CN-121978925-A - Press segmentation closed track generation method based on consistency safety Bayesian optimization

CN121978925ACN 121978925 ACN121978925 ACN 121978925ACN-121978925-A

Abstract

The invention provides a press sectional closed track generation method based on consistency safety Bayesian optimization, which utilizes a force-displacement curve and a mold internal pressure signal obtained by one-time pressure test, extracts a compaction response and permeability parameter range under the constraint of an upper limit of a press, establishes a combined calculation model of extrusion flow and gas discharge under the boundary conditions of a mold and a cavity, determines the starting and finishing displacement of a sealed edge and the displacement point that the central pressure reaches a descending threshold value, introduces the physical constraint into a scoring network, adopts the safety Bayesian optimization to determine the speed and the residence time of each section in the track parameter range, generates a target sectional closed track which meets the central pressure before the sealed edge and reduces to the threshold value without triggering the upper limit of equipment, and outputs a corresponding control instruction.

Inventors

CHEN YIWEI
FAN CONGZE
ZHU SHU
SONG WENZHE
ZHENG JINGHUA
LI YUE

Assignees

南京航空航天大学
东华大学

Dates

Publication Date: 20260505
Application Date: 20260114

Claims (9)

1. The press segmentation closed track generation method based on consistency safe Bayesian optimization is characterized by comprising the following steps of: s1, acquiring a force displacement curve and an in-mold pressure signal of one-time pressure test, acquiring upper limits of pressure, speed and travel, performing time alignment on the force displacement curve and the in-mold pressure signal, determining a track parameter range and a center pressure drop threshold according to the upper limits of the pressure, speed and travel, and obtaining an aligned force displacement curve, an aligned in-mold pressure signal, a track parameter range and the center pressure drop threshold; S2, constructing a parameter combination range according to a force displacement curve of Ji Hou and a in-mold pressure signal, namely determining a compaction response range according to the force displacement curve, determining a permeability range according to the in-mold pressure signal, jointly restricting compaction response and permeability, checking the combination of the compaction response range and the permeability range one by one, and eliminating the combination which cannot simultaneously explain the change trend of the force displacement curve and the in-mold pressure signal to obtain the parameter combination range; S3, establishing joint calculation of extrusion flow and gas discharge under the geometric and cavity boundary conditions of the die, converting a track parameter range into cavity thickness change, calculating change of central pressure along with displacement, selecting a parameter combination unfavorable to central exhaust in a parameter combination range, determining edge sealing starting displacement and edge sealing finishing displacement, and determining a displacement point of the central pressure reaching a central pressure reduction threshold under the unfavorable parameter combination to obtain edge sealing starting displacement, edge sealing finishing displacement and unfavorable parameter combination; s4, determining the segment speed and the stay in the track parameter range by adopting safe Bayesian optimization according to the parameter combination range, the edge sealing starting displacement and the edge sealing finishing displacement, and generating a first candidate track; S5, carrying out joint calculation on the first candidate track, judging whether the first candidate track reaches a central pressure reduction threshold before the edge sealing starts to shift under the condition of unfavorable parameter combination, obtaining a target track when the first candidate track is satisfied, and carrying out iterative optimization and updating on the target track within a track parameter range when the first candidate track is not satisfied; S6, generating a sectional closing control instruction according to the target track, wherein the sectional closing control instruction comprises displacement sections, section speed and residence time.
2. The press segment closed trajectory generation method based on the consistent safety bayesian optimization of claim 1, wherein S1 is specifically: Acquiring a force displacement curve of primary pressure test and a die internal pressure signal, acquiring the upper limit of the pressure, the speed and the travel of the press, and performing time alignment on the force displacement curve and the die internal pressure signal, wherein the common starting moment and the displacement zero point are used as alignment references, so as to obtain an aligned force displacement curve and an aligned die internal pressure signal; Defining a track parameter range as a value boundary of displacement segmentation, segment speed and residence time according to the upper limit of the pressure, speed and travel, setting the upper limit and the lower limit of three values of three-segment closed displacement segmentation four values, segment speed three values and residence time three values, so that the track execution does not trigger the upper limit of the pressure, speed and travel, and forming the track parameter range; Determining a numerical standard of a central pressure drop threshold according to the platform pressure and the decay time of the aligned in-mold pressure signals and combining the slope of the ascending section of the aligned force displacement curve, and outputting the central pressure drop threshold; And taking the aligned force displacement curve, the aligned in-mold pressure signal, the track parameter range and the central pressure drop threshold value as input preparation items of a scoring network, so that the physical constraint fusion layer can start displacement with the edge sealing and finish combination fusion of the displacement and unfavorable parameters.
3. The press segment closed trajectory generation method based on the consistent safety bayesian optimization of claim 1, wherein S2 is specifically: Taking the aligned force displacement curve as input, extracting the slope of the initial section, the slope of the nonlinear section and the position of the load platform, taking the deformation rebound of the unloading section as a supplementary feature, establishing a sensitive interval of thickness variation to load, and defining a compaction response range; Taking the aligned in-mold pressure signals as input, extracting rising point time, pressure peak value, platform pressure and decay time, establishing a throughput capacity interval, and defining a permeability range; on the premise of fixed pressure test temperature and die geometry, a compaction response range and a penetration capacity range are combined one by one to generate candidate combinations, and the sequence, slope and peak time of a predicted force displacement curve in an ascending section and a platform section and the sequence of a predicted in-die pressure signal in starting, peak and platform are calculated according to the mapping relation between a sensitive section of thickness change to load and a passing capacity section; Applying a change trend consistency judgment to each candidate combination, wherein consistency standards are that the sequence is consistent, the slopes are coincident in a set interval, the peak time difference value does not exceed a set threshold value, eliminating combinations which do not meet consistency, and reserving the combinations which meet consistency to form a parameter combination range; Encoding the parameter combination range into two numerical values of the position index of compaction response and permeability in the parameter combination range, taking the numerical values as input items of a scoring network, and fusing a physical constraint fusion layer with edge sealing starting displacement and edge sealing finishing displacement with a central pressure drop threshold value, so that the scoring network takes the parameter combination range as constraint when generating a first candidate track; determining unfavorable parameter combinations in the parameter combination range by using the combination corresponding to the minimum value of the permeability range and the maximum value of the compaction response range, and outputting the parameter combination range and the unfavorable parameter combinations to provide constraints for iterative optimization of segment speed and stay and advance displacement calculation of scoring network safety heads in the track parameter range.
4. A press segment closed trajectory generation method based on consistent safety bayesian optimization according to claim 3, wherein the trend consistency decisions applied to each candidate combination are quantified using a consistency scoring formula: ; Wherein, the For consistency scoring, dimensionless, For predicted force displacement curve The slope of the rising section is such that, For aligned force displacement curves The slope of the rising section is such that, For predicted force displacement curve Is set at the peak time of (a), For aligned force displacement curves Is set at the peak time of (a), For predicted intra-mode pressure signals Is set at the peak time of (a), For aligned in-mold pressure signals Is set at the peak time of (a), As a threshold value for the slope difference, Is the peak time difference threshold value, For order consistency identification, the value is Or (b) , For the sign of slope superposition, take the value as Or (b) , 、、 The value of the weight coefficient is , 、 The penalty coefficient is taken as the value To For screening criteria, remain satisfied And eliminating unsatisfied combinations.
5. The press segment closed trajectory generation method based on the consistent safety bayesian optimization of claim 1, wherein S3 is specifically: Establishing joint calculation of extrusion flow and gas discharge under the boundary conditions of the die geometry and the die cavity, taking displacement segmentation, segment speed and stay in the track parameter range as input, and generating a continuous change curve of the die cavity thickness along with time according to a segmentation time sequence to form thickness change; Calculating the advancing process of edge contact from displacement point to displacement point based on thickness change, judging whether the edge gap is zero, recording the position of the first zero as edge sealing starting displacement, and continuously calculating the displacement of the contact line closing to form a continuous ring as edge sealing finishing displacement; generating candidate unfavorable combinations in a parameter combination range by combining the lower boundary of the permeability range with the upper boundary of the compaction response range, and selecting the candidates from the combinations judged by the change trend consistency as unfavorable parameter combinations for central exhaust judgment; under the unfavorable parameter combination, mapping the value in the thickness change and compaction response range into a load relation, mapping the value in the thickness change and permeability range into discharge capacity, and coupling to obtain the change of the central pressure along with displacement; Scanning the change of the central pressure along with the displacement along a displacement axis by taking the central pressure drop threshold as a judgment standard, positioning a displacement point which firstly meets the central pressure drop threshold, and outputting the displacement point of which the central pressure reaches the central pressure drop threshold; Performing time-displacement sequence comparison on the initial displacement of the edge sealing, the completion displacement of the edge sealing and the displacement point of the center pressure reaching the center pressure drop threshold value, wherein the displacement point of the center pressure reaching the center pressure drop threshold value is required to be earlier than the initial displacement of the edge sealing, so that time-displacement sequence constraint for track screening is formed; And taking the displacement points of the edge sealing starting displacement, the edge sealing finishing displacement and the unfavorable parameter combination and the central pressure reaching the central pressure drop threshold value together with the central pressure drop threshold value as input items of a scoring network, fusing the two points at the physical constraint fusion layer, and defining the advance displacement by the difference between the edge sealing starting displacement and the displacement point of the central pressure reaching the central pressure drop threshold value, wherein the advance displacement is used by a safety head for generating a first candidate track in the track parameter range.
6. The method for generating a segmented closed path of a press based on consistent safety bayesian optimization according to claim 5, wherein the step of scanning the change of the center pressure along the displacement axis with the displacement by using the center pressure drop threshold as a decision criterion and locating the displacement point which satisfies the center pressure drop threshold for the first time adopts a scanning formula to perform quantization decision: ; ; Wherein, the The unit is displacement amount for the displacement point that the central pressure reaches the central pressure drop threshold value, At discrete moments for displacement versus time The numerical value of the position is expressed as displacement, The minimum time step index for satisfying the central pressure not higher than the threshold value is a dimensionless integer, Is a time step index, is a dimensionless integer, At discrete moments for central pressure The numerical value of the pressure is expressed as pressure, Is the central pressure drop threshold, in units of pressure, Is the first A discrete time, in units of time, As a non-negative integer set, as a dimensionless set, the definition being uniquely determined in a minimized form of discrete index And is generated by joint calculation And Direct correspondence, avoiding uncertainty brought by continuous domain interpolation.
7. The press segment closed trajectory generation method based on the consistent safety bayesian optimization of claim 1, wherein S4 is specifically: encoding a parameter combination range, edge sealing starting displacement, edge sealing finishing displacement, a central pressure drop threshold value, a pressure machine force, a speed and a travel upper limit together with displacement segmentation, segment speed and stay in a track parameter range into track feature vectors, and inputting the track feature vectors into a scoring network; fusing the track feature vector and the constraint in a physical constraint fusion layer, outputting the average value and the uncertainty of the displacement of the lead by a safety head, and outputting the average value and the uncertainty of the sum of the residence time by a beat head; The acquisition value calculation module adopts safe Bayesian optimization, calculates a safety lower bound and a time index according to a safety head and a beat head, searches for the speed and the stay of a section in a track parameter range, and ensures that the safety lower bound is larger than zero and does not trigger the upper limit of the pressure, the speed and the travel; And sequencing according to time indexes in the safety set, selecting the trace feature vector to generate a first candidate trace, and outputting displacement segmentation, segment speed and residence time.
8. The press segment closed trajectory generation method based on the consistent safety bayesian optimization of claim 1, wherein S5 is specifically: The displacement segmentation, segment speed and stay input of the first candidate track are calculated in a combined mode, the displacement segmentation, segment speed and stay input are converted into thickness change under the boundary conditions of the die geometry and the cavity, the change of the central pressure along with the displacement is calculated under the unfavorable parameter combination, and a displacement point that the central pressure reaches a central pressure drop threshold value is located; Comparing the displacement point with the edge sealing starting displacement in time-displacement sequence, checking the upper limit of the non-triggering pressure, speed and travel at the same time, judging whether the central pressure drop threshold is met before the edge sealing starting displacement, and confirming the first candidate track as a target track when the central pressure drop threshold is met; when the safety head and the beat head of the grading network are not satisfied, the safety head and the beat head of the grading network are output as the input of an acquisition value calculation module, and displacement segmentation, segment speed and stay are adjusted within the track parameter range according to the safety lower bound and the time index, so that a new track feature vector is generated; and repeating the joint calculation and the sequence judgment on the new track feature vector, and sequencing and selecting the updated track feature vector according to the time index in the safety set until the central pressure drop threshold is reached before the edge sealing starts to shift, and outputting the target track.
9. The press segment closed trajectory generation method based on the consistent safety bayesian optimization of claim 1, wherein S6 is specifically: the displacement segmentation, the segment speed and the residence time of the target track are arranged according to the segmentation time sequence, and a basic field of a segmentation closing control instruction is generated; The edge sealing starting displacement and the edge sealing finishing displacement are combined into a sectional closing control instruction, time-displacement order constraint is set, the displacement of the advance is adopted as a judging item, the displacement of the advance is required to be larger than zero, and a displacement point corresponding to the central pressure reaching the central pressure drop threshold value is earlier than the edge sealing starting displacement; The unfavorable parameter combinations in the parameter combination range are used for executing judgment, and time-displacement order constraint judgment is carried out on the segmented closure control instruction under the constraint of the pressure, the speed and the upper limit of the stroke; after the time-displacement sequence constraint determination passes, confirming the displacement segment, segment speed and residence time of the segment closure control command, and using the displacement segment, segment speed and residence time for press segment closure execution.

Description

Press segmentation closed track generation method based on consistency safety Bayesian optimization Technical Field The invention relates to the technical field of control of a thermoplastic composite material compression forming process, in particular to a method for generating a segmented closed track of a press based on consistency safety Bayesian optimization. Background After heating to a molten state, the thermoplastic composite is shaped by closing the mold with a press. In the closing process, the material is subjected to thickness compaction, resin and gas flow in the cavity and are discharged through the exhaust channel, and the pressure in the cavity changes along with displacement. In actual production, a multistage displacement closing mode is often adopted, and the compaction force, the in-mold pressure, the exhaust effect and the forming beat are coordinated by setting closing speeds and residence time in different stages. However, material batch fluctuation, die geometry difference and equipment capacity limit make the setting of the segmented closed trajectory highly dependent on experience, and it is difficult to quantitatively evaluate whether the central area is sufficiently exhausted and the rationality of the edge sealing time sequence in time. In the prior art, aiming at the control of the closing process of a thermoplastic composite material press, a method is generally adopted, wherein one method is based on a force-displacement curve or a die internal pressure curve obtained by limited pressure test, the closing speed and the stroke are manually adjusted to ensure that the maximum load does not exceed the upper limit of the press, and technological parameters are gradually corrected by observing defects of a test piece, the other method is used for establishing an extrusion flow or simplified cavity pressure model under the given geometric condition of the die, the closing stroke and the speed are used as input, the change of the load and the pressure with time is predicted, the sectional track is set according to the predicted change, and the method is combined with a die internal pressure sensor to set a pressure threshold value or a platform section interval, so that simple feedback control or staged stay is implemented on the closing process. These schemes generally avoid overload to some extent and monitor the pressure level in the die for guiding the formulation of the segmented closure process. However, the prior art is often directly based on a single pressure test curve setting process, lacks parameterization description of uncertainty of compaction response and permeability, is difficult to form a parameter combination range covering batch fluctuation, meanwhile, extrusion flow and gas discharge are usually separated or simplified, the relation between edge sealing start-stop displacement and central pressure evolution is not uniformly calculated under given die geometry and cavity boundary conditions, the position of ' central pressure is difficult to accurately position in a displacement domain to fall to a safe level ', and a systematic method for simultaneously incorporating the time sequence constraint of equipment force, speed, stroke upper limit and ' exhaust before sealing is also lacking when a sectional closed track is generated. Thus, a method of creating a segmented closed path for a press that addresses the above-described deficiencies of the prior art is a problem that one skilled in the art would need to address. Disclosure of Invention The invention aims to provide a method for generating a segmented closed track of a press based on consistency safety Bayesian optimization, which aims to solve the core technical problems of how to generate a segmented closed track on the premise of considering material compaction response and permeability fluctuation, die geometry, die cavity boundary conditions and upper limit of press force/speed/travel in the forming process of a thermoplastic composite material press, so that gas in the central area of the die cavity is reliably discharged before edge sealing, and a segmented closed instruction which can be directly used for press control is formed according to the segmented closed track. According to the embodiment of the invention, the press subsection closed track generation method based on consistency safe Bayesian optimization comprises the following steps: s1, acquiring a force displacement curve and an in-mold pressure signal of one-time pressure test, acquiring upper limits of pressure, speed and travel, performing time alignment on the force displacement curve and the in-mold pressure signal, determining a track parameter range and a center pressure drop threshold according to the upper limits of the pressure, speed and travel, and obtaining an aligned force displacement curve, an aligned in-mold pressure signal, a track parameter range and the center pressure drop threshold; S2, constructing a parameter combi