Press segmented closed track generation method based on consistency safety Bayesian optimization

Abstract

The invention provides a press segmented closed track generation method based on consistency safety Bayesian optimization, and the method comprises the steps: extracting a compaction response and infiltration capacity parameter range under the upper limit constraint of a press through a force-displacement curve and an in-mold pressure signal which are obtained through one-time pressure test; establishing an extrusion flow and gas discharge combined calculation model under mould geometry and cavity boundary conditions, and determining displacement points when the edge sealing starts and finishes displacement and the central pressure reaches a drop threshold value; introducing the physical constraints into a scoring network, determining the speed and the residence time of each section in a trajectory parameter range by adopting secure Bayesian optimization, generating a target section closed trajectory which meets the condition that the central pressure is reduced to a threshold value before edge sealing and does not trigger the upper limit of equipment, and outputting a corresponding control instruction; according to the method, under the condition of considering material and working condition fluctuation, the exhaust reliability and the forming quality in the pressing forming process are improved, and takt time and equipment safety are both considered.

Description

Technical Field

[0001] This invention relates to the field of thermoplastic composite material compression molding process control technology, and in particular to a method for generating segmented closed trajectories of a press based on consistent safety Bayesian optimization. Background Technology

[0002] After thermoplastic composite materials are heated to a molten state, they are formed by closing a mold using a press. During the closing process, the material undergoes thickness compaction, resin and gas flow within the cavity and are expelled through the venting channel, and the pressure within the cavity changes with displacement. In actual production, a multi-stage displacement closing method is often used, coordinating the compaction force, mold pressure, venting effect, and forming cycle by setting different closing speeds and dwell times for each stage. However, batch fluctuations in material, differences in mold geometry, and limitations in equipment capacity make the setting of segmented closing trajectories highly dependent on experience, making it difficult to quantitatively assess in a timely manner whether the venting in the central area is sufficient and the rationality of the edge sealing sequence.

[0003] In existing technologies, the control of the closing process of thermoplastic composite material presses typically employs the following approaches: One method relies on force-displacement curves or in-mold pressure curves obtained from a limited number of press trials, manually adjusting the closing speed and stroke segments to ensure the maximum load does not exceed the press's upper limit, and gradually correcting process parameters by observing specimen defects; another method establishes an extrusion flow or simplified cavity pressure model under given mold geometry, using the closing stroke and speed as inputs to predict load and pressure changes over time, and accordingly adjusting the segmented trajectory; yet another method combines in-mold pressure sensors, setting pressure thresholds or plateau intervals to implement simple feedback control or staged pauses in the closing process. These solutions generally can avoid overload to some extent and monitor in-mold pressure levels to guide the development of segmented closing processes.

[0004] However, the aforementioned existing technologies often directly set the process based on a single test pressure curve, lacking a parameterized description of the uncertainty of compaction response and permeability, making it difficult to form a parameter combination range that covers batch fluctuations. At the same time, extrusion flow and gas exhaust are usually treated separately or simplified, without uniformly calculating the relationship between the edge sealing start and end displacements and the evolution of the center pressure under given mold geometry and cavity boundary conditions, making it difficult to accurately locate the position where "the center pressure drops to a safe level" within the displacement domain. When generating segmented closed trajectories, there is also a lack of a systematic method that simultaneously incorporates equipment force, speed, stroke upper limit, and the timing constraint of "sufficient venting before sealing".

[0005] Therefore, a method for generating segmented closed trajectories of a press that can overcome the shortcomings of the existing technology is a problem that needs to be solved by those skilled in the art. Summary of the Invention

[0006] One objective of this invention is to propose a method for generating segmented closed trajectories for presses based on consistent safety Bayesian optimization. The core technical problem to be solved is: in the process of press forming of thermoplastic composite materials, how to generate a segmented closed trajectory under the premise of considering the material compaction response and permeability fluctuation, mold geometry and cavity boundary conditions, and the upper limit of press force / speed / stroke, so that the gas in the central area of ​​the cavity can be reliably discharged before the edge sealing, and a segmented closing command that can be directly used for press control can be formed accordingly.

[0007] A method for generating a segmented closed trajectory of a press based on consistent safe Bayesian optimization according to an embodiment of the present invention includes:

[0008] S1. Obtain the force-displacement curve and in-mold pressure signal of a single pressure test, obtain the press force, speed and upper limit of stroke, perform time alignment on the force-displacement curve and in-mold pressure signal, determine the trajectory parameter range and the center pressure drop threshold based on the press force, speed and upper limit of stroke, and obtain the aligned force-displacement curve, aligned in-mold pressure signal, trajectory parameter range and center pressure drop threshold.

[0009] S2. Construct the parameter combination range based on the aligned force-displacement curve and the in-mold pressure signal: First, determine the compaction response range based on the force-displacement curve, and then determine the permeability range based on the in-mold pressure signal. With compaction response and permeability constrained together, examine each combination of compaction response range and permeability range, and eliminate combinations that cannot simultaneously explain the changing trends of force-displacement curve and in-mold pressure signal to obtain the parameter combination range.

[0010] S3. Under the conditions of mold geometry and cavity boundary, establish the joint calculation of extrusion flow and gas exhaust, transform the trajectory parameter range into cavity thickness variation, calculate the change of center pressure with displacement, select the parameter combination that is unfavorable to center exhaust within the parameter combination range, determine the edge sealing start displacement and edge sealing completion displacement, and determine the displacement point where the center pressure reaches the center pressure drop threshold under this unfavorable parameter combination, thus obtaining the edge sealing start displacement, edge sealing completion displacement and unfavorable parameter combination;

[0011] S4. Based on the parameter combination range, the starting displacement of the sealing edge and the finishing displacement of the sealing edge, safe Bayesian optimization is used to determine the segment speed and dwell time within the trajectory parameter range to generate the first candidate trajectory.

[0012] S5. Perform joint calculations on the first candidate trajectory, and determine whether the center pressure drop threshold is reached before the edge sealing displacement begins under unfavorable parameter combinations. If the threshold is met, the target trajectory is obtained; otherwise, iteratively optimize and update the target trajectory within the trajectory parameter range.

[0013] S6. Generate segmented closure control commands based on the target trajectory. The segmented closure control commands include displacement segments, segment velocities, and dwell times.

[0014] Optionally, S1 is as follows:

[0015] Obtain the force-displacement curve and the in-mold pressure signal of a single pressure test, obtain the press force, speed and upper limit of stroke, and perform time alignment on the force-displacement curve and the in-mold pressure signal, using the common start time and displacement zero point as the alignment reference, to obtain the aligned force-displacement curve and the aligned in-mold pressure signal;

[0016] Based on the press force, speed, and stroke upper limit, the trajectory parameter range is defined as the boundary of the values ​​of displacement segment, segment speed, and dwell time. The upper and lower limits are set with four values ​​for the three closed displacement segments, three values ​​for the segment speed, and three values ​​for the dwell time, so that the trajectory execution does not trigger the press force, speed, and stroke upper limit, thus forming the trajectory parameter range.

[0017] Based on the platform pressure and decay time of the aligned intramold pressure signal, combined with the slope of the rising segment of the aligned force-displacement curve, the numerical standard of the center pressure drop threshold is determined, and the center pressure drop threshold is output.

[0018] The aligned force-displacement curve, aligned intramold pressure signal, trajectory parameter range, and center pressure drop threshold are used as input preparation items for the scoring network, enabling the physical constraint fusion layer to be integrated with the subsequent edge sealing displacement, edge sealing completion displacement, and unfavorable parameter combinations.

[0019] Optionally, S2 is as follows:

[0020] Using the aligned force-displacement curve as input, the slope of the initial segment, the slope of the nonlinear segment, and the position of the load platform are extracted. The deformation rebound of the unloading segment is used as a supplementary feature to establish the sensitive range of thickness change to load and define the compaction response range.

[0021] Using the aligned intramold pressure signal as input, the rising start time, pressure peak, plateau pressure and decay time are extracted to establish the penetration capacity range and define the permeability range.

[0022] Under the premise of fixed test temperature and mold geometry, the compaction response range and the permeability range are combined one by one to generate candidate combinations. The order, slope and peak time of the predicted force-displacement curve in the rising segment and plateau segment are calculated based on the mapping relationship between the thickness change sensitive range of load and the permeability range, as well as the order of the predicted in-mold pressure signal in the start, peak and plateau segments.

[0023] For each candidate combination, a consistency judgment of the trend of change is applied. The consistency criteria are consistent order, overlapping slope within a set interval, and peak time difference not exceeding a set threshold. Combinations that do not meet the consistency criteria are eliminated, and combinations that meet the consistency criteria are retained to form the parameter combination range.

[0024] The parameter combination range is encoded as two values: the position index of the compaction response and the permeability within the parameter combination range. These values ​​are used as inputs to the scoring network. The scoring network is then fused with the physical constraint fusion layer, the edge start displacement, the edge completion displacement, and the center pressure drop threshold. This allows the scoring network to use the parameter combination range as a constraint when generating the first candidate trajectory.

[0025] Within the parameter combination range, the unfavorable parameter combination is determined by the combination corresponding to the minimum value of the permeability range and the maximum value of the compaction response range. The parameter combination range and the unfavorable parameter combination are output, providing constraints for iterative optimization of segment velocity and dwell time within the trajectory parameter range and calculation of the lead displacement of the scoring network head.

[0026] Optionally, the consistency judgment of the changing trend applied to each candidate combination can be quantified using a consistency scoring formula:

[0027] ;

[0028] in, For consistency scoring, dimensionless. For the predicted force-displacement curve The slope of the rising segment, Force-displacement curves after alignment The slope of the rising segment, For the predicted force-displacement curve peak time, Force-displacement curves after alignment peak time, For the predicted intramold pressure signal peak time, For the aligned in-mold pressure signal peak time, The slope difference threshold, The peak time difference threshold, This is a sequence consistency identifier, with a value of [value to be filled in]. or , This is an indicator of slope overlap, with a value of [value missing]. or , , , The weighting coefficient has a value of [value]. , , The penalty coefficient has a value of [value missing]. ,by Based on the screening criteria, retain those that meet the requirements. Combine the given combinations and eliminate those that do not meet the requirements.

[0029] Optionally, S3 specifically refers to:

[0030] Under the conditions of mold geometry and cavity boundary, a joint calculation of extrusion flow and gas exhaust is established. The displacement segment, segment velocity and dwell time in the trajectory parameter range are used as inputs to generate a continuous change curve of cavity thickness over time according to the segment time sequence, thus forming the thickness change.

[0031] The process of advancing edge contact is calculated point by point based on thickness change. It is determined whether the edge gap is zero. The first position where it is zero is recorded as the start of edge sealing displacement. The displacement of the contact line closing to form a continuous loop is recorded as the edge sealing completion displacement.

[0032] Within the parameter combination range, candidate unfavorable combinations are generated by combining the lower bound of the permeability range and the upper bound of the compaction response range. Among the combinations that pass the consistency judgment of the change trend, the candidate is selected as the unfavorable parameter combination for the determination of central venting.

[0033] Under unfavorable parameter combinations, the thickness variation is mapped to the values ​​in the compaction response range as a load relationship, and the thickness variation is mapped to the values ​​in the permeability range as an outward discharge capacity, thus coupling the change of central pressure with displacement.

[0034] Using the central pressure drop threshold as the criterion, the change of central pressure with displacement is scanned along the displacement axis to locate the displacement point that first meets the central pressure drop threshold, and the displacement point where the central pressure reaches the central pressure drop threshold is output.

[0035] The time-displacement sequence of the edge sealing start displacement, edge sealing completion displacement, and center pressure reaching the center pressure drop threshold is compared. The displacement point where the center pressure reaches the center pressure drop threshold is required to be earlier than the edge sealing start displacement, thus forming a time-displacement sequence constraint for trajectory screening.

[0036] The edge sealing start displacement, edge sealing completion displacement, unfavorable parameter combination, and the displacement point where the center pressure reaches the center pressure drop threshold, along with the center pressure drop threshold, are used as input items for the scoring network. They are fused in the physical constraint fusion layer, and the difference between the edge sealing start displacement and the displacement point where the center pressure reaches the center pressure drop threshold is defined as the lead displacement, which is used by the safety head to generate the first candidate trajectory within the trajectory parameter range.

[0037] Optionally, the step of scanning the change of central pressure with displacement along the displacement axis and locating the displacement point that first meets the central pressure drop threshold, using the central pressure drop threshold as the criterion, is quantitatively determined using a scanning formula:

[0038] ;

[0039] ;

[0040] in, The displacement point at which the central pressure reaches the threshold for a decrease in central pressure, expressed in units of displacement. The displacement versus time curve at discrete times The value at that location is expressed in units of displacement. To ensure the minimum time step index that the central pressure does not exceed the threshold, a dimensionless integer is used. The time step index is a dimensionless integer. For the central pressure at discrete time points The value at that location is in pressure. The threshold for the decrease in central pressure, expressed in pressure. For the first Discrete moments, in units of time. Let be a set of non-negative integers, and let be a dimensionless set. The above definition is uniquely determined by minimizing the discrete index. and generated by joint computation and Direct correspondence avoids the uncertainty caused by continuous domain interpolation.

[0041] Optionally, S4 specifically refers to:

[0042] The parameter combination range, edge sealing start displacement, edge sealing completion displacement, center pressure drop threshold, press force, speed and stroke upper limit are encoded together with the displacement segments, segment speeds and dwell times in the trajectory parameter range into a trajectory feature vector, which is then input into the scoring network.

[0043] The trajectory feature vector is fused with the above constraints in the physical constraint fusion layer. The mean and uncertainty of the lead displacement are output by the safety head, and the mean and uncertainty of the sum of dwell times are output by the beat head.

[0044] The data acquisition calculation module adopts safe Bayesian optimization, calculates the safety lower bound and time index based on the safety head and the beat head, and searches for segment speed and dwell within the trajectory parameter range to make the safety lower bound greater than zero and not trigger the upper limit of press force, speed and stroke.

[0045] Sort the trajectory within the safety set by time index, select the trajectory feature vector to generate the first candidate trajectory, and output the displacement segments, segment velocity and dwell time.

[0046] Optional, S5 specifically includes:

[0047] The displacement segment, segment velocity and dwell input of the first candidate trajectory are calculated together and transformed into thickness change under the conditions of mold geometry and cavity boundary. Under unfavorable parameter combination, the change of center pressure with displacement is calculated, and the displacement point where the center pressure reaches the center pressure drop threshold is located.

[0048] The displacement point is compared with the edge sealing start displacement in time-displacement sequence. At the same time, it is checked whether the upper limit of press force, speed and stroke is not triggered. It is determined whether the center pressure drop threshold is met before the edge sealing start displacement. If it is met, the first candidate trajectory is confirmed as the target trajectory.

[0049] If the conditions are not met, the safety head and beat head outputs of the scoring network are used as inputs to the acquisition value calculation module. Based on the safety lower bound and time index, the displacement segment, segment velocity and dwell time are adjusted within the trajectory parameter range to generate a new trajectory feature vector.

[0050] Repeatedly perform joint calculations and order determination on the new trajectory feature vectors, and select the updated trajectory feature vectors in the safety set according to the time index until the center pressure drop threshold is reached before the edge starts to displace, and output the target trajectory.

[0051] Optional, S6 specifically includes:

[0052] The displacement, segment velocity, and dwell time of the target trajectory are arranged in segment sequence to generate the basic fields of the segmented closure control command;

[0053] The edge sealing start displacement and edge sealing completion displacement are incorporated into the segmented closure control command. Time-displacement sequence constraints are set, and advance displacement is used as the judgment item. It is required that the advance displacement is greater than zero and the displacement point corresponding to the center pressure reaching the center pressure drop threshold is earlier than the edge sealing start displacement.

[0054] The unfavorable parameter combinations in the parameter combination range are used for execution judgment. Under the constraints of press force, speed and stroke upper limit, the time-displacement sequence constraint judgment of the segmented closed control command is performed.

[0055] After the time-displacement sequence constraint is passed, the displacement segment, segment speed and dwell time of the segmented closure control command are confirmed and used for the execution of the press segmented closure.

[0056] The beneficial effects of this invention are:

[0057] 1. This proposal presents an improved method for generating segmented closed trajectories in thermoplastic composite presses. Based on the joint calculation of extrusion flow and gas venting, it introduces a scoring network with a physical constraint fusion layer and safety Bayesian optimization. The edge sealing start displacement, edge sealing completion displacement, center pressure drop threshold, and unfavorable parameter combinations are integrated and encoded as constraints for trajectory optimization. Compared to existing methods that rely solely on force-displacement curves or in-mold pressure curves to empirically tune segment speed and dwell time, this proposal uses "advance displacement" as a safety indicator and the total dwell time as a cycle time indicator within the trajectory parameter range. A safety lower bound is constructed using the mean and uncertainty of the safety head output. Under the premise that the center pressure drops to the threshold before edge sealing begins without triggering the press's force, speed, and stroke limits, the segment speed and dwell time are searched. Thus, without changing the press hardware, explicit displacement domain constraints on the edge sealing timing and center venting process are achieved, which is beneficial for consistently obtaining segmented closed trajectories with sufficient venting before edge sealing in mass production.

[0058] 2. This proposal presents a novel method for constructing parameter combination ranges and determining unfavorable parameter combinations. By aligning the force-displacement curve and in-mold pressure signal from a single pressure test over time, it extracts the initial slope, nonlinear slope, load plateau position, deformation rebound, and pressure rise start point, peak value, plateau pressure, and decay time. Under the premise of fixed test temperature and mold geometry, it establishes the compaction response range and permeability range. A consistency scoring formula for variation trends is used to screen parameter combination ranges that can simultaneously explain the force-displacement and in-mold pressure variation trends. Furthermore, within this parameter combination range, this proposal constructs unfavorable parameter combinations using an upper bound on the compaction response and a lower bound on permeability, ensuring that the joint calculation and scoring network operates under the most unfavorable conditions for material and flow characteristics. Compared to existing technologies that treat a single pressure test curve as a deterministic input and ignore compaction and permeability fluctuations, this technique explicitly transforms batch fluctuations into a parameter space without adding extra experimental burden. This ensures the physical rationality of the model input and provides a boundary covering actual fluctuations for subsequent safety optimization, thus improving the robustness of trajectory design to changes in materials and operating conditions.

[0059] 3. This proposal presents a holistic control method for segmented closure execution of presses. It encapsulates the displacement segments, segment speeds, and dwell times of the target trajectory, along with the edge sealing start displacement, edge sealing completion displacement, center pressure drop threshold, displacement point where the center pressure reaches the threshold, and unfavorable parameter combinations, into a unified segmented closure control command. The command embeds time-displacement sequence constraints and equipment upper limit verification logic. During the generation phase, the attainability of the center pressure threshold is discretized using joint calculations and scanning formulas, with advance displacement as the criterion. A control command is only generated when the displacement point corresponding to the center pressure reaching the threshold is earlier than the edge sealing start displacement and does not exceed the force, speed, and stroke upper limits. Unlike existing solutions that only provide trajectory parameters or simple pressure threshold control without forming an integrated control command directly applicable to the production line, this method uses unified time-displacement sequence constraints and unfavorable parameter combinations in the design, verification, and execution stages. This ensures consistent implementation of offline optimization results on the field equipment, facilitating continuous constraint of the relative timing of edge sealing and venting during actual forming, and improving the controllability and executability of the segmented closure process. Attached Figure Description

[0060] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0061] Figure 1 The flowchart shows a method for generating segmented closed trajectories of a press based on consistent and secure Bayesian optimization proposed in this invention.

[0062] Figure 2 The flowchart shows the data preparation and center pressure drop threshold setting for a press segment closed trajectory generation method based on consistent safety Bayesian optimization proposed in this invention.

[0063] Figure 3 This is a flowchart illustrating the parameter combination range construction and unfavorable parameter combination determination of a press segmented closed trajectory generation method based on consistent safety Bayesian optimization proposed in this invention.

[0064] Figure 4 The flowchart shows the joint calculation of extrusion flow and gas exhaust, along with time-displacement sequence constraints, for a compressor segmented closed trajectory generation method based on consistent safety Bayesian optimization proposed in this invention.

[0065] Figure 5 The scoring network and flowchart of generating the first candidate trajectory using secure Bayesian optimization are presented in the present invention for a method of generating segmented closed trajectories of a press based on consistent secure Bayesian optimization.

[0066] Figure 6The flowchart of the joint calculation and iterative optimization of the press segmented closed trajectory generation method based on consistent safe Bayesian optimization proposed in this invention is as follows:

[0067] Figure 7 This is a schematic diagram of the scoring network structure based on physical constraint fusion for a press segmented closed trajectory generation method based on consistent and secure Bayesian optimization proposed in this invention.

[0068] Figure 8 This is a schematic diagram of the mold cross-section under the traditional closed trajectory (unoptimized);

[0069] Figure 9 This is a schematic diagram of the lower die cross-section of the segmented closed trajectory (after optimization) of the press segmented closed trajectory generation method based on consistent safety Bayes optimization proposed in this invention.

[0070] Figure 10 This is a time-displacement diagram of the segmented closed trajectory of a press segmented closed trajectory generation method based on consistent safe Bayesian optimization proposed in this invention. Detailed Implementation

[0071] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0072] refer to Figures 1 to 10 A method for generating segmented closed trajectories of a press based on consistent and secure Bayesian optimization, characterized by comprising:

[0073] S1. Obtain the force-displacement curve and in-mold pressure signal of a single pressure test, obtain the press force, speed and upper limit of stroke, perform time alignment on the force-displacement curve and in-mold pressure signal, determine the trajectory parameter range and the center pressure drop threshold based on the press force, speed and upper limit of stroke, and obtain the aligned force-displacement curve, aligned in-mold pressure signal, trajectory parameter range and center pressure drop threshold.

[0074] S2. Construct the parameter combination range based on the aligned force-displacement curve and the in-mold pressure signal: First, determine the compaction response range based on the force-displacement curve, and then determine the permeability range based on the in-mold pressure signal. With compaction response and permeability constrained together, examine each combination of compaction response range and permeability range, and eliminate combinations that cannot simultaneously explain the changing trends of force-displacement curve and in-mold pressure signal to obtain the parameter combination range.

[0075] S3. Under the conditions of mold geometry and cavity boundary, establish the joint calculation of extrusion flow and gas exhaust, transform the trajectory parameter range into cavity thickness variation, calculate the change of center pressure with displacement, select the parameter combination that is unfavorable to center exhaust within the parameter combination range, determine the edge sealing start displacement and edge sealing completion displacement, and determine the displacement point where the center pressure reaches the center pressure drop threshold under this unfavorable parameter combination, thus obtaining the edge sealing start displacement, edge sealing completion displacement and unfavorable parameter combination;

[0076] S4. Based on the parameter combination range, the starting displacement of the sealing edge and the finishing displacement of the sealing edge, safe Bayesian optimization is used to determine the segment speed and dwell time within the trajectory parameter range to generate the first candidate trajectory.

[0077] S5. Perform joint calculations on the first candidate trajectory, and determine whether the center pressure drop threshold is reached before the edge sealing displacement begins under unfavorable parameter combinations. If the threshold is met, the target trajectory is obtained; otherwise, iteratively optimize and update the target trajectory within the trajectory parameter range.

[0078] S6. Generate segmented closure control commands based on the target trajectory. The segmented closure control commands include displacement segments, segment velocities, and dwell times.

[0079] In this embodiment, step S1 specifically includes:

[0080] The force-displacement curve of a single pressure test is denoted as... The in-mold pressure signal of one pressure test is recorded as The upper limit of the press force is denoted as The upper limit of the press speed is denoted as The upper limit of the press stroke is recorded as The common start time is recorded as The zero point of displacement is denoted as ,use and right and Perform time alignment, unifying the starting points of the time axis and the displacement axis to the same point. and The aligned force-displacement curves were obtained. Aligned intramold pressure signal Time alignment and Using this as the sole reference, the two signals were calibrated and sampled in the order recorded during a single pressure test to ensure that... and Maintain a consistent time-displacement sequence within the same pressure test;

[0081] according to , and Define the range of trajectory parameters Taking a three-segment closed loop as an example, the displacement is divided into four segments and a vector is formed. Combine the three values ​​of segment speed into a vector. The three values ​​of dwell time are combined into a vector. The minimum stable speed of the press is denoted as ,Will Set as And Set to no more than The value is required. , will each The lower limit is set to and each The upper limit is set to , will each Set the lower limit to zero and each The upper limit is set so as not to cause Exceed The value is set to ensure that it does not trigger. The upper limit of the load for the corresponding displacement segment is denoted as ,according to exist Peak value determination within the interval ,when Exceed The interval The upper limit is and Adjusting between to not make Exceed The value is formed by the above constraints. This allows for segmented displacement, segmented velocity, and dwell time. , and It takes values ​​within the boundary;

[0082] use and Determine the threshold for central pressure drop Record the platform pressure as ,exist The mean value is defined within the stable platform segment. The decay time is denoted as After the plateau phase ended, the pressure was lower than the level for the first time. Time definition Let the slope of the rising segment be denoted as ,by Perform linear fitting on the ascending interval and define the slope of the fitted line. The threshold for determining the slope of the rising segment is denoted as... Defined by the median of the slope fitted during the rising segment of a single pressure test. The lower bound of platform pressure is denoted as Defined by the minimum value within the platform segment ,Will Initial value set to exist The corresponding pressure level, when Less than When Downgraded to ,when Not less than Keep for exist The corresponding pressure level, obtained through the above process, provides a single value on the alignment signal of the same pressure test that can be used for time-displacement sequence constraints. ;

[0083] Will , , and As a rating network The input preparation items will In , and The segmented temporal encoding is used to generate trajectory feature vectors. ,Will , and Incorporate the upper limit value ,Will Incorporating the central pressure drop threshold value The trajectory coding layer of the scoring network is The physical constraint fusion layer is denoted as the input trajectory encoding. When receiving the initial displacement of the sealing edge, the final displacement of the sealing edge, and unfavorable parameter combinations, and... To integrate, so that and , and Enter with a unified representation Through the above implementation, the aligned force-displacement curve is obtained. Aligned intramold pressure signal Establish a consistent data source to ensure the range of trajectory parameters. by , and The upper and lower limits are entered into the trajectory encoding layer of the scoring network, causing the central pressure to decrease by a threshold. A single numerical value is entered into the physical constraint fusion layer for fusion calculation with the edge sealing start displacement, edge sealing completion displacement, and unfavorable parameter combinations. It also supports the determination of whether the center pressure drop threshold is reached before the edge sealing start displacement in subsequent joint calculations using time-displacement order constraints. The above implementation completes data preparation with unified test pressure input, unified equipment upper limit, and unified threshold setting, enabling the scoring network and the data acquisition calculation module to perform iterative optimization of segment speed and dwell time within the trajectory parameter range, and maintain the attainability determination of the center pressure drop threshold under unfavorable parameter combinations.

[0084] In this embodiment, step S2 specifically includes:

[0085] The aligned force-displacement curve is denoted as... The aligned intramold pressure signal is recorded as The range of trajectory parameters is denoted as The threshold for the decrease in central pressure is denoted as Under the premise that the test pressure temperature and mold geometry are fixed, the following is adopted: and Feature extraction and combination testing were performed to construct the compaction response range and permeability range, and a parameter combination range was formed for use by the scoring network.

[0086] Will As input, the slope of the initial segment is extracted and denoted as... ,by The slope is obtained by performing linear fitting on the initial linear interval, and the slope of the nonlinear segment is extracted and denoted as... ,by The slope is obtained by piecewise fitting within the nonlinear enhancement interval, and the position of the load platform is extracted and denoted as... ,by The displacement position is defined upon entering the stable platform segment, and the deformation springback of the unloading segment is extracted and denoted as... Defined by the difference between the termination displacement of the unloading section and the residual displacement after load release, the thickness variation-sensitive range to the load is denoted as... ,by , and Determined within the effective displacement range of a single pressure test The upper and lower bounds are defined to form a directly usable numerical boundary, and the compaction response range is denoted as . ,set up ,set up For nonlinear segments in The slope value of the last segment, and in Boundary checks are performed only when... The above values ​​are retained when the value is greater than zero;

[0087] Will As input, the starting point time of ascent is extracted and denoted as... ,by Defined as the first moment when the baseline enters the rising phase, the peak pressure is extracted and denoted as... Based on the definition of maximum pressure in the rising segment and the plateau segment, the plateau pressure is extracted and denoted as... Defined by the mean of the stable plateau segment, the decay time is extracted and denoted as... After the plateau phase ended, the pressure was lower than the level for the first time. The time definition will be denoted by the capability range as ,by , , and Determining the relationship between timing and amplitude The upper and lower bounds are defined to form directly usable numerical boundaries, and the penetration capability range is denoted as . ,set up ,set up ;

[0088] exist Without participating in the adjustment, and Candidate combinations are generated one by one, in order to and The value of is mapped to calculate the predicted curve, and the predicted force-displacement curve is denoted as . Take from each candidate combination As the slope of the rising segment, we obtain The predicted intramold pressure signal is denoted as follows, based on the order and slope of the rising and plateau segments. Take from each candidate combination As the platform amplitude and maintain and The order of events, obtained To support the judgment, the order of startup, peak value, and platform was analyzed, and the following parameters were extracted. The slope of the rising segment is denoted as ,extract The peak time is recorded as ,extract The peak time is recorded as Synchronous extraction The slope of the rising segment is denoted as The peak time is denoted as ,as well as The peak time is recorded as ;

[0089] Let the slope difference threshold be denoted as The slope quantile difference of the sampling point set between the rising segment and the plateau segment of a single pressure test is used as the threshold value. Let the peak time difference threshold be denoted as... The peak time quantile difference of a single pressure test is used as the setting, and the sequence consistency identifier is denoted as... The order of the rising phase and the platform phase, and the startup and platform phases, should be consistent. Inconsistent settings Let the slope coincidence be denoted as With slope difference not exceeding Set as Exceeding the set The consistency scoring formula is:

[0090] ;

[0091] in, For consistency scoring, dimensionless. For the predicted force-displacement curve The slope of the rising segment, Force-displacement curves after alignment The slope of the rising segment, For the predicted force-displacement curve peak time, Force-displacement curves after alignment peak time, For the predicted intramold pressure signal peak time, For the aligned in-mold pressure signal peak time, The slope difference threshold, The peak time difference threshold, This is a sequence consistency identifier, with a value of [value to be filled in]. or , This is an indicator of slope overlap, with a value of [value missing]. or , , , The weighting coefficient has a value of [value]. , , The penalty coefficient has a value of [value missing]. ,by Based on the screening criteria, retain those that meet the requirements. Combine the given combinations and eliminate those that do not satisfy the condition;

[0092] All combinations that meet the filtering criteria are grouped into a parameter combination range, denoted as . To support the input of the scoring network, the response will be compacted. The position index in is denoted as ,by to The linear normalized value definition will determine the penetration capability in The position index in is denoted as ,by to The linear normalized value is defined as follows: the scoring network is denoted as... Let the physical constraint fusion layer be denoted as ,Will and As two numerical input items, in The displacement at the start of edge sealing and the displacement upon completion of edge sealing. To integrate, so that The parameter combination range is used as a constraint when generating the first candidate trajectory;

[0093] exist The unfavorable parameter combination is determined internally, and denoted as . ,by maximum value and minimum value Corresponding combination definition The index of unfavorable parameter combinations is denoted as ,by and The corresponding values ​​are composed of, , and The target setting for the safety head advance displacement used in subsequent joint calculations and scoring networks, so that in During iterative optimization of segment speed and dwell time, any generated first candidate trajectory is based on... and To constrain the process, it can be determined whether the condition has been met before the edge sealing begins to shift. Through the above implementation, the same pressure test and Common constraints form a framework that covers fluctuations in production batches. It provides directly usable numerical inputs for scoring networks and joint computation with explicit indexes and unfavorable parameter combinations.

[0094] In this embodiment, step S3 specifically includes:

[0095] A joint calculation of extrusion flow and gas exhaust is established under the conditions of mold geometry and cavity boundary. The mold geometry is denoted as... Let the cavity boundary conditions be denoted as The range of trajectory parameters is denoted as The piecewise displacement vector is denoted as The segment velocity vector is denoted as The dwell time vector is denoted as The curve of displacement over time is denoted as The continuous variation curve of cavity thickness over time is denoted as... ,use , and Construct a timeline based on segmented time sequences, calculate the duration of each segment based on the displacement difference and corresponding segment velocity, insert dwell times between segments, and sum the results to obtain the final timeline. ,by and Based on the given initial cavity thickness and upper and lower mold distance mapping, The corresponding closed-loop stroke is converted into a thinning amount, which is calculated point by point along the time axis. This serves as the input for subsequent contact propulsion and center pressure calculations;

[0096] based on The advancement process of edge contact is calculated point-by-point, and the edge gap is denoted as... ,by The given cavity edge geometric spacing and Definition of calculated local thickness difference Scan along the time axis to determine If the value is zero, record the displacement at the first zero value. The displacement at the start of the sealing process is recorded as... Continue sampling and calculating the connectivity of the contact line along the circumferential direction until the gap between all edge sampling points is simultaneously zero. Record the displacement corresponding to this moment. The displacement after sealing is completed is denoted as... The above judgment is based on The execution is based on circumferential discrete points, ensuring that edge sealing events are quantized by displacement;

[0097] Within the parameter combination range, determine the unfavorable parameter combinations used for center exhaust determination, and denote the parameter combination range as... The compaction response range is denoted as The penetration range is denoted as In combinations judged by consistency of changing trends, with and The corresponding combinations are considered as candidate unfavorable combinations, and the unfavorable parameter combinations are denoted as... This choice enables the joint calculation to run under conditions where loads are generated quickly and the outflow capacity is weak, in order to form the most stringent time-displacement sequence constraints;

[0098] exist By considering the relationship between the coupled load and the outward displacement capacity, the change in central pressure with displacement is obtained. This relationship between central pressure and displacement is denoted as... The threshold for the decrease in central pressure is denoted as ,use and The load increment caused by thickness reduction is calculated at discrete time steps, and the load history is formed by accumulating the increments in time sequence. Simultaneously, [the following is used:] and The discharge capacity per unit time is calculated, and this discharge capacity is subtracted from the internal pressure corresponding to the load. This process is then used to progressively update the center pressure level at each time step. Map the center pressure of the time step to the index. This forms a displacement domain representation for threshold reachability. The above calculations are performed within the same discrete time step system to ensure a consistent correspondence between displacement, thickness, and central pressure.

[0099] by To determine the criterion, the earliest position where the central pressure reaches the threshold is identified within the displacement domain. The time axis is discretized with a fixed time step, and the discretization time is denoted as . The discrete value of the central pressure is denoted as The discrete values ​​of displacement are denoted as Instead of continuous scanning, discrete search is used to directly locate the earliest displacement achieved by the threshold at discrete points. The scanning formula is defined as follows:

[0100] ;

[0101] ;

[0102] in, The displacement point at which the central pressure reaches the threshold for a decrease in central pressure, expressed in units of displacement. The displacement versus time curve at discrete times The value at that location is expressed in units of displacement. To ensure the minimum time step index that the central pressure does not exceed the threshold, a dimensionless integer is used. The time step index is a dimensionless integer. For the central pressure at discrete time points The value at that location is in pressure. The threshold for the decrease in central pressure, expressed in pressure. For the first Discrete moments, in units of time. Let be a set of non-negative integers, and let be a dimensionless set. The above definition is uniquely determined by minimizing the discrete index. and generated by joint computation and Direct correspondence avoids the uncertainty caused by continuous domain interpolation;

[0103] Based on time-displacement order constraints, and To make a comparison, the requirements are as follows: Earlier If the conditions are not met, it is determined that there is a risk of the trajectory being blocked, and the trajectory parameter range is backtracked to adjust the segment speed and dwell time. If the conditions are met, it is recorded. , and For subsequent generation, the starting displacement of the sealing edge, the displacement upon completion of the sealing edge, the unfavorable parameter combination, and the displacement point where the center pressure reaches the center pressure drop threshold, along with the center pressure drop threshold, are used as inputs to the scoring network. The scoring network is denoted as... Let the physical constraint fusion layer be denoted as ,exist China Integration , , , and The lead displacement is defined as the difference between the initial displacement of the sealing edge and the displacement point at which the center pressure reaches the threshold of center pressure drop. The lead displacement is denoted as... ,Will As the direct target of the safety head, Within the trajectory parameter range When generating the first candidate trajectory, it explicitly follows the time-displacement order constraint and ensures that the center pressure drop threshold is reached before the edge sealing begins to move under unfavorable parameter combinations. Ultimately, this supports the formation of segmented closure control commands that can be directly issued on the production site.

[0104] In this embodiment, step S4 specifically includes:

[0105] Let the range of parameter combinations be denoted as The initial displacement of the sealing edge is denoted as... The displacement after edge sealing is recorded as The threshold for the decrease in central pressure is denoted as The upper limit of the press force is denoted as The upper limit of the press speed is denoted as The upper limit of the press stroke is recorded as The range of trajectory parameters is denoted as The piecewise displacement vector is denoted as The segment velocity vector is denoted as The dwell time vector is denoted as The index of the parameter combination range in the scoring network is denoted as... and , representing the compaction response and permeability respectively. The values ​​in the above data are encoded into a trajectory feature vector, which is denoted as [the value in the vector is missing here]. ,Include , , , , , , , , , and The segments are arranged in a fixed order according to their timing sequence and constraints for use as network input.

[0106] The scoring network is denoted as Let the physical constraint fusion layer be denoted as ,Will Input the trajectory encoding layer to obtain the trajectory code, then send it into... The displacement at the start of edge sealing, the displacement at the end of edge sealing, the center pressure drop threshold, and the position index of the parameter combination range are integrated to form a joint representation including geometric, material, and equipment boundaries. This joint representation is then fed into the safety assessment layer and the dual-output layer. The dual-output layer includes a safety head and a cycle head. The safety head outputs the mean and uncertainty of the lead displacement, denoted as follows: and The lead displacement is denoted as The mean and uncertainty of the sum of the dwell times of the beat head output are defined as the difference between the displacement at the start of the sealing process and the displacement point where the center pressure reaches the threshold of the center pressure drop. and , used for clock timing within safety constraints;

[0107] The module for calculating collected values ​​is combined with secure Bayesian optimization to achieve the following: and Calculate the lower bound of safety, denoted as . The judgment value is obtained by subtracting uncertainty from the mean. and Calculate time indicators, with time indicators marked as The ranking values ​​were obtained using the mean plus uncertainty method. Using segment speed and dwell time as search variables, a strategy combining segmented increment and local backoff is employed to generate candidate updates, denoted as . For each Calling the scoring network yields , , , ,calculate and and with , and The test load, speed, and stroke do not trigger the upper limit, which will satisfy the requirements. And does not trigger the upper limit Join the safe set, and the safe set contains... Sort the data and keep the smallest. corresponding ;

[0108] After the safety set is determined, fix And select the one with the best sorting. The first candidate trajectory is generated by combining the data, and the displacement segments, segment velocities, and dwell times of the first candidate trajectory are output as follows: , and Before outputting, use again , and Consistency checks are performed to ensure that displacement segments do not exceed the upper limit of travel, any segment speed does not exceed the upper limit of velocity, and the predicted load does not exceed the upper limit of force. Through the above implementation, the scoring network generates a determinable safe output at the physical constraint fusion layer with the displacement starting at the edge, the displacement completing at the edge, the center pressure drop threshold, and the position index of the parameter combination range as constraints. The acquisition value calculation module searches for segment speed and dwell time within the trajectory parameter range based on the safe lower bound entry condition and time index sorting principle. Finally, the first candidate trajectory that satisfies the positive lead displacement and is executable under unfavorable parameter combinations is formed, and the displacement segments, segment speeds, and dwell times are output for subsequent joint calculation and control command generation.

[0109] In this embodiment, step S5 specifically includes:

[0110] Let the piecewise vector of the displacement of the first candidate trajectory be denoted as... Let the segment velocity vector be denoted as Let the dwell time vector be denoted as Let the geometry of the mold be denoted as Let the cavity boundary conditions be denoted as The unfavorable parameter combinations within the parameter combination range are denoted as The upper bound of the compaction response range is denoted as . The lower bound of the penetration range is denoted as The relationship between central pressure and displacement is denoted as: The threshold for the decrease in central pressure is denoted as The displacement point where the central pressure reaches the threshold of central pressure drop is denoted as... The initial displacement of the sealing edge is denoted as... The displacement after edge sealing is recorded as The upper limit of the press force is denoted as The upper limit of the press speed is denoted as The upper limit of the press stroke is recorded as ;

[0111] Will , and Input joint calculation, in and Under constraints, a displacement versus time curve is generated and transformed into a continuous curve of cavity thickness changing with time. The thickness change is denoted as... ,exist General and Mapping to load relationships, and Mapped to outward discharge capacity, the central pressure level is updated hourly and obtained using displacement as an index. Locate the displacement point along the displacement domain that first satisfies the central pressure drop value, and record it as... ;

[0112] Will and Perform a time-displacement order comparison, requiring Earlier Simultaneously, an upper limit check is performed: the load check is based on the maximum load history value obtained from the joint calculation and... For comparison, the maximum load is required to be no greater than Speed ​​verification and The comparison requires that the speed of all segments is no greater than [a certain value]. Itinerary verification and For comparison, the maximum displacement must not exceed [a certain value]. When both the order comparison and the three upper limit checks are met, the first candidate trajectory is confirmed as the target trajectory, and the displacement segments, segment velocities, and dwell times are output as follows: , and ;

[0113] If the condition is not met, the output of the scoring network is used for iterative optimization. The scoring network is denoted as... The mean and uncertainty of the lead displacement of its safety head output are denoted as follows: and The mean and uncertainty of the sum of the beat head output dwell times are denoted as follows: and The lower bound of safety is denoted as ,Will Set as Mark the time indicator as ,Will Set as Within the trajectory parameter range, segment speed and dwell time are used as search variables to generate a new trajectory feature vector, denoted as . ,in To divide the displacement into segments, For segment speed, Duration of stay;

[0114] For each Repeated joint calculations and order determination will New results were obtained through joint calculations and ,and Perform time-displacement sequence comparison, and simultaneously use the maximum value of the load history, and Corresponding to , and Perform upper limit verification by calling the scoring network to obtain... , , , Calculated based on this and , will satisfy And does not trigger the upper limit Join the security set, and in the security set, press Sort the results and select the best sorted result. It may advance to the next round or be the first candidate trajectory in the updated version;

[0115] The process continues iterating until the center pressure drop threshold is reached before the edge sealing begins to displace. Once this condition is met, the target trajectory is output. The displacement segments, segment velocities, and dwell times of the target trajectory are denoted as follows: , and Before outputting, use again , and Consistency checks are performed to ensure that the displacement segments do not exceed the upper limit of the stroke, the speed of any segment does not exceed the upper limit of the velocity, and the predicted load does not exceed the upper limit of the force. Through the above iterative closed loop, the joint calculation, time-displacement sequence constraints, and scoring network output work together within the trajectory parameter range to form a target trajectory that can be directly used for production control.

[0116] In this embodiment, step S6 specifically includes:

[0117] The displacement vector of the target trajectory is denoted as... Each of them For the first The displacement of the segment is denoted by the segment velocity vector as... Each of them For the first The speed of the segment, denoted by the dwell time vector. Each of them The inter-segment dwell time is denoted as the starting displacement of the sealing edge. The displacement after edge sealing is recorded as The threshold for the decrease in central pressure is denoted as The displacement point where the central pressure reaches the threshold of central pressure drop is denoted as... The lead displacement is denoted as Displacement begins with the sealing edge. The difference is defined, and the unfavorable parameter combination in the parameter combination range is denoted as . Used to perform the judgment, the upper limit of the press force is recorded as The upper limit of the press speed is denoted as The upper limit of the press stroke is recorded as ;

[0118] The segmented closure control command is denoted as... Generate basic fields under segmented time series, and The displacement field is denoted as ,by Arranged by segment order, The speed field is denoted as ,by Arranged by segment order, The dwell field is denoted as ,by Arranged in paragraph order, and Calculate the duration for each segment and insert intervals between adjacent segments. The corresponding stops yield a segmented start and end time sequence, denoted as . Write the above basic fields in a fixed order. This forms the basic part of the segmented closed control command that can be sent to the press;

[0119] The starting displacement of the edge sealing and the finishing displacement are combined. The edge sealing field is denoted as ,Include and Set time-displacement order constraints, and denote the constraint field as follows: ,Include , and ,by For the decision item, it is required that... Greater than zero, with For the displacement determination point, the following requirements are required Earlier The above constraints are uniformly encapsulated as Used to determine the order of execution;

[0120] Incorporate unfavorable parameter combinations into the decision-making process, and denote the unfavorable parameter combination field as... ,by When performing the judgment, the equipment upper limit field is recorded as the material and channel parameter selection. ,Include , and During the execution phase, , , , , , and For the input, perform time-displacement order constraint judgment and upper limit check. The rule for upper limit check is: the speed of any segment does not exceed The cumulative displacement is not greater than The maximum value of the load history obtained by joint calculation is no greater than The order constraint rules are as follows: Located in time sequence Previously, and ;

[0121] After the time-displacement sequence constraint is passed, the displacement segments, segment velocities, and dwell times of the segmented closed control command are confirmed, and the confirmed displacement segments are recorded as follows. The confirmed segments are shorthanded as follows: Record the confirmed length of stay as ,by As a reference for execution timing, , and Write the final instruction body, denoted as ,Will The data is sent to the press segment closure execution interface to control the segment speed and dwell time according to the segment start and end times, monitor boundary events according to the sealing field, and make online judgments and abnormal stops according to the constraint field to ensure that even under unfavorable parameter combinations, the requirement that the displacement point corresponding to the center pressure reaching the center pressure drop threshold is earlier than the start of the sealing displacement is still met.

[0122] To ensure that instructions can be executed directly on-site, the field verification process and the issuance process are completed together during the generation phase. Verify equipment capability matching and refuse generation if any limits are exceeded. It also returns the segment indexes of displacement segments and segment velocities for on-site adjustment. Verify the time-displacement order constraint and refuse generation if the constraint is not met. and return and The displacement difference is used for on-site adjustment, to A consistency check is performed between the initial displacement and the final displacement of the edge sealing process; if no valid boundary event is generated between the two, the process is rejected. The timing index of the sealing event is returned for on-site adjustment. The above process is executed in a closed loop during the generation of control instructions, so that the segmented closing control instructions and the unfavorable parameter combinations, time-displacement sequence constraints and equipment upper limits in the parameter combination range form a unified executable object, which directly supports the segmented closing execution of the press.

[0123] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for generating segmented closed trajectories of a press based on consistent, safe Bayesian optimization, characterized in that, include: S1. Obtain the force-displacement curve and in-mold pressure signal of a single pressure test, obtain the press force, speed and upper limit of stroke, perform time alignment on the force-displacement curve and in-mold pressure signal, determine the trajectory parameter range and the center pressure drop threshold based on the press force, speed and upper limit of stroke, and obtain the aligned force-displacement curve, aligned in-mold pressure signal, trajectory parameter range and center pressure drop threshold. S2. Construct the parameter combination range based on the aligned force-displacement curve and the in-mold pressure signal: First, determine the compaction response range based on the force-displacement curve, and then determine the permeability range based on the in-mold pressure signal. With compaction response and permeability constrained together, examine each combination of compaction response range and permeability range, and eliminate combinations that cannot simultaneously explain the changing trends of force-displacement curve and in-mold pressure signal to obtain the parameter combination range. S3. Under the conditions of mold geometry and cavity boundary, establish the joint calculation of extrusion flow and gas exhaust, transform the trajectory parameter range into cavity thickness variation, calculate the change of center pressure with displacement, select the parameter combination that is unfavorable to center exhaust within the parameter combination range, determine the edge sealing start displacement and edge sealing completion displacement, and determine the displacement point where the center pressure reaches the center pressure drop threshold under this unfavorable parameter combination, thus obtaining the edge sealing start displacement, edge sealing completion displacement and unfavorable parameter combination; S4. Based on the parameter combination range, the starting displacement of the sealing edge and the finishing displacement of the sealing edge, safe Bayesian optimization is used to determine the segment speed and dwell time within the trajectory parameter range to generate the first candidate trajectory. S5. Perform joint calculations on the first candidate trajectory, and determine whether the center pressure drop threshold is reached before the edge sealing displacement begins under unfavorable parameter combinations. If the threshold is met, the target trajectory is obtained; otherwise, iteratively optimize and update the target trajectory within the trajectory parameter range. S6. Generate segmented closure control commands based on the target trajectory. The segmented closure control commands include displacement segments, segment velocities, and dwell times.

2. The method for generating a segmented closed trajectory of a press based on consistent, safe Bayesian optimization according to claim 1, characterized in that, S1 specifically refers to: Obtain the force-displacement curve and the in-mold pressure signal of a single pressure test, obtain the press force, speed and upper limit of stroke, and perform time alignment on the force-displacement curve and the in-mold pressure signal, using the common start time and displacement zero point as the alignment reference, to obtain the aligned force-displacement curve and the aligned in-mold pressure signal; Based on the press force, speed, and stroke upper limit, the trajectory parameter range is defined as the boundary of the values ​​of displacement segment, segment speed, and dwell time. The upper and lower limits are set with four values ​​for the three closed displacement segments, three values ​​for the segment speed, and three values ​​for the dwell time, so that the trajectory execution does not trigger the press force, speed, and stroke upper limit, thus forming the trajectory parameter range. Based on the platform pressure and decay time of the aligned intramold pressure signal, combined with the slope of the rising segment of the aligned force-displacement curve, the numerical standard of the center pressure drop threshold is determined, and the center pressure drop threshold is output. The aligned force-displacement curve, aligned intramold pressure signal, trajectory parameter range, and center pressure drop threshold are used as input preparation items for the scoring network, enabling the physical constraint fusion layer to be integrated with the subsequent edge sealing displacement, edge sealing completion displacement, and unfavorable parameter combinations.

3. The method for generating a segmented closed trajectory of a press based on consistent, safe Bayesian optimization according to claim 1, characterized in that, S2 specifically refers to: Using the aligned force-displacement curve as input, the slope of the initial segment, the slope of the nonlinear segment, and the position of the load platform are extracted. The deformation rebound of the unloading segment is used as a supplementary feature to establish the sensitive range of thickness change to load and define the compaction response range. Using the aligned intramold pressure signal as input, the rising start time, pressure peak, plateau pressure and decay time are extracted to establish the penetration capacity range and define the permeability range. Under the premise of fixed test temperature and mold geometry, the compaction response range and the permeability range are combined one by one to generate candidate combinations. The order, slope and peak time of the predicted force-displacement curve in the rising segment and plateau segment are calculated based on the mapping relationship between the thickness change sensitive range of load and the permeability range, as well as the order of the predicted in-mold pressure signal in the start, peak and plateau segments. For each candidate combination, a consistency judgment of the trend of change is applied. The consistency criteria are consistent order, overlapping slope within a set interval, and peak time difference not exceeding a set threshold. Combinations that do not meet the consistency criteria are eliminated, and combinations that meet the consistency criteria are retained to form the parameter combination range. The parameter combination range is encoded as two values: the position index of the compaction response and the permeability within the parameter combination range. These values ​​are used as inputs to the scoring network. The scoring network is then fused with the physical constraint fusion layer, the edge start displacement, the edge completion displacement, and the center pressure drop threshold. This allows the scoring network to use the parameter combination range as a constraint when generating the first candidate trajectory. Within the parameter combination range, the unfavorable parameter combination is determined by the combination corresponding to the minimum value of the permeability range and the maximum value of the compaction response range. The parameter combination range and the unfavorable parameter combination are output, providing constraints for iterative optimization of segment velocity and dwell time within the trajectory parameter range and calculation of the lead displacement of the scoring network head.

4. The method for generating a segmented closed trajectory of a press based on consistent, safe Bayesian optimization according to claim 3, characterized in that, The consistency of the changing trends applied to each candidate combination is quantified using a consistency scoring formula: ； in, For consistency scoring, dimensionless. For the predicted force-displacement curve The slope of the rising segment, Force-displacement curves after alignment The slope of the rising segment, For the predicted force-displacement curve peak time, Force-displacement curves after alignment peak time, For the predicted intramold pressure signal peak time, For the aligned in-mold pressure signal peak time, The slope difference threshold, The peak time difference threshold, This is a sequence consistency identifier, with a value of [value to be filled in]. or , This is an indicator of slope overlap, with a value of [value missing]. or , , , The weighting coefficient has a value of [value]. , , The penalty coefficient has a value of [value missing]. ,by Based on the screening criteria, retain those that meet the requirements. Combine the given combinations and eliminate those that do not meet the requirements.

5. The method for generating a segmented closed trajectory of a press based on consistent, safe Bayesian optimization according to claim 1, characterized in that, S3 specifically refers to: Under the conditions of mold geometry and cavity boundary, a joint calculation of extrusion flow and gas exhaust is established. The displacement segment, segment velocity and dwell time in the trajectory parameter range are used as inputs to generate a continuous change curve of cavity thickness over time according to the segment time sequence, thus forming the thickness change. The process of advancing edge contact is calculated point by point based on thickness change. It is determined whether the edge gap is zero. The first position where it is zero is recorded as the start of edge sealing displacement. The displacement of the contact line closing to form a continuous loop is recorded as the edge sealing completion displacement. Within the parameter combination range, candidate unfavorable combinations are generated by combining the lower bound of the permeability range and the upper bound of the compaction response range. Among the combinations that pass the consistency judgment of the change trend, the candidate is selected as the unfavorable parameter combination for the determination of central venting. Under unfavorable parameter combinations, the thickness variation is mapped to the values ​​in the compaction response range as a load relationship, and the thickness variation is mapped to the values ​​in the permeability range as an outward discharge capacity, thus coupling the change of central pressure with displacement. Using the central pressure drop threshold as the criterion, the change of central pressure with displacement is scanned along the displacement axis to locate the displacement point that first meets the central pressure drop threshold, and the displacement point where the central pressure reaches the central pressure drop threshold is output. The time-displacement sequence of the edge sealing start displacement, edge sealing completion displacement, and center pressure reaching the center pressure drop threshold is compared. The displacement point where the center pressure reaches the center pressure drop threshold is required to be earlier than the edge sealing start displacement, thus forming a time-displacement sequence constraint for trajectory screening. The edge sealing start displacement, edge sealing completion displacement, unfavorable parameter combination, and the displacement point where the center pressure reaches the center pressure drop threshold, along with the center pressure drop threshold, are used as input items for the scoring network. They are fused in the physical constraint fusion layer, and the difference between the edge sealing start displacement and the displacement point where the center pressure reaches the center pressure drop threshold is defined as the lead displacement, which is used by the safety head to generate the first candidate trajectory within the trajectory parameter range.

6. The method for generating a segmented closed trajectory of a press based on consistent, safe Bayesian optimization according to claim 5, characterized in that, The step of using a central pressure drop threshold as the criterion, scanning the change of central pressure along the displacement axis, and locating the displacement point that first meets the central pressure drop threshold is quantified using a scanning formula: ； ； in, The displacement point at which the central pressure reaches the threshold for a decrease in central pressure, expressed in units of displacement. The displacement versus time curve at discrete times The value at that location is expressed in units of displacement. To ensure the minimum time step index that the central pressure does not exceed the threshold, a dimensionless integer is used. The time step index is a dimensionless integer. For the central pressure at discrete time points The value at that location is in pressure. The threshold for the decrease in central pressure, expressed in pressure. For the first Discrete moments, in units of time. Let be a set of non-negative integers, and let be a dimensionless set. The above definition is uniquely determined by minimizing the discrete index. and generated by joint computation and Direct correspondence avoids the uncertainty caused by continuous domain interpolation.

7. The method for generating a segmented closed trajectory of a press based on consistent, safe Bayesian optimization according to claim 1, characterized in that, S4 specifically refers to: The parameter combination range, edge sealing start displacement, edge sealing completion displacement, center pressure drop threshold, press force, speed and stroke upper limit are encoded together with the displacement segments, segment speeds and dwell times in the trajectory parameter range into a trajectory feature vector, which is then input into the scoring network. The trajectory feature vector is fused with the above constraints in the physical constraint fusion layer. The mean and uncertainty of the lead displacement are output by the safety head, and the mean and uncertainty of the sum of dwell times are output by the beat head. The data acquisition calculation module adopts safe Bayesian optimization, calculates the safety lower bound and time index based on the safety head and the beat head, and searches for segment speed and dwell within the trajectory parameter range to make the safety lower bound greater than zero and not trigger the upper limit of press force, speed and stroke. Sort the trajectory within the safety set by time index, select the trajectory feature vector to generate the first candidate trajectory, and output the displacement segments, segment velocity and dwell time.

8. The method for generating a segmented closed trajectory of a press based on consistent, safe Bayesian optimization according to claim 1, characterized in that, S5 specifically refers to: The displacement segment, segment velocity and dwell input of the first candidate trajectory are calculated together and transformed into thickness change under the conditions of mold geometry and cavity boundary. Under unfavorable parameter combination, the change of center pressure with displacement is calculated, and the displacement point where the center pressure reaches the center pressure drop threshold is located. The displacement point is compared with the edge sealing start displacement in time-displacement sequence. At the same time, it is checked whether the upper limit of press force, speed and stroke is not triggered. It is determined whether the center pressure drop threshold is met before the edge sealing start displacement. If it is met, the first candidate trajectory is confirmed as the target trajectory. If the conditions are not met, the safety head and beat head outputs of the scoring network are used as inputs to the acquisition value calculation module. Based on the safety lower bound and time index, the displacement segment, segment velocity and dwell time are adjusted within the trajectory parameter range to generate a new trajectory feature vector. Repeatedly perform joint calculations and order determination on the new trajectory feature vectors, and select the updated trajectory feature vectors in the safety set according to the time index until the center pressure drop threshold is reached before the edge starts to displace, and output the target trajectory.

9. The method for generating a segmented closed trajectory of a press based on consistent, safe Bayesian optimization according to claim 1, characterized in that, S6 specifically refers to: The displacement, segment velocity, and dwell time of the target trajectory are arranged in segment sequence to generate the basic fields of the segmented closure control command; The edge sealing start displacement and edge sealing completion displacement are incorporated into the segmented closure control command. Time-displacement sequence constraints are set, and advance displacement is used as the judgment item. It is required that the advance displacement is greater than zero and the displacement point corresponding to the center pressure reaching the center pressure drop threshold is earlier than the edge sealing start displacement. The unfavorable parameter combinations in the parameter combination range are used for execution judgment. Under the constraints of press force, speed and stroke upper limit, the time-displacement sequence constraint judgment of the segmented closed control command is performed. After the time-displacement sequence constraint is passed, the displacement segment, segment speed and dwell time of the segmented closure control command are confirmed and used for the execution of the press segmented closure.

Images