Method for optimizing the compression molding process of carbon fiber reinforced composite thin materials

Abstract

The present application relates to the technical field of composite material forming, and discloses a molding process optimization method for carbon fiber reinforced composite material thin materials. The method collects multi-dimensional process parameters such as temperature gradient distribution data, pressure load change data and resin flow rate data in the molding process in real time through a distributed sensing network; establishes a process state evaluation model to identify initial defect characteristics of these parameters, and generates an initial defect characteristic spectrum; performs defect mode matching on the initial defect characteristic spectrum based on a material characteristic database, separates a main defect data stream from a secondary defect data stream; constructs a real-time compensation instruction set according to the main defect data stream, synchronously drives a molding equipment actuator to perform dynamic process parameter correction; feeds back the corrected secondary defect data stream to the material characteristic database to update the defect mode matching rules; iteratively performs dynamic process parameter correction until the process convergence condition is met, and outputs a quality control report of the final formed thin material.

Description

Technical Field

[0001] This invention relates to the field of composite material molding technology, specifically to an optimized method for compression molding of carbon fiber reinforced composite thin films. Background Technology

[0002] Carbon fiber reinforced composite thin sheets, with their high specific strength, excellent fatigue resistance, and strong design flexibility, have been widely used in precision manufacturing fields such as aerospace, new energy vehicles, and high-end electronic equipment. The molding quality of these materials directly determines the structural safety and service life of the end product. In particular, in thin-walled components such as drone wing skins, power battery casings, and chip carriers, stringent requirements are placed on the material's thickness uniformity, resin content distribution, and interfacial bonding strength.

[0003] Current mainstream compression molding processes face multiple technical challenges during production. Process parameters are highly coupled, with complex nonlinear relationships between temperature gradients, pressure loads, and resin flow. Traditional, empirically-based parameter settings struggle to adapt to the characteristic fluctuations of different batches of materials, easily leading to defects such as bubbles, wrinkles, and disordered fiber orientation. For example, when the temperature distribution deviation in the mold cavity exceeds ±5℃, the local resin curing degree difference in thin materials can reach over 20%, directly affecting the uniformity of the material's mechanical properties.

[0004] Defect detection exhibits significant lag, with existing technologies largely relying on offline inspection after molding, failing to capture the dynamic evolution of defects during the molding process in real time. For ultra-thin components less than 1 mm thick, even minute pressure fluctuations can trigger interlayer delamination, and post-processing not only increases production costs but also makes it difficult to restore structural integrity. Statistical data shows that traditional processes have high defect rates, and in high-reliability fields such as aerospace, this indicator has become a key bottleneck restricting industrial applications.

[0005] Furthermore, there is insufficient process adaptability: with the diversification of composite material systems, the curing kinetics of different materials vary significantly. Existing molding equipment lacks an intelligent adjustment mechanism based on material properties. When changing material types or adjusting component dimensions, a large number of trial-and-error experiments are required to recalibrate process parameters, leading to extended production cycles and difficulty in ensuring batch stability.

[0006] Data utilization efficiency is low. Process data such as temperature and pressure generated during molding are mostly stored in isolated form, without forming a knowledge base associated with defect patterns, leading to the recurrence of similar defects. At the same time, the single-variable feedback mechanism of traditional control systems cannot cope with complex defects caused by multi-parameter coupling. For example, resin enrichment and fiber exposure are often accompanied by temperature field distortion and uneven pressure distribution, and adjusting a single parameter is difficult to achieve fundamental improvement.

[0007] Developing an intelligent optimization method that can perceive multi-dimensional process parameters in real time, accurately identify defect characteristics, and dynamically correct process parameters is of great significance for improving the molding quality of carbon fiber reinforced composite thin films, reducing production costs, and promoting technological upgrading in the high-end manufacturing field. Summary of the Invention

[0008] The purpose of this invention is to provide an optimized method for the compression molding process of carbon fiber reinforced composite thin sheets, so as to solve the problems mentioned in the background art.

[0009] To achieve the above objectives, the present invention provides an optimized method for the compression molding process of carbon fiber reinforced composite thin sheets, the method comprising:

[0010] Multi-dimensional process parameters during the compression molding process are collected in real time through a distributed sensor network. These process parameters include temperature gradient distribution data, pressure load change data, and resin flow rate data.

[0011] A process status assessment model is established to identify initial defect features of the multi-dimensional process parameters and generate an initial defect feature map.

[0012] Based on the material property database, defect pattern matching is performed on the initial defect feature map to separate the main defect data stream and the secondary defect data stream.

[0013] A real-time compensation instruction set is constructed based on the main defect data stream to synchronously drive the actuator of the molding equipment to perform dynamic process parameter correction.

[0014] The corrected sub-defect data stream is fed back to the material property database to update the defect pattern matching rules;

[0015] Iteratively execute dynamic process parameter corrections until the process convergence conditions are met, and output the final quality control report of the formed thin material.

[0016] Preferably, the establishment of the process status assessment model includes:

[0017] Set the initial number of topological layers in the model and the feature extraction threshold;

[0018] The multi-dimensional process parameters are input into a convolutional neural network to generate a primary defect feature vector.

[0019] The model optimization coefficients are calculated based on the difference between the primary defect feature vector and the pre-stored standard feature library.

[0020] An adaptive genetic algorithm is used to iteratively optimize the initial number of topological layers and the feature extraction threshold of the model to obtain the target number of topological layers and the target feature extraction threshold.

[0021] The convolutional neural network structure is reconstructed by the target topology layer number and the target feature extraction threshold, and the optimized defect feature vector is output as the initial defect feature map.

[0022] Preferably, the step of calculating the model optimization coefficients based on the difference between the primary defect feature vector and the pre-stored standard feature library includes:

[0023] The primary defect feature vector is divided into N groups of feature sub-vectors;

[0024] Extract the corresponding N sets of standard feature sub-vectors from the pre-stored standard feature library;

[0025] Calculate the multidimensional similarity between each set of feature vectors and the standard feature vectors;

[0026] A comprehensive similarity index is generated by fusing N sets of multidimensional similarities.

[0027] The comprehensive similarity index is compared with the preset critical value to generate the difference value of the model optimization coefficient;

[0028] The convolutional neural network weight parameters are updated in reverse based on the difference in the model optimization coefficients.

[0029] Preferably, the calculation of the multidimensional similarity between each set of feature vectors and the standard feature vectors includes:

[0030] Extract the spatial distribution feature values ​​and temporal variation feature values ​​of the feature sub-vectors respectively;

[0031] The spatial distribution feature value and the temporal variation feature value are normalized and weighted and fused to generate a composite feature value;

[0032] Local similarity is generated based on the Euclidean distance between composite feature values ​​and standard feature vectors;

[0033] Multidimensional similarity is generated by multiplying local similarity with a preset weight matrix.

[0034] Preferably, the separation of the primary defect data stream and the secondary defect data stream includes:

[0035] Establish a defect feature correlation matrix in the material property database;

[0036] The initial defect feature map is mapped to the defect feature correlation matrix to generate defect impact weight values;

[0037] Feature regions whose defect impact weight value exceeds the preset main defect threshold are identified as the core regions of the main defect.

[0038] Extract the data stream corresponding to the core region of the main defect to form the main defect data stream;

[0039] Data streams of feature regions whose defect impact weight values ​​are lower than the primary defect threshold are marked as secondary defect data streams.

[0040] The secondary defect data stream is input into the defect feature correlation matrix to update the secondary defect impact factor.

[0041] Preferably, the step of constructing a real-time compensation instruction set based on the main defect data stream includes:

[0042] Analyze the temperature deviation parameters, pressure fluctuation parameters, and resin flow rate anomaly parameters in the main defect data stream;

[0043] Generate a temperature compensation gradient function based on the temperature deviation parameters;

[0044] A pressure compensation step size sequence is constructed based on pressure fluctuation parameters;

[0045] The resin flow rate compensation ratio is calculated by driving abnormal resin flow rate parameters.

[0046] A multi-parameter collaborative compensation instruction set is generated by integrating the temperature compensation gradient function, the pressure compensation step size sequence, and the resin flow rate compensation ratio.

[0047] The multi-parameter collaborative compensation instruction set is transmitted to the main controller of the molding equipment.

[0048] Preferably, the dynamic process parameter correction includes:

[0049] The main controller of the molding equipment analyzes the multi-parameter collaborative compensation instruction set;

[0050] Adjust the heating plate zone temperature control strategy according to the temperature compensation gradient function;

[0051] The pressure loading curve of the hydraulic actuator is reconstructed based on the pressure compensation step sequence; the resin flow compensation ratio is used to correct the opening control parameters of the glue injection valve.

[0052] Synchronously record device response delay data during the correction process;

[0053] The device response delay data is fed back to the real-time compensation instruction set for response delay compensation.

[0054] Preferably, the step of feeding back the corrected secondary defect data stream to the material property database to update the defect pattern matching rules includes:

[0055] Extract defect feature evolution trend data from the secondary defect data stream;

[0056] Establish a mapping table between defect feature evolution trend data and process parameters in the material property database;

[0057] When a new defect feature pattern appears in the defect data stream, a new defect feature pattern record is created in the association mapping table;

[0058] The influence coefficient threshold of existing defect feature patterns in the association mapping table is corrected based on the historical secondary defect data stream.

[0059] The updated association mapping table will be used as the new version of the defect pattern matching rules.

[0060] Preferably, the process convergence conditions include:

[0061] Real-time monitoring of the characteristic intensity decay rate of the main defect data stream;

[0062] When the feature intensity decay rate is lower than the preset convergence threshold for M consecutive iterations, the sub-defect data stream is determined to be stable within the allowable fluctuation range.

[0063] Synchronously detect the frequency of correction actions of the actuator in the molding equipment;

[0064] A process termination command is triggered when the frequency of corrective actions drops below the equipment's baseline maintenance frequency.

[0065] Output the thickness uniformity test data and fiber orientation distribution data of the final shaped thin material as a quality control report.

[0066] Preferably, the iterative execution process includes:

[0067] When the feature intensity decay rate does not reach the convergence threshold, the current defect data stream is input into the process status assessment model to generate a new generation of initial defect feature map.

[0068] Defect patterns are rematched based on the updated material property database to the new generation of initial defect feature maps;

[0069] The primary and secondary defect data streams are re-separated and an enhanced real-time compensation instruction set is generated;

[0070] Execute the enhanced real-time compensation instruction set until the process convergence condition is met.

[0071] Compared with the prior art, the beneficial effects of the present invention are:

[0072] This process optimization method uses a distributed sensor network to collect multi-dimensional process parameters in real time, overcoming the limitations of traditional single-point monitoring. It comprehensively reflects key information such as temperature gradient distribution, pressure load changes, and resin flow rate during the compression molding process, making process status monitoring more accurate and comprehensive. This real-time acquisition of multi-dimensional parameters provides a rich data foundation for subsequent defect identification and process correction, helping to promptly detect anomalies in the molding process.

[0073] A process status assessment model was established to identify initial defect features of multi-dimensional process parameters and generate initial defect feature maps, achieving automated and intelligent defect identification and reducing over-reliance on human experience. Compared with traditional offline inspection methods, it can quickly capture initial defect features during the molding process, saving time for subsequent defect processing and reducing the possibility of defect expansion.

[0074] Defect pattern matching is performed on the initial defect feature map based on a material property database to separate the primary defect data stream from the secondary defect data stream, thus improving the accuracy of defect classification. By clearly defining the different data streams of primary and secondary defects, targeted processing measures can be taken, avoiding the waste of resources and low processing efficiency caused by treating all defects the same.

[0075] A real-time compensation instruction set is constructed based on the main defect data stream, synchronously driving the actuator of the molding equipment to dynamically correct process parameters, thus achieving real-time adjustment of process parameters. This dynamic correction can respond promptly to changes in the molding process, ensuring that process parameters match the material molding requirements, reducing product quality problems caused by parameter deviations, and improving the stability of the molding process.

[0076] The corrected secondary defect data stream is fed back to the material property database to update the defect pattern matching rules, forming a closed-loop optimization system. As the database is continuously updated, the accuracy of defect pattern matching gradually improves, making subsequent defect identification and processing more precise, and further enhancing the effectiveness of process optimization.

[0077] Iteratively executing dynamic process parameter corrections until the process convergence conditions are met ensures the quality of the final molded product. Through multiple iterations and continuous optimization of process parameters, defects in the molding process can be gradually eliminated, enabling the product quality to meet expected requirements. Simultaneously, the output quality control report facilitates product quality traceability and analysis. Attached Figure Description

[0078] Figure 1 This is a schematic diagram illustrating the working principle of the optimized compression molding process for carbon fiber reinforced composite thin materials according to the present invention.

[0079] Figure 2 A flowchart for calculating the optimization coefficients of the model;

[0080] Figure 3 A flowchart for separating primary and secondary defect data streams;

[0081] Figure 4 This is a flowchart for dynamic process parameter correction. Detailed Implementation

[0082] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0083] Please see Figure 1 This invention provides an optimized method for the compression molding process of carbon fiber reinforced composite thin films, the method comprising:

[0084] Multi-dimensional process parameters during the compression molding process are collected in real time through a distributed sensor network, including temperature gradient distribution data, pressure load variation data, and resin flow rate data. A process status assessment model identifies initial defect features based on these multi-dimensional process parameters, generating an initial defect feature map. Defect pattern matching is performed on the initial defect feature map based on a material property database, separating the primary defect data stream from the secondary defect data stream. A real-time compensation instruction set is constructed based on the primary defect data stream, synchronously driving the compression molding equipment actuators to dynamically correct process parameters. The corrected secondary defect data stream is fed back to the material property database to update the defect pattern matching rules. The dynamic process parameter correction is iteratively executed until the process convergence conditions are met, outputting a quality control report for the final molded thin material.

[0085] Example 1: See Figure 2 The establishment of the process condition assessment model begins with setting the initial number of topological layers and feature extraction thresholds. The initial number of topological layers is determined based on the complexity of the molding process, typically using a multi-layered convolutional neural network architecture to accommodate defect feature extraction at different scales. The feature extraction threshold is used to control the model's sensitivity to input data and avoid misjudgments caused by noise interference. After the initial parameters are set, multi-dimensional process parameters are input into the convolutional neural network for processing. These parameters include temperature gradient distribution data, pressure load variation data, and resin flow rate data, which are collected in real time through a sensor network and transmitted to the data processing unit.

[0086] A convolutional neural network extracts features layer by layer from the input multi-dimensional process parameters, generating primary defect feature vectors. These vectors contain key information such as the spatial distribution of abnormal temperature regions, the temporal variation of pressure fluctuations, and the non-uniformity of resin flow. These feature vectors reflect the types and severity of defects that may exist in the current compression molding process. Subsequently, the primary defect feature vectors are compared with a pre-stored standard feature library, which stores the range of normal process parameters and typical defect patterns accumulated during historical production.

[0087] The difference calculation is achieved by dividing the primary defect feature vector into N sets of feature sub-vectors. Each set of feature sub-vectors corresponds to a process parameter dimension, such as temperature, pressure, and resin flow. N sets of standard feature sub-vectors from a pre-stored standard feature library are extracted and used for similarity analysis with the current feature sub-vector. The similarity calculation comprehensively considers spatial distribution feature values ​​and temporal variation feature values, generating a composite feature value through normalized weighted fusion. The Euclidean distance between the composite feature value and the standard feature sub-vectors is used to quantify local similarity, reflecting the degree of deviation between the current process state and the standard state.

[0088] Local similarity scores are multiplied by a preset weight matrix to generate multidimensional similarity. The weight matrix is ​​dynamically adjusted based on the impact of each process parameter on the final product quality; for example, temperature fluctuations may have a significant impact on the resin curing process, thus receiving a higher weight. N sets of multidimensional similarity scores are fused to generate a comprehensive similarity index, used to assess the overall deviation of the current process state. The comprehensive similarity index is compared with a preset critical value to generate a model optimization coefficient difference value. This difference value reflects the degree of matching between the current model parameters and actual process requirements, guiding subsequent model optimization.

[0089] Model optimization employs an adaptive genetic algorithm for iterative adjustments. The initial population of the genetic algorithm consists of multiple combinations of different topological layer numbers and feature extraction thresholds, with each set of parameters corresponding to a candidate model. The performance of each candidate model in the defect feature recognition task is evaluated by calculating its fitness function. The fitness function comprehensively considers both similarity metrics and model computational efficiency, ensuring a balance between accuracy and real-time performance in the optimized model. The genetic algorithm progressively optimizes the model parameters through selection, crossover, and mutation operations, ultimately obtaining the target topological layer number and target feature extraction threshold.

[0090] The target topology layer number and target feature extraction threshold are used to reconstruct the convolutional neural network structure. The reconstructed network architecture can extract key defect features more accurately while reducing redundant computation. The optimized convolutional neural network outputs an optimized defect feature vector, which serves as the initial defect feature map. The initial defect feature map is presented in a visual form, showing the defect distribution in the current molding process, including areas of abnormal temperature, uneven pressure, and resin flow hindrance points.

[0091] The process status assessment model was trained using offline validation with historical production data. This historical data covered molding processes under different conditions, including normal operating conditions and typical defect conditions. Cross-validation was used to evaluate the model's generalization ability, ensuring it could accurately identify defect characteristics even with new process parameters. After training, the model was deployed to the real-time control system, working in conjunction with the distributed sensor network and the molding equipment actuators to achieve dynamic process optimization.

[0092] The model's performance is continuously monitored during operation. When process conditions change or new defect modes emerge, model parameters can be fine-tuned through an online learning mechanism. Online learning utilizes real-time acquired process data and defect detection results to dynamically update the standard feature library and weight matrix, enabling the model to adapt to changes in the production environment. The output of the process condition assessment model is not only used for defect identification but also provides feedback information for updating the material property database, forming a closed-loop optimization system.

[0093] The entire process assessment emphasizes the fusion and analysis of multi-dimensional data. The synergistic processing of temperature, pressure, and resin flow parameters can reveal the formation mechanisms of composite defects; for example, high-temperature regions may cause premature resin curing, leading to uneven flow. By comprehensively analyzing the interactions between these parameters, the model can more comprehensively assess the process status, providing precise guidance for subsequent dynamic compensation. The establishment and optimization of the process status assessment model is a key step in the intelligentization of compression molding processes, and its output directly affects the quality stability and production efficiency of the final product.

[0094] Example 2: See Figure 3 In the optimization of the compression molding process for carbon fiber reinforced composite thin films, multidimensional similarity calculation and separation of primary and secondary defect data streams are key aspects of process status assessment. This implementation method achieves accurate identification and classification management of various defects during the compression molding process by establishing a refined feature analysis mechanism and defect classification strategy.

[0095] Multidimensional similarity calculation begins with the extraction of spatial distribution feature values ​​and temporal variation feature values ​​from the feature vectors. Spatial distribution feature values ​​reflect the distribution of parameters such as temperature and pressure in different areas of the molding die; gridding transforms the discrete data collected by sensors into a continuous spatial distribution map. Temporal variation feature values ​​record the dynamic characteristics of each process parameter fluctuating over time; a sliding window method is used to extract the parameter change trends over different time periods. After normalization, the two types of feature values ​​are fused according to a preset weight ratio to form a composite feature value that can simultaneously characterize spatial non-uniformity and temporal instability.

[0096] When comparing the composite feature values ​​with reference features in the standard feature library, a distance-based analysis method is used. This method considers not only the absolute differences in feature values ​​but also the consistency of feature change trends. When the composite feature values ​​are close to the standard features in both spatial distribution and temporal evolution, the process status of the region is determined to be normal; otherwise, it is marked as a potential defect region. The similarity calculation results are weighted and summarized using a preset weight matrix, which is dynamically adjusted according to material properties and process requirements to ensure that changes in key process parameters are fully considered.

[0097] The separation of primary and secondary defect data streams is based on the defect feature correlation matrix in the material property database. This matrix is ​​trained using historical production data and records the correlation between different types of defects and their impact on the final product quality. After the initial defect feature map is mapped to this matrix, the system automatically calculates the global influence weight value of each defect feature. The calculation of the influence weight value comprehensively considers the physical characteristics of the defect, its frequency of occurrence, and its coupling effect with other defects, and uses a multi-factor weighted algorithm to derive a quantitative defect severity index.

[0098] The identification of the core area of ​​the primary defect employs a dynamic threshold determination method. The system automatically adjusts the primary defect threshold based on the quality requirements of the current production batch. When the defect impact weight value of a certain area exceeds this threshold, it is identified as the core area of ​​the primary defect. The primary defect data stream not only includes the original process parameters of the core area but also extracts extended features from its surrounding related areas to fully reflect the formation mechanism and development trend of the primary defect. The secondary defect data stream contains feature data that have not yet met the primary defect criteria but still require monitoring. This data is labeled and input into the defect feature correlation matrix to update the secondary defect impact factor.

[0099] The updating of the defect feature correlation matrix is ​​a continuous optimization process. Potential defect patterns contained in the secondary defect data stream are automatically analyzed by the system. When new feature combinations are detected or the evolution trend of existing features changes, the correlation weights and influencing factors in the matrix are adjusted accordingly. This dynamic updating mechanism enables the system to adapt to fluctuations in material properties and changes in process conditions, maintaining the accuracy of defect classification. The criteria for separating primary and secondary defects are not fixed but are continuously refined as the production process progresses, forming a progressive defect management system.

[0100] In practical applications, this implementation demonstrates adaptability to complex process conditions. When changes in raw material batches or environmental conditions cause baseline drift in process parameters, the system can quickly establish new quality control benchmarks through recalculation of feature similarity and automatic adjustment of defect classification criteria. This adaptive characteristic is particularly important for maintaining long-term production stability, enabling the process optimization system to evolve alongside the company's actual production environment, forming a virtuous cycle of continuous improvement.

[0101] The effectiveness of multidimensional similarity calculation and the separation of primary and secondary defect data streams is reflected in multiple aspects of the process optimization. By accurately identifying key defect areas, the system can adjust molding parameters in a targeted manner, avoiding resource waste caused by global process adjustments. Continuous monitoring of secondary defects provides data support for preventative maintenance, helping to identify potential equipment problems or material anomalies. This hierarchical and refined defect management model provides a reliable technical guarantee for the high-quality molding of carbon fiber composite materials.

[0102] The technological innovation of this implementation lies in the establishment of a dynamically adjustable defect assessment system. Unlike traditional fixed-threshold alarm mechanisms, this method achieves quantitative assessment and intelligent classification of defect severity through feature correlation matrices and influence weight calculations. The system can not only identify current process problems but also predict potential defect development trends, providing a basis for proactive process adjustments. This predictive analysis capability shifts the control of compression molding processes from passive response to proactive optimization, significantly improving the level of intelligence in the production process.

[0103] In terms of system integration, this implementation method achieves deep integration with the control system of the molding equipment. Defect classification results are directly converted into equipment control commands, forming a closed-loop control link from detection to execution. The human-machine interface intuitively displays the distribution and changing trends of primary and secondary defects, helping operators understand the basis of system decisions and to intervene manually when necessary. This human-machine collaborative working mode leverages the efficiency of automation while retaining the judgment ability of human experts, making the entire optimization system more robust and reliable.

[0104] Example 3: See Figure 4 In the optimization of the carbon fiber reinforced composite thin-film compression molding process, the construction of a real-time compensation instruction set based on the main defect data stream and the dynamic correction of process parameters are the core links to achieve closed-loop process control. This implementation method, by establishing a multi-parameter collaborative compensation mechanism and an equipment response feedback system, transforms defect identification results into precise process adjustment actions, forming a complete control link from detection to execution.

[0105] The main defect data stream analysis begins with the extraction of temperature deviation parameters, pressure fluctuation parameters, and resin flow rate anomaly parameters. Temperature deviation parameters reflect the difference between the actual temperature and the set value in each zone of the mold. These are collected through a distributed temperature sensor array, and the data is preprocessed to generate a three-dimensional temperature field distribution map. Pressure fluctuation parameters include real-time pressure readings of each actuator in the hydraulic system, as well as the pressure gradient over time. Resin flow rate anomaly parameters are obtained through joint measurement using a flow meter and viscosity sensor, while also considering changes in the position of the resin flow front within the mold cavity. After normalization, these parameters are input into the compensation algorithm to generate corresponding adjustment commands.

[0106] The temperature compensation gradient function is constructed using a piecewise compensation strategy based on regional differences. For the core region with large temperature deviations, a steeper adjustment curve is used; for the peripheral regions, a gentler compensation method is employed. This function can be expressed as:

[0107]

[0108] Where, ΔT c This represents the temperature compensation value, κ is the regional sensitivity coefficient, and T errLet σ be the temperature deviation and σ be the smoothing factor. This function form ensures the non-linear growth characteristics of the compensation amount, avoiding sudden changes in command near the critical point.

[0109] The generation of the pressure compensation step sequence takes into account the spatiotemporal characteristics of pressure fluctuations. For periodic pressure fluctuations, the system identifies their frequency characteristics and generates a phase-matched compensation waveform; for sudden pressure anomalies, a gradual compensation strategy is adopted. The compensation step size is automatically graded according to the fluctuation amplitude, with large fluctuations corresponding to large step sizes for rapid correction and small fluctuations using fine-tuning. The response characteristics of the hydraulic actuator are pre-modeled, and the compensation command includes a pre-compensation amount for the equipment response delay.

[0110] The resin flow compensation ratio calculation incorporates a viscosity-temperature coupling correction factor. The system monitors resin viscosity changes in real time, and automatically adjusts the flow compensation ratio when temperature compensation alters the resin's rheological properties. The control parameters of the dispensing valve are generated using a fuzzy logic algorithm, comprehensively considering multiple factors such as current flow rate, cavity filling rate, and viscosity change trends. The valve opening adjustment employs a gradual approximation method to avoid instability in the flow front caused by sudden flow changes.

[0111] The synthesis of multi-parameter collaborative compensation instruction sets is based on an understanding of the interactive effects of process parameters. Complex coupling relationships exist between temperature, pressure, and resin flow parameters; the instruction set generation algorithm quantifies these interactive effects using a decoupling matrix. Compensation instructions are ordered according to execution priority, with adjustments to critical parameters sent first, followed by minor parameter fine-tuning. The instruction set transmission employs a timestamp synchronization mechanism to ensure that all actuators receive coordinated and consistent control commands.

[0112] The instruction parsing process of the molding equipment's main controller includes a security verification step. Each compensation instruction is accompanied by a validity verification code to prevent malfunctions caused by transmission errors. Temperature compensation instructions are converted into power adjustment signals for each zone of the heating plate, and the zone temperature control strategy adopts a feedforward-feedback composite control mode. Pressure compensation instructions drive the hydraulic system actuators to adjust the pressure parameters according to a predetermined curve, and the system monitors the deviation between the actual pressure trajectory and the theoretical curve in real time. Resin flow compensation is converted into opening control signals for the injection valves, and the valve response characteristics are pre-calibrated to ensure the accuracy of flow regulation.

[0113] The acquisition and processing of equipment response delay data forms a crucial feedback loop. The system records the time delay characteristics from command issuance to actual parameter changes, establishing dynamic response models for each actuator. Delay data is categorized and stored, including communication delay, mechanical response delay, and sensor detection delay. This data is used to optimize the timing of subsequent compensation command transmission, achieving predictive pre-compensation. For periodic fluctuation patterns, the system learns its phase characteristics and automatically sends compensation commands in advance, synchronizing actual process parameters with target values.

[0114] The interaction of parameters during the dynamic compensation process is continuously monitored. When temperature adjustments affect resin flowability, the system automatically triggers a viscosity compensation process; when pressure changes cause micro-deformation of the mold, the temperature field distribution is adjusted accordingly. This multi-parameter collaborative adaptation mechanism enables the system to maintain the stability of the process window and avoids the loss of control of other parameters due to the adjustment of a single parameter. The execution effect of all compensation actions is verified in real time through a sensor network, forming a closed-loop quality control.

[0115] The adaptive capability of the compensation system is reflected in its ability to track changes in equipment performance. As the molding equipment is used for an extended period, the response characteristics of the actuators may drift. By analyzing historical response data, the system automatically updates the equipment model parameters to maintain the accuracy of compensation commands. For sudden declines in equipment performance, such as aging heating elements or hydraulic system leaks, the system can identify abnormal response patterns and trigger maintenance alarms.

[0116] The human-machine interface provides operators with visual monitoring of the compensation process. Real-time displays of compensation command execution status include: a temperature compensation progress bar, a pressure adjustment trajectory graph, and a marker indicating the resin flow front position. Operators can view historical adjustment records for each parameter and intervene manually when necessary. The system records the basis and results of all compensation decisions, creating a traceable process adjustment log.

[0117] The optimized generation of compensation commands relies on a continuously running simulation engine. Before sending actual compensation commands, the system performs virtual execution on a digital twin model to predict potential effects. The simulation results are used to correct compensation parameters, avoiding over-adjustment in actual production. The digital twin model is continuously updated as production data accumulates, gradually improving prediction accuracy.

[0118] Safety protection mechanisms are implemented throughout the entire dynamic compensation process. Each compensation command has an allowable adjustment range limit to prevent accidental overshoot. The system monitors the equipment's operating status in real time, and immediately suspends the compensation process and activates safety protocols when abnormal vibration, overheating, or sudden pressure changes are detected. All compensation actions follow a gradual principle to avoid shocks to equipment and materials caused by abrupt changes.

[0119] The scalable design of the compensation system supports the gradual addition of new features. When new sensors or actuators are added, the system architecture allows for the seamless integration of new compensation dimensions. The modular design of the control algorithm facilitates the development of dedicated compensation strategies for specific problems, which can be flexibly combined and applied. This open architecture enables the system to continuously evolve alongside the development of production processes.

[0120] The quality traceability system records compensation data for each production batch throughout the entire process. This data is correlated with the final product's quality inspection results to continuously optimize compensation strategies. When a correlation is found between a specific compensation pattern and the yield rate, the system automatically adjusts the importance weights of relevant parameters. This evidence-based continuous improvement mechanism leads to a continuous improvement in process control.

[0121] Example 4: In the optimization process of carbon fiber reinforced composite thin-film compression molding, the feedback of corrected secondary defect data streams to the material property database and the updating of defect pattern matching rules are key steps in the continuous optimization of the system. This implementation method establishes a dynamically evolving defect feature knowledge base, enabling the system to adapt to changes in material properties and the emergence of new defect patterns, and gradually improve the process control strategy.

[0122] The extraction of defect feature evolution trend data from the sub-defect data stream employs time series analysis. The system identifies recurring feature change patterns from historical sub-defect records; for example, a certain temperature distribution anomaly is often accompanied by a specific form of resin flow stagnation. These correlations are quantified as feature correlation coefficients and stored in a correlation mapping table in the material property database. The correlation mapping table uses a multi-dimensional data structure to record the complex interactions between defect features and process parameters. When a new feature combination appears in the sub-defect data stream, the system automatically creates a new record entry and assigns an initial influence coefficient to that feature combination.

[0123] Taking the molding production of a batch of carbon fiber prepreg as an example, the system detected a new sub-defect pattern: periodic temperature fluctuations occurred in the edge area of ​​the mold, with an amplitude within ±3℃ and a frequency of approximately once every 15 minutes. This pattern was not recorded in the previous defect feature library, so the system created a new entry for it in the correlation mapping table and began tracking the development trend of the defect. As production continued, the system discovered a time correlation between this temperature fluctuation and the pressure regulation of the hydraulic system, and established a correlation record between the two in the correlation mapping table.

[0124] Table 1 shows the contents of the updated defect feature association mapping table during the production process of this batch.

[0125]

[0126]

[0127] The impact coefficient threshold correction for existing defect feature patterns adopts a gradual adjustment strategy. The system analyzes the frequency, duration, and subsequent development of similar defects in historical secondary defect data streams, and dynamically updates their weight values ​​in the association mapping table. For example, a certain resin flow unevenness defect was initially classified as a minor problem, but with the accumulation of multiple batches of production data, the system found that this defect often develops into a product thickness unevenness problem in the later stages of molding. Therefore, it automatically increases its impact coefficient threshold in the association mapping table, making it more likely to be identified as a primary defect.

[0128] The analysis of defect feature evolution trend data employs a multi-timescale observation method. The system simultaneously tracks short-term fluctuations (minute-level), medium-term trends (hour-level), and long-term evolution (batch-level) patterns of defect features. This multi-scale analysis helps distinguish between random fluctuations and systematic changes, avoiding overreactions to temporary anomalies. For example, occasional noise from the mold temperature sensor is identified as a short-term fluctuation and the correlation mapping table is not updated, while persistent regional temperature anomalies are recorded as valid defect patterns.

[0129] The identification of new defect feature patterns employs a similarity-based clustering algorithm. The system compares anomalous features in the sub-defect data stream with known patterns across multiple dimensions. When the similarity is below a preset threshold, it is identified as a new pattern. The initial parameter settings for new patterns are based on empirical values ​​for similar materials and are gradually adjusted during subsequent observations. For example, after a new type of carbon fiber fabric was introduced into production, the system detected unique flow front instability during its resin impregnation process. This pattern was identified as a new defect type and assigned a unique code.

[0130] Version management of the association mapping table adopts an incremental update mechanism. Each modification generates a new version record, preserving a complete change history. This design facilitates the retrospective analysis of the development process of specific defect patterns and provides a reference for process debugging. When significant changes occur in production conditions, such as changing raw material suppliers or modifying mold structures, the system can quickly roll back to a specific version of the association mapping table, maintaining the continuity of process control.

[0131] The feedback processing of secondary defect data streams employs a priority queuing mechanism. Based on the potential impact and urgency of the defect characteristics, the system categorizes feedback data into multiple priority levels. High-priority data immediately triggers an update to the association mapping table, medium-priority data enters a batch processing queue, and low-priority data is cached before analysis. This hierarchical processing strategy optimizes system resource allocation and ensures timely responses to critical defect information.

[0132] The update process for the material property database includes multiple verification steps. Before each modification to the association mapping table, the system checks the consistency between the new data and existing knowledge to avoid contradictions or redundant records. For major modifications, such as adjustments to the impact coefficient threshold, simulation verification is required before the changes take effect. After the database is updated, the system automatically triggers parameter readjustment in the relevant process control modules to ensure that defect identification standards and control strategies remain synchronized.

[0133] The optimization of defect pattern matching rules employs a feedback mechanism based on actual quality results. The system correlates the development of secondary defects with the quality inspection data of the final product to evaluate the effectiveness of existing matching rules. When a secondary defect, though not identified as a primary defect, frequently causes quality problems, the system adjusts its matching priority accordingly. This quality-oriented optimization makes defect classification more aligned with actual production needs.

[0134] The analysis of historical secondary defect data provides directional guidance for process improvement. The system regularly analyzes the frequency and trends of various secondary defects to identify potential material-process mismatches. For example, the occurrence of the same type of edge area defect in multiple consecutive batches may indicate that the mold heating system needs to be redesigned. These analytical results form process improvement recommendations for engineers to reference.

[0135] The human-machine collaborative interface facilitates engineers' participation in defect knowledge management. Key modifications to the association mapping table generate visual reports, highlighting changes and their potential impact. Engineers can view the complete evolution of defect characteristics and make manual corrections when necessary. The system also marks automatically updated records with questionable entries, requesting manual confirmation. This human-machine collaborative knowledge update mechanism balances automation efficiency with the value of human experience.

[0136] The storage of sub-defect data streams employs a combination of structured and unstructured methods. Numerical process parameters are stored in a time-series database for easy retrieval and analysis; unstructured data such as defect feature images and waveforms are managed using a dedicated file system. This hybrid storage scheme satisfies the need for efficient processing while preserving complete defect feature information. All data is accurately timestamped and marked with production batch identifiers, supporting multi-dimensional traceability analysis.

[0137] The update and propagation of defect pattern matching rules employs a distributed notification mechanism. When significant modifications occur to the association mapping table, all relevant process control modules receive real-time notifications and automatically synchronize the latest rules. This instant update mechanism ensures that the entire system always makes decisions based on a unified understanding of defects, avoiding control conflicts caused by information asynchrony. Data consistency between modules is guaranteed through a version verification mechanism.

[0138] In practical applications, this implementation demonstrates its adaptability to complex production environments. Faced with batch-to-batch variations in raw materials and fluctuations in equipment status, the system maintains the accuracy of process control through continuous knowledge updates. When unprecedented anomalies occur, the system's learning mechanism can quickly establish new defect perceptions, shortening the process debugging cycle. This adaptability is of significant value in maintaining production stability and product quality consistency.

[0139] In terms of system resource management, this implementation adopts a flexible computing architecture to cope with fluctuating data processing demands. During peak periods of secondary defect data flow, computing resources are automatically expanded to ensure timely processing; during relatively idle periods, deep data mining and knowledge extraction are performed. This intelligent resource allocation strategy optimizes operating costs while ensuring system responsiveness.

[0140] Data security measures are implemented throughout the entire knowledge update process. All secondary defect data is transmitted and stored in encrypted form, and modifications to the association mapping table require authorization verification. The system maintains a complete operation log, recording the content, time, and executor of each data change. These security designs protect the company's core process knowledge and meet the stringent data security requirements of modern manufacturing.

[0141] Example 5: In the optimization process of carbon fiber reinforced composite thin-film compression molding, the determination of process convergence conditions and the iterative execution mechanism constitute the closed-loop terminal of the entire quality control system. This implementation method achieves adaptive optimization control of the compression molding process by establishing a multi-dimensional and multi-level process stability evaluation system, ensuring that the final product meets the predetermined quality standards.

[0142] Process convergence monitoring begins with the analysis of the characteristic intensity decay rate of the main defect data stream. The system tracks the changing trends of the characteristic parameters of the identified main defects in real time and calculates the decay curve of its intensity with the number of process adjustments. The characteristic intensity decay rate reflects the effectiveness of process correction measures. By comparing the current decay rate with the preset convergence threshold, it is determined whether the process state is stabilizing. When the characteristic intensity decay rate is below the convergence threshold for M consecutive iterations, it indicates that the main process problem has been effectively controlled, and the system enters the secondary defect stability verification stage.

[0143] The allowable fluctuation range of the secondary defect data stream is dynamically set based on material properties and product requirements. The system analyzes the statistical distribution characteristics of secondary defect parameters and establishes a fluctuation range assessment method based on a probability model. Short-term fluctuations and long-term trends of secondary defect parameters are treated differently to avoid misjudging normal process fluctuations as systemic problems. When the secondary defect data stream remains within the allowable range for multiple consecutive process cycles, the system determines that the process has reached a stable state.

[0144] The monitoring of the frequency of actuator correction actions in the molding equipment employs a time window statistical method. The system records the number of adjustments made by each actuator per unit time, including temperature control commands, pressure regulation signals, and resin flow correction operations. The equipment baseline maintenance frequency is customized based on the equipment model and usage history, taking into account practical factors such as mechanical wear and component aging. When the actual correction frequency consistently falls below the baseline value, it indicates that the process parameters have entered a stable maintenance phase, and the system prepares to trigger the process termination procedure.

[0145] Thickness uniformity testing data is acquired using high-precision laser scanning technology. Multi-point thickness measurements of the formed thin material create a two-dimensional distribution map, and the system calculates the overall thickness variance and maximum local deviation. Fiber orientation distribution data is obtained through microscopic image analysis, quantitatively assessing the consistency of fiber arrangement within the matrix. These final quality parameters are correlated with adjustment records during the process to generate a complete quality control report, documenting product compliance and process optimization progress.

[0146] The iterative execution process is initiated based on non-converged defect features. When the attenuation rate of the main defect feature intensity fails to meet the convergence criterion, the system re-inputs the current defect data stream into the process state evaluation model. The model generates a new generation of initial defect feature maps based on the latest data, reflecting the changes in process state after previous adjustments. Comparative analysis between the new generation map and the original map reveals the actual effect of the process adjustments and guides subsequent optimization directions.

[0147] An updated version of the material property database was applied to pattern matching in the next-generation defect map. The system employs a progressive matching strategy, prioritizing the matching of verified defect patterns before processing newly emerging feature combinations. During defect pattern re-matching, historical correction records are used as a reference to avoid repeating ineffective adjustment schemes. The newly separated primary and secondary defect data streams reflect the latest characteristics of the current process state, providing a basis for generating enhanced compensation instructions.

[0148] The enhanced real-time compensation instruction set was constructed taking into account the accumulated experience from previous adjustments. The system analyzes the effectiveness data of historical compensation measures to optimize the combination of instruction parameters and adjustment ranges. The temperature compensation strategy incorporates a regional coordination factor to ensure that adjustments in adjacent temperature zones are coordinated; the pressure compensation scheme adds a dynamic buffer mechanism to prevent oscillations caused by over-adjustment; and the resin flow control employs predictive regulation to compensate for impending flow changes in advance. These enhancements aim to improve correction efficiency and accelerate the process convergence.

[0149] The triggering of process termination commands follows a tiered verification principle. After detecting that the convergence conditions are met, the system initiates a multi-level verification process: first, it verifies the stability of key process parameters; second, it checks the idle status of equipment actuators; and finally, it pre-checks the compliance of product quality parameters. Once all verifications are successful, the system systematically shuts down each control loop, generates a final process report, and completes the entire optimization cycle. Failure to pass verification will trigger a re-test within a specified range or initiate a manual intervention process.

[0150] The quality control report is organized in a structured manner. The main body of the report comprises three levels: a summary of the process adjustment process, trends in key parameter changes, and final quality inspection results. The adjustment process is presented chronologically, showing the main corrective measures and their effects; the parameter trend graphs visually display the convergence process of core variables such as temperature and pressure; and the quality inspection section details the compliance status of indicators such as thickness distribution and fiber orientation. The report is automatically archived after generation and stored in conjunction with production batch information.

[0151] The dynamic convergence threshold setting reflects the intelligent nature of process control. The system automatically adjusts the convergence criteria based on factors such as batch material variations and environmental fluctuations, achieving a balance between stringent quality requirements and actual production conditions. A relatively lenient threshold is used in the initial iteration phase to accelerate the resolution of key issues, while the criteria are gradually tightened in later stages for fine-tuning. This adaptive threshold management makes the process optimization process more efficient and rational.

[0152] Equipment status monitoring and process convergence determination are integrated. The system not only focuses on the numerical convergence of process parameters but also monitors the actual response of actuators. When a decline in equipment performance is detected that may affect control accuracy, the convergence determination strategy is automatically adjusted or a maintenance alarm is triggered. This equipment-process collaborative evaluation mechanism improves the reliability of the entire control system and avoids misjudgments caused by hardware problems.

[0153] The assessment of secondary defect stability employs a multi-index comprehensive analysis method. The system simultaneously examines the fluctuation amplitude, rate of change, and pattern consistency of secondary defect parameters, and determines their actual impact through weighted scoring. Short-term abnormal fluctuations in secondary parameters do not immediately negate process convergence, but persistent minor deviations are still recorded and marked for reference in subsequent batches. This comprehensive assessment avoids extreme judgments that are either overly sensitive or overly lenient.

[0154] Data management during iterative execution employs an incremental storage strategy. Key data generated in each iteration is selectively retained, forming a concise yet complete record of the optimization process. The system automatically identifies and stores representative snapshots of process states, including typical defect characteristics, effective corrective measures, and inflection point parameters. This intelligent data reduction technology optimizes storage efficiency without losing important information.

[0155] The knowledge extraction process following process convergence enriches the system's experience base. Successful optimization cases are abstracted into typical process adjustment patterns and stored in the material property database. These patterns contain information such as key feature identification points, effective countermeasures, and expected improvement effects, providing a reference for subsequent production. As cases accumulate, the system gradually establishes a set of optimization strategies for different materials and molds, enhancing its adaptability to new production tasks.

[0156] The human-machine interface provides decision support information during the process convergence phase. The system visually displays the current process status relative to the convergence criteria, helping operators understand the basis for automatic judgments. In boundary cases, it provides an entry point for manual confirmation or intervention, retaining final control. All automatic judgment results are accompanied by detailed explanatory information, including reference data, calculation logic, and confidence level assessments, ensuring process transparency and reliability.

[0157] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0158] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An optimized method for the compression molding process of carbon fiber reinforced composite thin materials, characterized in that, include: Multi-dimensional process parameters during the compression molding process are collected in real time through a distributed sensor network. These process parameters include temperature gradient distribution data, pressure load change data, and resin flow rate data. A process status assessment model is established to identify initial defect features of the multi-dimensional process parameters and generate an initial defect feature map. Based on the material property database, defect pattern matching is performed on the initial defect feature map to separate the main defect data stream and the secondary defect data stream. A real-time compensation instruction set is constructed based on the main defect data stream to synchronously drive the actuator of the molding equipment to perform dynamic process parameter correction. The corrected sub-defect data stream is fed back to the material property database to update the defect pattern matching rules; Iteratively execute dynamic process parameter corrections until the process convergence conditions are met, and output the quality control report of the final formed thin material. The separation of the primary defect data stream and the secondary defect data stream includes: Establish a defect feature correlation matrix in the material property database; The initial defect feature map is mapped to the defect feature correlation matrix to generate defect impact weight values; Feature regions whose defect impact weight value exceeds the preset main defect threshold are identified as the core regions of the main defect. Extract the data stream corresponding to the core region of the main defect to form the main defect data stream; Data streams of feature regions whose defect impact weight values ​​are lower than the primary defect threshold are marked as secondary defect data streams. The secondary defect data stream is input into the defect feature correlation matrix to update the secondary defect impact factor; The construction of the real-time compensation instruction set based on the main defect data stream includes: Analyze the temperature deviation parameters, pressure fluctuation parameters, and resin flow rate anomaly parameters in the main defect data stream; Generate a temperature compensation gradient function based on the temperature deviation parameters; A pressure compensation step size sequence is constructed based on pressure fluctuation parameters; The resin flow rate compensation ratio is calculated by driving abnormal resin flow rate parameters. A multi-parameter collaborative compensation instruction set is generated by integrating the temperature compensation gradient function, the pressure compensation step size sequence, and the resin flow rate compensation ratio. Transmit the multi-parameter collaborative compensation instruction set to the main controller of the molding equipment; The step of feeding back the corrected sub-defect data stream to the material property database to update the defect pattern matching rules includes: Extract defect feature evolution trend data from the secondary defect data stream; Establish a mapping table between defect feature evolution trend data and process parameters in the material property database; When a new defect feature pattern appears in the defect data stream, a new defect feature pattern record is created in the association mapping table; The influence coefficient threshold of existing defect feature patterns in the association mapping table is corrected based on the historical secondary defect data stream. The updated association mapping table will be used as the new version of the defect pattern matching rules.

2. The optimized molding process method for carbon fiber reinforced composite thin materials according to claim 1, characterized in that, The establishment of the process status assessment model includes: Set the initial number of topological layers in the model and the feature extraction threshold; The multi-dimensional process parameters are input into a convolutional neural network to generate a primary defect feature vector. The model optimization coefficients are calculated based on the difference between the primary defect feature vector and the pre-stored standard feature library. An adaptive genetic algorithm is used to iteratively optimize the initial number of topological layers and the feature extraction threshold of the model to obtain the target number of topological layers and the target feature extraction threshold. The convolutional neural network structure is reconstructed by the target topology layer number and the target feature extraction threshold, and the optimized defect feature vector is output as the initial defect feature map.

3. The optimized molding process method for carbon fiber reinforced composite thin materials according to claim 2, characterized in that, The calculation of model optimization coefficients based on the difference between the primary defect feature vector and the pre-stored standard feature library includes: The primary defect feature vector is divided into N groups of feature sub-vectors; Extract the corresponding N sets of standard feature sub-vectors from the pre-stored standard feature library; Calculate the multidimensional similarity between each set of feature vectors and the standard feature vectors; A comprehensive similarity index is generated by fusing N sets of multidimensional similarities. The comprehensive similarity index is compared with the preset critical value to generate the difference value of the model optimization coefficient; The convolutional neural network weight parameters are updated in reverse based on the difference in the model optimization coefficients.

4. The optimized molding process method for carbon fiber reinforced composite thin films according to claim 3, characterized in that, The calculation of the multidimensional similarity between each set of feature vectors and the standard feature vectors includes: Extract the spatial distribution feature values ​​and temporal variation feature values ​​of the feature sub-vectors respectively; The spatial distribution feature value and the temporal variation feature value are normalized and weighted and fused to generate a composite feature value; Local similarity is generated based on the Euclidean distance between composite feature values ​​and standard feature vectors; Multidimensional similarity is generated by multiplying local similarity with a preset weight matrix.

5. The optimized molding process method for carbon fiber reinforced composite thin materials according to claim 4, characterized in that, The dynamic process parameter correction includes: The main controller of the molding equipment analyzes the multi-parameter collaborative compensation instruction set; Adjust the heating plate zone temperature control strategy according to the temperature compensation gradient function; The pressure loading curve of the hydraulic actuator is reconstructed based on the pressure compensation step sequence; the resin flow compensation ratio is used to correct the opening control parameters of the glue injection valve. Synchronously record device response delay data during the correction process; The device response delay data is fed back to the real-time compensation instruction set for response delay compensation.

6. The optimized molding process method for carbon fiber reinforced composite thin materials according to claim 1, characterized in that, The process convergence conditions include: Real-time monitoring of the characteristic intensity decay rate of the main defect data stream; When the feature intensity decay rate is lower than the preset convergence threshold for M consecutive iterations, the sub-defect data stream is determined to be stable within the allowable fluctuation range. Synchronously detect the frequency of correction actions of the actuator in the molding equipment; A process termination command is triggered when the frequency of corrective actions drops below the equipment's baseline maintenance frequency. Output the thickness uniformity test data and fiber orientation distribution data of the final shaped thin material as a quality control report.

7. The optimized molding process method for carbon fiber reinforced composite thin films according to claim 6, characterized in that, The iterative execution process includes: When the feature intensity decay rate does not reach the convergence threshold, the current defect data stream is input into the process status assessment model to generate a new generation of initial defect feature map. Defect patterns are rematched based on the updated material property database to the new generation of initial defect feature maps; The primary and secondary defect data streams are re-separated and an enhanced real-time compensation instruction set is generated; Execute the enhanced real-time compensation instruction set until the process convergence condition is met.

Images