A method for multi-station continuous bending and forming of rigid PC sheets and precision compensation control
By using in-situ online detection and ternary digital twin construction, combined with stress-oriented decoupling planning and three-temperature gradient preheating, the problem of insufficient precision compensation in multi-station continuous bending of rigid PC sheets was solved, achieving high precision and stable bending effect, and reducing stress whitening and thermal deformation defects.
Patent Information
- Application Number
- CN202611463301.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-09-20
- Publication Date
- 2026-11-17
AI Technical Summary
In the existing multi-station continuous bending process of rigid PC sheets, the precision compensation method does not fully consider the viscoelastic deformation characteristics, which leads to fluctuations in material properties between batches, causing angle deviations, stress residues and thermal deformation accumulation, making it difficult to meet the requirements of high-precision processing. In addition, the cold bending and hot bending processes have quality problems such as stress whitening, cracking and large heat-affected zones, and cannot dynamically respond to environmental changes and mold wear.
By acquiring thickness and stress distribution data of the sheet material through in-situ online detection, a ternary digital twin is constructed to predict the rebound amount and stress evolution law. Stress-oriented decoupling planning and three-temperature gradient preheating are adopted, combined with dual closed-loop drive and cascade compensation mechanism to achieve accurate error tracing and dynamic compensation, thereby reducing thermal deformation and stress concentration.
It significantly improves the accuracy of bending angle prediction for rigid PC sheets, reduces cold bending stress whitening and hot bending defects, shrinks the heat-affected zone, improves molding quality and dimensional stability, and achieves dynamic compensation for adaptive material fluctuations and mold wear.
Smart Images

Figure CN123104041A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of sheet metal bending and forming technology, specifically a method for multi-station continuous bending and forming of rigid PC sheets and precision compensation control. Background Technology
[0002] Rigid polycarbonate (PC) sheets are widely used in electromechanical protection, architectural lighting, and automotive interiors. Complex structural components typically require multiple bending processes to form. Currently, the industry mainly divides rigid PC sheet bending into two processes: cold bending and hot bending. While multi-station continuous bending can significantly improve production efficiency, the following technical problems still exist: Existing bending accuracy compensation methods mostly directly follow the springback compensation logic of metal sheets, without fully considering the viscoelastic deformation characteristics of PC sheets. The springback amount is affected by multiple factors such as temperature, bending rate, and holding time. The compensation method with fixed over-bending amount can only cover a single working condition. The fluctuation of material properties between batches can easily cause large angle deviations, making it difficult to meet the requirements of high-precision processing.
[0003] During continuous bending, the residual bending stress, positioning offset, and thermal deformation generated at the previous station will be transmitted and superimposed at each station. The conventional single-station independent compensation mode cannot decouple the sources of error. After multiple bends, the cumulative effect of overall dimensional deviation is significant, which cannot meet the tolerance requirements of precision assembly.
[0004] Cold bending of rigid PC sheets is prone to defects such as stress whitening and bending cracks; hot bending processes often use overall or wide-area heating methods, resulting in a large heat-affected zone, which can easily cause quality problems such as yellowing and orange peel on the surface of the sheets. In addition, there is a lack of temperature control mechanisms between workstations, and residual heat from previous processes can change the material's mechanical state at subsequent bending workstations, further amplifying the fluctuation in precision.
[0005] Existing technologies mostly adopt an open-loop compensation mode of batch correction after first-piece inspection, which cannot dynamically respond to real-time factors such as changes in ambient temperature, mold wear, and fluctuations in sheet material properties during processing. This results in poor dimensional consistency in batch production and a high product rework rate. Summary of the Invention
[0006] The purpose of this invention is to provide a method for multi-station continuous bending and forming of rigid PC sheets and precision compensation control, so as to solve one or more problems mentioned in the background art.
[0007] To achieve the above objectives, the present invention provides the following technical solution: a method for multi-station continuous bending forming and precision compensation control of rigid PC sheets, comprising the following specific steps: Furthermore, in-situ online testing is performed on the rigid PC sheet to be processed. The full-width thickness distribution data of the sheet is obtained through an array thickness gauge. A micro online creep detection unit applies a small constant load to the sheet before processing and collects the stress relaxation curve at the current ambient temperature in real time, extracting viscoelastic characteristic parameters including yield strength, relaxation modulus, creep compliance, and glass transition range. The vector distribution data of the original residual stress of the sheet is obtained through a scanning stress detection device. By binding all parameters to the unique identifier of the sheet material, a ternary digital twin of material, process, and working condition is constructed, establishing a ternary mapping relationship between intrinsic material parameters, environmental working condition parameters, and process response parameters. With a built-in viscoelastic constitutive model, it can directly predict the springback and stress evolution under different heating temperatures, bending rates, and holding times.
[0008] Furthermore, based on the ternary digital twin and the target forming structure, stress-oriented decoupling planning is carried out for multiple bending processes. The stress vector arrangement method is adopted to quantify the magnitude and direction of the residual stress vector generated by each bending. By optimizing the bending sequence, the included angle of the principal stress vectors of adjacent workstations is controlled within the range of 120°~180°. A stress relay station is inserted between the two large deformation bending processes. The bending action is not performed, but only a small amount of constant temperature and pressure is applied. Simultaneously, the bending line position, heating zone boundary, and mold feed path of each station are planned, and the minimum time interval between stations is dynamically calculated based on the preceding stress attenuation curve.
[0009] Furthermore, for the bending area of each workstation, a three-temperature gradient preheating is implemented using a line-focused infrared heating device; with the bending line as the center, the temperature of the core heating area is controlled within the optimal segment of the glass transition temperature range based on the optimal molding viscosity window matched by the ternary digital twin. The transition zones on both sides adopt a stepped cooling gradient to strictly control the width of the heat-affected zone within the range of 0.8 to 1.2 times the bending radius; a low-temperature constraint zone is set outside the transition zone to constrain the temperature of the non-deformation zone through air cooling, suppressing the diffusion of thermal deformation to the non-bending zone, and forming a self-homogenizing effect of internal stress induced by the temperature gradient. The heating power is adjusted by real-time feedback of the temperature field distribution through an infrared temperature measurement array and pulse power regulation.
[0010] Furthermore, during the continuous feeding of the sheet metal, the bending dies at each station adopt a dual closed-loop follow-up synchronous drive mode of position and speed. The dies and the sheet metal feed speed maintain the same direction and speed, while compensating for the position fluctuation of the sheet metal feed in real time. The entire process of bending, pressing down, holding pressure, and releasing pressure is completed in a relatively static state. Each workstation independently controls the bending rate and pressure holding parameters. The bending rate adopts a variable speed matching mode. During the pressing stage, the rate matches the material's viscoelastic deformation rate. When the forming angle is approached, it switches to low-speed creep forming. During the pressure holding stage, a micro-amplitude reciprocating compaction action is superimposed on a constant pressure. Small-amplitude high-frequency vibrations promote molecular chain rearrangement and accelerate the uniform release of residual stress. The pressure holding time is matched with the stress relaxation characteristics of PC material at the corresponding temperature.
[0011] Furthermore, a vision and infrared temperature measurement coupled detection unit is set after each bending station to simultaneously collect five types of characteristic data: bending angle, bending line position, sheet material contour, surface defects, and temperature field distribution. A multi-station deviation transmission chain model is established. The total error is decomposed into four independent components: positioning error, springback error, thermal deformation error, and mold wear error through a deviation decoupling algorithm. The error generated by the current station itself is distinguished from the cumulative error transmitted from the previous station. The error transmission coefficient of each level from the first station to the current station is calculated to quantify the error contribution of each station. Simultaneously, by coupling the temperature field with the dimensional data, three sub-components—thermoelastic deformation, permanent plastic deformation, and springback deformation—are separated from the total deformation; the value, direction, and confidence level of each error component are output.
[0012] Furthermore, based on the results of deviation decoupling and source tracing, a cascaded feedforward and station feedback dual-loop closed-loop compensation mechanism is constructed. For springback error, the bending over-bending amount and holding time of the next station are corrected in real time; for positioning error, the back gauge position and mold centering amount of the subsequent station are adjusted; for thermal deformation error, the heating temperature and cooling air volume of the subsequent station are finely adjusted. The feedforward loop is based on a multi-station deviation transmission chain model to pre-correct the process parameters of all subsequent stations; the feedback loop corrects its own parameters based on the detection results of this station. At the same time, a Gaussian process regression incremental self-learning model is established. After each workpiece is completed, the material, process, and deviation mapping relationship is updated incrementally, and the slow variable error is automatically adapted.
[0013] Furthermore, after all bending processes are completed, the workpiece enters the final stabilization station and adopts a zoned gradient annealing-full surface straightening integrated process. Gradient annealing is used for the bending area, with the temperature decreasing in steps from the center of the bending line to the edge to match the residual stress distribution gradient. Gradient pressing stress is applied simultaneously during the annealing process, with the highest pressing force in the area with the highest temperature, thus utilizing the high-temperature creep characteristics of the material to simultaneously correct the flatness of the plate. After completion, a full range of tests on final dimensions, angles, and residual stress are conducted online. A dual pass / fail rule of dimensional accuracy and stress level is adopted to mark unqualified workpieces. At the same time, the final inspection data is sent back to the Gaussian process regression incremental self-learning model.
[0014] The beneficial effects of this invention are as follows: 1. This invention enables in-situ online detection of rigid PC sheets to be processed. It acquires full-width thickness distribution data using an array thickness gauge, applies a small-amplitude constant load using a miniature online creep detection unit, collects stress relaxation curves, and extracts viscoelastic characteristic parameters such as yield strength, relaxation modulus, and creep compliance. It also acquires original residual stress vector distribution data using a scanning stress detection device. All data is bound to a unique identifier on the sheet, constructing a ternary digital twin of material, process, and working condition, establishing a ternary mapping relationship for the three types of parameters, and incorporating a generalized Maxwell viscoelastic constitutive model. This model can predict the springback and stress evolution under different heating temperatures, bending rates, and holding times; it avoids angle deviations caused by batch-to-batch material performance fluctuations, significantly improving the accuracy of bending angle prediction.
[0015] 2. This invention employs a three-temperature gradient preheating process for the bending areas at each workstation. A core heating zone, two transition zones on either side, and an outer low-temperature constraint zone are sequentially set around the bending line. The core zone temperature is matched to the optimal forming viscosity window and controlled within the optimal glass transition range. The transition zone uses a stepped cooling gradient to control the width of the heat-affected zone to 0.8~1.2 times the bending radius. The low-temperature constraint zone uses air cooling to suppress thermal deformation diffusion and induce internal stress self-homogenization, combined with pulsed power regulation to maintain a stable temperature field. The bending die uses a dual closed-loop follow-up drive for position and speed, matching the sheet metal feed speed to achieve relatively static bending. The bending rate adopts a variable speed matching mode, with micro-amplitude reciprocating compaction actions superimposed during the pressure holding stage. This reduces defects such as whitening and cracking due to cold bending stress, and yellowing and orange peel defects due to hot bending, shrinks the heat-affected zone, weakens the stress concentration phenomenon during forming, and significantly improves the bending forming quality and dimensional stability.
[0016] 3. This invention sets up a vision and infrared temperature measurement coupled detection unit after each bending station to simultaneously collect five types of data: bending angle, bending line position, contour, surface defects, and temperature field. A multi-station deviation transmission chain model is established, and a deviation decoupling algorithm is used to separate four error components: positioning, springback, thermal deformation, and mold wear. The contribution of each station's error is traced, and three deformation sub-components—thermoelastic, plastic permanent, and springback—are separated. A cascaded feedforward and station feedback dual-loop compensation mechanism is constructed, allocating compensation weights according to error type and transmission level. A graded compensation mode is used to digest excessive errors, and a Gaussian process regression incremental self-learning model is simultaneously established to update parameter mapping relationships piece by piece. A zoned gradient annealing stabilization station is set at the end. This achieves accurate error tracing and dynamic closed-loop compensation, adapting to slow variable deviations such as material fluctuations and mold wear, thus improving the dimensional accuracy and stress control level of the finished product. Attached Figure Description
[0017] Figure 1 This is a flowchart of the overall process of the method of the present invention. Detailed Implementation
[0018] 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.
[0019] like Figure 1 As shown, this embodiment of the invention provides a method for multi-station continuous bending forming and precision compensation control of rigid PC sheets, including the following specific steps: In this embodiment of the invention, the array thickness gauge, the miniature online creep detection unit, and the scanning stress detection device are arranged sequentially at the feeding station along the feeding direction of the sheet material. All of them adopt non-contact online detection, and the detection cycle is matched with the sheet material feeding speed. The detection time of a single sheet material does not affect the overall production cycle.
[0020] When the sheet material enters the first end of the feeding station, a unique identification code is engraved on the side of the sheet material by a laser marking unit. Each detection unit is equipped with a barcode scanning device, and each set of data collected is associated with the corresponding sheet material identification code: the full-width thickness distribution data of the sheet material is obtained by an array thickness gauge, a micro online creep detection unit applies a small constant load to the sheet material, and the stress relaxation curve at the current ambient temperature is collected in real time to extract viscoelastic characteristic parameters such as yield strength, relaxation modulus, creep compliance, and glass transition range. The vector distribution data of the original residual stress of the sheet material is obtained by a scanning stress detection device. All data are simultaneously uploaded to the material parameter library of the ternary digital twin.
[0021] Based on the full range of detection parameters bound with unique identifiers, a ternary digital twin of materials, processes, and operating conditions is constructed. A ternary mapping relationship is established between intrinsic material parameters, environmental operating condition parameters, and process response parameters. With a built-in viscoelastic constitutive model, the springback and stress evolution can be directly predicted under different heating temperatures, bending rates, and holding times.
[0022] The ternary digital twin is configured with three core modules: a material parameter library, a working condition sensing layer, and a process mapping layer. The material parameter library stores the thickness distribution, viscoelastic characteristic parameters, and residual stress vector data corresponding to the unique identifier of the sheet material. The working condition sensing layer receives real-time online working condition data such as ambient temperature, feed rate, and mold wear. The process mapping layer carries the viscoelastic constitutive model and establishes a parameter mapping interface.
[0023] The viscoelastic constitutive model is constructed using a generalized Maxwell model. The input parameters include heating temperature, bending rate, holding time, and initial residual stress value. The output parameters include bending springback angle, stress relaxation curve, and residual stress distribution vector. The model is pre-trained using an offline calibrated PC material viscoelastic dataset. During online operation, the current plate material and environmental parameters are injected in real time by the material parameter library and the working condition sensing layer. The output prediction results are synchronously sent to the bending sequence planning module and the temperature control module.
[0024] In this embodiment of the invention, based on the ternary digital twin and the target forming structure, stress-oriented decoupling planning is performed for multiple bending processes. Before each sheet enters each processing station, the station controller reads the unique identification code on the side of the sheet and retrieves the material intrinsic parameters, previous processing data and current stress state of the corresponding sheet from the ternary digital twin. This automatically generates the target process parameters for the station, achieving personalized parameter matching for the entire process of a single sheet. There is no need for manual batch-by-batch process adjustment, adapting to individual differences in material properties.
[0025] The stress vector arrangement method is adopted, taking the normal direction of each bending line of the target formed part as the reference, and combining the bending depth and the thickness of the sheet metal to quantitatively calculate the magnitude and direction of the residual stress vector generated by each bend; the bending sequence is optimized, and bending processes with the included angle of the principal stress direction in the range of 120°~180° are arranged adjacently to make use of the opposite stress components to cancel each other out and reduce the cumulative effect of the same stress; for bending processes with principal stress directions that are close to the same direction, adjacent processing is not arranged.
[0026] The insertion of stress relay stations is triggered by two-dimensional conditions: first, a single bending angle exceeding a set threshold is considered a large deformation bend, and an insertion must be made between two consecutive large deformation bends; second, insertion is forced when the angle between the principal stress vectors of two adjacent bends is less than 120°. The number and layout of the stations are dynamically determined based on the total number of bending operations and stress distribution optimization results, adapting to the processing requirements of different workpiece structures. Stress relay stations do not perform bending actions; they only apply a small amount of constant temperature and pressure, prompting the premature attenuation and release of residual stress from the preceding process.
[0027] Simultaneously, the bending line position, heating zone boundary, and mold feed path of each station are planned, and the minimum time interval between stations is dynamically calculated based on the previous stress attenuation curve to avoid interference from previous residual heat and residual stress to subsequent stations.
[0028] The stress relay station is equipped with the same constant temperature heating module and pressure holding mold as the bending station. The pressure holding mold adopts a planar structure and does not apply bending deformation to the sheet metal. The pressure holding temperature is consistent with the pressure holding stage temperature of the preceding bending station, and the pressure holding pressure is set to 15% to 25% of the preceding bending pressure, maintaining only surface contact between the sheet metal and the mold.
[0029] The holding time is dynamically calculated based on the residual stress peak value of the previous station. The time is set based on the residual stress decaying to less than 30% of the peak value. No additional forced cooling is applied during the holding process. Natural heat dissipation and stress relaxation are carried out simultaneously to ensure the stability of the material mechanical state in the subsequent bending station.
[0030] In this embodiment of the invention, a line-focused infrared heating device is used to implement dynamic gradient preheating in three temperature zones for the bending area of each workstation. The three temperature zones are symmetrically distributed around the bending line: the width of the core heating zone matches the radius of the bending fillet, and the temperature is controlled within the optimal segment of the glass transition temperature range based on the optimal forming viscosity window matched by the ternary digital twin. The width of the transition zone is 1.2 to 1.5 times that of the core heating zone. A stepped cooling gradient is adopted to strictly control the width of the heat-affected zone within the range of 0.8 to 1.2 times the bending radius. The low-temperature constraint zone covers the entire non-deformable area from the outer side of the transition zone to the edge of the plate. By using air cooling, the temperature of the non-deformable area is constrained, which inhibits the spread of thermal deformation to the non-bending area and at the same time forms a self-homogenizing effect of internal stress induced by temperature gradient.
[0031] The preheating adopts a follow-up heating mode and runs synchronously with the sheet material feeding. The heating area moves synchronously with the sheet material bending line. The sheet material is preheated before entering the bending station. When the bending line runs directly under the mold, the temperature of the core heating area accurately reaches the target value and remains stable. The temperature gradient of the three temperature zones is continuously maintained throughout the bending and holding process. Heating is stopped after the sheet material is released from pressure and moved out of the heating area, ensuring that the temperature field is stable throughout the deformation process.
[0032] The internal stress self-homogenization effect is achieved through the continuous temperature gradient between the core heating zone and the low-temperature constraint zone. The internal stress generated by bending deformation gradually diffuses and relaxes from the high-temperature zone to the low-temperature zone under the action of the temperature gradient, avoiding stress concentration at the bending line. The slope of the temperature gradient is precisely controlled by the step cooling gradient in the transition zone, taking into account both the stress homogenization effect and the control of the width of the heat-affected zone. If the slope is too large, it will easily cause additional thermal stress, and if the slope is too small, the self-homogenization effect will be insufficient.
[0033] The infrared temperature measurement array provides real-time feedback on the temperature field distribution, and pulse power regulation is used to dynamically adjust the heating power: the real-time temperature of the core area is used as the feedback signal. When the temperature exceeds the upper limit of the set range, the heating power ratio is reduced, and when it is below the lower limit, the power ratio is increased. The power adjustment step size is positively correlated with the temperature deviation amplitude, maintaining the temperature fluctuation of the core area within ±2℃ of the set value, and avoiding local overheating that could cause yellowing and blistering of the board.
[0034] The low-temperature confinement zone uses a side-blowing air curtain to supply air, with the airflow direction parallel to the surface of the board and pointing outward from the bending line. The temperature difference between the airflow and the ambient temperature is controlled within 5℃. By adjusting the airflow, the temperature of the confinement zone is maintained below the lower limit of the glass transition temperature of the PC material, forming a stable temperature gradient boundary.
[0035] In this embodiment of the invention, during the continuous feeding of the sheet metal, the bending dies at each station adopt a position and speed dual closed-loop follow-up synchronous drive mode: the speed loop takes the sheet metal feeding linear speed as the reference value, and collects the dies running speed in real time through the encoder and dynamically adjusts the output of the drive motor to ensure that the difference between the linear speed of the dies and the sheet metal is controlled within the set threshold. The position loop uses the bending line positioning mark on the sheet metal as the feedback reference. The visual positioning unit captures the relative position deviation between the bending line and the die in real time, and dynamically corrects the die's feed start position and running phase. The die and sheet metal feed speed maintain the same direction and speed, and compensate for the position fluctuation of the sheet metal feed in real time, completing the entire process of bending, pressing, holding, and releasing pressure in a relatively static state.
[0036] Each workstation independently controls the bending rate, adopting a variable speed matching mode: the rate during the pressing stage matches the upper limit of the viscoelastic deformation rate of the material at the current temperature, ensuring that the internal stress of the material rises steadily during the deformation process without any sudden stress changes; When the bending angle approaches the forming angle, the speed is switched to low-speed creep forming. The angle threshold for the speed switching is calculated in real time by the viscoelastic constitutive model in the ternary digital twin. The switching point is the set ratio of the bending angle to the target forming angle. The switching point angle is positively correlated with the degree of glass transition of the material. During the low-speed creep stage, the creep rate gradually decreases as the angle approaches the target value, and drops to zero at the beginning of the pressure holding stage, achieving a smooth transition from deformation to pressure holding and avoiding additional internal stress caused by mechanical impact.
[0037] During the pressure holding stage, a small-amplitude reciprocating compaction action along the bending normal is superimposed on the constant pressure. The small-amplitude high-frequency vibration promotes molecular chain rearrangement and accelerates the uniform release of residual stress. The pressure holding time is matched with the stress relaxation characteristics of PC material at the corresponding temperature to ensure that the rebound amount is stable and controllable.
[0038] The vibration parameters are automatically set by matching the current plate thickness and holding temperature with the ternary digital twin. The vibration amplitude is controlled within the range of 0.1 to 0.3 mm. The larger the plate thickness, the greater the vibration amplitude. The vibration frequency decreases as the holding temperature increases.
[0039] The compaction action is initiated in the middle of the pressure holding stage and stopped before the end of the pressure holding stage. It is not initiated during the initial stage of rapid stress rise in the pressure holding stage, but is started after the stress enters the stable relaxation stage. It is stopped at the end of the pressure holding stage after the stress has basically stabilized, so as to avoid vibration from interfering with the dimensional accuracy of the molding.
[0040] In this embodiment of the invention, a vision and infrared temperature measurement coupled detection unit is coaxially arranged after each bending station. The acquisition surface covers the entire bending area and the transition areas on both sides, and simultaneously acquires five types of characteristic data: bending angle, bending line position, sheet material contour, surface defects, and temperature field distribution.
[0041] The acquisition trigger signal is provided by the station feed encoder, corresponding to the moment when the pressure holding and release of the board is completed and the board moves directly below the detection unit, ensuring that the visual dimension data and infrared temperature data are synchronous acquisition results of the same time and the same area; each set of acquired data is automatically associated with the unique identification code of the corresponding board and the current station number, and synchronously transmitted to the multi-station deviation transmission chain model.
[0042] The multi-station deviation transmission chain model constructs a hierarchical chain structure according to the bending station sequence. Each station corresponds to an error transmission node. The node input includes the cumulative error of the previous node and the process parameters of the current station. The node output is the total error of the current station and is transmitted to the next level. The model interacts with the detection unit of each station in real time. The node weight is updated once after each bending process is completed, realizing the dynamic correction of the error transmission relationship.
[0043] Error source tracing is carried out through a multi-source deviation decoupling algorithm: taking the five types of feature data output by the coupled detection unit as input, the calculation logic combining independent component analysis and error source tracing matrix is adopted. First, the feature signals of four independent error sources, namely positioning error, springback error, thermal deformation error and mold wear error, are separated from the total dimensional deviation. Then, combined with the hierarchical weight matrix of the deviation transmission chain, the transmission coefficient of the error of each previous station to the current station is calculated. Finally, the proportion and value of the current station's own error and the previous accumulated error are separated to quantify the error contribution of each station.
[0044] Simultaneously, by coupling temperature field and dimensional data, three types of deformation sub-components are separated from the total deformation: thermoelastic deformation is calculated based on real-time temperature field distribution and material thermal expansion coefficient, corresponding to reversible dimensional changes caused by temperature changes; plastic permanent deformation is calculated by combining the pressure, duration and material creep characteristics of the holding stage, corresponding to irreversible deformation generated during processing; and springback deformation is the remaining part after deducting the first two from the total deformation, corresponding to the elastic recovery of the material after unloading.
[0045] The three types of deformation components correspond to different compensation strategies: thermoelastic deformation is corrected by temperature control, plastic permanent deformation is corrected by pressure holding parameters, and springback deformation is corrected by bending amount.
[0046] After each workstation inspection is completed, the actual measured bending dimensions, residual stress distribution, and temperature field data are synchronously updated to the corresponding ternary digital twin of the board material. The original predicted parameters are overwritten with the actual processing state, and the stress state and morphological characteristics of the board material are corrected in real time. All subsequent unprocessed workstations calculate process parameters based on the updated actual state, ensuring that the parameters of each process match the actual situation of the current board material, rather than relying solely on the predicted values of the initial inspection.
[0047] In this embodiment of the invention, based on the results of deviation decoupling and source tracing, a cascaded feedforward and station feedback dual-loop closed-loop compensation mechanism is constructed. The dual-loop compensation allocates compensation weights according to the error type and transmission level: for the cumulative error of the preceding sequence, the feedforward loop mainly bears the burden, and the compensation amount covers all subsequent unprocessed stations. Based on the deviation transmission chain model, the subsequent process parameters are corrected in advance to offset the error transmission effect; for the error generated by the current station itself, the feedback loop mainly bears the burden, and only the corresponding process parameters of this station are corrected to suppress the error of the current station.
[0048] The compensation priority is sorted in the order of springback error, positioning error, and thermal deformation error, with priority given to ensuring bending angle accuracy: for springback error, the bending over-bending amount and holding time are corrected; for positioning error, the back gauge position and mold alignment are adjusted; and for thermal deformation error, the heating temperature and cooling air volume are finely adjusted.
[0049] The feedforward compensation amount is allocated according to the error propagation coefficient ratio of each subsequent station. The closer the subsequent station is to the error source station, the higher the compensation ratio is allocated. It decreases step by step along the processing direction. The compensation amount is issued using a station-by-station delayed triggering mechanism. It is issued to the controller of the subsequent station immediately after the current station completes the detection. The parameters are pre-updated before the workpiece reaches the target station. When the workpiece enters the station, the corrected process parameters are directly called to avoid deformation instability caused by parameter mutations during processing. It ensures that the compensation action is seamlessly connected with the processing cycle. The feedback compensation completes the parameter correction before the next bending action at this station. The response cycle matches the processing cycle of a single station.
[0050] Each compensation amount has upper and lower limits. The upper limit corresponds to the maximum adjustable range of the process parameters, and the lower limit corresponds to the minimum adjustment accuracy of the equipment. This avoids sudden changes in process parameters caused by excessive compensation in a single operation. When the detection error exceeds the compensation limit of a single station, the hierarchical compensation mode is automatically triggered to distribute the error exceeding the limit to multiple subsequent stations for gradual digestion. This completes the error correction of the entire process without affecting the process stability of a single station.
[0051] At the same time, a Gaussian process regression incremental self-learning model is established. After each workpiece is completed, the material, process, and deviation mapping relationship is updated incrementally. It automatically adapts to slow variable errors such as environmental temperature drift and slow mold wear, and improves the adaptation accuracy in small sample scenarios.
[0052] The model has a three-layer structure: input layer, kernel function layer, and output layer. The input layer has an 8-dimensional dimension, and the input features include material viscoelastic parameters, ambient temperature, process parameters of each station, and the values of each error component. The kernel function layer uses a combination of square exponential kernel and linear kernel to fit the nonlinear mapping relationship between material, process, and deviation. The output layer outputs the compensation and correction amounts for bending over-bending amount, holding time, and heating temperature of each station.
[0053] The model adopts an incremental training method: before the first piece is processed, the model is initialized using an offline small sample dataset; after each piece is processed, the actual deviation data obtained from the final inspection and the corresponding process parameters are used as new samples, and the hyperparameters and covariance matrix of the model are updated through the sparse Gaussian process incremental update algorithm, without the need for full retraining; the model update cycle is the single piece processing cycle, the initial value of the noise term in the hyperparameter is set to 0.02, and it is adaptively adjusted once every 50 new samples. The compensation correction amount output by the model is synchronously input into the dual-loop compensation mechanism.
[0054] In this embodiment of the invention, after all bending processes are completed, the workpiece enters the final stabilization stage, where a zoned gradient annealing-full-surface straightening integrated process is adopted. The zoned gradient annealing is set with three temperature steps: the central area of the bending line is the first level, with the temperature maintained at the lower-middle limit of the glass transition range; extending to both sides are the second and third levels, with the temperature decreasing step by step, and the temperature difference between levels controlled at 8 to 12°C. The temperature distribution matches the residual stress distribution gradient, so that the residual stress is fully and uniformly released, avoiding the generation of new internal stress by uniform annealing.
[0055] Gradient compressive stress is applied simultaneously during the annealing process. The compressive stress is set in stages according to the temperature gradient. The higher the temperature, the greater the compressive stress. The compressive stress in the central area is 2 to 3 times that in the edge area. The flatness of the plate is corrected simultaneously by utilizing the high-temperature creep characteristics of the material.
[0056] The total annealing time is set according to the thickness of the sheet and the final residual stress requirements. The greater the thickness, the longer the time. The temperature gradient is kept stable during the annealing process. After the annealing is completed, uniform air cooling is used to slowly cool down to room temperature to avoid generating new internal stress during the cooling process.
[0057] After completion, a full range of tests on final dimensions, angles, and residual stress are conducted online. A dual pass / fail judgment rule is adopted for both dimensional accuracy and stress level: dimensional accuracy is based on the design tolerance requirements of each bending angle and bending line position. All dimensional parameters are judged to be dimensionally qualified if they are within the tolerance range. The stress level is judged as qualified if the peak residual stress in the bending area is lower than the material yield strength by a set ratio and the uniformity of residual stress distribution meets the requirements. If both conditions are met, the finished product is considered qualified; if either condition is not met, it is marked as unqualified.
[0058] The final inspection data of both qualified and unqualified workpieces are fed back to the Gaussian process regression incremental self-learning model. Unqualified data are marked with quality anomaly labels to ensure that the model samples cover the entire range of working conditions.
[0059] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.
[0060] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.
Claims
1. A method for multi-station continuous bending and forming of rigid PC sheets and precision compensation control, characterized in that, Includes the following steps: In-situ online detection of rigid PC sheets to be processed is carried out to collect thickness, viscoelasticity and residual stress data, bind the unique identifier of the sheet, construct a three-dimensional digital twin of material, process and working condition, establish parameter mapping and build a built-in viscoelastic constitutive model; By combining the ternary digital twin with the target forming structure, the bending residual stress vector is calculated, the bending sequence is optimized, the principal stress angle between adjacent stations is controlled, stress relay stations are inserted between large deformation bending processes and constant temperature and pressure are maintained, and process parameters and timing intervals are planned. Three-temperature gradient preheating is performed on the bending area of each station, matching the molding viscosity window, setting the cooling gradient and low temperature constraint, collecting the temperature field and using pulsed power control. Each bending die at each station adopts a dual closed-loop follow-up drive for position and speed, independently controlling the bending rate. During the pressure holding stage, a micro-amplitude reciprocating compaction action is superimposed, and the pressure holding time is set. A coupling detection unit is set up after each bending station to collect bending dimensions, morphology and temperature data, establish a deviation transmission chain model, and decompose error components through a deviation decoupling algorithm to trace the error contribution of each station. Based on the deviation tracing results, a cascaded feedforward and station feedback dual-loop compensation mechanism is constructed to correct the corresponding station process parameters, establish an incremental self-learning model, and update the material, process, and deviation mapping relationship for each piece. The workpiece enters the final stabilization stage, where it undergoes zoned gradient annealing and is simultaneously subjected to gradient compressive stress. After completion, a full range of tests are performed and the data is transmitted back.
2. The method for multi-station continuous bending and forming of rigid PC sheets and precision compensation control according to claim 1, characterized in that, In the in-situ online detection, the full-width thickness distribution data of the plate is obtained by an array thickness gauge; a micro-amplitude constant load is applied to the plate by a micro online creep detection unit, and the stress relaxation curve at the current ambient temperature is collected in real time to extract viscoelastic characteristic parameters including yield strength, relaxation modulus, creep compliance, and glass transition range. Vector distribution data of the original residual stress in the plate are obtained by scanning stress detection device.
3. The method for multi-station continuous bending and forming of rigid PC sheets and precision compensation control according to claim 2, characterized in that, By binding all testing parameters to the unique identifier of the sheet material, a ternary digital twin is constructed to establish a ternary mapping relationship between the intrinsic parameters of the material, environmental conditions, and process response parameters. The built-in viscoelastic constitutive model can directly predict the rebound amount and stress evolution law under different heating temperatures, bending rates, and holding times.
4. The method for multi-station continuous bending and forming of rigid PC sheets and precision compensation control according to claim 1, characterized in that, The magnitude and direction of the residual stress vector generated by each bend are quantitatively calculated using the stress vector arrangement method. The bending sequence is optimized so that the included angle of the principal stress vectors of adjacent workstations is controlled within the range of 120° to 180°. The stress relay workstation does not perform bending action, but only applies a small amount of constant temperature and pressure to promote stress attenuation.
5. The method for multi-station continuous bending and forming of rigid PC sheets and precision compensation control according to claim 4, characterized in that, The bending line position, heating zone boundary, and mold feed path of each station are planned synchronously. The minimum time interval between stations is dynamically calculated based on the stress attenuation curve of the previous station to ensure that stress is fully released without affecting the overall production cycle and molding consistency.
6. The method for multi-station continuous bending and forming of rigid PC sheets and precision compensation control according to claim 1, characterized in that, Centered on the bending line, a core heating zone, two transition zones on both sides, and an outer low-temperature constraint zone are set sequentially. The temperature of the core heating zone is matched with the optimal forming viscosity window and controlled within the optimal segment of the glass transition temperature range. The transition zone adopts a stepped cooling gradient to control the width of the heat-affected zone within the range of 0.8 to 1.2 times the bending radius. The low-temperature constraint zone is vented with air cooling to constrain the temperature of the non-deformation zone, suppressing the diffusion of thermal deformation to the non-bending zone, and at the same time forming a self-homogenizing effect of internal stress induced by the temperature gradient.
7. The method for multi-station continuous bending and forming of rigid PC sheets and precision compensation control according to claim 1, characterized in that, During the continuous feeding of the sheet metal, the die and the sheet metal feed speed maintain the same direction and speed, and the position fluctuation of the sheet metal feed is compensated in real time. The entire process of bending, pressing, holding and releasing pressure is completed in a relatively static state. The bending speed adopts a variable speed matching mode. During the pressing stage, the speed matches the viscoelastic deformation rate of the material. When the forming angle is approached, it switches to low-speed creep forming.
8. The method for multi-station continuous bending and forming of rigid PC sheets and precision compensation control according to claim 7, characterized in that, During the pressure holding stage, a micro-amplitude reciprocating compaction action is superimposed on a constant pressure. Small-amplitude high-frequency vibrations promote molecular chain rearrangement and accelerate the uniform release of residual stress. The pressure holding time is matched with the stress relaxation characteristics of PC material at the corresponding temperature to ensure molding accuracy and stability.
9. The method for multi-station continuous bending and forming of rigid PC sheets and precision compensation control according to claim 1, characterized in that, The coupled detection unit is a vision and infrared temperature measurement coupled detection unit, which simultaneously collects five types of data: bending angle, bending line position, sheet material contour, surface defects, and temperature field distribution. The deviation decoupling algorithm decomposes the total error into four independent components: positioning, springback, thermal deformation, and mold wear. It distinguishes between the current station's own error and the previous accumulated error, traces the source to calculate the error transmission coefficient, and quantifies the error contribution of each station. At the same time, it couples the temperature field and dimensional data to separate three types of deformation sub-components from the total deformation: thermoelastic, plastic permanent, and springback.
10. The method for multi-station continuous bending and forming of rigid PC sheets and precision compensation control according to claim 1, characterized in that, The final stabilization process adopts a zoned gradient annealing-full-surface straightening integrated process. The temperature in the bending area decreases stepwise from the center of the bending line to the edge to match the residual stress distribution gradient. Gradient pressing stress is applied simultaneously during annealing, with the pressing force being the largest in the area with the highest temperature. The flatness of the sheet is corrected by utilizing the high-temperature creep characteristics. After completion, full-item inspection of dimensions, angles, and residual stress is carried out. A dual pass / fail judgment rule of dimensional accuracy and stress level is adopted to mark unqualified workpieces. The final inspection data is fed back to the incremental self-learning model.
