Intelligent optimization method and system for hot pressing process parameters of anti-crack finish panel

Abstract

The application relates to the technical field of hot-pressing process optimization, and discloses an intelligent optimization method and system for hot-pressing process parameters of anti-crack facing plates, wherein the method comprises the following steps: acquiring adhesive characteristic parameters and calculating a partition initial solidification degree vector at a cooling starting moment; performing time-step-by-time-step combined calculation of solidification degrees and cumulative damage degrees in a cooling stage, replacing a fixed complete solidification strength value with an instantaneous shear strength value as a normalization reference, and calculating a dynamic normalized cumulative damage degree according to a modified Miner criterion; training a solidification-damage coupling prediction network; solving an optimal partition cooling rate parameter vector under double constraints; generating a cooling medium temperature setting value sequence after verifying and correcting the prediction network precision; and integrating a generated hot-pressing process parameter execution scheme.

Description

Technical Field

[0001] This invention relates to the field of hot pressing process optimization technology, specifically to an intelligent optimization method and system for hot pressing process parameters of crack-resistant decorative panels. Background Technology

[0002] In the hot-pressing production of crack-resistant decorative panels, the degree of adhesive curing continues to evolve during the cooling stage, and microscopic damage at the interlayer interface accumulates simultaneously. In the early stage of cooling, the degree of adhesive curing is usually in a nonlinear range where shear strength is highly sensitive to the degree of curing. The incremental microscopic damage caused by stress cycles driven by thermal shrinkage differences in this range is significantly greater than the incremental damage under the same stress amplitude in the fully cured state.

[0003] Existing process parameter optimization methods employ two main strategies. The cumulative damage assessment strategy uses a fixed fully cured shear strength as a normalization benchmark, overestimating the adhesive layer's load-bearing capacity in the early cooling phase, leading to a systematic underestimation of the calculated cumulative damage. The cure-sensing strategy only constrains transient stress at each moment to not exceed a dynamic threshold, failing to assess the impact of rapid damage accumulation from high-frequency stress cycles in the low cure-degree range on the total cumulative damage at the end of cooling. When these two strategies operate independently, a blind spot in damage assessment is created in the early cooling phase. The optimized cooling parameters trigger rapid accumulation of microscopic damage in the adhesive layer, ultimately leading to interlayer cracking during product service. Summary of the Invention

[0004] This invention provides an intelligent optimization method and system for hot pressing process parameters of crack-resistant decorative panels, which solves the technical problems in related technologies such as distortion of interlayer fatigue damage assessment benchmark caused by dynamic changes in the curing state of adhesives during the cooling stage and low efficiency of collaborative optimization of cooling rate parameters in multiple zones.

[0005] This invention discloses an intelligent optimization method for hot-pressing process parameters of crack-resistant decorative panels, comprising: acquiring the curing kinetic parameters, nonlinear mapping curve data of curing degree and interlaminar shear strength, and fatigue damage parameters of the adhesive; calculating the curing degree value of the adhesive layer in each zone at the end of the pressure holding time; generating the initial curing degree vector of each zone; discretizing the cooling stage into multiple time steps; performing joint calculations on the cooling rate parameter samples of each component zone step by step; calculating the curing degree increment based on the current time step temperature value and curing kinetic equation and accumulating it to obtain the current cumulative curing degree value; obtaining the instantaneous shear strength value through the nonlinear mapping curve; calculating the transient interlaminar shear stress value based on the interlaminar temperature difference and the difference in thermal expansion coefficients; and optimizing the cooling process. The stress time history from the start to the current time step is analyzed by rainflow counting. The instantaneous shear strength value is used instead of the fixed fully cured strength value as the normalization benchmark to calculate the current cumulative damage value. The curing damage coupling prediction network is trained with the partition cooling rate parameter vector and the initial curing degree vector of the partition as inputs and the final cumulative damage value, final curing value and warpage value of each partition as outputs. The optimal partition cooling rate parameter vector is solved with the constraints that the final cumulative damage value of each partition output by the curing damage coupling prediction network is lower than the safe damage threshold, the final curing value reaches the minimum curing degree standard, and the warpage is lower than the flatness standard, and the goal is to minimize the total cooling time.

[0006] Further, the step of using the instantaneous shear strength value instead of the fixed fully cured strength value as a normalization benchmark to calculate the current cumulative damage value includes: for each stress cycle identified by the rainflow counting analysis, the ratio of the stress amplitude of the stress cycle to the instantaneous shear strength value corresponding to the occurrence time of the stress cycle is used as the normalized stress ratio, and the number of stress cycles corresponding to fatigue failure under the normalized stress ratio is obtained by substituting it into the fatigue life function of the adhesive, and the reciprocal of the number of stress cycles is taken as the damage increment of the stress cycle; the damage increments of all identified stress cycles from the start of cooling to the current time step are summed to obtain the current cumulative damage value.

[0007] Furthermore, the curing damage coupling prediction network is a multi-task regression network, including an input layer, a shared hidden layer, a time-step attention weight layer, and a multi-task output head. The input layer receives the concatenated features of the partition cooling rate parameter vector and the initial curing degree vector of the partition and passes them to the shared hidden layer. The shared hidden layer performs a nonlinear transformation on the concatenated features and outputs the hidden layer feature vector corresponding to each time step. The time-step attention weight layer calculates the attention score for the hidden layer feature vector of each time step, normalizes the attention score, and uses it as a weight to perform weighted aggregation of the hidden layer feature vectors of each time step, outputting an aggregated feature vector. The multi-task output head includes a cumulative damage output head, a curing degree output head, and a warpage output head. Each output head independently receives the aggregated feature vector and outputs the predicted values ​​of the final cumulative damage value, the final curing degree value, and the warpage value of each partition, respectively.

[0008] Furthermore, the attention score assigned to the time step attention weight layer for the time step corresponding to the low curing degree interval is higher than the attention score assigned to the time step corresponding to the high curing degree interval, thereby increasing the feature contribution ratio of the high damage sensitivity stage in the early cooling stage in the aggregated feature vector.

[0009] Furthermore, the training of the curing damage coupling prediction network adopts a multi-task mean square error loss function. The mean square error of the predicted value of each task output head and the corresponding label value jointly calculated step by step are calculated separately and then weighted and summed. The total loss is minimized by the Adam optimization algorithm and the training is continued until the error of each task on the validation set converges.

[0010] Furthermore, the optimal partition cooling rate parameter vector is solved using the interior point method, with the cooling rate of each partition at each time step as the decision variable. In each iteration, the degree of satisfaction of the cumulative damage constraint and the curing degree constraint is evaluated simultaneously through the curing damage coupling prediction network.

[0011] Furthermore, after solving the optimal partition cooling rate parameter vector, the method further includes: inputting the temperature time series corresponding to the optimal partition cooling rate parameter vector into the curing kinetic equation and rainflow counting analysis, and performing precise calculations step by step throughout the entire process; comparing the deviation between the precisely calculated dynamic normalized cumulative damage value and the predicted value of the curing damage coupling prediction network step by step; if the deviation exceeds the allowable range, adding the verification data to the training set to incrementally train the curing damage coupling prediction network, and re-executing the step of solving the optimal partition cooling rate parameter vector.

[0012] Furthermore, if the deviation is within the allowable range, the cooling rate sequence corresponding to the optimal partition cooling rate parameter vector is segmented and subjected to ramp transition processing to generate a cooling medium temperature setpoint sequence for each partition in each time period; wherein, the segmented step processing refers to dividing the continuously changing cooling rate sequence into several constant rate segments according to time periods, and the ramp transition processing refers to inserting a linear transition segment between adjacent constant rate segments; the cooling medium temperature setpoint sequence is integrated with the temperature, pressure, and time parameters of the pressure holding stage to generate a hot pressing process parameter execution scheme covering the entire pressure holding to cooling stage.

[0013] Furthermore, the cooling rate parameter samples for each component region are generated by Latin hypercube sampling within the allowable range of cooling rate at each time step.

[0014] This invention discloses an intelligent optimization system for hot-pressing process parameters of crack-resistant decorative panels, used to execute the aforementioned intelligent optimization method for hot-pressing process parameters of crack-resistant decorative panels. The system includes: an initial curing degree calculation module, used to acquire the curing kinetic parameters of the adhesive, the nonlinear mapping curve data of curing degree and interlayer shear strength, and fatigue damage parameters; calculate the curing degree value of the adhesive layer in each zone at the end of the pressure holding period; and generate a zone initial curing degree vector. A time-step joint calculation module is used to discretize the cooling stage into multiple time steps, and perform time-step calculation of curing degree increment, acquisition of instantaneous shear strength value, calculation of transient interlayer shear stress value, and rainflow counting for the cooling rate parameter samples of each component zone. The analysis uses the instantaneous shear strength value instead of the fixed fully cured strength value as a normalization benchmark to calculate the current cumulative damage value; the network training module is used to train the curing damage coupling prediction network with the partition cooling rate parameter vector and the initial curing degree vector of the partition as inputs, and the final cumulative damage value, final curing value and warpage value of each partition as outputs; the optimization solution module is used to solve the optimal partition cooling rate parameter vector with the constraints that the final cumulative damage value of each partition output by the curing damage coupling prediction network is lower than the safe damage value threshold, the final curing value reaches the minimum curing degree standard, and the warpage is lower than the flatness standard, and the objective is to minimize the total cooling time.

[0015] This invention performs simultaneous calculations of the curing kinetic equation and rainflow counting damage analysis at each discrete time step during the cooling stage, using the instantaneous shear strength value corresponding to the current cumulative degree of curing instead of the fixed fully cured strength value as the normalization benchmark for the cumulative damage degree. In the early stages of cooling when the degree of curing is in a low range, the instantaneous shear strength value is significantly lower than the fully cured strength value, resulting in a correspondingly larger normalized stress ratio. This leads to a larger damage increment obtained from each stress cycle calculation, eliminating the bias of existing cumulative damage assessments that systematically underestimate the damage degree in the early stages of cooling.

[0016] The time-step joint calculation synchronously tracks the dynamic evolution of cumulative damage based on the constraint of transient stress at each moment, so that the rapid increase of cumulative damage caused by high-frequency stress cycling in the low solidification range can be fully captured, eliminating the damage assessment blind spot formed in the early stage of cooling when the two existing strategies are run independently.

[0017] The time-step attention weight layer in the curing damage coupling prediction network automatically assigns higher weights to time steps corresponding to low curing degree intervals when fitting the mapping relationship between cooling rate parameters and final cumulative damage degree, thereby improving the fitting accuracy of parameter response patterns in the high damage sensitivity stage during the initial cooling phase.

[0018] The optimization solution with cumulative damage degree and final curing degree as dual constraints automatically limits the cooling rate in the early stage of cooling due to the tightening of the cumulative damage degree constraint, which reduces the transient stress amplitude and extends the time window for residual curing. When the curing degree is sufficiently increased to the strength-insensitive range, the cumulative damage degree constraint is relaxed, and the cooling rate is released, thereby shortening the cooling cycle while ensuring the bonding reliability of the product. Attached Figure Description

[0019] Figure 1 This is a flowchart of the intelligent optimization method for hot pressing process parameters of crack-resistant decorative panels provided in this embodiment of the invention; Figure 2 This is a schematic diagram of the initial curing degree of each zone at the cooling start time provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of the evolution of the curing degree and instantaneous shear strength of partition 4 step by step according to the embodiment of the present invention; Figure 4 This is a schematic diagram of the time-step cumulative damage and transient shear stress evolution of partition 4 provided in the embodiment of the present invention; Figure 5 This is a schematic diagram comparing the training and verification errors of the solidification damage coupling prediction network provided in an embodiment of the present invention; Figure 6 This is a schematic diagram comparing the cumulative damage of each partition before and after optimization, provided in an embodiment of the present invention. Figure 7 This is a schematic diagram of the final curing degree of each partition after optimization, provided in an embodiment of the present invention; Figure 8 This is a schematic diagram illustrating the verification of the deviation between the accurate calculated value and the network predicted value of partition 4 provided in an embodiment of the present invention; Figure 9 This is a schematic diagram of the segmented distribution of cooling rate in the fourth cooling stage of the present invention. Detailed Implementation

[0020] In the hot-pressing production of crack-resistant decorative panels, the cooling stage is a unique period in which the adhesive curing degree continuously evolves and interlayer interface micro-damage accumulates simultaneously. In the initial cooling phase, the adhesive curing degree is typically in a non-linear range where shear strength is highly sensitive to curing degree. The micro-damage increment caused by stress cycles driven by thermal shrinkage differences within this range is significantly greater than the damage increment under the same stress amplitude in a fully cured state. Existing process parameter optimization methods use a fixed fully cured shear strength as a normalization benchmark, overestimating the adhesive layer's load-bearing capacity in the initial cooling phase, leading to a systematically low calculated cumulative damage degree. Curing degree-sensing strategies only constrain transient stress at each moment to not exceed a dynamic threshold, failing to assess the impact of rapid damage accumulation from high-frequency stress cycles in the low curing degree range on the total cumulative damage degree at the end of cooling. When these two strategies operate independently, a blind spot in damage assessment is created in the initial cooling phase. The optimized cooling parameters trigger rapid accumulation of micro-damage in the adhesive layer, ultimately leading to interlayer cracking during product service.

[0021] The hot pressing equipment involved in this embodiment is a multi-layer hot press equipped with a zoned temperature control module and a pressure control module. Each zone temperature control module can independently adjust the temperature of the cooling medium. The curing kinetic parameters and fatigue damage parameters of the adhesive are obtained in advance using differential scanning calorimetry and fatigue testing.

[0022] A method for intelligent optimization of hot-pressing process parameters for crack-resistant decorative panels according to an embodiment of the present invention, such as... Figure 1 As shown, it includes the following steps: Step 1: Obtain the adhesive property parameters and calculate the initial curing degree vector of the partition at the start of cooling; Acquire the curing kinetic parameters, nonlinear mapping curve data of degree of cure and interlaminar shear strength, and fatigue damage parameters of the adhesive used in the hot pressing process. Input the temperature-time history of the holding pressure stage into the curing kinetic equation for numerical integration, calculate the degree of cure value of each partition adhesive layer at the end of the holding pressure stage, and generate the initial degree of cure vector of each partition at the start of cooling.

[0023] It should be noted that the initial curing degree vector of a zone refers to the vector composed of the curing degree values ​​of the adhesive layer in each zone at the end of the pressure holding stage after the hot press plate surface is divided into multiple temperature-controlled zones. Since there may be differences in temperature distribution among the zones during the pressure holding stage, the initial curing degree values ​​of each zone are not the same. The initial curing degree vector of a zone reflects the differences in the curing state of the adhesive layer in each zone at the beginning of the cooling stage.

[0024] Step 2: Perform a time-step joint calculation of the degree of solidification and the cumulative damage during the cooling stage; Discretize the time axis of the cooling phase as follows: Each time step, of which This represents the total number of time steps. Within the allowable range of the cooling rate at each time step, Latin hypercube sampling is used to generate... Sample of cooling rate parameters for component regions, among which This represents the number of sample groups.

[0025] For each group of cooling rate parameter samples, starting from the first time step, the following time-step joint calculations are performed sequentially: The degree of cure increment is calculated based on the temperature values ​​of each zone and the curing kinetic equation at the current time step; this increment is then added to the cumulative degree of cure value from the previous time step to obtain the current cumulative degree of cure value; the current cumulative degree of cure value is substituted into the nonlinear mapping curve to obtain the instantaneous shear strength value of each zone at the current moment. Simultaneously, the transient interlaminar shear stress value is calculated based on the interlayer temperature difference and the difference in thermal expansion coefficients of the materials in each layer at the current time step.

[0026] Step 2 further includes: performing rainflow counting analysis on the complete stress time history from the start of cooling to the current time step to identify the stress amplitude of each stress cycle. The instantaneous shear strength value of each zone at the current moment is used instead of the fixed fully cured strength value as the normalization benchmark for each stress cycle, and the current cumulative damage value normalized to dynamic strength is calculated according to the modified Miner criterion.

[0027] It should be noted that using the instantaneous shear strength value instead of the fixed fully cured strength value as the normalization benchmark means that when calculating the contribution of each stress cycle to the cumulative damage, the damage increment of each stress cycle is obtained in the following way: the stress amplitude of the stress cycle is... Instantaneous shear strength value corresponding to the moment when the stress cycle occurs Substituting the ratio into the fatigue damage parameters, the damage increment for that stress cycle is obtained. Specifically, the increase in damage. Calculate using the following formula: ; in, The stress cycle amplitude identified by rainflow counting analysis, For a moment The instantaneous shear strength value is obtained from the cumulative degree of cure through a nonlinear mapping curve. The fatigue life function of the adhesive is expressed as the normalized stress ratio. The number of stress cycles required for the adhesive to reach fatigue failure is determined by pre-acquired fatigue damage parameters. This represents the incremental contribution of this stress cycle to the cumulative damage. The cumulative damage value at the end of the current time step. It is the sum of the damage increments from the start of cooling to the current time step, i.e. ,in This represents the total number of stress cycles identified from the start of cooling to the current time step. For the first The damage increment per stress cycle. In the low-curing-degree range during the initial cooling phase. Significantly lower than the fully cured strength, resulting in a normalized stress ratio Increase The corresponding reduction reflects the physical fact that a single stress cycle causes a larger increase in microscopic damage under low curing conditions.

[0028] After the time-step joint calculation is completed, the solidification degree sequence, dynamic intensity sequence and dynamic normalized cumulative damage degree sequence of the cooling rate parameter samples of each component region are generated at each time step.

[0029] Step 3: Train the solidification damage coupling prediction network; For each group of cooling rate parameter samples, the final cumulative damage value, final curing value, and final warpage value of each zone at the end of cooling are extracted. Using the concatenation of the zone cooling rate parameter vector and the zone initial curing value vector as input, and the final cumulative damage value, final curing value, and warpage value of each zone as multi-task output, the curing damage coupling prediction network is trained until the error of each task on the validation set converges.

[0030] The curing damage coupling prediction network is a multi-task regression network, comprising an input layer, a shared hidden layer, a time-step attention weight layer, and a multi-task output head. The data transfer relationships between these components are as follows: The input layer receives the concatenated features of the partition cooling rate parameter vector and the partition initial curing degree vector, and passes these features to the shared hidden layer; the shared hidden layer performs a nonlinear transformation on the concatenated features and outputs the result. The hidden layer feature vectors corresponding to each time step, where The hidden feature vectors are passed to the time-step attention weight layer to determine the total number of time steps. The time-step attention weight layer calculates an attention score for each hidden feature vector at each time step, and then normalizes the attention scores using softmax before using them as weights. The hidden layer feature vectors are weighted and aggregated to output an aggregated feature vector that incorporates temporal importance information. The aggregated feature vector is then passed to the multi-task output head. The multi-task output head consists of three independent fully connected layers: a cumulative damage output head, a curvature output head, and a warpage output head. Each output head independently receives the aggregated feature vector and outputs the predicted values ​​of the final cumulative damage value, the final curvature value, and the warpage value for each partition.

[0031] The training uses a multi-task mean squared error loss function. The mean squared error of the predicted value of each task output head and the corresponding label value generated in step 2 are calculated separately and then weighted and summed. The Adam optimization algorithm is used to minimize the total loss.

[0032] Step 4: Perform a cooling rate optimization solution under dual constraints; Using the cooling rate of each partition at each time step as the decision variable, and constrained by the following conditions: the final cumulative damage value of each partition output by the curing damage coupling prediction network is lower than the safe damage threshold; the final curing degree value of each partition reaches the minimum curing degree standard; and the warpage is lower than the flatness standard, the optimization objective is to minimize the total cooling time. An interior-point method is used to solve this problem. In each iteration, the curing damage coupling prediction network simultaneously evaluates the degree to which the cumulative damage constraint and the curing degree constraint are satisfied, outputting the optimal partition cooling rate parameter vector that satisfies both the curing degree and cumulative damage constraints.

[0033] It should be noted that the safety damage threshold, minimum curing standard, and flatness standard are all engineering criteria pre-set according to product quality requirements. They correspond to the maximum allowable cumulative damage degree that the adhesive layer will not crack during service, the minimum degree of curing required to ensure bonding reliability, and the maximum allowable warpage required for the product to meet appearance requirements.

[0034] In this embodiment of the application, in order to improve the reliability of the optimization results, the following steps are also included in addition to step 4: Step 5: Verify and correct the prediction accuracy of the curing damage coupling prediction network; The temperature time series corresponding to the optimal partition cooling rate parameter vector is input into the curing kinetics equation and rainflow counting analysis, and a precise calculation is performed step-by-step throughout the entire process. The deviation between the precisely calculated dynamic normalized cumulative damage value and the predicted value of the curing damage coupling prediction network is compared step-by-step. If the deviation exceeds the allowable range, the validation data is added to the training set to incrementally train the curing damage coupling prediction network, and the optimization solution in step 4 is re-executed. If the deviation is within the allowable range, the cooling rate sequence corresponding to the optimal partition cooling rate parameter vector is segmented into steps and subjected to a ramp transition to generate a sequence of cooling medium temperature setpoints for each partition in each time period.

[0035] It should be noted that segmented stepping refers to dividing the continuously changing cooling rate sequence into several constant rate segments according to time periods, and ramp transition processing refers to inserting a linear transition segment between adjacent constant rate segments to adapt to the actual adjustment characteristics of the temperature control module of the hot pressing equipment.

[0036] In this embodiment of the application, in order to generate a complete executable process solution, the following steps are also included: Step 6: Generate the hot pressing process parameter execution plan; By integrating the setpoint sequence of cooling medium temperature with the temperature, pressure, and time parameters of the holding stage, a hot pressing process parameter execution plan covering the entire holding to cooling stage is generated.

[0037] This implementation performs simultaneous calculations of the curing kinetics equation and rainflow counting damage analysis at each discrete time step during the cooling stage. The instantaneous shear strength value corresponding to the current cumulative degree of cure is used instead of the fixed fully cured strength value as the normalized benchmark for cumulative damage. In the early cooling stage, when the adhesive curing degree is in a low range, the instantaneous shear strength value is significantly lower than the fully cured strength value, resulting in a correspondingly larger normalized stress ratio. This leads to a larger damage increment obtained from each stress cycle calculation, eliminating the bias of existing cumulative damage assessments that systematically underestimate damage in the early cooling stage when using a fixed fully cured strength as the benchmark. Simultaneously, the time-step joint calculation tracks the dynamic evolution of cumulative damage while constraining the transient stress at each moment. This ensures that even if the transient stress at each moment does not exceed the dynamic threshold, the rapid increase in cumulative damage caused by high-frequency stress cycles in the low curing degree range can still be fully captured. The time-step attention weight layer enables the curing damage coupling prediction network to automatically assign higher weights to the time steps corresponding to the low curing degree range when fitting the mapping relationship between the cooling rate parameter and the final cumulative damage, improving the fitting accuracy of the parameter response law in the high-damage-sensitive stage of the early cooling stage. The optimization solution, constrained by both cumulative damage and final cure degree, automatically limits the cooling rate in the early stages of cooling due to the tightening of the cumulative damage degree constraint, reducing transient stress amplitude and extending the time window for residual curing. Once the cure degree has sufficiently increased to the strength-insensitive range, the cumulative damage degree constraint relaxes, allowing the cooling rate to be released and shortening the total cooling cycle. This simultaneous calculation and constraint of cure degree and cumulative damage at the time step level eliminates the damage assessment blind spot that occurs in the early stages of cooling when the two existing strategies operate independently, shortening the cooling cycle while ensuring product bonding reliability.

[0038] The following is an example of an application of the present invention, such as Figure 2-9 As shown, the implementation process is as follows: A board manufacturer produces three-layer crack-resistant decorative panels using a phenolic-modified urea-formaldehyde resin system as the adhesive. The hot-pressed panel surface is divided into four temperature-controlled zones (zones 1 to 4). After the pressure holding stage, the cooling stage begins. The factory reports that after 6 to 12 months of service, interlayer cracks appear in the corner areas, suspecting that improper cooling stage parameter settings led to excessive accumulation of micro-damage to the adhesive layer. The factory had previously obtained curing kinetic parameters using differential scanning calorimetry and fatigue damage parameters through fatigue testing. The nonlinear mapping curve between the degree of curing and interlayer shear strength had also been calibrated. The cooling stage time axis was discretized as follows: There are 1 time step, each step lasts 90 seconds, and the total cooldown window is 30 minutes.

[0039] Step 1: Calculate the initial degree of curing of each zone at the start of cooling; After inputting the temperature-time history of each zone into the curing kinetic equation during the pressure holding stage, numerical integration was performed on each zone to calculate the degree of curing of the adhesive layer at the end of the pressure holding stage. Zone 1, located in the center of the hot press plate, has the most uniform temperature and the highest degree of curing; Zone 4, located at the edge of the plate, dissipates heat faster and has a relatively lower degree of curing. The generated initial degree of curing vectors for each zone reflect the differences in the curing state of each zone at the start of cooling.

[0040] Table 1 Initial Curing Degree Vectors of Each Zone at Cooling Start Time

[0041] Step 2: Calculate the degree of curing and cumulative damage step by step; Generated using Latin hypercube sampling Sample cooling rate parameters for each component zone. Taking zone 4 as an example, the data flow process of time-step joint calculation is illustrated.

[0042] At the third time step (270 seconds after the start of cooling), the current temperature of partition 4 is 121℃. Based on the curing kinetics equation, the degree of cure increment at this time step is calculated to be 0.03, and the cumulative degree of cure value is updated to 0.76. Substituting 0.76 into the nonlinear mapping curve, the instantaneous shear strength value is obtained. MPa. At this point, the fully cured strength is 9.20 MPa, and the instantaneous strength is only 63.5% of the fully cured strength.

[0043] Within the same time step, based on the interlayer temperature difference (approximately 18°C) and the difference in thermal expansion coefficients between partition 4 and adjacent layers, the transient interlayer shear stress value was calculated to be 2.31 MPa. Rainflow counting analysis was performed on the complete stress time history from the start of cooling to the third time step, identifying three stress cycles. The stress was then analyzed based on the time of occurrence of each cycle. As a normalization baseline, the damage increment for each cycle is calculated using the following formula: ; in, This represents the damage increment corresponding to a single stress cycle. Let be the fatigue life function with normalized stress ratio as the independent variable. This represents the stress cycle amplitude. This represents the instantaneous shear strength at the moment the cycle occurs.

[0044] Taking the second stress cycle as an example, MPa MPa, normalized stress ratio is 0.39, corresponding to Damage increment If the fully cured strength of 9.20 MPa is used as a benchmark, the normalized stress ratio for the same cycle is only 0.24. The increase is significant, and the incremental damage will be severely underestimated.

[0045] Table 2 shows the key results of joint calculations at each time step in partition 4 (partial time steps).

[0046] As shown in Table 2, the instantaneous shear strength is in the low range during the initial cooling stage (steps 1 to 3), and the incremental increase in cumulative damage per step is significantly greater than that during the later cooling stage (steps 8 to 15), which is consistent with the physical fact that damage sensitivity is high in the low curing degree range.

[0047] Step 3: Train the solidification damage coupling prediction network; For all 200 samples, the final cumulative damage value, final solidification value, and warpage value of each partition at the end of cooling were extracted to construct training labels. The input features are the cooling rate parameter vectors of the four partitions (total... The concatenation of the 4-dimensional initial curing degree vector and the 5-dimensional initial curing degree vector results in a total of 84 dimensions.

[0048] The time-step attention weight layer calculates attention scores for each of the 20 time steps and normalizes them using softmax. After training convergence, the average attention weight for the initial cooling phase (steps 1 to 5) is 0.073, and the average attention weight for the later cooling phase (steps 16 to 20) is 0.031. The weights obtained in the initial cooling phase are approximately 2.4 times those in the later phase, reflecting the higher contribution of the low solidification range to the final cumulative damage.

[0049] Table 3 Training and validation errors of the solidification damage coupling prediction network

[0050] Step 4: Optimization of cooling rate under multiple constraints; Using the cooling rate of each zone at each time step as the decision variable, the safety damage threshold is set to 0.0015, the minimum curing standard to 0.92, and the flatness standard (maximum warpage) to 1.2 mm. The interior point method is used iteratively to minimize the total cooling time.

[0051] In each iteration, the current cooling rate parameter vector is concatenated with the initial curing degree vector of the partition and input into the curing damage coupling prediction network. The cumulative damage prediction value and the curing degree prediction value are obtained simultaneously to determine whether the dual constraints are satisfied. The optimization results show that in the initial cooling stage (steps 1 to 6), the cooling rate of each partition is tightened by the cumulative damage constraint, with the cooling rate of partition 4 limited to below 0.31℃ / s. After step 10, the curing degree generally exceeds 0.90, the cumulative damage constraint is relaxed, and the cooling rate can be released up to 0.68℃ / s. The total cooling time is reduced from 28 minutes in the original process to 23 minutes.

[0052] Table 4 Comparison of key indicators for each partition before and after optimization.

[0053] Before optimization, the cumulative damage level of partition 4 (0.00187) exceeded the safe damage level threshold (0.0015), which directly corresponds to the corner cracking phenomenon reported by the factory. After optimization, the damage level of partition 4 decreased to 0.00138, meeting the constraint requirements.

[0054] Step 5: Precise verification and segmented step-by-step processing; The temperature time series corresponding to the optimal partition cooling rate parameter vector is input into the curing kinetics equation and rainflow counting analysis. Precise calculations are performed step by step for 20 time steps. Taking partition 4 as an example, the deviation between the precise value and the predicted network output value is compared step by step.

[0055] Table 5. Verification of the discrepancy between the exact calculated values ​​and the predicted values ​​in partition 4 (partial time steps).

[0056] The relative deviations at each time step are all below the allowable upper limit (5%), so no incremental training is required, and the process can proceed directly to segmented step processing. The optimal cooling rate sequence is divided into constant rate segments by four time periods, and a 30-second linear ramp transition is inserted between adjacent segments to generate the cooling medium temperature setpoint sequence for each zone.

[0057] Step 6: Generate a hot pressing process parameter execution plan; The setpoint sequence of cooling medium temperature for each zone is integrated with the pressure holding stage parameters (temperature 148℃, pressure 2.8MPa, pressure holding time 12 minutes) to generate a hot pressing process parameter execution plan covering the entire stage from pressure holding to cooling, which is then sent to the zone temperature control module and pressure control module of the hot pressing equipment.

[0058] Table 6 Hot pressing process parameter execution plan (cooling stage zone 4 temperature setpoint sequence)

[0059] Throughout the implementation process, the data starts from the initial curing degree vector of the partition in step 1. In step 2, it drives the joint evolution of curing degree and cumulative damage degree step by step, generating dynamic sequence data that reflects the true damage sensitivity in the low curing degree range. This sequence data is used as training labels in step 3, causing the time-step attention weight layer of the curing-damage coupled prediction network to automatically tilt towards the high-damage-sensitivity period in the early cooling stage. Step 4 utilizes this network to simultaneously evaluate the dual constraints in the interior-point method iteration. The output optimal cooling rate parameter vector is automatically constrained in the early cooling stage and released in the later stage, eliminating the potential risk of exceeding the cumulative damage limit in the original process of partition 4. Accurate verification in step 5 confirms that the prediction deviation is within the allowable range. Step 6 transforms the continuous cooling rate sequence into segmented temperature setpoints that the equipment can directly execute, completing the complete data flow from numerical optimization results to an executable process scheme.

[0060] The embodiments of the present invention have been described above. However, the embodiments are not limited to the specific implementation methods described above. The specific implementation methods described above are merely illustrative and not restrictive. Those skilled in the art can make more equivalent embodiments under the guidance of the present embodiments, and all of them are within the protection scope of the present embodiments.

Claims

1. A method for intelligent optimization of hot-pressing process parameters for crack-resistant decorative panels, characterized in that, include: Acquire the curing kinetic parameters, nonlinear mapping curve data of curing degree and interlaminar shear strength, and fatigue damage parameters of the adhesive; calculate the curing degree value of each partition adhesive layer at the end of the pressure holding time; and generate the initial curing degree vector of each partition. The cooling stage is discretized into multiple time steps, and joint calculations are performed on the cooling rate parameter samples of each component region step by step: the degree of curing increment is calculated based on the current time step temperature value and the curing kinetic equation, and the current cumulative degree of curing value is obtained by summing them. The instantaneous shear strength value is obtained through the nonlinear mapping curve; the transient interlayer shear stress value is calculated based on the interlayer temperature difference and the difference in thermal expansion coefficients; rainflow counting analysis is performed on the stress time history from the start of cooling to the current time step, and the instantaneous shear strength value is used to replace the fixed fully cured strength value as the normalization benchmark to calculate the current cumulative damage value. Using the partition cooling rate parameter vector and the initial curing degree vector of the partition as inputs, and the final cumulative damage value, final curing value and warpage value of each partition as outputs, a curing damage coupling prediction network is trained. With constraints that the final cumulative damage value of each partition output by the curing damage coupling prediction network is lower than the safe damage threshold, the final curing value reaches the minimum curing standard, and the warpage is lower than the flatness standard, and with the goal of minimizing the total cooling time, the optimal partition cooling rate parameter vector is solved.

2. The intelligent optimization method for hot-pressing process parameters of crack-resistant decorative panels according to claim 1, characterized in that, The calculation of the current cumulative damage value by replacing the fixed fully cured strength value with the instantaneous shear strength value as the normalization benchmark includes: For each stress cycle identified by the rainflow counting analysis, the ratio of the stress amplitude of the stress cycle to the instantaneous shear strength value corresponding to the occurrence time of the stress cycle is used as the normalized stress ratio. Substitute this ratio into the fatigue life function of the adhesive to obtain the number of stress cycles corresponding to fatigue failure under the normalized stress ratio. Take the reciprocal of the number of stress cycles as the damage increment of the stress cycle. Sum the damage increments of all identified stress cycles from the start of cooling to the current time step to obtain the current cumulative damage value.

3. The intelligent optimization method for hot-pressing process parameters of crack-resistant decorative panels according to claim 1, characterized in that, The solidification damage coupling prediction network is a multi-task regression network, including an input layer, a shared hidden layer, a time-step attention weight layer, and a multi-task output head; The input layer receives the concatenated features of the partition cooling rate parameter vector and the initial curing degree vector of the partition and transmits them to the shared hidden layer. The shared hidden layer performs a nonlinear transformation on the concatenated features and outputs the hidden layer feature vectors corresponding to each time step. The time step attention weight layer calculates the attention score for the hidden layer feature vector of each time step, normalizes the attention score, and uses it as a weight to perform weighted aggregation of the hidden layer feature vectors of each time step, outputting an aggregated feature vector. The multi-task output head includes a cumulative damage output head, a curing degree output head, and a warpage output head. Each output head independently receives the aggregated feature vector and outputs the predicted values ​​of the final cumulative damage value, the final curing degree value, and the warpage value of each partition, respectively.

4. The intelligent optimization method for hot-pressing process parameters of crack-resistant decorative panels according to claim 3, characterized in that, The attention weight layer of the time step assigns a higher attention score to the time step corresponding to the low curing degree range than to the time step corresponding to the high curing degree range, thereby increasing the feature contribution ratio of the high damage sensitivity stage in the early cooling stage in the aggregated feature vector.

5. The intelligent optimization method for hot-pressing process parameters of crack-resistant decorative panels according to claim 3, characterized in that, The training of the solidification damage coupling prediction network adopts a multi-task mean square error loss function. The mean square error of the predicted value of each task output head and the corresponding label value generated by joint calculation at each time step are calculated separately and then weighted and summed. The total loss is minimized by the Adam optimization algorithm and the training is continued until the error of each task on the validation set converges.

6. The intelligent optimization method for hot-pressing process parameters of crack-resistant decorative panels according to claim 1, characterized in that, The optimal partition cooling rate parameter vector is solved using the interior point method, with the cooling rate of each partition at each time step as the decision variable. In each iteration, the degree of satisfaction of the cumulative damage constraint and the curing degree constraint is evaluated simultaneously through the curing damage coupling prediction network.

7. The intelligent optimization method for hot-pressing process parameters of crack-resistant decorative panels according to claim 1, characterized in that, After solving for the optimal partition cooling rate parameter vector, the process also includes: The temperature time series corresponding to the optimal partition cooling rate parameter vector is input into the curing kinetic equation and rainflow counting analysis, and the entire process is performed with precise calculations step by step. The deviation between the precisely calculated dynamic normalized cumulative damage value and the predicted value of the curing damage coupling prediction network is compared step by step. If the deviation exceeds the allowable range, the verification data is added to the training set to incrementally train the curing damage coupling prediction network, and the step of solving the optimal partition cooling rate parameter vector is re-executed.

8. The intelligent optimization method for hot-pressing process parameters of crack-resistant decorative panels according to claim 7, characterized in that, If the deviation is within the allowable range, the cooling rate sequence corresponding to the optimal partition cooling rate parameter vector is segmented and subjected to ramp transition processing to generate a cooling medium temperature setpoint sequence for each partition in each time period; wherein, the segmented step processing refers to dividing the continuously changing cooling rate sequence into several constant rate segments according to time period, and the ramp transition processing refers to inserting a linear transition segment between adjacent constant rate segments; the cooling medium temperature setpoint sequence is integrated with the temperature, pressure and time parameters of the holding stage to generate a hot pressing process parameter execution scheme covering the entire holding to cooling stage.

9. The intelligent optimization method for hot-pressing process parameters of crack-resistant decorative panels according to claim 1, characterized in that, The cooling rate parameter samples for each component region are generated by Latin hypercube sampling within the allowable range of cooling rate at each time step.

10. A smart optimization system for hot-pressing process parameters of crack-resistant decorative panels, used to execute the smart optimization method for hot-pressing process parameters of crack-resistant decorative panels according to any one of claims 1 to 9, characterized in that, include: The initial curing degree calculation module is used to obtain the curing kinetic parameters of the adhesive, the nonlinear mapping curve data of curing degree and interlayer shear strength, and fatigue damage parameters. It calculates the curing degree value of each partition adhesive layer at the end of the pressure holding time and generates the initial curing degree vector of the partition. The time-step joint calculation module is used to discretize the cooling stage into multiple time steps, and perform curing degree increment calculation, instantaneous shear strength value acquisition, transient interlayer shear stress value calculation, and rainflow counting analysis on the cooling rate parameter samples of each component area on a time-step basis. The instantaneous shear strength value is used to replace the fixed fully cured strength value as the normalization benchmark to calculate the current cumulative damage value. The network training module is used to train a curing damage coupling prediction network by taking the partition cooling rate parameter vector and the partition initial curing degree vector as inputs and the final cumulative damage value, final curing value and warpage value of each partition as outputs. The optimization solution module is used to solve for the optimal partition cooling rate parameter vector with the constraints that the final cumulative damage value of each partition output by the curing damage coupling prediction network is lower than the safe damage threshold, the final curing value reaches the minimum curing standard, and the warpage is lower than the flatness standard, and with the goal of minimizing the total cooling time.

Images