A method of forming a recycled plastic pallet by zoned cooling

Abstract

The application discloses a regenerative plastic tray partition cooling forming method and belongs to the technical field of regenerative plastic injection molding control. In order to solve the problem of deformation and excessive forming waiting caused by different cooling of each part of the tray, the application obtains the flow characteristics and thermal shrinkage mapping of mixed batches, estimates the state of the skin layer and the core of each cooling partition, predicts the residual shrinkage, shrinkage difference and accumulated deformation of the region and converts them into the adjusted functional size, determines the executable cooling action and the earliest feasible common demolding time under the constraints of size, residual shrinkage and ejection bearing, selects the mold times with consistent thermal observation, corrects the batch shrinkage parameters by using the geometric increment of two times after the demolding of the same tray, re-calculates after verification and feedback, and realizes the technical effects of reducing the differential shrinkage affecting the support and assembly, reducing the unnecessary in-mold waiting and reducing the feedback misadjustment.

Description

Technical Field

[0001] This invention relates to the field of recycled plastic injection molding control technology, and in particular to a method for partitioned cooling molding of recycled plastic trays. Background Technology

[0002] Plastic pallets are used for cargo carrying, stacking, and assembly with side panels. Their panels, reinforcing ribs, and legs vary in thickness, local volume, and heat dissipation paths. When recycled plastics and reclaimed plastic sprues are mixed for injection molding, the flow characteristics and shrinkage properties of the mixed batch vary from batch to batch. This can further cause asynchronous shaping processes in thin-walled panels, rib junctions, and thick-walled areas such as legs. Therefore, the cooling process must simultaneously consider raw material properties, mold heat exchange conditions, and the final functional dimensions of the pallet.

[0003] When using a uniform cooling temperature and time, thin-walled areas have already formed a load-bearing structure, while thick-walled nodes may continue to shrink. Differential shrinkage after demolding can easily cause panel warping, non-coplanarity of the support base surfaces, and dimensional misalignment of the enclosure. If only the overall mold cooling time is extended, areas that have already met the shaping requirements may continue to wait. Judging demolding solely based on the mold surface temperature or regional temperature difference is insufficient to directly reflect the core condition, accumulated deformation, and assembly and support dimensions under specified adjustment conditions.

[0004] Dimensional feedback across mold cycles is also affected by changes in medium transport, loop heat transfer, and mold temperature measurements. If dimensional deviations caused by thermal anomalies are directly attributed to material shrinkage parameters, or if the material model is corrected using only single-moment geometric deviations, thermal faults and early geometric biases may be incorporated into subsequent cooling decisions. Therefore, it is necessary to verify the consistency of thermal observations before updating material parameters and ensure that the measured increments after demolding correspond to the predicted increments under the same thermal history and external mold conditions.

[0005] Therefore, there is a need for a partitioned cooling molding method for recycled plastic pallets that can overcome the shortcomings of the existing technology. Summary of the Invention

[0006] One objective of this invention is to propose a method for partitioned cooling molding of recycled plastic pallets. Addressing the problems of asynchronous cooling, functional dimension deviation, and increased in-mold waiting when using uniform cooling conditions in existing technologies, this invention proposes a technical solution that uses the flow characteristics of mixed batches and thermal shrinkage mapping as material basis, forms cooling partitions according to structure and controllable loops, estimates the skin and core states based on thermal observation, and predicts the adjusted functional dimensions. Under the constraints of functional dimensions, residual shrinkage, and ejection bearing capacity, the earliest feasible common demolding time and partitioned cooling action are determined. Subsequent batch predictions are corrected by screening for consistency in thermal observation and comparing the geometric increments of two moments for the same pallet. This invention has the technical effects of reducing support and assembly-related differential shrinkage, reducing unnecessary in-mold waiting, and reducing cross-mold feedback misadjustment.

[0007] This invention provides a method for partitioned cooling molding of recycled plastic pallets, comprising:

[0008] S1. Obtain the flow characteristics and thermal shrinkage mapping of the mixed batch of recycled plastic and sprue material, form cooling zones according to the tray structure and controllable medium circuit, and establish the relationship between regional shrinkage and panel deformation, coplanarity of the bottom surface of the support legs and the matching dimensions of the enclosure.

[0009] S2. Utilize the mold temperature, pressure, and medium thermal observation during the injection molding process, combined with the circuit thermal process parameters, to estimate the skin and core state of each zone;

[0010] S3. Based on the state, pressure history and thermal shrinkage mapping, predict the remaining shrinkage, shrinkage difference and accumulated deformation of the region corresponding to the candidate cooling action and demolding time, and obtain the functional dimension index under the specified adjustment conditions through the correlation.

[0011] S4. Under the constraints of functional dimensions, residual shrinkage and ejection load, determine the earliest feasible common demolding time and the medium inlet temperature, flow rate and switching sequence of each zone from the executable cooling actions, execute them segment by segment and re-estimate and predict, and open the mold when the release conditions are met.

[0012] S5. Save the batch shrinkage parameters and loop thermal process parameters separately. For the module where the thermal observation and prediction are consistent, compare the measured geometric increments at two moments after the same tray is demolded with the corresponding increments predicted based on the thermal history of that module. Correct the shrinkage parameters after the batch. After verification, use them to feed back the prediction of the remaining shrinkage of the same batch after arrival and the cooling action and common demolding time for recalculation.

[0013] Optionally, S1 includes: obtaining the flow characteristics under specified material temperature and shear conditions according to the batch number of the mixed batch, and establishing a thermal shrinkage mapping based on the recorded holding pressure history and temperature and dimensional changes under cooling conditions for samples of different thicknesses, and mapping the temperature, pressure history, thickness and cooling time to the shrinkage amount and the change of shrinkage amount after demolding over time.

[0014] The cooling zones are determined based on the thickness of the panel, rib intersection, support legs and enclosure mating parts, the heat dissipation path and the independently adjustable range of the circuit. The common circuit coverage parts that cannot be independently adjusted are treated as common action constraints.

[0015] By utilizing the tray geometry and heat shrinkage calibration results, a correspondence between regional shrinkage and functional measuring point displacement is established. The influence of shrinkage differences between upper and lower panels, adjacent areas, and associated supports on functional measuring point displacement is retained as a correlation between regional shrinkage and panel deformation, support bottom coplanarity, and enclosure fit dimensions.

[0016] Furthermore, the heat shrinkage mapping selects the applicable thickness, holding pressure history, and cooling condition calibration range according to the flow characteristics of the mixed batch, and records the sample size change in the same reference direction to determine the shrinkage curve under each calibration condition.

[0017] The shrinkage at the moment of demolding and the shrinkage over time after demolding are parameterized separately. The shrinkage parameter after batch is limited to the shrinkage amplitude and time scale parameter after demolding, and the thickness and thermal history applicable range corresponding to the parameter are recorded.

[0018] For mixed batches or molding conditions that exceed the calibration range, supplementary calibration should be performed first, and then the applicable heat shrinkage mapping should be generated based on the supplementary calibration results.

[0019] Optionally, S2 includes: collecting the regional mold temperature of each cooling zone, the pressure at the representative location, and the supply water temperature, return water temperature, and flow rate of each medium loop and aligning them according to the sampling time;

[0020] The medium transport delay is determined based on the effective volume and flow rate of the circuit, and the supply water temperature and return water temperature are converted to the corresponding mold heat exchange period.

[0021] The heat exchange obtained from the flow rate and the temperature difference between the supply and return water is combined with the heat storage of the mold, the thermal contact between the plastic and the mold, and the heat transfer between the zones and included in the stratified heat balance. The initial state of the zone after filling is determined by combining the flow characteristics, and the predicted state of the stratified heat balance is corrected by the zone mold temperature. The temperature of the skin and core of each zone and the degree of curing are output.

[0022] The circuit thermal process parameters include the effective volume for stratified thermal balance, mold thermal capacity, thermal contact parameters, and heat transfer parameters between zones.

[0023] Optionally, S3 includes: for candidate cooling actions and demolding times, starting from the current state of the skin and core of each zone, combining pressure history, thermal shrinkage mapping and candidate cooling actions to deduce the in-mold thermal shrinkage process, and combining the ambient temperature and support conditions after demolding to deduce the specified adjustment conditions, wherein the specified adjustment conditions are the predetermined adjustment temperature, adjustment duration and support method.

[0024] The predicted shrinkage of each region from the candidate demolding time to the end of the specified adjustment conditions is taken as the region's residual shrinkage. The absolute amount of each region is constrained, and the shrinkage difference between related regions and between the upper and lower layers of the panel is calculated.

[0025] Based on the accumulated deformation caused by cumulative shrinkage and the additional shrinkage after demolding, the final coordinates of the functional measuring points are obtained by relating the area shrinkage with the panel deformation, the coplanarity of the bottom surface of the support, and the fit dimensions of the enclosure. This avoids replacing the accumulated deformation with only the remaining shrinkage difference, and the functional dimension indicators are calculated accordingly.

[0026] Furthermore, the functional measuring points include measuring points on the panel surface, measuring points at the bottom of each support leg, and measuring points at the mating parts of the enclosure. The final coordinates of the functional measuring points are uniformly represented in the coordinate system defined by the pallet design datum.

[0027] The panel deformation index is obtained by the normal offset of the panel surface measuring points relative to the design panel reference plane; the coplanarity of the bottom surface of the support is obtained by the minimum distance between two parallel planes that encompass all the bottom measuring points of the support; and the enclosure mating dimensions are obtained by the spacing of the corresponding measuring points of the enclosure mating parts along the assembly direction.

[0028] The three types of indicators are compared with the tolerances given in the pallet design, and the residual shrinkage of the area is compared with the upper limit of the residual shrinkage of the area calibrated according to the adjusted dimensional stability requirements.

[0029] Optionally, S4 includes: generating executable cooling actions based on the adjustable range, rate of change, valve switching time, shared water supply capacity, and loop limitations that cannot be adjusted independently for each loop inlet temperature and flow rate, and forming a candidate demolding time sequence according to the control cycle;

[0030] Inspect the actions and time combinations one by one in sequence, retain the combination that simultaneously satisfies the panel deformation tolerance, the coplanarity tolerance of the support bottom surface, the dimensional tolerance of the enclosure, the upper limit of the remaining shrinkage of each area, and the ejection load constraint, and determine the first time with the retained combination as the common demolding time.

[0031] The ejection bearing constraint is determined by the material bearing capacity corresponding to the skin and core states, the ejection force on the pallet, and the allowable ejection deformation.

[0032] The inlet temperature, flow rate and switching sequence of the corresponding combination are executed segment by segment. The status and combination are updated in the next control cycle. When the common demolding time is reached and the update result still meets all constraints, the mold opening is released.

[0033] Furthermore, the candidate demolding time sequence and executable cooling action are regenerated in each control cycle according to the current state of the equipment, and the earliest feasible common demolding time is limited to the first feasible time within the current candidate demolding time sequence and executable cooling action range;

[0034] If no combination of constraints is simultaneously satisfied, keep the mold closed, execute the next control cycle according to the cooling action verified by the trial mold within the allowable range of the equipment, and then re-evaluate after expanding the candidate demolding time sequence;

[0035] If no feasible combination can be found after the maximum dwell time specified in the equipment process procedure, automatic mold opening is prohibited and a process abnormality signal is issued.

[0036] Optionally, S5 includes: forming thermal observation prediction values ​​based on loop thermal process parameters, comparing the difference between the observed values ​​and the corresponding prediction values ​​of each loop flow rate, supply and return water temperature difference and regional module temperature with the allowable deviation of the thermal process calibration item by item, and only using modules that are all within the allowable deviation as batch learning candidates.

[0037] For the same pallet of the candidate mold, at two predetermined times after demolding, the signed coordinates or spacing of the panel, the bottom of the support legs and the mating parts of the enclosure are measured according to the same physical reference and coordinate direction, forming the measured geometric increment of the later time relative to the previous time.

[0038] The thermal history of candidate modules and the external temperature and support conditions between two predetermined times are used to predict the corresponding increments at the same location, in the same direction, and at the same time. The residuals between the measured geometric increments and the corresponding increments are used to correct the batch shrinkage parameters in the thermal shrinkage mapping, while keeping the loop thermal process parameters out of the residual fitting.

[0039] After the correction is completed within the physical boundaries and update limits of the parameters, the geometric increment prediction error is verified using the retained samples that were not involved in the fitting. Only after the verification is passed can the correction parameters be used, and the absolute size check under the specified adjustment conditions is retained.

[0040] Furthermore, before correction, trial changes are applied to the pre-selected batch shrinkage parameters respectively, and a sensitivity matrix is ​​formed by the ratio of the corresponding incremental change to the trial change. Correction is only performed when the sensitivity matrix is ​​full column rank and the number of conditions does not exceed the allowable number of conditions calibrated according to the measurement noise; otherwise, the batch shrinkage parameters are frozen and subsequent candidate modes are accumulated.

[0041] The correction is achieved by performing constrained least squares fitting on the geometric increment residual. The physical boundaries of the parameters and the single update limit are taken from the calibration range of the batch sample. The sample verification requirement is retained to ensure that the corrected prediction error does not exceed the preset allowable error and is not greater than the error before correction.

[0042] When thermal observations and predictions are inconsistent, the corresponding module is prohibited from participating in the batch shrinkage parameter update. The medium transport, heat exchange and module temperature measurements are checked, and the batch learning candidate screening is resumed after the loop thermal process parameters are recalibrated and the thermal observation consistency test is passed.

[0043] The verified parameter version is bound to the batch number, data source module number, and enabled module number for prediction and calling in the same batch after feedback arrives. When the absolute size check fails, the corresponding pallet is isolated and the process review is triggered.

[0044] The beneficial effects of this invention are:

[0045] 1. This invention combines batch heat shrinkage mapping with the state of the skin and core of the region and pressure history to predict the absolute residual shrinkage, the shrinkage difference of the associated region and the accumulated deformation, and then converts them into panel deformation, coplanarity of the bottom surface of the support feet and enclosure fit dimensions under specified adjustment conditions. This makes the cooling decision correspond to the support and assembly requirements of the tray, and reduces the omissions caused by judging the shaping state by only using mold temperature or temperature difference.

[0046] 2. Within the operable actions and ejection bearing conditions of the equipment, the present invention combines and filters the inlet temperature, flow rate and switching sequence, selects the first feasible common demolding time in the candidate time sequence, and updates the prediction after each segment is executed, so that the molding process that has met the functional constraints can be released in time, reducing unnecessary in-mold dwell time caused by waiting for local thick-walled nodes.

[0047] 3. This invention stores the loop thermal process parameters and batch shrinkage parameters separately, uses thermal observation consistency to screen material learning modules, uses the geometric increment of the same tray at two time points under the same physical reference for feedback, and verifies the constraint update through identifiability, parameter boundaries, amplitude limiting and retained samples, so that thermal faults and early geometric biases are not easily erroneously absorbed into the batch shrinkage parameters; at the same time, it retains absolute size checks to avoid incremental prediction meeting the requirements but masking absolute size non-compliance. Attached Figure Description

[0048] 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:

[0049] Figure 1 This is a flowchart illustrating the overall process of the recycled plastic pallet partition cooling molding method of the present invention.

[0050] Figure 2 This is a flowchart of the batch shrinkage parameter feedback correction in S5 of the present invention. Detailed Implementation

[0051] 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.

[0052] refer to Figures 1-2 A method for partitioned cooling molding of recycled plastic pallets, comprising:

[0053] S1, obtain the flow characteristics and thermal shrinkage mapping of the mixed batch of recycled plastic and sprue, form cooling zones according to the tray structure and controllable medium circuit, and establish the relationship between regional shrinkage and panel deformation, coplanarity of the bottom surface of the support legs and the matching dimensions of the enclosure.

[0054] S2, by using the mold temperature, pressure and medium thermal observation during the injection molding process, combined with the circuit thermal process parameters, to estimate the skin and core state of each region;

[0055] S3. Based on the state, pressure history and thermal shrinkage mapping, predict the remaining shrinkage, shrinkage difference and accumulated deformation of the area corresponding to the candidate cooling action and demolding time, and obtain the functional dimension index under the specified adjustment conditions through the correlation.

[0056] S4, under the constraints of functional dimensions, residual shrinkage and ejection load, determines the earliest feasible common demolding time and the medium inlet temperature, flow rate and switching sequence of each zone from the executable cooling actions, executes segment by segment and re-estimates and predicts, and opens the mold when the release conditions are met;

[0057] S5. The batch shrinkage parameters and loop thermal process parameters are saved separately. For the module where the thermal observation and prediction are consistent, the measured geometric increments at two moments after the same tray is demolded are compared with the corresponding increments predicted based on the thermal history of that module. The shrinkage parameters after the batch are corrected. After verification, they are used to feed back the prediction of the remaining shrinkage of the same batch after arrival, as well as the cooling action and the recalculation of the common demolding time.

[0058] In this specific embodiment, S1 includes:

[0059] This specific implementation method uses a nine-legged pallet injection molded from a mixture of recycled polypropylene and sprue of the same material at a mass ratio of 90:10 as the processing object. A structure with a length of 1200 mm, a width of 1000 mm, a nominal panel thickness of 4 mm, and an equivalent thickness of 8 mm at the rib intersection nodes is selected. The actual thickness of the mating parts of the legs and enclosure is retained according to the design model. The controller establishes molding records using the batch number of the mixed material, the mold number, and the mold sequence number, binding the weighing ratio, incoming material identification, flow characteristics, calibration range, and parameter version. If the mixing ratio or raw material source changes, a new batch is established, and the shrinkage parameters are not applied across batches.

[0060] The controller obtains the flow characteristics of the mixed batch material under specified material temperature and shear conditions according to the batch number, and first eliminates batch number missing and sensor abnormal records. Before sample preparation, the apparent viscosity corresponding to shear rates of 100, 500 and 1000 per second is obtained by capillary rheology measurement at a material temperature of 230 degrees Celsius. The viscosity unit is Pascal-second. Each condition is measured three times and the median value is taken. The three viscosity values ​​and their corresponding conditions constitute the flow characteristics. In this specific embodiment, the flow characteristics, along with the sample filling time and pressure arrival time, are stored in the batch calibration record. The same material temperature is used in the production filling process. The heat shrinkage mapping selects the applicable thickness, holding pressure history and cooling condition calibration range according to the flow characteristics of the mixed batch material. The corresponding heat shrinkage mapping is selected only when all viscosity values ​​are within the calibrated envelope and the mixing ratio is consistent. If any one of them exceeds the range, the automatic release based on the mapping is suspended and the sample calibration is supplemented.

[0061] Calibration was performed using specimens with thicknesses of 4, 6, 8, and 12 mm, each with a gauge length of 100 mm along the flow direction and perpendicular to the flow direction. Combinations of pressure holding platforms of 30, 45, and 60 MPa, held for 10 and 20 seconds, were set, along with inlet conditions of 25, 35, and 45 degrees Celsius for the cooling medium. At least three independent specimens were prepared for each combination and each release point described below. The pressure sampling sequence from the end of filling, surface and center temperatures, mold opening time, time of complete release of mold cavity constraints, the rapid measurement records described below, and the gauge length in the same direction at 10 minutes, 60 minutes, 4 hours, 24 hours, and 48 hours after demolding were recorded. The specified adjustment conditions were 23 degrees Celsius, 24 hours after demolding, and three low-friction supports, with a temperature control deviation not exceeding 1 degree Celsius. The support positions were recorded and remained consistent in subsequent measurements. These conditions pertain to the calibration scheme of this embodiment, and the specific mapping coefficients were fitted from the measurement data obtained by this scheme.

[0062] The controller uses the mold cavity gauge length as a fixed length reference. It subtracts the current specimen gauge length from the mold cavity gauge length and then divides it by the mold cavity gauge length to obtain a positive value representing a dimensionless dimensional shrinkage observation. This observation includes the combined effects of constraint release, elastic rebound, and subsequent shrinkage, and cannot be directly used as the inherent strain of the constrained structural model. Following the release test and inversion process described below, the inherent shrinkage curve within the mold is obtained separately. The cumulative inherent shrinkage within the mold, the dimensional baseline after release at the moment of demolding, and the newly added inherent shrinkage parameters outside the mold are saved and labeled separately. S3 only inputs the inherent shrinkage into the structural model. The baseline is used for verification; the in-mold curve takes thickness, two gauge length directions, local temperature, time elapsed after filling, holding pressure duration, and pressure history as inputs. In addition to saving the entire sequence, the pressure history saves the pressure-time integral in the intervals of 0 to 5 seconds, 5 to 10 seconds, 10 to 20 seconds after filling, and 20 seconds to gate freezing, in megapascals-seconds; piecewise linear interpolation between adjacent calibration nodes is used, and each input must be surrounded by a node. The pressure after freezing is no longer extrapolated to be the effective holding pressure. Candidate conditions that exceed the thickness, temperature, or any historical integral envelope are transferred to supplementary calibration;

[0063] To obtain cumulative shrinkage information at different times within the mold, for each of the above-mentioned thickness, holding pressure, and cooling combinations, independent release samples were arranged at 20, 30, 45, 60, 90, 120, and 180 seconds, with the end of filling as the zero point. Samples at 35, 75, and 150 seconds were used as a verification group not participating in the fitting process. Release must not occur earlier than the end of holding pressure and gate freezing, and must meet the load-bearing condition that allows the sample to be completely removed. Data points where the sample cannot be completely removed or where the demolding temperature exceeds the calibration range of the post-shrinkage time ratio are not considered acceptable. Two records were simultaneously recorded for each sample. The out-of-plane coordinates of the gauge length, center and four corners, and surface and center temperatures are measured using synchronous binocular high-speed non-contact measurement with a resolution of no more than 0.02 mm. Data is continuously acquired from the trigger signal that completely releases the mold cavity constraints for 3 seconds. The delay of the first measurement that can reliably determine the gauge length and out-of-plane coordinates without rigid body motion must not exceed 0.2 seconds. Subsequently, records are acquired for 10 minutes to 48 hours. The time stamp and camera scale are verified using a rigid scale and no-load motion. The time, temperature and coordinates of each frame are saved. The first frame is not directly recorded as the zero-delay demolding dimension.

[0064] The aforementioned inherent shrinkage within the mold was obtained through constraint release inversion of the same batch of samples: First, the elastic modulus, Poisson's ratio, relaxation time, thermophysical properties, and mold wall contact and friction parameters identified by independent mold trials were fixed. During the inversion process, these parameters were not allowed to be adjusted simultaneously to compensate for shrinkage errors. Starting from the initial hydrostatic pressure state corresponding to the mold cavity geometry and measured cavity pressure at the end of filling, the three-layer structural model described in this step was used to apply the mold cavity contact and pressure holding boundaries segment by segment according to the measured pressure and temperature history. Then, the constraints were released at the moment of complete release and the rebound was calculated. The value to be estimated is the cumulative intrinsic shrinkage value in two directions at the effective interpolation node of the thermal shrinkage mapping. The reference value at the end of filling is fixed at zero. The time nodes are selected as 0, 10, 20, 30, 60, 120 and 180 seconds. The temperature node takes the measured temperature of the corresponding sample and its layered thermal calculation value. Repeated nodes are merged and the linear interpolation weight of adjacent effective nodes is used. Each layer calls the common mapping according to its own temperature and pressure history. The average length change measured by the whole sample is not repeatedly assigned to the three layers, and interpolation is not performed on the temperature and history combination that is not surrounded by effective nodes.

[0065] The inversion is performed jointly on the two gauge lengths and five out-of-plane coordinates of all fitted groups. The goal is to minimize the sum of squares of the difference between the predicted and measured coordinates after normalization of the standard deviation of repeated measurements for each measurement channel. The cumulative shrinkage node is limited to the dimensionless range of 0 to 0.05 and does not decrease with the cooling process in the same actual cooling history. To handle the shrinkage that is still developing between release and the first frame, the amplitude and time scale of the newly added out-of-mold shrinkage are used as auxiliary unknowns. The equivalent time is calculated along the measured out-of-mold temperature. The following saturation relationship and the structural relaxation model are used to fit the record from the first frame to 48 hours. During the inversion, the data is extrapolated frame by frame from the actual release time. Therefore, the predicted shrinkage and rebound include the measurement delay. The zero gauge length at the demolding time is the output of the continuous forward model at zero equivalent time after release and in the state of unconstrained equilibrium, rather than taking the 10-minute dimension as the demolding dimension. If the residual of the first frame is significantly affected by inertial vibration, the corresponding frame is discarded and the earliest effective delay is checked to be no more than 0.2 seconds. Otherwise, the piece is not used for inversion.

[0066] The specific fitting method employs bounded nonlinear least squares. Each estimator is normalized within its permissible range before calculating the numerical difference sensitivity. The subsequent contraction amplitude range is 0 to 0.05, and the initial time scale range is 1 second to 10^7 seconds. If the parameter reaches the upper or lower bound and cannot form an independent confidence interval, it is considered unidentified. The normalized sensitivity matrix must have full column rank and a condition number not exceeding 100. Rank determination is based on a cutoff of one millionth of the largest singular value. If these conditions are not met, adjacent time or temperature nodes are first merged to reduce degrees of freedom, and then release time points or change the cooling boundaries of the upper and lower molds are added. To increase the difference in interlayer response, regularization alone cannot be used to declare that the in-mold curve has been identified; the original quantitative measurement and independent verification rules are still used under the supplementary boundary. If the identifiability cannot be restored, the automatic release will not be enabled for the thickness and thermal history range; for the curves obtained by refitting according to the positive and negative disturbances of measurement noise, the changes in the gauge length and out-of-plane coordinates of the two directions at the verification time point shall not exceed 0.10 mm, and the maximum residual of each coordinate of the effective release samples at 35, 75 and 150 seconds that did not participate in the fitting shall not exceed 0.10 mm. Otherwise, supplementary calibration is required. Interpolation curve smoothing cannot replace verification.

[0067] Based on the inversion results, the shrinkage at the demolding moment and the shrinkage over time after demolding are parameterized separately. The batch shrinkage parameters are limited to the shrinkage amplitude and time scale parameters after demolding. The in-mold curve is first determined and frozen through the aforementioned joint inversion and independent release verification. Then, under the conditions of freezing the curve, contact, and relaxation parameters, the out-of-mold amplitude and time scale are refitted using full-time records after the actual release moment of each piece, forming the formal out-of-mold parameters. Gauge length changes are first converted into observed increments based on a fixed mold cavity gauge length reference, and then relaxation springback and newly added inherent shrinkage are separated using the same structural model. Therefore, the following text... The post-contraction amount obtained from the gauge length curve refers to the amount after the structure correction is completed; the curve from zero to the first safe release point is the in-mold input combined with the subsequent release response inversion, and does not falsely claim to be a direct size measurement. Its nodes must also pass the above-mentioned rank, noise disturbance and reserved release test. Early nodes without constraints cannot be freely extrapolated; the original size, measurement delay, structural correction value, effective nodes, frozen in-mold version and out-of-mold fitting version are retained in the record respectively. S5 only reads the amplitude and time scale that are allowed to be updated in the last item, and does not modify the frozen in-mold curve or the zero-time constraint inversion method;

[0068] For each thickness, orientation, and demolding thermal history calibration unit, the additional shrinkage after demolding adopts a saturation development relationship. ,in This represents the dimensionless shrinkage increase relative to the instant of demolding, calculated from the change in gauge length in the same direction, with values ​​ranging from zero to the corresponding amplitude. This represents the shrinkage amplitude of the unit after demolding, a dimensionless, non-negative parameter, obtained by fitting the gauge length curve after demolding of the sample. This represents the cumulative equivalent settling time since demolding, in seconds, with an initial value of zero. This represents the positive time scale parameter of the unit, in seconds, fitted by the same curve; denoted as the natural exponential function, its independent variable is dimensionless, and the contraction amplitude and time scale are two types of batch post-contraction parameters that can be corrected by S5. The in-mold curve coefficients do not participate in this correction.

[0069] The equivalent settling time is accumulated segment by segment by multiplying the actual duration in seconds of each sampling period by the dimensionless time conversion factor corresponding to the material temperature of that period. At 23 degrees Celsius, this factor is fixed at 1. Isothermal dimensional samples of the same batch are supplemented at 40, 60, 80, and 100 degrees Celsius to make the positive time ratio fitting conversion factor so that the isothermal shrinkage curve coincides with the 23-degree Celsius curve. Linear interpolation is performed between nodes, and the demolding temperature must not exceed the calibration range of this ratio and thermal history. Each material parameter record includes thickness, orientation, pressure history range, demolding skin and core temperature range, in-mold curve, amplitude, time scale, time conversion factor, fitting residual, and version. The controller integrates the equivalent time on the skin and core temperature trajectories respectively, and then uses the post-shrinkage relationship of the corresponding unit.

[0070] Cooling zones are determined based on the design entity and water circuit diagram. The upper side of the panel, the lower side of the panel, the intersection of ribs, the support legs and the enclosure are respectively formed into thermal calculation areas. When the local thickness difference exceeds 2 mm or the heat dissipation path crosses the rib node, the thermal zone is split. The thermal zone and the physical loop are connected by an association table. The table records the area number, volume, layer thickness, adjacent area number, heat exchange area, loop number and independent valve group number. When two thermal zones share a branch that cannot be adjusted independently, their thermal states are still estimated separately, but their inlet temperature, main valve command and switching sequence must be the same. The branch flow is allocated according to the measured resistance. It is forbidden to mistakenly treat the thermal calculation zone as an independent execution channel.

[0071] The correlation between regional shrinkage and functional measuring points adopts a linear incremental structural model that preserves interlayer differences. The panel thickness direction is divided into upper skin layer, core layer and lower skin layer. The ribs and legs adopt the same three-layer discretization. The reference thickness of each layer is 20%, 60% and 20% of the local thickness, respectively. The inherent shrinkage in two directions calibrated by the following constraint inversion is used as the element inherent strain. The shortening is done with negative strain input. The laminate and solid element stiffness are assembled according to the tray design grid. The mold cavity contact constraint is applied in the in-mold stage. The mold cavity constraint is removed in the demolding stage and the specified support is applied. The displacement of the panel, legs and enclosure measuring points after equilibrium is solved. The elastic modulus, Poisson's ratio and stress relaxation time with temperature change are obtained by tensile, transverse strain and constant strain holding tests of the same batch of samples. Interpolation is performed according to temperature nodes. In a single relaxation period, the deviatoric stress of the previous period is first decayed according to the natural exponential, and then the stress caused by the elastic strain increment of the current period is added. The relaxation time is always positive. The volume stress is calculated by volume strain and volume modulus.

[0072] For online use, a dimensionless trial shrinkage of 0.0001 is applied to each layer in two directions. The structural model described above is used to solve for the change in displacement of the measuring points relative to the undisturbed state. The result is then divided by the trial shrinkage to establish a displacement sensitivity column in millimeters. Each column retains the coupled response of the upper and lower panels, adjacent areas, and associated supports, and cannot be compressed into the average temperature difference of each area. Online, the system is re-linearized at the current temperature, contact state, and cumulative stress. When the single predicted shrinkage increment exceeds the trial linear range, it is broken down into increments of no more than 0.0001 and solved piecewise. When the contact state changes, the corresponding sensitivity is re-established. The correlation is verified using the displacement of three types of measuring points in the independent test model. The system is only activated if the maximum coordinate residual does not exceed 0.2 millimeters. Otherwise, the mesh is refined, the corresponding thickness and temperature calibration are supplemented, and the system is re-verified.

[0073] Using the panel design plane as the height reference, the tray design length direction as the first axis, the width direction as the second axis, and the panel normal as the third axis, 25 measuring points are set in 5 rows and 5 columns on the panel. Five measuring points are set at the four corners and the center of each support bottom surface. Two pairs of measuring points are set on each mating edge of the enclosure, corresponding to each other along the assembly direction. The design coordinates of each measuring point, the design panel reference plane, the support position, and the mating distance tolerance are stored in the geometric reference record. In this implementation, the upper limit of the absolute value of the panel normal offset is set to 3 mm, the upper limit of the coplanarity of the support bottom surface is set to 2 mm, and the allowable deviation of the enclosure mating distance relative to the design value is ±1 mm. The controller encapsulates the material parameter record, the area loop association table, the geometric reference record, and the structural model into a batch configuration, which is stored in the local process database for S2 initialization and S3 prediction. All parameters are marked with the source identifier of the sample or mold data. Fields with incomplete calibration must not participate in the release judgment with default values.

[0074] In this specific embodiment, S2 includes:

[0075] The controller reads the S1 batch configuration and uses the filling end signal as the zero point of the current mold time. The regional mold temperature is obtained by embedded temperature sensors near the two sides of the panel, the rib intersection nodes, the support legs, and the enclosure area. The pressure at the representative position is obtained by the corresponding cavity pressure sensor. Each physical loop is set with supply water temperature, return water temperature, and flow rate acquisition channels. The thermal state update interval is 0.1 seconds. The pressure is sampled at 0.01 seconds and then integrated over time and summarized to the thermal state time grid. Each data entry includes the channel number, mold number, time stamp, value, unit, and validity flag. The temperature unit is degrees Celsius, and the flow rate acquisition unit is liters per minute. If a channel is missing for more than two consecutive thermal state intervals, the current prediction is invalidated and the mold remains closed.

[0076] Before the trial molding, the sensor was calibrated using a constant temperature reference and zero pressure conditions. Then, the supply water temperature and flow rate were stepped on the closed mold without plastic. The changes in supply and return water were recorded. Combined with the volume measurement after the loop was emptied and refilled, the effective volume of the three parts were identified: the supply water measurement point to the inlet of the heat exchange section, the heat exchange section body, and the heat exchange section outlet to the return water measurement point. When the flow rate changed, the historical measured volumetric flow rate was integrated over time to find the position where the cumulative transport volume reached the corresponding effective volume as the arrival time. At a constant flow rate, this time interval is the effective volume divided by the volumetric flow rate. The supply water temperature and return water temperature were converted to the corresponding mold heat exchange period so that the supply and return water were matched to the same fluid heat exchange period passing through the mold, instead of directly subtracting the two temperatures at the same sampling time. For the recent period when the return water measurement has not yet been obtained, only the model prediction is retained. After the return water arrives, it is backtracked, corrected, and recalculated to the current time.

[0077] For the aligned heat exchange period, the water-side heat extraction power is calculated according to... Calculation, where This represents the heat power transported by the medium in the physical circuit, calculated based on the aligned inlet and outlet temperature differences, in watts. It is calculated from observations of the aligned medium, and a positive value indicates an increase in the enthalpy of the return medium relative to the supply medium. This indicates the density of the medium, expressed in kilograms per cubic meter. The specific heat capacity of the medium is expressed in joules per kilogram of Kelvin. Both values ​​are obtained by interpolation from the physical property records of the medium at the average water temperature and are kept positive. This represents the effective volumetric flow rate, expressed in cubic meters per second (m³ / s). It is converted from a reading in liters per minute (L / min) by dividing by 60,000. During normal heat exchange, it is greater than zero. This represents the temperature difference between the return water temperature and the supply water temperature during the same heat exchange period, expressed in Kelvin, and its value is equal to the corresponding Celsius temperature difference. Multiplying this power by the length of the heat exchange period in seconds yields the heat transported by the medium, expressed in Joules. In steady state and when the heat storage of the medium is negligible, the corresponding mold heat extraction occurs. During a step transition, the heat storage changes at the medium nodes and the inlet and outlet enthalpy flows must be used together to calculate the wall heat extraction. The transport temperature difference power cannot be directly used as the instantaneous outflow term of all mold walls. This measured calculation is only used for the verification of processes that have already occurred. Future heat extraction is generated by the forward heat exchange model.

[0078] The thermal state model treats the total enthalpy of the upper skin layer, core, lower skin layer, and adjacent mold nodes of each region as independent states. The temperature is derived from the enthalpy-temperature relationship of each node. Thermal connections are established between plastic nodes, between plastic and mold, and between adjacent mold nodes. The heat flow of each connection is calculated by multiplying the temperature difference by the effective thermal conductivity, with the unit of thermal conductivity being watts per Kelvin. The effective thermal conductivity between plastic and mold is obtained by multiplying the contact heat transfer coefficient by the contact area. The contact heat transfer coefficient is calibrated segmentally according to the holding pressure and interface temperature. Every 0.1-second interval, the net heat obtained by subtracting the output heat flow from the input heat flow of the node is added to the node enthalpy at the previous moment using the backward Euler method, and the new temperature is calculated by reverse calculation from the material enthalpy-temperature curve. The heat capacity of the mold node is calculated by multiplying the mass by the specific heat capacity. The actual heat transfer flow between the mold and the medium is used as the outflow term of the mold node. The water-side heat is not deducted again for plastic nodes.

[0079] Each physical loop's in-mold heat exchange section is divided into six series-connected, fully mixed medium nodes along the flow direction. The sum of the effective volumes of each node equals the identified heat exchange section volume. The medium mass is calculated by multiplying the density by the effective volume. The total enthalpy is obtained by integrating the medium's physical properties with a fixed reference temperature as zero. The inlet temperature is first transported to the first node via the water supply pipeline. The outlet temperature of the upstream node serves as the inlet temperature of the next node. The net heat flow at each node is composed of the mass flow rate multiplied by the difference between the inlet and the specific enthalpy of this node, plus the heat flow transferred from each connected mold wall. The mass flow rate is calculated by multiplying the medium density by the volumetric flow rate. The heat flow from each mold wall to the medium node... The heat transfer at each point is calculated by subtracting the medium temperature from the mold wall temperature and then multiplying by the effective thermal conductivity. The unit of thermal conductivity is watts per Kelvin. The heat transfer is recorded in the mold wall and medium nodes with opposite signs. The same contact area is recorded only once. The heat flow of each associated mold wall in a shared loop is added at the medium node without being recorded repeatedly for the entire loop heat. The total enthalpy of the mold, plastic, and medium is updated using the same backward Euler step. Finally, the medium node temperature is used as the outlet temperature of the heat exchange section. The return water measurement point is then predicted by the return water volume transport. The pipe volume is only used for pure transport, and the mixed volume of the heat exchange section is only used for heat storage at the medium node. The two are not repeatedly included in the time delay.

[0080] The thermal conductivity from the mold wall to the medium is identified in the empty mold step under inlet conditions of 6, 9, and 12 liters per minute and 25, 35, and 45 degrees Celsius. The goal is to minimize the combined timing error of the mold temperature and return water temperature. The area distribution of the six nodes is fixed according to the waterway geometry. Each thermal conductivity is positive and interpolated piecewise according to the flow rate and medium temperature. Independent molds with material are used for verification. The thermal conductivity, medium volume, and pipeline transport parameters belong to the loop thermal process version and cannot be modified in S5 material learning. If the candidate flow rate or temperature exceeds the calibration envelope or the simultaneous thermal balance iteration fails to converge, the candidate becomes invalid. Normal candidates maintain positive flow rates. If a flow interruption is detected, automatic prediction and release are stopped and the mold is kept closed. After doubling the number of discrete nodes to twelve and halving the time step, it is required that the difference between the mold temperature and the return water temperature does not exceed their respective measurement resolutions. Otherwise, further refinement and re-verification are performed, and then the discrete specifications used for control are locked.

[0081] The enthalpy-temperature curve and degree of curing were obtained from differential scanning calorimetry (DSC) tests of the same batch. Test nodes of 5, 10, and 20 Kelvin per minute were set according to the sample cooling rate, and supplementary tests were conducted to cover the local cooling rate range of the production process. The degree of curing was defined as the proportion of the latent heat released in the exothermic crystallization stage to the total latent heat of that stage, and was limited to between zero and one. Zero was taken at the liquid end and one was taken at the end of crystallization. Sensible heat and latent heat were included in the enthalpy-temperature curve to avoid adding latent heat again. When the local cooling rate was between two test nodes, the corresponding enthalpy-temperature curve and curing curve were interpolated. If the rate was outside the range, the candidate state was not included in the automatic release and calorimetric calibration was performed first.

[0082] Effective thermal conductivity, mold thermal capacity, and interlayer thermal conductivity are identified by combining the cooling processes of the empty mold step, the sample with temperature measurement, and the temperature decay curves at different inlet temperatures. The goal is to minimize the weighted square error of the predicted mold temperature, sample internal temperature, and loop heat relative to the measured values. Each residual is first divided by its respective repeated measurement standard deviation to eliminate unit differences. The parameter constraints are that the effective volume, heat capacity, and thermal conductivity are all positive, and the contact area and layer mass are fixed to the design geometric values. The mold used for identification is separated from the verification mold. The loop thermal process record is stored by mold number and loop number, including effective volume, heat capacity, contact parameters, interval thermal conductivity, temperature measurement location, and version. It is not mixed with the shrinkage amplitude and time scale after batch.

[0083] The initial temperature distribution after filling is derived by the same layered model along the measured filling sequence. Flow characteristics are used to select adjacent viscosity calibration conditions and interpolate to obtain the melt arrival delay and local inlet temperature correction values ​​for each zone. The arrival delay is checked when the cavity pressure first continuously exceeds 1 MPa. The initial melt temperature is the measured material temperature plus the corresponding correction value. The nodes not reached by the melt maintain the mold temperature before filling. During filling, each layer continuously exchanges heat with the mold. The enthalpy, temperature and degree of solidification are retained as the initial state until the end of filling. All layers cannot be directly set to the same material temperature. The pressure history is independently saved by the original pressure sequence for S3 to call.

[0084] Regional mold temperature correction employs extended Kalman state estimation. The state vector is uniformly composed of the total enthalpy of each plastic layer, mold node, and medium heat transfer node relative to their respective fixed reference temperatures, all in joules. Temperature and degree of curing are quantities derived from the enthalpy-temperature curve; temperature is no longer treated as a separate independent state. First, the node enthalpy and its error covariance are predicted based on the heat dissipation equilibrium. Then, the predicted enthalpy is used to obtain the mold temperature prediction through enthalpy-temperature transformation and measurement point interpolation. The observation residual in Kelvin units is formed by subtracting the predicted value at the same location from the measured value of the embedded mold temperature. Numerical difference is performed on the enthalpy state transition to obtain the dimensionless state transition Jacobian. The observation Jacobian includes the derivative of temperature with respect to node enthalpy and the measurement point interpolation weights, in Kelvin per joule. The state covariance and process noise covariance are in joule squares, and the measurement noise covariance is in Kelvin squares. The variance is first projected onto the measurement space by the observed Jacobian and its transpose, and then the measurement noise covariance is added. The gain per joule (Kelvin) is obtained by multiplying the cross-covariance of state and measurement by the inverse of the sum matrix. The gain is multiplied by the temperature residual to obtain the enthalpy correction per joule. The covariance is updated using the Joseph form according to the same observed Jacobian. If the inverse matrix cannot be stably obtained, no observation update is performed and the state is marked as invalid. The initial and process enthalpy errors are converted into joules by independent initialization verification and one-step predicted temperature difference through the local derivative of the total enthalpy of each node with respect to temperature. The measurement noise is statistically analyzed by repeated isothermal measurements. The square of the temperature resolution is only used as the lower limit of the diagonal element of the measurement covariance. If it is necessary to set the lower limit of the enthalpy state covariance, the temperature resolution is first multiplied by the derivative of the total enthalpy of the corresponding node with respect to temperature and then squared. After updating, the temperature and curing degree are recalculated from the node enthalpy.

[0085] After each update, check whether the enthalpy-temperature conversion of all nodes is within the calibration range, and compare the algebraic sum of the medium heat, mold heat storage and plastic enthalpy change with the allowable thermal balance residuals of the empty mold and the sample calibration. If the state exceeds the upper limit, it is marked as failing. The controller writes the area temperature, degree of curing, cumulative pressure history, estimation error range, effective observation cutoff time and loop parameter version into the thermal state cache of this mold. S3 only uses the effective cache to make predictions. S5 saves the one-step prediction value before correction for thermal observation consistency screening. The correction result that has absorbed the same observation cannot replace the independent prediction residual.

[0086] In this specific embodiment, S3 includes:

[0087] The controller copies the current enthalpy, temperature, curing degree, and pressure history of each layer from the S2 thermal state cache, reads the S1 material parameter record and structural model, and establishes an independent prediction copy for each cooling action and candidate demolding time requested by S4. The thermal and pressure history that has occurred in the copy is fixed. The future inlet temperature and flow rate are deduced according to the candidate actions. The unfinished pressure holding continues according to the current process setting trajectory and is limited to the calibrated pressure history range. The candidate path does not modify the actual cache. The prediction step size in the mold is kept at 0.1 seconds. After demolding outside the mold, 5 seconds are taken in the first hour, and 60 seconds are taken thereafter until the end of the adjustment. Before the step size is enabled, it is compared with the prediction of the step size that is halved in each segment. If the difference of any functional index exceeds 0.01 mm or the difference of regional shrinkage exceeds 1% of the limit of that region, it continues to be halved. The verified step size is saved according to the corresponding thermal history range. The calculation results are stored with the current mold number, candidate number, prediction start point, and parameter version as indexes.

[0088] Before the candidate demolding moment, the future temperature of each zone is calculated according to the thermal balance of the plastic-mold-medium nodes and pipeline transportation in S2. The future inlet temperature and flow rate are only derived from the candidate action and the response of the calibrated actuator. The heat flow of the mold wall, the heat storage of the medium and the return water temperature are all calculated backward from the current buffer state. The measured value of the future return water is not used. The temperature, thickness, pressure holding integral and cooling time of each layer are input into the S1 in-mold thermal shrinkage mapping. The shrinkage increment of each time period is saved from the end of filling and accumulated segment by segment in the structural model. Under the mold cavity contact constraint, the deformation, contact state and residual stress are saved at the same time. The shrinkage effect that has not yet been freely presented is carried by the accumulated stress. The accumulated state cannot be zeroed because the displacement of the measuring point under the mold constraint is very small. When the candidate demolding is performed, the mold cavity contact is released, and the static balance is re-evaluated with the accumulated stress to obtain the instantaneous elastic rebound and the initial coordinates after the candidate demolding. Then the external shrinkage is continued.

[0089] The out-of-mold process is based on the actual support method and ambient temperature records. The plan is to predict that the three-point low-friction support at 23 degrees Celsius will be executed until 24 hours after demolding. When predicting the temperature of each layer, the heat exchange between the plastic and the mold is changed to the heat exchange between the plastic surface and the air and the support contact surface. The effective air heat transfer coefficient includes the combined effect of convection and radiation in this temperature range and is fitted by the temperature decay curve of the same batch of samples after demolding. The support heat transfer coefficient is calibrated using samples with the same support material and contact area. Both are saved in the fixed out-of-mold parameter record. The predicted material temperature is input into the S1 time conversion coefficient curve, and the equivalent adjustment time is accumulated layer by layer and the new shrinkage is calculated. The out-of-mold heat transfer parameters must not be included in the batch shrinkage residual fitting without recalibration.

[0090] The residual shrinkage of a region is the additional shrinkage in two directions within the layer from the candidate demolding to the end of the specified adjustment. Each region retains the signed values ​​of each layer and each direction, and the maximum absolute value is used as the constraint quantity of that region, with the unit being the dimensionless length ratio. The shrinkage difference between related regions is subtracted pairwise according to the correlation table. The shrinkage difference between the upper and lower skin layers of the panel is stored separately. The absolute quantities of two adjacent regions are not considered qualified just because their residual shrinkage is the same. The upper limit of a region is calibrated by grouping and determining whether the dimensional change does not exceed 0.2 mm after adjustment in independent mold trials. The allowable residual shrinkage envelope is taken from the samples that meet the stability requirements, and the measurement error margin is deducted. The specific upper limit is recorded in the record of each region. Regions that have not obtained stability calibration are not allowed to be released.

[0091] The structural model updates displacements based on the increased contraction in each layer during the current time period and the cumulative stress from the previous time period. The final functional coordinates are adopted. ,in This represents the column vector of coordinates of all functional measuring points at the end of the specified adjustment, in millimeters. This represents the sequential design coordinate column vectors in the S1 geometric datum record, in millimeters. This represents the cumulative displacement column vector from the design coordinates to the instantaneous equilibrium coordinates after candidate demolding, in millimeters. It includes cumulative shrinkage within the mold, mold size compensation, and instantaneous springback after releasing mold cavity constraints. The displacement column vector represents the displacement generated by the new contraction and subsequent stress relaxation from the instantaneous equilibrium state to the end of the adjustment, in millimeters; each vector is arranged in the order of fixed measuring points of panel, support, and enclosure, each measuring point occupies three design axis direction components, with the same dimension, and the positive and negative values ​​of displacement are taken according to the design coordinate direction, and the calculation range is limited to the effective range of S1 calibration and structure association;

[0092] The fixed compensation between the design coordinates and the cavity coordinates is determined by the actual measured dimensions of the mold. The initial offset as the cumulative displacement is added only once. The structural model solves all areas simultaneously during the same period, so that the support connection, the stiffness of adjacent areas and the difference between the upper and lower skin layers jointly affect each measuring point. When storing the final coordinates, the S1 design reference direction and the origin are still maintained. The current point cloud is not re-aligned to minimize the panel deformation, nor is the panel coordinate changed to a deviation relative to its best fitting plane, so as to retain the absolute functional offset.

[0093] The panel deformation index is taken as the maximum value among the absolute values ​​of the offset of 25 surface measuring points along the design panel normal; the coplanarity of the support bottom surface is calculated by the minimum region method. The bearing surface direction of the convex hull of all 45 support bottom measuring points is enumerated, and the difference between the maximum and minimum projections in each direction is calculated. Then, continuous optimization is performed in the unit normal to obtain the global minimum containment width. The numerical solution tolerance is 0.01 mm. If the solution does not converge, the candidate is rejected; the spacing of the corresponding measuring points of the enclosure mating part along the assembly direction is used as the enclosure mating dimension. After calculating the coordinates of each pair of opposing measuring points, the projection along the pre-fixed assembly direction is calculated by subtracting the coordinates of the previous measuring point. The direction is given by the geometric reference record. It is forbidden to take the absolute value successively to cover up the reversed pairing of measuring points; the three types of indexes are compared with the tolerances given by the pallet design. In this embodiment, they are compared with the limits of 3 mm, 2 mm and ±1 mm of the corresponding design spacing, respectively. The remaining shrinkage of each area is compared with the upper limit of the stability calibration of that area.

[0094] Candidate screening first checks whether there are corresponding records for the calibration range and the external support boundary. If any necessary boundary is missing, the candidate calculation is stopped and the infeasibility reason is written. If the geometric solution fails or any dimension does not meet the limit, the candidate is removed. Passing flags are not generated without using non-converged coordinates. Each prediction copy retains two items: cumulative deformation and post-shrinkage displacement, so that the current displacement after re-estimation and the future increment use the same reference. Displacements that have been predicted by the previous round of candidates but have not yet actually occurred are not carried over to the actual state of the next round.

[0095] The prediction error margin is the sum of the maximum absolute error of the corresponding index in the independent verification sample and the measurement resolution. The margin and the index use the same unit. Panel deformation, coplanarity and regional residual shrinkage are judged according to the prediction value plus the margin not exceeding the upper limit. The double-sided mating dimensions require the prediction range to fall within the tolerance zone. In this specific implementation, each prediction value, error margin, limit value, whether it passes and the reason for failure are stored in the candidate result record. At the same time, the temperature of each layer, the degree of curing and the cumulative stress at the time of candidate demolding are written for S4 to check the ejection bearing capacity. The summary qualified mark should not be transmitted and the absolute shrinkage of each area should not be lost.

[0096] In this specific embodiment, S4 includes:

[0097] The controller uses a 2-second decision cycle to read the current effective thermal state of S2, the feedback from the execution valve group, and the candidate result calculation interface of S3. Starting from the boundary of the current control cycle, it generates candidate demolding times for the next 2, 4, 6 seconds up to 60 seconds. The candidate time only represents the search range in the current discrete sequence. In this specific embodiment, the longest in-mold dwell time of the equipment is set to 180 seconds after the filling is completed, and the candidate sequence is truncated at this upper limit. Each rolling calculation regenerates the action and candidate time with the latest state and does not use the earliest demolding judgment that has not been updated in the previous cycle.

[0098] This embodiment sets up six independently operable valve groups with inlet temperature settings of 25, 35, and 45 degrees Celsius, and flow rate settings of 6, 9, and 12 liters per minute for each group. The shared water supply capacity is capped at 48 liters per minute. The equipment capacity record saves the actual maximum inlet temperature change of 1 degree Celsius per second, the maximum flow rate change of 3 liters per minute per second, the valve switching completion time of 0.5 seconds, and the correspondence between the valve group and the hot zone. These values ​​are used as the equipment configuration for this embodiment. Before activation, a step command is used to check the actual arrival time. Any settings or rates that cannot be reached are deleted from the capacity record.

[0099] The six valve groups correspond sequentially to the upper and lower sides of the panel, the two sets of thick joints and supports, and the two sets of enclosure areas. They operate online using six pre-defined target configurations: all groups at 35 degrees Celsius and 6 liters per minute each; all groups at 25 degrees Celsius and 6 liters per minute each; all groups at 45 degrees Celsius and 6 liters per minute each; the two sets of thick joints and supports at 25 degrees Celsius and 12 liters per minute each, while the remaining groups are at 35 degrees Celsius and 6 liters per minute each; the two panel groups at 35 degrees Celsius and the remaining groups at 25 degrees Celsius and 6 liters per minute each; and the two panel groups at 45 degrees Celsius and the remaining groups at 25 degrees Celsius and 6 liters per minute each. Each configuration opens uniformly at 0, 2, or 4 seconds. The system starts to change towards the target level and maintains it until the candidate time, forming 18 action trajectories. The current holding trajectory is also added. Branches that cannot be adjusted independently are merged into the same execution component. Temperature and flow rate change along the allowable rate ramp starting from the current measured value. During valve switching, the system predicts based on the measured switching trajectory. Combinations where the total flow rate exceeds the shared limit or valve timing conflicts are first eliminated. Then, S3 is called according to the remaining combinations. The executable set includes the current holding action. Heating and cooling capabilities, loop transport delay, and temperature level arrival time all enter the future boundary. The target inlet temperature cannot be regarded as an immediate effect on the mold wall.

[0100] The actions and candidate time combinations are checked one by one in chronological order. For each combination, the three functional dimensions of S3, the remaining shrinkage of all areas, and the ejection load are checked simultaneously. The ejection load is checked by referring to the load records of the same batch of materials using the temperature and curing degree of each layer. The records are stored with temperature and curing degree as keys, storing the allowable compressive stress, elastic modulus and their verification range. The temperature nodes are consistent with the S1 material test. The allowable stress is the stress before the irreversible deformation occurs in the compression test and then divided by the safety factor of 1.5 set in this embodiment. The curing degree is obtained from the corresponding calorimetric curve under the same conditions. If the key value is not covered by the calibrated envelope, the candidate is not passed.

[0101] The force applied during tray ejection is recorded using independent mold tests with force-measuring ejector rods. The loads are grouped according to mold opening temperature, holding pressure history, and demolding speed. The maximum measured peak load of each ejector rod is multiplied by 1.2 to determine the upper limit of the predicted load, with the load unit being Newtons. Local compressive stress is calculated based on the contact area of ​​the ejector rods and compared with the corresponding allowable stress. Simultaneously, using the accumulated stress of S3 as the initial state, the upper limit of the ejection load and the tray's self-weight are applied to the laminate and solid model. The mass is first multiplied by 9.81 meters per second squared to convert to Newtons, and the instantaneous displacement during ejection is obtained. This implementation requires that the compressive stress in all contact areas does not exceed the allowable stress of the material, the equivalent stress of the structure does not exceed the allowable value given by the material test at the same temperature, and the instantaneous additional displacement of the panel during ejection does not exceed 0.5 mm. Candidates not covered by the load records are not approved based on this.

[0102] If a candidate time point has a combination of actions that all pass, then that time point is taken as the earliest feasible time point for both the current sequence and the current executable set to be de-demolded. The passing combinations at that time point are first sorted by the cumulative water supply volume from the current time point to that time point in ascending order, and then sorted by the number of valve switching times in ascending order. If they are still the same, the first item is selected according to the fixed coding order of temperature, flow rate and switching start point. This sorting is only used to select actions within the same feasible time point. Later times cannot be selected on the grounds of saving water, nor can it be claimed that the earliest global value of all cooling strategies outside the continuous time or set is obtained.

[0103] Once selected, only the current 2-second action segment is issued, writing the inlet temperature target, flow rate target, ramp limit, switching start time, candidate demolding time, and source prediction version into the execution register. S2 continuously receives feedback and estimates the state, and the solution is recalculated in the next cycle. The cycle preparing for release must also check the current time, replace the previous cycle's prediction history with the measured execution history, and then complete S3 and ejection constraint calculations. A common mold opening permission signal is sent only when the promised time has been reached, the latest state is valid, and all constraints are still passed. Each zone does not demold prematurely. If the latest state makes the original time invalid, the permission signal is immediately revoked and the mold remains closed, and a subsequent time is reselected.

[0104] If no feasible combination exists, a pre-verified cooling holding action is performed for one cycle within the equipment's allowable range. In this implementation, the six-valve group at 35 degrees Celsius, each group at 6 liters per minute, and the allowable ramp are used as the holding action to be verified. It is only registered and activated after the trial molding verification is completed within the applicable range of the current batch and mold. The mold is closed again and the predicted time domain is extended by 20 seconds. The calculation is recalculated in the next cycle. If the calculation is not completed within the cycle, it is also treated as no proven feasible combination exists. If no feasible combination exists after 180 seconds, automatic mold opening is prohibited and a process abnormality signal with mold number, failure constraint, and status record is issued. The equipment process abnormality handling process takes over. The controller does not interpret the time limit as a forced mold opening condition.

[0105] The execution process's action instructions, actual inlet temperature and flow rate, actual demolding time, various allowances at release, and the versions of parameters used are written into the record for this mold cycle segment by segment. The demolding tray is bound to this record through an identifier. S5 uses this real execution history and subsequent measurement feedback to conduct screening and correction. When feedback arrives, the cooling instructions for completed mold cycles are not written back. Mold cycles in the same batch that are still forming are re-predicted and recalculated in the next control cycle according to the verified new material version.

[0106] In this specific embodiment, S5 includes:

[0107] In this specific embodiment, a method for partitioned cooling molding of recycled plastic pallets involves a controller performing cross-mold updates. The controller has a material version library that stores only batch shrinkage parameters and a thermal process version library indexed by mold number and circuit number. It reads the actual instructions, execution history, and parameter versions from the current mold record in S4, and uses the predicted value from the previous step saved in S2 to verify the thermal process. The flow prediction is calculated from the first-order response identified by valve commands, water supply pressure difference, and step flow in the empty mold. The steady-state flow is taken from the calibration values ​​of each level, and the response time constant is taken as the total change of 6 when the step is reached. The required time is 3.2%. The supply and return water temperature difference is generated by the S2 mold wall-medium forward heat exchange node and pipeline transportation. For a certain return water sampling moment to be inspected, only the previously received state and known inlet boundary and valve command are used to form and save the return water and mold temperature at that moment for one-step prediction. The return water temperature to be inspected or the temperature difference power obtained from it is not read first. The residual is calculated and verified only after the actual return water arrives. Subsequent observation updates or historical retrospectives must not overwrite the previously saved predictions. The regional mold temperature is generated by the same thermal state prediction and cannot be directly copied from the observation value to be inspected into its prediction value.

[0108] For each loop flow rate, supply and return water temperature difference, and regional model temperature, the observed minus predicted signed residual is calculated. The absolute value of each residual is compared with the allowable deviation for all sampling points that are time-aligned and valid. The allowable deviation is determined by adding the sensor resolution to the 99.5 percentile of the absolute value of the residual of the independent thermal process verification module and is fixed before learning. In this implementation, the upper limits for enabling verification are set to 0.5 liters per minute, 0.5 degrees Celsius, and 1 degree Celsius. If the statistically obtained allowable deviation exceeds the upper limit, the module is recalibrated without relaxing the threshold. The module is written into the batch learning candidate pool only when all channels are within their respective allowable deviations for all valid time periods and there are no missing time periods. Any inconsistency in any channel will prevent the module from participating in material parameter updates.

[0109] Candidate records are deduplicated by a combination of batch number, module number, and pallet identity key, and then sorted by actual demolding time. For measurements with the same identity arriving repeatedly, only versions with consistent equipment signatures and complete sampling timestamps are retained, and repeated measurements of the same pallet are not counted as independent modules. When the order of two measurement times is reversed or any temperature or support record is missing, the incremental fitting of that pallet is stopped and the raw material version is retained. When the material feedback version switch and a prediction occur simultaneously, the current prediction retains the parameter snapshot at the beginning, and is updated uniformly in the next control cycle to avoid mixing parameters at two different times for the same candidate.

[0110] For the same pallet in each candidate mold, the coordinates or spacing of the 25 points on the panel, the 45 points on the legs, and the mating points of the enclosure are obtained using the same measuring equipment at two predetermined times, 10 minutes and 60 minutes after demolding. The actual sampling time, ambient temperature, support position, and pallet identity are recorded. The measuring fixture uses the same set of three physical reference positioning points to fix the origin and direction. The positioning only eliminates the rigid body changes caused by repeated placement. The three-point positioning sequence is predefined. No best fit is made for all measuring points, and the point cloud is not scaled. The predicted motion of the physical reference points themselves is processed by the same positioning transformation. The actual measurement and prediction use the same coordinate convention.

[0111] A signed geometric increment is generated by subtracting the 10-minute value from the 60-minute value for the same measuring point and coordinate components. The spacing of the enclosure is also calculated by subtracting the previous value from the subsequent value according to the fixed pairing and assembly direction. The increment unit is uniformly millimeters. In this implementation method, repeated positioning measurement is used to confirm that the standard deviation of a single coordinate does not exceed 0.02 millimeters. The standard deviation of the incremental noise is obtained by the incremental distribution of two complete positioning measurements of the same pallet. Data that does not meet the measurement accuracy requirements are not included in the fitting. The controller retains the original absolute coordinates at two times. The increment is only used for batch shrinkage parameter learning and cannot replace the absolute size check after the specified adjustment.

[0112] The actual in-mold thermal history and actual demolding time of the same tray in the candidate pool are reproduced. The S3 structural model is deduced based on the ambient temperature and support conditions recorded between 10 minutes and 60 minutes. The prediction increment is obtained at the same measurement point, in the same direction and at the same time period under the same measurement and positioning transformation.

[0113] The predicted subtraction order is consistent with the actual measurement. The measured increment is subtracted from the predicted increment to form the fitting residual. The geometric offset of the same tray 10 minutes ago is used as the initial state shared by the two predictions. The S1 in-mold curve, loop thermal process parameters, out-of-mold heat transfer parameters, geometric reference and measurement positioning rules are all fixed. The loop thermal process parameters are not involved in the fitting of this residual. It is not allowed to reduce the residual of this time through these parameters.

[0114] In this embodiment, the post-shrinkage amplitude and time scale of the cover panel thickness unit and thick node unit are selected in advance to participate in the candidate correction. The layer direction and thermal history unit are selected according to the S1 index. Each unit participating in the fitting must have a candidate mode falling within its applicable range. First, subtract the calibrated physical lower bound from each parameter to be corrected and then divide by the difference between the upper and lower bounds to convert it into a dimensionless coordinate between zero and one. The physical upper and lower bounds are taken as the envelope of the fitting results of the batch of repeated samples and confirmed by independent samples. The lower bound of the amplitude is not less than zero and the lower bound of the time scale is strictly greater than zero. Parameters with the same upper and lower bounds are kept fixed and are not included in the parameters to be estimated.

[0115] Apply trial changes of 1% of the calibration interval width sequentially near the current parameter. When approaching the boundary, change to change towards the inside of the interval. Keep all parameters unchanged except the parameter being tested. Divide the predicted increment change by the dimensionless trial change of the parameter to form a sensitivity column. Then divide each row by the noise standard deviation of the measurement increment to obtain the dimensionless sensitivity matrix.

[0116] The singular value decomposition test is performed to check for full column rank. One millionth of the largest singular value is used as the zero limit of the value. The condition number is obtained by dividing the largest non-zero singular value by the smallest singular value. Only when the matrix has full column rank and the condition number does not exceed the allowable value verified by repeated measurement noise disturbance is the fit performed. In this implementation, the allowable value is preset to 100 and the parameter change under noise disturbance is required not to exceed the single update limit. Otherwise, the set of participating parameters is reduced or all parameters are frozen and subsequent candidate modules are accumulated. After the parameter set is reduced, this test must be performed again.

[0117] Before correction, the candidate pool is divided into a fitting group and a retention and validation group according to the module number. The fifth of every five modules in the time sequence of the module number is used as the retention sample. The measurement points of the same tray are not split into two groups. Correction is only started after at least 20 candidate modules have been accumulated and each estimated unit is covered in both groups.

[0118] The objective of constrained least squares is to minimize the sum of squares of the incremental residuals of all fitted measurement points divided by their noise standard deviation. It uses bounded trust region iteration, and each time the complete prediction is called to calculate the increment. The magnitude and time scale must not exceed the calibration physical boundary. The change in a single activation must not exceed 5% of the width of the corresponding calibration interval. The stopping condition is that the relative change of the objective function between two adjacent times is less than one part per million or the iteration reaches 50 times. If the iteration limit is reached and the convergence is not achieved, the update will not be activated.

[0119] After correction, the geometric increment is re-predicted using the retained modulus that was never used in the fitting. The root mean square of the residuals of the panel, legs and enclosure measurement components is calculated separately. In the calculation, the average value of the squares of the millimeter residuals is first calculated and then the square root is taken to restore the millimeter unit. In this implementation, the allowable error for all three types is set at 0.10 mm. The error after each type of correction must not exceed the allowable error and must not be greater than the error of the same group recalculated with the original parameters. The absolute value of each residual must not exceed 0.20 mm. Only after all of them pass can an usable material version be formed.

[0120] For any type of failure, the original parameters are retained, and subsequent candidate modes within the applicable range are added before fitting. The retained groups that have been used to select parameters or repeatedly tune parameters will not be used as independent validation groups in the next round. In the next round, subsequent modes will be selected as the retained groups.

[0121] When thermal observations are inconsistent, check the supply and return water timescales, effective transport volume, flow sensor, loop blockage, and mold temperature probe, and re-identify the affected loop thermal process parameters according to the S2 empty mold and material sample procedure; after the new thermal process version is independently verified and five consecutive molds of the same batch pass all thermal observation thresholds, resume learning candidate screening, permanently retain the mark of abnormal molds not participating in the current material residual fitting, restore the abnormal history to qualified without retrospection;

[0122] Each validated material version is bound to the mixed batch number, fitting source module number, retained validation module number, parameter physical range, modification amount, receiving time, and enabled module number. It is stored by first writing the version to be enabled and then atomically switching the valid version flag. After the feedback arrives, for the same batch of trays still in the mold, only the post-shrinkage parameters within the applicable range are replaced in the next control cycle. The thermal history, cumulative in-mold shrinkage, and actual action records remain unchanged. S3 is called again and S4 recalculates the cooling action and common demolding time. Modules that are outside the applicable range of the new version continue to use their validated applicable version.

[0123] At 23 degrees Celsius, under three-point low-friction support, and after 24 hours of demolding, each tracked pallet is inspected according to the S3 fixed design benchmark for panel deformation, coplanarity of all support bottom surfaces, and enclosure fit dimensions. If any absolute dimension fails to meet the standard, the corresponding pallet is isolated, the failure index is recorded, and the process review is triggered. Even if the incremental prediction of the pallet at two time points is qualified, the isolation is not revoked. The controller archives the absolute dimension inspection results, material version, and loop version together as the source for subsequent batch applicability verification, completing the feedback process of batch prediction update and product dimension inspection being independent of each other.

[0124] 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.

[0125] This invention uses regional thermal states as a connection, incorporating the shrinkage characteristics of mixed batches, in-mold cooling actions, and adjusted functional dimensions into a single decision-making process. The common demolding time is jointly limited by the three types of functional dimensions, regional residual shrinkage, and ejection bearing conditions. Inlet temperature, flow rate, and switching sequence are adjusted in zones within the equipment's executable range, ensuring that differential shrinkage control and molding cycle control have consistent judgment criteria.

[0126] The cross-module feedback adopts a separate processing structure for thermal process verification and material parameter correction, and establishes a prediction comparison between the geometric increments at two time points of the same pallet and the actual thermal history and external conditions. Only parameters that have been screened, constrained and corrected, and validated with retained samples are used for prediction and cooling decisions in the same batch after feedback arrives, thereby reducing the interference of thermal process anomalies and early geometric biases on subsequent module adjustments.

Claims

1. A method for partitioned cooling molding of recycled plastic pallets, characterized in that, include: S1. Obtain the flow characteristics and thermal shrinkage mapping of the mixed batch of recycled plastic and sprue material, form cooling zones according to the tray structure and controllable medium circuit, and establish the relationship between regional shrinkage and panel deformation, coplanarity of the bottom surface of the support legs and the matching dimensions of the enclosure. S2. Utilize the mold temperature, pressure, and medium thermal observation during the injection molding process, combined with the circuit thermal process parameters, to estimate the skin and core state of each zone; S3. Based on the state, pressure history and thermal shrinkage mapping, predict the remaining shrinkage, shrinkage difference and accumulated deformation of the area corresponding to the candidate cooling action and demolding time, and obtain the functional dimension index under the specified adjustment conditions by correlation. S4. Under the constraints of functional dimensions, residual shrinkage and ejection load, determine the earliest feasible common demolding time and the medium inlet temperature, flow rate and switching sequence of each zone from the executable cooling actions, execute them segment by segment and re-estimate and predict, and open the mold when the release conditions are met. S5. Save the batch shrinkage parameters and loop thermal process parameters separately. For the module where the thermal observation and prediction are consistent, compare the measured geometric increments at two moments after the same tray is demolded with the corresponding increments predicted based on the thermal history of that module. Correct the shrinkage parameters after the batch. After verification, use them to feed back the prediction of the remaining shrinkage of the same batch after arrival and the cooling action and common demolding time for recalculation.

2. The method for partitioned cooling molding of recycled plastic pallets according to claim 1, characterized in that, S1 includes: The flow characteristics under specified material temperature and shear conditions were obtained according to the batch number of the mixed batch. A thermal shrinkage mapping was established using the recorded holding pressure history and temperature and dimensional changes under cooling conditions for samples of different thicknesses. The temperature, pressure history, thickness, and cooling time were mapped to the shrinkage amount and the shrinkage amount after demolding over time. Cooling zones were determined according to the thickness, heat dissipation path, and independently adjustable range of the panel, rib intersection, support, and enclosure mating parts. The common and non-independently adjustable loop coverage parts were used as common action constraints. The correspondence between regional shrinkage and functional measuring point displacement was established using the tray geometry and thermal shrinkage calibration results. The influence of the shrinkage difference between the upper and lower layers of the panel, adjacent areas, and associated supports on the functional measuring point displacement was retained as the correlation between regional shrinkage and panel deformation, support bottom coplanarity, and enclosure mating dimensions.

3. The method for partitioned cooling molding of recycled plastic pallets according to claim 1, characterized in that, S2 includes: Collect the regional mold temperature, representative location pressure, and supply water temperature, return water temperature, and flow rate of each medium circuit for each cooling zone and align them according to the sampling time. The medium transport delay is determined based on the effective volume and flow rate of the loop. The supply water temperature and return water temperature are converted to the corresponding mold heat exchange period. The heat exchange obtained from the flow rate and the supply and return water temperature difference is combined with the mold heat storage, the thermal contact between the plastic and the mold, and the heat transfer between zones into the stratified heat balance. The initial state of the area after filling is determined by combining the flow characteristics, and the predicted state of the stratified heat balance is corrected by the zone mold temperature. The skin and core temperatures and the degree of curing of each zone are output. The loop thermal process parameters include the effective volume used for stratified heat balance, the mold heat capacity, the thermal contact parameters, and the heat transfer parameters between zones.

4. The method for partitioned cooling molding of recycled plastic pallets according to claim 1, characterized in that, S3 includes: for candidate cooling actions and demolding times, starting from the current skin and core states of each zone, combining pressure history, thermal shrinkage mapping, and candidate cooling actions, the in-mold thermal shrinkage process is deduced, and combined with the ambient temperature and support conditions after demolding, the specified adjustment conditions are deduced to the specified adjustment conditions, which are predetermined adjustment temperatures, adjustment durations, and support methods; the predicted shrinkage amount of each zone from the candidate demolding time to the end of the specified adjustment conditions is taken as the residual shrinkage of the region, and the absolute amount of each zone is constrained, and the shrinkage difference between related regions and between the upper and lower layers of the panel is calculated; based on the accumulated deformation caused by cumulative shrinkage and the new shrinkage after demolding, the final coordinates of the functional measurement points are obtained through the correlation between regional shrinkage and panel deformation, coplanarity of the bottom surface of the support, and the fit dimensions of the enclosure, avoiding the use of the residual shrinkage difference to replace the accumulated deformation, and the functional dimension indicators are calculated accordingly.

5. The method for partitioned cooling molding of a recycled plastic pallet according to claim 1, characterized in that, S4 includes: Based on the adjustable range, rate of change, valve switching time, shared water supply capacity, and limitations of loops that cannot be independently adjusted for each loop inlet temperature and flow rate, executable cooling actions are generated, and a candidate demolding time sequence is formed according to the control cycle. The action and time combination are checked one by one in sequence, and the combination that simultaneously satisfies the panel deformation tolerance, the coplanarity tolerance of the support bottom surface, the dimensional tolerance of the enclosure, the upper limit of the remaining shrinkage of each zone, and the ejection load constraint is retained. The first time with the retained combination is determined as the common demolding time. The ejection load constraint is determined by the material load-bearing capacity corresponding to the skin and core states, the ejection force of the tray, and the allowable ejection deformation. The inlet temperature, flow rate, and switching sequence of the corresponding combination are executed segment by segment. The state and combination are updated in the next control cycle. When the common demolding time is reached and the updated result still satisfies all constraints, the mold is opened.

6. The method for partitioned cooling molding of recycled plastic pallets according to claim 1, characterized in that, S5 includes: forming thermal observation prediction values ​​based on loop thermal process parameters; comparing the differences between the observed values ​​and corresponding prediction values ​​of each loop flow rate, supply and return water temperature difference, and regional mold temperature with the allowable deviation of the thermal process calibration item by item; only selecting all molds within the allowable deviation as batch learning candidates; for the same tray of the candidate molds, measuring the signed coordinates or spacing of the panel, bottom of the support legs, and the mating parts of the enclosure at two predetermined times after demolding according to the same physical reference and coordinate direction to form the measured geometric increment of the later time relative to the previous time; using the thermal history of the candidate molds and the external temperature and support conditions between the two predetermined times to predict the corresponding increment at the same position, in the same direction, and in the same time period; using the residual between the measured geometric increment and the corresponding increment to correct the batch post-shrinkage parameters in the thermal shrinkage mapping, while keeping the loop thermal process parameters out of the current residual fitting; after completing the correction within the parameter physical boundary and update limit, verifying the geometric increment prediction error with the retained samples that did not participate in the fitting; only after the verification is passed will the correction parameters be enabled, and the absolute size check under the specified adjustment conditions will be retained.

7. The method for partitioned cooling molding of a recycled plastic pallet according to claim 2, characterized in that, In step S1, the heat shrinkage mapping selects the applicable thickness, holding pressure history, and cooling condition calibration range according to the flow characteristics of the mixed batch. The shrinkage curve of each calibration condition is determined by recording the sample size change in the same reference direction. The shrinkage amount at the demolding moment and the shrinkage amount that develops over time after demolding are parameterized separately. The shrinkage parameter after batching is limited to the shrinkage amplitude and time scale parameter after demolding, and the applicable range of thickness and thermal history corresponding to the parameter is recorded. Mixed batches or molding conditions that exceed the calibration range are first supplemented with calibration, and then the applicable heat shrinkage mapping is generated based on the supplementary calibration results.

8. The method for partitioned cooling molding of a recycled plastic pallet according to claim 4, characterized in that, The functional measuring points in step S3 include panel surface measuring points, bottom measuring points of each support leg, and measuring points of the enclosure mating parts. The final coordinates of the functional measuring points are uniformly represented in the coordinate system defined by the pallet design datum. The panel deformation index is obtained by the normal offset of the panel surface measuring points relative to the design panel datum plane. The coplanarity of the support leg bottom surface is obtained by the minimum distance between two parallel planes that encompass all the bottom measuring points of the support legs. The enclosure mating dimensions are obtained by the spacing of the corresponding measuring points of the enclosure mating parts along the assembly direction. The three types of indicators are compared with the tolerances given in the pallet design. The remaining shrinkage in the area is compared with the upper limit of the remaining shrinkage in the area calibrated according to the adjusted dimensional stability requirements.

9. A method for partitioned cooling molding of a recycled plastic pallet according to claim 5, characterized in that, In step S4, the candidate demolding time sequence and executable cooling action are regenerated in each control cycle based on the current state of the equipment. The earliest feasible common demolding time is limited to the first feasible time within the current candidate demolding time sequence and executable cooling action range. If there is no combination that simultaneously satisfies the constraints, the mold is kept closed, and the next control cycle is executed according to the cooling action that has been verified by the trial mold within the allowable range of the equipment. The candidate demolding time sequence is then expanded and re-evaluated. If there is still no feasible combination after the longest in-mold dwell time specified by the equipment process specification, automatic mold opening is prohibited and a process abnormality signal is issued.

10. A method for partitioned cooling molding of a recycled plastic pallet according to claim 6, characterized in that, In step S5, before correction, trial changes are applied to the pre-selected post-batch shrinkage parameters. A sensitivity matrix is ​​formed by the ratio of the corresponding increment change to the trial change. Correction is performed only when the sensitivity matrix has full column rank and the number of conditions does not exceed the allowable number of conditions calibrated based on the measurement noise. Otherwise, the post-batch shrinkage parameters are frozen and subsequent candidate modes are accumulated. Correction is completed by performing constrained least squares fitting on the geometric increment residuals. The physical boundaries of the parameters and the single update limit are taken from the calibration range of the batch samples. The sample verification requirement is retained, and the prediction error after correction does not exceed the preset allowable error and is not greater than the error before correction. When thermal observations and predictions are inconsistent, the corresponding module is prohibited from participating in the batch shrinkage parameter update. The medium transport, heat exchange and module temperature measurement are checked, and the batch learning candidate screening is resumed after the loop thermal process parameters are recalibrated and the thermal observation consistency test is passed. The verified parameter version is bound to the batch number, data source module number and enabled module number for prediction in the same batch after feedback arrives. When the absolute size check is unqualified, the corresponding tray is isolated and the process review is triggered.

Images