strain energy transport streamline density distribution map. The maximum and minimum streamline density refers to the minimum streamline density among all possible paths, meaning that this path is the channel with the widest bottleneck and the lowest overall transport resistance during the transport process. The sequence of nodes connected by this main path is defined as the plastic flexibility ridge, which is the main channel for force energy transfer during the coordinated compensation process.

[0051] Finally, along the plastic flexibility ridge, the anisotropic rheological traction vector at each node and the strain energy diffusion fugacity of the corresponding edge are vector synthesized. During the synthesis process, the rheological traction vector provides the directional reference, and the diffusion fugacity provides the amplitude correction. The synthesis result is the compensation deflection angle at each node. The control system projects this compensation deflection angle onto the blank holder force adjustment direction and the anti-jacking force adjustment direction, respectively, to obtain the coordinated compensation increment coefficient of each channel. This forms the minimum energy distribution path along the plastic flexibility ridge for the coordinated compensation spectrum of blank holder force and anti-jacking force, ensuring that the execution of the coordinated compensation command physically comes at the cost of minimal strain energy dissipation, thereby guiding the strain state at the critical point towards the safe domain.

[0052] In a preferred embodiment of the present invention, it further includes: Extract the rigid rotation component of the microtexture orientation distribution tensor that represents the orientation principal axis from the extreme decomposition of the microtexture orientation distribution tensor, and spatially register it with the local tangent vector at the discrete target point of the high tolerance deformation corridor. Solve to obtain the microtexture compliance quaternion at each discrete target point that makes the material slip direction conform to the microtexture direction, and form an initial discrete attitude sequence. Among them, the extreme decomposition of the microtexture orientation distribution tensor refers to performing extreme decomposition operation on the microtexture orientation distribution tensor (a second-order symmetric tensor) constructed in step one, decomposing it into the product of an orthogonal tensor representing pure rigid rotation and a symmetric positive definite tensor representing pure tensile deformation. The orthogonal tensor is the rigid rotation component of the microtexture, and its physical meaning is the rigid body rotation change of the main orientation direction of the material microtexture in space at different positions. The initial discrete attitude sequence is mapped to a unit quaternion hypersphere, the logarithmic mapping distance between adjacent microtexture conforming quaternions is calculated, and based on the cumulative rate of change of the logarithmic mapping distance along the ordinal number of the discrete target point, a singularity metric functional for quantifying the severity of attitude jumps is constructed, and attitude discontinuities exceeding a preset smoothness tolerance threshold are marked. Using the attitude discontinuity points as boundaries, the initial discrete attitude sequence is adaptively divided into multiple attitude smoothing clusters. Within each attitude smoothing cluster, quaternion geodesics are constructed with the first and last microtexture compliant quaternions as endpoints. The arc length parameter of the quaternion geodesics is weighted by the acoustic-mechanical principal response direction in the acoustic-mechanical anisotropic coupled fingerprint to obtain anisotropic constrained geodesic skeleton. For the connection region between adjacent attitude smoothing clusters, quaternion spherical linear interpolation is performed with the endpoints of the anisotropic constraint geodesic skeleton as constraints. The singularity metric functional is used as the objective function to adjust the interpolation step size until the quaternion change gradient of the connection region is less than the smoothing tolerance threshold. This yields the singularity-resolved attitude connection segment, which is then spliced ​​with the anisotropic constraint geodesic skeleton inside each attitude smoothing cluster to generate a complete non-singularity attitude transition sequence. The three-dimensional attitude adjustment increment relative to the mold reference system is calculated point by point from the non-singularity attitude transition sequence. The attitude adjustment increment is decomposed into punch attitude correction amount and blank holder normal tilt angle correction amount, and injected as feedforward command into the blank holder force and anti-top force coordination compensation spectrum to realize the parallel and coordinated execution of attitude transition and force control.

[0053] It should be noted that the rigid rotation component of the microtexture is extracted from the extreme decomposition of the microtexture orientation distribution tensor. Simultaneously, local tangent vectors are obtained at each discrete target point in the high-tolerance deformation corridor. These local tangent vectors are defined within the tangent space at each point on the workpiece surface manifold. The rigid rotation component of the microtexture is spatially registered with the corresponding local tangent vector at the discrete target point; that is, the optimal spatial rotation that makes both oriented in the local coordinate system is obtained. This optimal rotation is expressed as a unit quaternion, called the microtexture compliance quaternion, which represents the optimal posture expression that makes the material slip direction and the microtexture orientation conform at the discrete target point. After solving point-by-point along all discrete target points of the high-tolerance deformation corridor, an initial discrete posture sequence is formed.

[0054] Subsequently, the initial discrete attitude sequence is mapped to a unit quaternion hypersphere. A unit quaternion hypersphere is a hypersphere with radius 1 in a four-dimensional space composed of all unit quaternions, with each attitude corresponding to a point on the hypersphere. The logarithmic mapping distance between adjacent microtexture conforming quaternions is calculated. The logarithmic mapping distance is defined as the geodesic distance between adjacent attitudes on the hypersphere, reflecting the smoothness of attitude changes. Based on the cumulative rate of change of the logarithmic mapping distance along the ordinal number of the discrete target point, a singularity metric functional is constructed to quantify the severity of attitude jumps. When the rate of change of the logarithmic mapping distance at a certain location significantly deviates from the global mean, it indicates that there is attitude singularity or directional discontinuity at that location. A preset smoothness tolerance threshold can be set as the mean of the logarithmic mapping distance between adjacent attitudes plus twice the standard deviation; points exceeding this threshold are marked as attitude discontinuities.

[0055] Using the attitude discontinuities as boundaries, the initial discrete attitude sequence is adaptively divided into multiple attitude smoothing clusters. Within each attitude smoothing cluster, attitude changes are relatively smooth, with no singularity issues. Using the first and last microtexture compliant quaternions of each cluster as endpoints, quaternion geodesics connecting the two endpoints are constructed on a unit quaternion hypersphere, representing the shortest path curve between two points on the hypersphere. Simultaneously, the arc length parameters of the quaternion geodesics are weighted by the acoustic-mechanical principal response directions in the acoustic-mechanical anisotropic coupling fingerprint, resulting in denser arc length sampling for components along the principal response directions and sparser arc length sampling for components in orthogonal directions, thus obtaining an anisotropic constrained geodesic skeleton that preserves the tensor directionality characteristics of microtexture orientation. For the connection regions between adjacent attitude smoothing clusters, i.e., singular segments with attitude discontinuities, quaternion spherical linear interpolation is performed using the endpoints of the anisotropic constrained geodesic skeletons of adjacent clusters as constraints. During interpolation, the singularity metric functional is used as the objective function for feedback. The singularity metric value at each point in the interpolation sequence is monitored, and the interpolation step size is dynamically adjusted. When the singularity metric value is too high, the step size is reduced; when it is too low, the step size is appropriately increased, until the quaternion change gradient of all interpolation points in the connection region is less than the smoothing tolerance threshold. The singularity-resolving attitude connection segment obtained in this way is spliced ​​with the anisotropic constraint geodesic skeleton inside each attitude smoothing cluster to generate a complete singularity-free attitude transition sequence. The three-dimensional attitude adjustment increment relative to the mold reference system is calculated point by point from the singularity-free attitude transition sequence. The relative rotation at each point is obtained by performing a difference operation between each attitude quaternion and the reference attitude of the mold reference system. This rotation is decomposed into three components: the deflection angle around the punch feed direction, the torsion angle around the plate surface normal, and the tilt angle around the transverse axis. Among them, the deflection angle and the torsion angle are combined into the punch attitude correction amount, and the transverse axis tilt angle is decomposed into the pressure ring normal tilt angle correction amount. Both are injected as feedforward commands into the coordinated compensation spectrum of the edge pressure force and the counter-pressure force described in step five, so as to realize the parallel and coordinated execution of attitude transition and force energy regulation on the same spatiotemporal scale, and further improve the smoothness and effectiveness of strain state fallback during the high tolerance deformation corridor guidance process.

[0056] like Figure 2 As shown, the second aspect of the present invention discloses an intelligent stamping equipment control system with adaptive adjustment function. The system includes a memory and a processor. The memory stores an intelligent stamping equipment control method program. When the intelligent stamping equipment control method program is executed by the processor, the steps of the intelligent stamping equipment control method described in any one of the claims are implemented.

[0057] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. A control method for an intelligent stamping equipment with adaptive adjustment function, characterized in that, Includes the following steps: Step 1: Obtain the spatial pose of the stamped workpiece and extract the anisotropic features of the surface micro-texture to generate a multimodal feeding feature that includes macro-geometric positioning data and micro-texture orientation distribution tensor. Step 2: During the stamping process, the servo torque signal and the acoustic emission spectrum envelope signal extracted by phase lock with the punch movement are collected in real time to construct an acoustic-mechanical anisotropic coupling fingerprint. Step 3: Based on the aforementioned acoustic-mechanical anisotropic coupling fingerprint, predict the critical driving force for necking propagation in the transient forming limit stress space considering the strain path. Step 4: When the precursor of the high-order buckling mode of wrinkling is identified, a nonlinear compliant switching strategy is triggered to switch the servo actuator from the position-force composite master control mode to the stiffness guided control mode. Step 5: In the stiffness-guided control mode, using the real-time updated transient forming limit boundary as a constraint, a coordinated compensation command is generated to guide the strain state of the danger point back to the safe domain along the preset high-tolerance deformation corridor trajectory. Step 6: After the forming is completed, a damage tolerance model is constructed based on the trajectory deviation data of the high tolerance deformation corridor, and the damage tolerance model is recursively fed back to the multimodal feeding feature and nonlinear compliant switching strategy to update the initial weights of the microtexture orientation distribution tensor and the identification threshold of the high-order buckling mode.

2. The control method according to claim 1, characterized in that, Step one specifically includes: The multi-angle sequence lighting unit is activated to project polarized structured light onto the stamping workpiece located at the pre-calibrated station from multiple azimuth angles, and simultaneously triggers the multi-view vision system to acquire polarization state image sequences at different azimuth angles. The spatial pose of the workpiece is calculated from the polarization state image sequences to generate macroscopic geometric positioning data. The polarization state image sequence is analyzed pixel by pixel to calculate the surface normal azimuth angle and polarization phase angle of each pixel, and a normal-polarization fusion gradient field reflecting the micro-undulation characteristics of the workpiece surface is constructed. Extract the spatial distribution pattern of gradient direction in the normal-polarization fused gradient field, and use the reference coordinate system established by the macroscopic geometric positioning data as a reference to segment and identify microtexture primitives with consistent orientation, and assign each microtexture primitive a microtexture orientation vector describing its spatial orientation. All microtexture orientation vectors are statistically mapped and tensorized in the reference coordinate system to construct a microtexture orientation distribution tensor with the macroscopic geometric coordinate system as the reference. The macroscopic geometric positioning data and the microtexture orientation distribution tensor are encapsulated into a multimodal feeding feature package, and the optimal placement angle of the stamping workpiece is determined based on the matching relationship between the principal direction component of the microtexture orientation distribution tensor and the force direction of the stamping die.

3. The control method according to claim 1, characterized in that, Step two specifically includes: Using the punch movement cycle trigger sequence generated when the punch moves to the preset phase-locked window as the time reference, the servo torque sensor and acoustic emission sensor are synchronously triggered to acquire the servo torque time-domain sequence and acoustic emission waveform stream locked with the punching cycle, respectively. The acoustic emission waveform stream is segmented with the punch movement beat trigger sequence as the frame boundary. Time-frequency transformation is performed on each frame to extract the acoustic emission signal envelope that crosses the forming frequency band, thus forming a phase-locked acoustic emission spectrum envelope signal that reflects the spectral evolution characteristics of the material's micro-deformation. The servo torque time-domain sequence is subjected to sliding window filtering to separate the dynamic torque fluctuation component associated with the fluctuation of the material's plastic flow resistance. The phase-locked acoustic emission spectrum envelope signal and the dynamic torque fluctuation component are dynamically time-normalized and aligned using the punch motion beat trigger sequence. The aligned dynamic torque fluctuation components and the phase-locked acoustic emission spectrum envelope signal are correlated and mapped in the time domain and frequency domain to construct an acoustic-force time-varying coupling matrix with time as the index and torque fluctuation amplitude and acoustic emission radio frequency band energy distribution as the two axes. The acoustic-force time-varying coupling matrix is ​​defined as an acoustic-force anisotropic coupling fingerprint.

4. The control method according to claim 1, characterized in that, Step three specifically includes: The acoustic emission main frequency energy migration trajectory and the directional characteristics of the dynamic torque fluctuation component are decoupled from the acoustic-mechanical anisotropic coupling fingerprint. The main axis direction of the energy distribution in the acoustic-mechanical time-varying coupling matrix is ​​identified as the acoustic-mechanical main response direction of the material anisotropic flow under the current strain path. The acoustic-force principal response direction is mapped to the transient forming limit stress space that takes into account the strain path, so as to construct an equivalent stress space distance field containing the current stress state point and the transient forming limit boundary. The transient forming limit stress space is updated in real time with the strain path ergodicity characterized by the acoustic emission principal frequency energy migration trajectory as a parameter. In the equivalent stress space distance field, the nearest neighbor failure point between the current stress state point and the transient forming limit boundary is searched along the acoustic-force principal response direction. Starting from the nearest neighbor failure point, a necking propagation mode vector pointing to the principal strain increment direction of the current stress state point is defined. The projection components of the necking propagation mode vector on each principal stress axis in the transient forming limit stress space are obtained. Combined with the energy ratio of the secondary axis orthogonal to the principal acoustic-force response direction in the acoustic-force time-varying coupling matrix, the anisotropic critical driving force that drives the necking to propagate along the anisotropic direction of the material is calculated. The anisotropic critical driving force is determined as the critical driving force for necking propagation, and the strain state corresponding to the nearest neighbor failure point is used as the instability critical criterion and output together.

5. The control method according to claim 1, characterized in that, Step four specifically includes: The high-frequency components of the phase-locked acoustic emission spectrum envelope signal are extracted from the acoustic-mechanical anisotropic coupling fingerprint. The time evolution curve of the energy ratio of each frequency band is monitored. When the energy ratio of the high-order buckling mode characteristic frequency band representing the stress wave reflection in the thickness direction of the plate is alternately reversed, it is determined to be a precursor of the high-order buckling mode of wrinkling. At the instant the precursor of the high-order buckling mode of wrinkling is identified, the end stiffness value of the current position-force composite master control mode of the servo actuator is recorded. Starting from the end stiffness value, a nonlinear decaying compliant switching transition trajectory is generated based on the growth rate of the proportion of secondary axis energy orthogonal to the main acoustic-force response direction in the acoustic-force time-varying coupling matrix. Based on the compliant switching transition trajectory, the position loop gain of the servo actuator is synchronously modulated to generate a position loop stiffness attenuation envelope that is synchronous with the nonlinear attenuation, so that the end stiffness gradually decreases along the compliant switching transition trajectory. During the end stiffness decay process, the dynamic response spectrum of the anti-buckling force sensor is collected synchronously. Based on the degree of coupling between the dynamic response spectrum and the characteristic frequency band of the higher buckling mode, the decay rate of the position loop stiffness decay envelope is adjusted in real time to form a closed-loop compliant switching control. When the end stiffness decays to a preset stiffness guidance threshold, the servo actuator disengages from the position-force composite master control mode. Based on the final stiffness of the compliant switching transition trajectory and the current reaction force value of the anti-top force sensor, a stiffness guidance vector field is constructed with the virtual elastic constraint of the workpiece as the target. The servo actuator then switches to a stiffness guidance control mode with the stiffness guidance vector field as the control reference.

6. The control method according to claim 5, characterized in that, Step five specifically includes: Using the stiffness-guided vector field as the force control reference, the dangerous strain state corresponding to the nearest neighbor failure point is mapped to the real-time updated transient forming limit boundary to determine the current instability margin position of the dangerous point. Using the transient forming limit boundary as a constraint, and combining the acoustic-mechanical principal response direction with the current energy proportion of the characteristic frequency band of the higher buckling mode, a high-tolerance deformation corridor is generated in the equivalent stress spatial distance field, extending from the danger point to the safe region and along the gradient direction of the decreasing energy proportion of the characteristic frequency band of the higher buckling mode. The high-tolerance deformation corridor is discretized into a series of intermediate target strain state points. Based on the direction of the stiffness guiding vector field and the secondary axis energy ratio weight in the acoustic-force time-varying coupling matrix, each intermediate target strain state point is transformed into a strain deviation compensation vector containing the increment of the blanking force, the increment of the anti-top force, and the punch attitude correction. A coordinated compensation spectrum is constructed with the strain deviation compensation vector as input and the blank holder force adjustment channel and the anti-jacking force adjustment channel as execution objects, so that each channel performs synchronous coordinated output according to the stiffness allocation weight corresponding to the compliant switching transition trajectory; During the guidance process along the high-tolerance deformation corridor, the alternating reversal of the energy proportion of the high-order buckling mode characteristic frequency band in the phase-locked acoustic emission spectrum envelope signal is continuously monitored to see if it disappears. When the alternating reversal state is detected to have disappeared and the acoustic emission main frequency energy migration trajectory returns to the steady-state frequency band, it is determined that the danger point has fallen back into the safe domain.

7. The control method according to claim 1, characterized in that, Step six specifically includes: The trajectory deviation data of the high tolerance deformation corridor is obtained, the residual offset between the actual fall trajectory of the danger point under the action of each adjustment channel of the coordinated compensation spectrum and the high tolerance deformation corridor is extracted, and the steady-state regression rate of the acoustic emission main frequency energy migration trajectory in the entire forming cycle is simultaneously traced back. Based on the residual offset and the steady-state regression rate of the acoustic emission main frequency energy migration trajectory, an acoustic emission b-value attenuation recursive sequence is established to correlate the accumulation of microscopic damage and the deviation of macroscopic strain in the material. Based on the projection weight of the necking propagation mode vector on each stress principal axis, the acoustic emission b-value attenuation recursive sequence is mapped to a damage tolerance model characterizing the anisotropic damage tolerance of the plate. The damage sensitivity of each principal direction component of the microtexture orientation distribution tensor along the damage tolerance model is calculated, and a microtexture weight correction factor is generated to adjust the initial contribution weight of each microtexture orientation vector in the statistical mapping. Meanwhile, the energy proportion offset corresponding to the b-value attenuation recursion process of the high-order buckling mode feature frequency band is extracted from the damage tolerance model, and the energy proportion offset is converted into the high-order buckling mode recognition threshold update amount. The microtexture weight correction factor is fed back to the multimodal feeding feature to update the initial weight of the microtexture orientation distribution tensor; and the high-order buckling mode recognition threshold update is fed back to the nonlinear compliant switching strategy to update the recognition threshold of the high-order buckling mode, thus completing cross-batch adaptive iteration. When the fluctuation range of the damage tolerance model across multiple consecutive batches is lower than the preset threshold convergence criterion, the control strategy is determined to have converged.

8. A smart stamping equipment control system with adaptive adjustment function, characterized in that, The system includes a memory and a processor. The memory stores a program for controlling an intelligent stamping equipment. When the program for controlling an intelligent stamping equipment is executed by the processor, the steps of the intelligent stamping equipment control method as described in any one of claims 1 to 7 are implemented.

Images