Carbon fiber prepreg paving real-time positioning calibration method and system

Abstract

The invention provides a carbon fiber prepreg paving real-time positioning calibration method and system. The method comprises the following steps: acquiring paving process parameters which influence the optical characteristics of the surface of a prepreg strip; collecting visual information of the edge of the prepreg strip, and extracting visual characteristics reflecting the optical characteristics of the surface of the prepreg strip from the visual information; presetting an association relationship between the paving process parameters and the visual features; receiving a visual measurement result indicating that the paving position of the prepreg strip has deviation; according to the paving process parameters, the visual features and the incidence relation, the source of paving position deviation is judged; and executing calibration or compensation operation according to the judgment result of the deviation source. According to the invention, invalid calibration or accumulative errors caused by misjudgment can be avoided.

Description

Technical Field

[0001] This application relates to the field of carbon fiber prepreg laying technology, and more specifically, to a real-time positioning calibration method and system for carbon fiber prepreg laying. Background Technology

[0002] Automated carbon fiber prepreg placement technology is crucial in high-end manufacturing fields such as aerospace, requiring the precise placement of prepreg strips onto complex mold surfaces. To achieve high precision, the equipment needs to monitor and calibrate the prepreg position in real time.

[0003] However, due to the microscopic inhomogeneity of the prepreg material itself, the transient physical changes caused by heating and compaction during the laying process, and the mechanical deformation caused by laying complex curved surfaces, the image quality acquired by the vision sensor is severely affected, leading to errors in edge feature extraction. These minute measurement errors accumulate over long periods and in multiple layers of laying operations, eventually potentially causing defects in composite material components that exceed design tolerances, resulting in significant economic losses and production delays. Existing systems struggle to distinguish between measurement drift and actual physical deviations, making accurate calibration or compensation impossible. Summary of the Invention

[0004] This application discloses a real-time positioning and calibration method and system for carbon fiber prepreg laying, which aims to solve the technical problem that the accumulation of visual measurement errors caused by the dynamic changes in the optical properties of the prepreg surface during the automated carbon fiber prepreg laying process, thereby affecting product quality and causing economic losses.

[0005] In a first aspect, this application discloses a real-time positioning and calibration method for carbon fiber prepreg laying, comprising the following steps: Obtain the laying process parameters that affect the optical properties of the prepreg strip surface; Visual information of the edge of the prepreg strip is collected, and visual features reflecting the optical properties of the prepreg strip surface are extracted from the visual information; The relationship between preset tiling process parameters and visual features; Receive visual measurement results indicating deviations in the placement of prepreg strips; Based on the tiling process parameters, visual characteristics, and correlations, determine the source of tiling position deviation; Based on the determination of the source of the deviation, perform calibration or compensation operations, specifically including: When measurement drift is detected, the visual measurement results are compensated or the visual information processing method is adjusted. When a physical deviation is determined to be real, a physical correction command is generated to adjust the position of the tiling head.

[0006] Secondly, this application also discloses a real-time positioning calibration system for carbon fiber prepreg laying, the system comprising: The parameter acquisition module is used to acquire laying process parameters, which affect the surface optical properties of the prepreg strip. The visual information acquisition module is used to acquire visual information of the edge of the prepreg strip and extract visual features reflecting the optical properties of the prepreg strip surface from the visual information. The association preset module is used to preset the association between tiling process parameters and visual features; The measurement result receiving module is used to receive visual measurement results indicating that there is a deviation in the laying position of the prepreg strip; The deviation source judgment module is used to determine the source of the tiling position deviation based on tiling process parameters, visual features, and correlation relationships. A calibration or compensation execution module is used to perform calibration or compensation operations based on the determination of the source of deviation. This calibration or compensation execution module specifically includes: The measurement drift processing unit is used to compensate for visual measurement results or adjust the visual information processing method when measurement drift is detected. The physical correction instruction generation unit is used to generate physical correction instructions to adjust the position of the tiling head when a real physical deviation is determined.

[0007] Compared with the prior art, this application has at least the following beneficial effects: This application can effectively distinguish whether the deviation in the tiling position is due to visual measurement drift or actual physical deviation, thereby avoiding invalid calibration or cumulative error caused by misjudgment, and solving the problem of calibration failure caused by the inability to distinguish the source of deviation in the prior art. Attached Figure Description

[0008] Figure 1 This is a flowchart illustrating a real-time positioning and calibration method for laying carbon fiber prepreg, as provided in this application.

[0009] Figure 2 This is a schematic diagram of a real-time positioning and calibration system for carbon fiber prepreg laying provided in this application. Detailed Implementation

[0010] The technical solutions in this application will now be clearly and completely described in conjunction with the accompanying drawings.

[0011] This application proposes a real-time positioning and calibration method for carbon fiber prepreg laying, such as... Figure 1 As shown, it includes the following steps: Obtain the laying process parameters that affect the optical properties of the prepreg strip surface; Visual information of the edge of the prepreg strip is collected, and visual features reflecting the optical properties of the prepreg strip surface are extracted from the visual information; The relationship between preset tiling process parameters and visual features; Receive visual measurement results indicating deviations in the placement of prepreg strips; Based on the tiling process parameters, visual characteristics, and correlations, determine the source of tiling position deviation; Based on the determination of the source of the deviation, perform calibration or compensation operations, specifically including: When measurement drift is detected, the visual measurement results are compensated or the visual information processing method is adjusted. When a physical deviation is determined to be real, a physical correction command is generated to adjust the position of the tiling head.

[0012] This application, through comprehensive analysis of tiling process parameters, visual characteristics, and the relationships between them, can effectively determine whether the source of tiling position deviation is measurement drift or actual physical deviation, and perform targeted calibration or compensation operations accordingly, thus avoiding product defects caused by error accumulation in traditional methods.

[0013] To better understand the real-time positioning and calibration method for carbon fiber prepreg laying proposed in this application, it is necessary to explain some key terms and implementation environments involved.

[0014] Prepreg strips refer to carbon fiber strips pre-impregnated with a resin matrix, and are the basic unit in composite material manufacturing. Their surface optical properties, such as reflectivity, texture, and color, are crucial for visual inspection. These properties are affected by various factors, including material batch, storage conditions, ambient humidity, and temperature.

[0015] Laying process parameters refer to various process variables that can be monitored and controlled during the laying of carbon fiber prepreg, such as laying speed, laying head temperature, compaction pressure, ambient temperature, ambient humidity, and the distance between the laying head and the die. These parameters directly or indirectly affect the surface optical properties of the prepreg strip. For example, excessively high laying head temperature may cause localized softening of the resin, altering the surface gloss; excessive compaction pressure may cause the fibers to be more densely packed, affecting reflective properties.

[0016] Visual information refers to image or video data of the edges of prepreg strips acquired through visual sensors (such as industrial cameras). This information includes visually identifiable features of the prepreg strips, such as their geometry, surface texture, and color distribution.

[0017] Visual features are data extracted from visual information that quantifies the optical properties of prepreg strip surfaces. Examples include edge contrast, edge gradient, texture mean, texture variance, local brightness, and color channel distribution. Changes in these features can reflect changes in the optical properties of the prepreg strip surface.

[0018] The correlation refers to a pre-established mathematical model or mapping rule between the laying process parameters and visual features. This correlation can be obtained through experimental data, physical modeling, or machine learning methods, and is used to describe the expected changes in the optical properties (manifested through visual features) of the prepreg strip surface under different laying process parameters.

[0019] Visual measurement results refer to the deviation data obtained by using a vision system to measure the prepreg strip laying position in real time. This result typically represents the offset between the actual position and the target position of the prepreg strip.

[0020] Measurement drift refers to the phenomenon where changes in the optical properties of the prepreg strip surface cause deviations in the recognition of the prepreg strip's edge by the vision sensor or image processing algorithm, resulting in a visual measurement result deviating from the actual physical position. This deviation is not due to a change in the actual physical position of the prepreg strip, but rather an error inherent in the measurement system itself.

[0021] Actual physical deviation refers to the physical offset between the actual placement position and the target position of the prepreg strip. This may be caused by inaccurate control of the laying head movement, mold positioning errors, or deformation of the prepreg strip itself.

[0022] The implementation environment of this application typically includes an automated prepreg tape laying system equipped with a high-precision laying head, vision sensors, a motion control system, a data processing unit, and a human-machine interface. The vision sensors are usually installed near the laying head to monitor the laying of the prepreg tape in real time.

[0023] The core of the real-time positioning calibration method for carbon fiber prepreg laying proposed in this application lies in accurately determining the source of the laying position deviation and adopting differentiated calibration or compensation strategies accordingly.

[0024] First, it is necessary to obtain the laying process parameters that affect the optical properties of the prepreg strip surface. These parameters can be manually input, such as the laying speed and temperature set by the operator based on experience; or they can be automatically acquired through sensors, such as temperature sensors monitoring the laying head temperature in real time and pressure sensors monitoring the compaction pressure in real time. For example, a parameter acquisition module can be set up, which integrates temperature sensors, pressure sensors, and speed sensors to acquire the laying head temperature, compaction pressure, and laying speed in real time.

[0025] Secondly, it is necessary to acquire visual information of the prepreg strip edges and extract visual features reflecting the optical properties of the prepreg strip surface from this information. Visual information acquisition can be accomplished using an industrial camera; for example, a high-resolution industrial camera can be installed in front of the layup head to capture images of the prepreg strip edges in real time. Visual feature extraction can be achieved using image processing algorithms; for example, edge detection algorithms (such as the Canny operator) can be used to extract the edges of the prepreg strip and calculate features such as edge contrast and gradient.

[0026] Furthermore, it is necessary to pre-define the correlation between the laying process parameters and visual features. This correlation can be established through offline experiments. For example, standard prepreg strips can be laid at different laying speeds and temperatures, and the changes in their visual features can be recorded. Then, a mapping model can be obtained through regression analysis or machine learning training. For instance, a database can be established to store the correspondence between different laying speeds, laying head temperatures, and the edge contrast and texture smoothness of the prepreg strips. Specifically, this could be as follows: Before the installation begins, the system pre-defines a series of feature association rules based on the material properties of the carbon fiber prepreg and the physical principles of the installation process. These rules are not derived through data learning, but rather based on an understanding of materials science, optical physics, and thermodynamics principles, and are deterministic relationships established through prior experimental verification. For example: The relationship between heating temperature and optical properties: As the heating temperature increases, the epoxy resin on the surface of the prepreg gradually softens, its viscosity decreases, and its surface tension changes, resulting in a smoother surface, increased specular reflection, and decreased diffuse reflection. In images, this manifests as an increase in the local grayscale mean in edge areas, enhanced edge contrast, and blurred texture details. The specific correlation rule can be expressed as: ΔG = f1(T, T0) ΔC = f2(T, T0) Where ΔG represents the change in the mean gray level, ΔC represents the change in edge contrast, T is the current heating temperature, T0 is the reference temperature, and f1 and f2 are nonlinear functions determined based on physical experiments.

[0027] The relationship between compaction roller pressure and optical properties: When the pressure applied by the compaction roller increases, the prepreg strip adheres more tightly to the mold surface, the internal fiber arrangement may become more regular, and the surface micromorphology changes. This may lead to local changes in surface reflectivity. For example, in the compacted area, the fiber gaps decrease, light scattering decreases, thus affecting texture features and edge sharpness. This relationship can be expressed as: ΔT = g1(P, P0) ΔE = g2(P, P0) Where ΔT represents the change in texture features (such as LBP value), ΔE represents the change in edge gradient intensity, P is the current compaction pressure, P0 is the reference pressure, and g1 and g2 are functions determined based on physical experiments. The relationship between local curvature of the mold and optical properties: In areas with large mold curvature, prepreg strips may experience local stretching or compression. Stretching may lead to increased fiber gaps, increased surface roughness, and enhanced light scattering, resulting in blurred edges and more pronounced texture details. Compression, on the other hand, may make the fibers denser and the surface smoother. These deformations affect the microstructure of the prepreg surface, thereby altering its optical reflection properties. This relationship can be expressed as: ΔS = h1(K, K0) ΔV = h2(K, K0) Where ΔS represents the change in edge sharpness (such as gradient intensity), ΔV represents the change in texture features (such as GLCM entropy), K is the current local curvature, K0 is the reference curvature, and h1 and h2 are functions determined based on physical experiments. These rules are stored in the system's decision module in the form of lookup tables or parameterized functions, used to predict in real time the expected changes in visual signal characteristics under specific process parameters. Subsequently, the system receives visual measurement results indicating deviations in the prepreg strip laying position. This is typically output by the laying system's visual measurement module; for example, the visual measurement module calculates lateral and longitudinal laying deviations by comparing the real-time acquired prepreg strip edge positions with the target laying path.

[0028] Based on the installation process parameters, visual features, and correlations, the system determines the source of installation position deviation. For example, when visual measurements indicate an installation deviation, the system simultaneously acquires the current installation process parameters and visual features. Then, using preset correlations, it predicts what the visual features of the prepreg strip should be under the current installation process parameters. If the actual extracted visual features differ significantly from the predicted visual features, it may indicate a change in the surface optical properties of the prepreg strip, leading to measurement drift. If the actual visual features are largely consistent with the predicted visual features, but the visual measurement results still show a deviation, it is more likely a true physical deviation.

[0029] Finally, based on the determination of the source of the deviation, calibration or compensation operations are performed. When it is determined to be measurement drift, the visual measurement results can be compensated, for example, by correcting the visual measurement results based on the change in visual features to eliminate measurement errors; or the visual information processing method can be adjusted, for example, by dynamically adjusting the parameters of the image processing algorithm to adapt to changes in the optical properties of the prepreg strip surface. When it is determined to be a real physical deviation, a physical correction command is generated to adjust the position of the laying head, for example, by sending a command to the motion control system of the laying head to move it a certain distance in the opposite direction of the deviation to correct the laying position.

[0030] The real-time positioning and calibration method for carbon fiber prepreg laying proposed in this application works by constructing an intelligent decision-making mechanism to cope with the complex dynamic changes during the prepreg laying process. Traditional methods often directly regard visual measurement results as actual physical deviations and perform mechanical corrections. This can lead to misjudgments and miscalibrations when the optical properties of the prepreg surface change, thus accumulating errors.

[0031] This application establishes a multi-dimensional data analysis framework by introducing tiling process parameters and visual features and their correlation. When the visual measurement system reports a deviation in the tiling position, this application does not immediately perform physical correction. Instead, it first comprehensively analyzes the current tiling process parameters (such as tiling speed, temperature, pressure, etc.) and visual features extracted from the prepreg strip edges (such as edge contrast, texture smoothness, etc.). Through preset correlations, the system can predict the visual characteristics of the prepreg strip under the current tiling process parameters.

[0032] If the actual extracted visual features differ significantly from the expected features, it indicates that the surface optical properties of the prepreg strip may have changed. For example, localized resin softening may lead to altered gloss, or fine-tuning of fiber arrangement may cause a shift in reflection direction. In this case, the visual measurement system may have misjudged the true edge of the prepreg strip, resulting in measurement drift. This application determines that the deviation originates from measurement drift and takes non-physical compensation or adjustment measures, such as correcting the visual measurement results or dynamically adjusting the parameters of the image processing algorithm, to improve the accuracy of visual recognition, thereby avoiding ineffective or even harmful physical corrections for non-existent physical deviations.

[0033] Conversely, if the actual extracted visual features are basically consistent with the expected features, but the visual measurement results still show a laying deviation, it indicates that the vision system's recognition of the prepreg strip edge is accurate, and the deviation is indeed caused by inaccurate movement of the laying head or physical deviation of the prepreg strip itself. In this case, this application will determine that the source of the deviation is a real physical deviation and generate a physical correction command to precisely adjust the position of the laying head to correct the actual laying error.

[0034] Through this intelligent judgment mechanism, this application can effectively distinguish between measurement drift and actual physical deviation, avoiding the error accumulation problem caused by the inability to distinguish the source of deviation in traditional methods. It ensures that physical correction is only performed when actual physical deviation occurs, while the problem is solved by adjusting visual processing or compensating for measurement results when measurement drift occurs, ultimately improving the manufacturing quality of composite material parts.

[0035] However, in the actual carbon fiber prepreg laying process, material variations in the prepreg strip itself (such as uneven resin content, differences in fiber arrangement, or changes in surface microstructure) may cause changes in its surface optical properties. These changes in optical response caused by material variations, when judged solely based on visual characteristics and laying process parameters, may be misinterpreted as measurement drift or actual physical deviation of the laying head, leading to inaccurate calibration or compensation operations.

[0036] In response, this application further proposes that the steps for determining the source of tiling position deviation based on tiling process parameters, visual characteristics, and correlations include: Obtain information on the micromechanical response of prepreg strips; Obtain transient thermal response information of prepreg strips; Preset anomaly patterns that include the effects of material variations on visual features, micromechanical response information, and transient thermal response information; Based on the current tiling process parameters and relationships, determine the expected changes in visual features; Obtain the actual changes in current visual features; Compare the actual changes in visual features with the expected changes; Based on the laying process parameters, visual characteristics, micromechanical response information, transient thermal response information, correlation, anomaly patterns, and comparison results, determine whether the laying position deviation is mainly caused by measurement drift due to changes in the optical properties of the prepreg strip surface, the actual physical deviation of the laying head, or the optical response caused by material variation.

[0037] Specifically, micromechanical response information can be understood as data such as deformation, vibration, or acoustic response of prepreg strips under minute stress, reflecting subtle changes in the material's internal structure and mechanical properties. For example, it can be acquired through ultrasonic testing, micro-strain sensors, or acoustic emission sensors, aiming to provide direct evidence of the material's physical state. Transient thermal response information refers to data such as temperature changes, thermal diffusion, or infrared radiation of prepreg strips under short-term thermal stimulation, revealing the material's thermal properties and internal defects. For example, it can be monitored in real-time using infrared thermal imagers or thermocouple arrays, aiming to assist in identifying material uniformity and potential defects. In practical applications, anomaly patterns refer to pre-defined, typical change patterns of visual features, micromechanical response information, and transient thermal response information associated with specific material variation types, such as: Based on previous experiments and materials science knowledge, a library of abnormal response patterns was established, including several typical material intrinsic variations (such as slight oxidation and local pre-curing) and their impact on visual signal characteristics and auxiliary detection signal characteristics under different process parameters. Example of implementation: The abnormal response pattern library is stored in the form of a structured data table. For example, the "slight oxidation" pattern is defined as follows: under heating temperature T and compaction pressure P, the deviation between the expected change in visual grayscale mean ΔG_exp and the actual ΔG_act exceeds the threshold εG. At the same time, the micro-interaction features (such as the pressure RMS value) are outside the normal range, and the transient thermal response (such as the thermal decay time constant τ) deviates from the normal value. This abnormal pattern also includes a description of the specific impact on visual features, such as "the grayscale mean does not increase but decreases, and the edge gradient intensity is lower than expected".

[0038] These models can be pre-established using experimental data, historical data analysis, or finite element simulation, with the aim of providing a reference benchmark for subsequent determination of the source of deviation.

[0039] Furthermore, the expected changes in visual features are predicted based on the current laying process parameters and a preset correlation model, forecasting the expected performance of visual features under ideal conditions (no deviation, no material variation). The actual changes in visual features are obtained through real-time visual information acquisition and feature extraction of the current visual feature data. By comparing these two, the degree of deviation in visual features can be quantified. This application, by introducing micromechanical response information and transient thermal response information, combined with preset anomaly patterns, enables a more comprehensive and in-depth analysis of the actual state of the prepreg strip. When visual measurement results indicate a deviation in the laying position, the judgment no longer relies solely on the correlation between visual features and laying process parameters. Instead, by simultaneously considering the micromechanical behavior and thermal response of the material, it is possible to effectively distinguish between optical response changes caused by material variations and purely measurement drift or physical deviation of the laying head. For example, if visual characteristics show abnormal changes, but micromechanical response and transient thermal response information both indicate that the material state is normal, it is more likely to be judged as measurement drift. If visual characteristics, micromechanical response information, and transient thermal response information all show features consistent with a certain preset abnormal material variation pattern, it can be accurately judged as an optical response caused by material variation, rather than a simple physical deviation. This multi-dimensional data fusion analysis significantly improves the accuracy and robustness of the deviation source judgment.

[0040] Through the above technical solution, this application effectively overcomes the limitations of traditional methods in distinguishing between optical responses caused by material variations and measurement drift or actual physical deviations. By comprehensively analyzing the visual characteristics, micromechanical response information, and transient thermal response information of the prepreg strips, and combining this with preset anomaly patterns, a refined judgment of the source of the laying position deviation can be achieved. This not only avoids invalid or erroneous calibration compensation operations caused by misjudgment, but also enables targeted measures. For example, when the optical response is determined to be caused by material variations, specific quality control processes can be triggered or laying strategies can be adjusted, thereby improving the intelligence level of the carbon fiber prepreg laying process and the stability of product quality.

[0041] However, in real-world, complex installation environments, the determination of the source of deviation can be affected by various uncertainties, such as sensor noise, batch-to-batch material variations, and ambient temperature fluctuations, making the determination of the source of deviation not always absolutely accurate. Performing calibration or compensation operations directly under such highly uncertain conditions may introduce new errors and even adversely affect the prepreg strips or the installation process.

[0042] To address this, this application further proposes a more robust mechanism for identifying and responding to deviations, in order to mitigate uncertainties in the judgment process. The steps outlined above, based on laying process parameters, visual characteristics, micromechanical response information, transient thermal response information, correlations, anomaly patterns, and comparison results, to determine whether the laying position deviation is primarily caused by measurement drift due to changes in the optical properties of the prepreg strip surface, the actual physical deviation of the laying head, or optical response due to material variations, include: Obtain the numerical deviation between the actual and expected changes in visual features; Obtain the numerical difference between the micromechanical response information and the normal baseline; Obtain the numerical difference between transient thermal response information and the normal baseline; The comprehensive uncertainty index is calculated based on the deviation value, the difference value, and the preset weighting factors. When the overall uncertainty index exceeds a preset threshold, a risk aversion strategy is implemented. This risk aversion strategy includes: Reduce the magnitude of the physical correction command; Adjust the tiling speed; Increase the frequency of data acquisition; and The manual review alert has been triggered.

[0043] Specifically, obtaining the deviation value between the actual and expected changes in visual features refers to quantifying the deviation value by comparing the currently collected visual features of the prepreg strip edge with the visual features predicted based on the laying process parameters and correlations. This deviation value can reflect the stability of the visual measurement results or whether there are any anomalies. Obtaining the difference value between micromechanical response information and the normal baseline involves monitoring the micromechanical behavior of the prepreg strip in real time during the laying process using sensors (such as strain sensors or acoustic emission sensors) and comparing it with a preset baseline of micromechanical response under normal and healthy conditions to obtain a quantified difference value. This difference value can indicate potential anomalies in the internal structure or stress state of the material. In practical applications, obtaining the difference value between transient thermal response information and the normal baseline involves using equipment such as infrared thermal imagers to acquire the transient temperature distribution or heat flow changes of the prepreg strip during the laying process and comparing it with the thermal response baseline under normal operating conditions to identify whether there is abnormal heating, uneven cooling, or changes in the material's thermal properties.

[0044] Furthermore, a comprehensive uncertainty index is calculated based on the deviation value, the difference value, and preset weighting factors. This comprehensive uncertainty index is an aggregated indicator that quantifies multiple differences from visual, micromechanical, and transient thermal response data. Through weighted summation and / or other multivariate fusion algorithms, it comprehensively reflects the overall uncertainty level or potential risk of the current tiling status. The preset weighting factors can be adjusted based on the importance or reliability of deviation judgments according to different types of data. When the comprehensive uncertainty index exceeds a preset threshold, the system determines that the current status has high uncertainty or risk and executes risk avoidance strategies. Risk avoidance strategies are a series of preventative measures designed to reduce risk and avoid potential losses. These include: reducing the magnitude of physical correction instructions to avoid excessive or erroneous physical intervention; adjusting the tiling speed, such as reducing the tiling speed to provide the system with more time for data collection and analysis, or to reduce the impact of erroneous decisions; increasing the frequency of data collection to obtain more intensive and real-time information, thereby improving the accuracy of subsequent judgments; and triggering manual review alarms to notify operators of the current anomaly for manual intervention and decision-making.

[0045] This application effectively addresses the uncertainty in determining the source of deviation in complex laying environments by introducing multi-source information fusion and uncertainty quantification mechanisms. Specifically, by acquiring deviation values ​​of visual features, differences in micromechanical response information, and differences in transient thermal response information, the system can comprehensively perceive the laying status of prepreg strips from multiple dimensions. These values ​​reflect the degree of deviation between the current state and the ideal state from optical, mechanical, and thermal perspectives, respectively. Given the limitations or susceptibility to interference of single sensors or single-type data, this application uses preset weighting factors to comprehensively calculate these multi-source difference values, thereby obtaining a more comprehensive and robust overall uncertainty index. Because this index can quantify the overall uncertainty of the current laying status, when it exceeds a preset threshold, the system can promptly identify potential risks or situations with low confidence levels. Based on this, the system no longer blindly performs potentially risky calibration or compensation operations but proactively initiates risk avoidance strategies. For example, reducing the magnitude of physical correction commands can effectively avoid over-correction or erroneous correction due to uncertain judgments, thereby protecting the prepreg strips and equipment; adjusting the laying speed provides the system with more time windows for data resampling or further analysis, while reducing the risks that rapid response may bring; increasing the data acquisition frequency helps to obtain richer and more timely information in situations of high uncertainty, supporting subsequent accurate judgments; and triggering manual review alarms introduces human experience and judgment into the decision-making chain, which is particularly suitable for complex or high-risk situations that automated systems cannot fully handle. Through this multi-level, multi-dimensional risk management mechanism, the solution in this application can ensure that the laying process maintains high stability and safety even in uncertain environments.

[0046] However, in the actual paving process, factors such as the geometry of the mold, the material state of the prepreg strip, and the production cycle requirements are dynamic. If the risk avoidance strategy is implemented in a fixed way, it may not be able to fully adapt to these complex and ever-changing working conditions, resulting in poor calibration results, or even excessive or insufficient intervention, affecting paving efficiency and product quality.

[0047] To address this, this application further proposes a scheme for dynamically adjusting risk avoidance strategies. This scheme aims to flexibly adjust the combination and intensity of risk avoidance strategies based on real-time tiling conditions and task requirements, thereby achieving more accurate, efficient, and adaptable calibration. The implementation scheme can be as follows: When the overall uncertainty index exceeds a preset threshold, the steps for implementing a risk aversion strategy include: Obtain the geometric shape information of the mold area where the current tiling head is located; Assess the degree of material abnormality in the current prepreg strip; Obtain the production cycle time requirement for the current tiling task; Based on geometric information, the degree of material anomaly, and production cycle requirements, dynamically adjust the combination of risk avoidance strategies; The intensity of risk avoidance strategies is dynamically adjusted based on geometric information, the degree of material anomalies, and production cycle requirements.

[0048] Specifically, obtaining the geometric shape information of the mold area where the current tiling head is located refers to acquiring geometric feature data such as curvature, slope, and unevenness of the mold surface corresponding to the current position of the tiling head through a pre-established 3D model of the mold or real-time scanning data. This information is crucial for judging the difficulty of tiling and potential stress concentration areas. Assessing the degree of material state anomaly of the current prepreg strip involves comprehensively analyzing the visual characteristics, micromechanical response information, and transient thermal response information of the prepreg strip to quantify its deviation from the normal material state. For example, an anomaly score can be calculated based on indicators such as fiber uniformity, resin content fluctuation, and local defects (such as wrinkles and bubbles). Obtaining the production cycle requirements of the current tiling task refers to obtaining the constraints on time efficiency of the current tiling operation, such as the allowable time for each layer and the total production cycle. This helps to achieve a balance between risk avoidance and production efficiency.

[0049] The dynamic adjustment of risk avoidance strategies refers to intelligently selecting and combining different risk avoidance measures based on the acquired geometric information, the degree of material anomaly, and production cycle requirements. For example, in complex geometric areas or when the degree of material anomaly is high, it may be necessary to simultaneously reduce the magnitude of physical correction instructions and increase the data acquisition frequency; while in flat areas with normal material conditions, only minor adjustments to the laying speed may be needed. The dynamic adjustment of the strength of risk avoidance strategies refers to further refining the execution degree of each strategy based on the specific working conditions, building upon the selected strategy combination. For example, in areas with extremely high stress concentration risk, the magnitude of physical correction instructions may need to be significantly reduced; while under tight production cycle requirements but with controllable risks, the reduction in laying speed may need to be moderate.

[0050] This application incorporates mold geometry information, the degree of prepreg material anomaly, and production cycle requirements as decision-making criteria, enabling risk mitigation strategies to move beyond a single, fixed pattern. When the overall uncertainty index exceeds a preset threshold, the system no longer simply executes pre-defined risk mitigation measures but first acquires detailed contextual information about the current laying environment. Geometric information is used to identify potential laying challenge areas; for example, high-curvature areas may be more prone to wrinkles or stress concentration. The degree of material anomaly directly reflects the quality or performance fluctuations of the prepreg itself, which may lead to unreliability in visual measurements or actual deformation during laying. Production cycle requirements provide time efficiency constraints for strategy adjustments. Based on this real-time acquired contextual information, the system can intelligently determine the most suitable combination of risk mitigation strategies. For example, when the mold geometry is complex and the degree of material anomaly is high, it may be necessary to simultaneously take multiple measures, such as reducing the magnitude of physical correction commands, adjusting the laying speed, and increasing the data acquisition frequency. Furthermore, the system can dynamically adjust the strength of various strategies based on the specific values ​​of this information. For example, when the degree of material anomaly is very high, the magnitude of the physical correction command may need to be significantly reduced to avoid erroneous corrections due to excessive uncertainty. This dynamic and adaptive adjustment mechanism enables risk avoidance strategies to respond more accurately to actual working conditions, avoiding the shortcomings of a one-size-fits-all approach.

[0051] Through the above technical solution, this application overcomes the limitations of traditional fixed risk avoidance strategies in complex and variable tiling environments. By acquiring and comprehensively considering mold geometry information, the degree of abnormality in prepreg strip material condition, and production cycle requirements in real time, the system can achieve intelligent and adaptive adjustment of risk avoidance strategies. This allows calibration or compensation operations to more accurately match the current working conditions, avoiding excessive or insufficient intervention.

[0052] However, in the actual process of laying carbon fiber prepreg, the instantaneous changes in the movement state of the laying head (such as shaking or oscillation) and the rapid fluctuations in the optical properties of the prepreg strip surface may cause simple dynamic adjustments to be unable to effectively cope with these sudden and dynamic risks, thereby affecting the accuracy and efficiency of laying.

[0053] In this regard, this application further proposes that when the aforementioned comprehensive uncertainty index exceeds a preset threshold, the steps for implementing a risk aversion strategy include: Continuously monitor the stability of the tiling head's movement; Evaluate the rate of change of the surface optical properties of the prepreg strip; When the tiling head moves with vibration or oscillation, and the rate of change of optical characteristics exceeds a preset threshold, a combination of strategies is selected to reduce the tiling speed and increase the data acquisition frequency. Limit the magnitude of physical correction commands; The strength of each strategy in the selected strategy combination is adjusted differently based on the complexity of the mold geometry.

[0054] Specifically, continuously monitoring the stability of the tiling head's motion can be understood as acquiring real-time data on the tiling head's position, velocity, and acceleration in three-dimensional space through a high-precision inertial measurement unit integrated into the tiling head or an external high-speed vision system. Analyzing this data can identify whether the tiling head exhibits excessive shaking, oscillation, or irregular motion, thus determining its motion stability. Shaking typically refers to high-frequency, small-amplitude random vibrations, while oscillation refers to reciprocating motion with a certain periodicity and larger amplitude.

[0055] Meanwhile, assessing the rate of change of the optical properties of the prepreg strip surface refers to extracting the changing trends and speeds of visual features (such as edge sharpness, reflectivity, and texture uniformity) by performing time-series analysis on continuously acquired visual information of the prepreg strip edges. For example, the derivatives or rates of change of visual feature parameters per unit time can be calculated to determine whether the optical properties are rapidly deteriorating or fluctuating.

[0056] When the system detects jitter or oscillation in the laying head movement and the rate of change of the optical properties of the prepreg strip surface exceeds a preset threshold, it will prioritize a strategy combination of reducing the laying speed and increasing the data acquisition frequency. Reducing the laying speed aims to provide the laying head with a longer response time to decrease the cumulative effect of physical deviations and provide more stable observation conditions for the vision system. Increasing the data acquisition frequency allows for the acquisition of denser, more real-time visual and sensor data, thereby improving the ability to perceive changes in the laying status and the accuracy of subsequent deviation judgments.

[0057] Furthermore, this application limits the magnitude of physical correction commands. This means that even if a genuine physical deviation is determined, in situations with high uncertainty (such as tiling head jitter and rapid changes in optical properties), the system will avoid issuing excessively large or overly aggressive physical correction commands to prevent the introduction of new instabilities or exacerbation of existing problems due to overcorrection.

[0058] Furthermore, the strength of each strategy in the selected strategy combination is adjusted differently based on the complexity of the mold geometry. The complexity of the mold geometry can be quantified using indicators such as local curvature, slope change rate, and the presence of sharp angles or small-radius areas. For example, in areas with greater curvature or drastic geometric changes, it may be necessary to significantly reduce the laying speed and collect data more frequently.

[0059] This application introduces real-time monitoring and evaluation of the stability of the tiling head's motion and the rate of change of the optical properties of the prepreg strip surface, enabling more timely and accurate identification of dynamic uncertainties during the tiling process. When these dynamic uncertainties reach a certain level, the system no longer relies solely on static or slowly varying parameters for strategy adjustments, but instead prioritizes a targeted combination of risk-avoidance strategies, namely reducing the tiling speed and increasing the data acquisition frequency. Reducing the tiling speed helps slow the accumulation of physical deviations of the tiling head in unstable states and provides the vision system with longer exposure time, thereby improving image quality and the stability of feature extraction. Increasing the data acquisition frequency captures more instantaneous data, providing richer and more real-time information for subsequent deviation judgment and calibration, effectively addressing measurement uncertainties caused by rapid changes in optical properties. Simultaneously, limiting the magnitude of physical correction commands is to avoid new instability or overshoot caused by aggressive physical intervention in highly uncertain environments, thus ensuring the stability and safety of the calibration process. Furthermore, by adjusting the strength of the strategy in accordance with the complexity of the mold geometry, the risk avoidance measures can be better adapted to the actual needs of different paving areas. For example, a more conservative strategy can be adopted in complex curved areas, which further improves the robustness and adaptability of calibration.

[0060] Through the above technical solution, this application effectively solves the real-time positioning and calibration challenges caused by the unstable movement of the laying head and the rapid changes in the optical properties of the prepreg strip surface during carbon fiber prepreg laying. Compared to adjusting strategies solely based on mold geometry, material condition, and production cycle time, this solution introduces real-time perception and assessment of dynamic risk factors, achieving intelligent, adaptive selection and intensity adjustment of risk avoidance strategy combinations. Therefore, when facing complex conditions such as laying head vibration, oscillation, or drastic changes in the optical properties of the prepreg strip, more precise and conservative countermeasures can be taken promptly. Simultaneously, limiting the amplitude of physical correction commands effectively avoids the negative impacts of over-correction, ensuring the stability and safety of the laying process.

[0061] However, in actual installation, simply adjusting based on the complexity of the mold geometry may not adequately address the risk of localized stress concentration when laying prepreg strips on complex curved surfaces, potentially leading to insufficient precision or poor effectiveness of the adjustment strategy. Without a thorough consideration of the material's inherent deformation characteristics and the actual mechanical response during installation, the corrective or compensatory operations may not effectively prevent defects such as wrinkles, deformation, or even delamination of the prepreg strips in complex geometric areas.

[0062] In this regard, this application further proposes the following steps for differentially adjusting the strength of each strategy in the selected strategy combination based on the complexity of the mold geometry: Obtain real-time strain information of the prepreg strip along its length and width directions; Obtain the principal direction of local curvature of the mold area where the current tiling head is located; Obtain the current laying direction of the prepreg strip; Calculate the angle between the prepreg strip laying direction and the principal direction of mold curvature; Based on the included angle, real-time strain information, and preset material anisotropic deformation characteristics, the risk of local stress concentration of prepreg strip in the current laying area is predicted; When the predicted risk of local stress concentration exceeds a preset threshold, the strength of each strategy in the selected strategy combination is adjusted differentially, specifically as follows: The magnitude of the reduction in physical correction instructions will be further reduced based on the degree of stress concentration risk. To adjust the laying speed, further reduce the laying speed based on the degree of stress concentration risk; To enhance the data acquisition frequency, the frequency should be further increased based on the degree of stress concentration risk.

[0063] Specifically, the steps described above, which differentiate the strength of each strategy in the selected strategy combination based on the complexity of the mold geometry, aim to achieve more precise control over the strength of risk aversion strategies by introducing more refined mechanical and geometric parameters.

[0064] Obtaining real-time strain information of the prepreg strip along its length and width directions refers to monitoring the tensile, compressive, or shear deformation experienced by the prepreg strip during the laying process in real time using, for example, fiber optic sensors, strain gauge arrays, or vision-based deformation measurement systems. The purpose is to quantify the actual deformation state of the prepreg strip in the laying area, providing crucial mechanical input for subsequent stress concentration risk prediction.

[0065] Furthermore, obtaining the principal direction of local curvature in the mold area where the paving head is currently located can be understood as determining the directions of maximum and minimum curvature on the mold surface where the paving head is currently situated, through methods such as CAD model data of the mold, laser scanning, or structured light measurement. This information is crucial for understanding the bending and twisting behavior of prepreg strips on complex curved surfaces.

[0066] Meanwhile, obtaining the current laying direction of the prepreg strip refers to determining the direction of movement of the prepreg strip relative to the mold surface. This can be obtained through the motion trajectory data of the laying head or a vision tracking system.

[0067] Based on this, the angle between the prepreg strip laying direction and the principal direction of mold curvature is calculated. The purpose is to assess the alignment of the prepreg strip with the geometric features of the mold surface during the laying process. When there is a large angle between the laying direction and the principal direction of curvature, the prepreg strip is more prone to shear deformation, thereby increasing the risk of stress concentration.

[0068] Based on the included angle, real-time strain information, and preset anisotropic deformation characteristics of the material, the risk of local stress concentration in the current laying area of ​​the prepreg strip is predicted. Specifically, the anisotropic deformation characteristics of the material refer to the different mechanical responses of the carbon fiber prepreg in different directions; for example, the stiffness along the fiber direction is much higher than the stiffness perpendicular to the fiber direction. By inputting real-time strain data, the included angle between the laying direction and the principal direction of curvature, and the inherent anisotropic characteristics of the material into a preset mechanical model or finite element analysis model, the potential local stress level of the prepreg strip in the current laying area can be assessed in real time, and the existence of stress concentration areas can be predicted.

[0069] When the predicted risk of localized stress concentration exceeds a preset threshold, the strength of each strategy in the selected strategy combination is adjusted differentially. Specifically, this means finely adjusting the strength of three strategies—reducing the magnitude of physical correction instructions, adjusting the laying speed, and increasing the data acquisition frequency—based on the severity of the stress concentration risk. For example, when the stress concentration risk is high, the magnitude of the physical correction instructions will be further reduced to avoid exacerbating material deformation due to over-correction; the laying speed will be further reduced to provide the material with more sufficient relaxation time; and the data acquisition frequency will be further increased to more intensively monitor the material condition and laying effect.

[0070] This application, by introducing and analyzing real-time strain information of prepreg strips, the principal direction of local curvature of the mold, and the laying direction, combined with the anisotropic deformation characteristics of the material, can accurately predict the risk of local stress concentration in complex geometric regions of prepreg strips. Because this risk can be quantified in real time, when the aforementioned comprehensive uncertainty index exceeds a preset threshold, the intensity adjustment of the risk avoidance strategy is no longer singular or coarse, but can be finely differentiated according to the degree of stress concentration risk. For example, by reducing the magnitude of physical correction commands, new defects can be avoided in stress-sensitive areas due to excessive mechanical intervention; by reducing the laying speed, a longer deformation relaxation time can be provided for the prepreg strip, thereby effectively alleviating local stress; by increasing the data acquisition frequency, laying status information can be obtained more timely and comprehensively, providing more reliable data support for subsequent decision-making. This fine adjustment mechanism based on actual mechanical response effectively compensates for the shortcomings of adjusting only based on the geometric complexity of the mold, ensuring the laying quality and stability of prepreg strips in complex laying environments.

[0071] However, in practical applications, simply reducing or lowering the stress concentration risk in a general way may not be sufficient to cope with the complex material properties, mold geometry, and dynamic process parameters involved in carbon fiber prepreg laying, resulting in insufficient precision and adaptability of strategy adjustments.

[0072] In this regard, this application further proposes the following steps for differentially adjusting the strength of each strategy in the selected strategy combination based on the complexity of the mold geometry: Obtain the material thickness information for the current paving area; Obtain mold surface roughness information; Obtain the resin flowability parameters of the prepreg strip; Based on material thickness information, mold surface roughness information, resin flow parameters, and the degree of stress concentration risk, the attenuation coefficient of the physical correction command amplitude is calculated through a preset nonlinear conversion rule. Based on material thickness information, mold surface roughness information, resin flow parameters, and the degree of stress concentration risk, the reduction ratio of laying speed is calculated through a preset nonlinear transformation rule. Based on material thickness information, mold surface roughness information, resin flow parameters, and the degree of stress concentration risk, the data acquisition frequency increase factor is calculated through preset nonlinear transformation rules. The calculated attenuation coefficient is applied to a strategy to reduce the amplitude of the physical correction command, thereby further reducing its amplitude; The calculated reduction ratio is applied to adjust the tiling speed strategy to further reduce the tiling speed; The calculated improvement factor is then applied to a strategy to enhance the data acquisition frequency, thereby further increasing the data acquisition frequency. Monitor the actual motion response of the tiling head after executing the calibration command; Monitor the actual deformation behavior of the prepreg strip at the adjusted laying speed; Based on the actual motion response of the laying head after executing the correction command and the actual deformation behavior of the prepreg strip at the adjusted laying speed, the nonlinear conversion rule is fine-tuned. Based on the fine-tuned nonlinear conversion rules, the attenuation coefficient of the physical correction command amplitude, the reduction ratio of the laying speed, and the increase factor of the data acquisition frequency are recalculated.

[0073] Specifically, when differentiating the strength of each strategy in the selected strategy combination, it is first necessary to obtain information on the material thickness of the current laying area, the surface roughness of the mold, and the resin flowability parameters of the prepreg strip. Material thickness information can be understood as the actual thickness of the prepreg strip at the current laying position, which can be measured in real time using a laser thickness gauge or ultrasonic sensor. Its purpose is to reflect the local stacking or thinning of the material, which directly affects the laying accuracy and stress distribution. Mold surface roughness information refers to the microscopic geometric features of the mold surface contacted by the laying head, which can be obtained using an optical profilometer or a stylus roughness meter. Its purpose is to evaluate the influence of the mold surface on the friction and adhesion of the prepreg strip. The resin flowability parameters of the prepreg strip reflect the flow properties of the resin under specific temperatures and pressures, which can be obtained using an online rheometer or a predictive model based on material batch data. Its purpose is to indicate the deformation and filling capacity of the prepreg during laying and curing.

[0074] Furthermore, the aforementioned material thickness information, mold surface roughness information, resin flowability parameters, and the degree of stress concentration risk are comprehensively considered. A preset nonlinear transformation rule is used to calculate the attenuation coefficient of the physical correction command amplitude, the reduction ratio of the laying speed, and the increase factor of the data acquisition frequency. The nonlinear transformation rule can be understood as a mathematical model or lookup table that maps multiple input parameters (such as material thickness, surface roughness, resin flowability, and stress concentration risk) to specific strategy strength adjustment values ​​(such as attenuation coefficient, reduction ratio, and increase factor). The preset rule aims to capture the complex nonlinear relationships between these parameters to achieve more accurate quantification of strategy strength. For example, when the stress concentration risk is high and the mold surface roughness is large, a larger physical correction command amplitude attenuation coefficient and a more significant reduction ratio of the laying speed may be required.

[0075] Furthermore, this application also includes monitoring the actual motion response of the laying head after executing the correction command and the actual deformation behavior of the prepreg strip at the adjusted laying speed. The actual motion response of the laying head after executing the correction command can be monitored in real time using a high-precision encoder, laser tracker, or vision system to assess whether the execution effect of the correction command meets expectations. The actual deformation behavior of the prepreg strip at the adjusted laying speed can be continuously acquired using a high-speed vision system or strain sensor to assess the actual impact of laying speed adjustment on material deformation (such as wrinkling or twisting).

[0076] Based on the above monitoring results, the nonlinear conversion rule was fine-tuned. Fine-tuning refers to making small, iterative corrections to the parameters or function forms in the preset nonlinear conversion rule according to actual feedback data, so as to better adapt it to the actual paving environment and material properties. Therefore, the fine-tuned nonlinear conversion rule was used to recalculate the attenuation coefficient of the physical correction command amplitude, the reduction ratio of the paving speed, and the increase factor of the data acquisition frequency, thus forming a closed-loop adaptive adjustment mechanism.

[0077] This application, by incorporating information on the material thickness of the current paving area, the surface roughness of the mold, and the resin flowability parameters of the prepreg strip, and combining this with the aforementioned degree of stress concentration risk, enables a more comprehensive assessment of potential risks and material response characteristics during the paving process. These multi-dimensional parameters are input into a preset nonlinear transformation rule, thereby enabling the calculation of a more precise attenuation coefficient for the physical correction command amplitude, the reduction ratio of paving speed, and the increase factor for data acquisition frequency. This overcomes the limitations of making general adjustments based solely on stress concentration risk, allowing for a more precise matching of strategy intensity adjustments to actual working conditions.

[0078] Furthermore, by real-time monitoring of the actual motion response of the laying head after executing the correction command and the actual deformation behavior of the prepreg strip at the adjusted laying speed, this application establishes a crucial feedback mechanism. This actual response data is used to fine-tune the aforementioned nonlinear transformation rules, ensuring that the rules can adaptively optimize with changes in the laying environment, material batch, or equipment status. Thus, the fine-tuned rules can more accurately guide subsequent strategy strength calculations, forming a continuously learning and optimizing closed-loop control system, effectively avoiding over- or under-correction and significantly improving the accuracy and reliability of the laying process.

[0079] Through the above technical solutions, this application enables refined and adaptive adjustment of the strength of risk avoidance strategies. Specifically, by introducing material thickness information, mold surface roughness information, and resin flow parameters, the strategy adjustment is no longer a single-dimensional response, but rather a comprehensive consideration of the complex interactions between materials, molds, and processes, thereby significantly improving the accuracy and targeting of strategy adjustments. Furthermore, by real-time monitoring of the laying head's motion response and prepreg strip deformation behavior, and accordingly fine-tuning the nonlinear conversion rules, this application establishes a dynamic feedback optimization mechanism. This allows the system to learn and correct itself based on actual laying results, ensuring optimal calibration and risk avoidance under varying working conditions, effectively reducing the incidence of laying defects, improving the quality and efficiency of carbon fiber prepreg laying, and avoiding unnecessary production cycle time losses.

[0080] However, in actual tiling operations, especially when dealing with molds with complex geometries or multi-layer tiling tasks, if the fine-tuning process lacks a refined data acquisition, processing, and feedback mechanism, it may lead to insufficient accuracy and real-time performance of rule corrections, thereby affecting the stability and consistency of the final tiling quality.

[0081] In response, this application further proposes a real-time positioning calibration method for carbon fiber prepreg laying. The method involves fine-tuning the nonlinear conversion rule based on the actual motion response of the laying head after executing the calibration command and the actual deformation behavior of the prepreg strip at the adjusted laying speed. After the tiling head executes the correction command, the high-speed sensor array is activated to perform high-frequency sampling of the instantaneous motion trajectory of the tiling head after correction. Using a high-frequency vision sensor, continuous image acquisition is performed on the local deformation area of ​​the prepreg strip after the laying speed is adjusted; Time-domain filtering is performed on the instantaneous motion trajectory data of the paving head sampled at high frequencies to remove high-frequency noise; Based on the data after time-domain filtering, motion trend analysis is performed to identify the actual motion direction and amplitude after the correction command is executed; Image sequence analysis was performed on the local deformation area data of prepreg strips acquired through continuous image acquisition to extract visual features of edge changes and texture smoothness of the deformation area. Temporal smoothing is performed on the extracted visual features to reduce random noise in the visual signal; Based on the visual features after temporal smoothing, the degree and rate of actual deformation are identified; The actual movement direction and amplitude of the laying head are compared with the expected effect of the preset correction command, and the actual deformation degree and rate of the prepreg strip are compared with the expected effect of the preset laying speed adjustment, so as to calculate the deviation of the correction effect. Based on the deviation of the correction effect, the preset feedback sensitivity threshold, and the adjustment step size, the relevant parameters in the nonlinear conversion rule are corrected in a small and gradual manner. When revising the rules, priority is given to adjusting parameters that directly affect the motion response of the laying head and the deformation behavior of the prepreg strip, and the revised rules are locally verified. In multi-layer tiling operations, the results of fine-tuning of non-linear transformation rules after each layer is completed are recorded in history, and trend analysis is performed on the rule changes between adjacent layers. An alarm is triggered when the trend of rule changes exceeds the preset stable range.

[0082] Specifically, after the laying head executes the correction command, a high-speed sensor array is activated to perform high-frequency sampling of the laying head's instantaneous motion trajectory after correction. The high-speed sensor array can include, but is not limited to, laser displacement sensors, inertial measurement units (IMUs), or high-precision encoders to ensure the capture of minute changes and transient responses in the laying head's motion. Simultaneously, a high-frequency vision sensor is used to continuously acquire images of the localized deformation areas of the prepreg strip after the laying speed is adjusted. The high-frequency vision sensor can be, for example, a high-speed industrial camera, capable of capturing the dynamic deformation of the prepreg strip during the laying process at extremely high frame rates, such as wrinkles, twists, or uneven edges.

[0083] The process involves temporal filtering of the high-frequency sampled instantaneous motion trajectory data of the tiling head to remove high-frequency noise. Various algorithms can be used for temporal filtering, such as mean filtering, median filtering, Gaussian filtering, or Kalman filtering, to smooth the data and preserve the true motion trend. Based on the time-domain filtered data, motion trend analysis is performed to identify the actual motion direction and amplitude after the correction command is executed. Motion trend analysis can employ curve fitting, feature point tracking, or machine learning-based pattern recognition methods.

[0084] Furthermore, image sequence analysis is performed on the locally deformed areas of the prepreg strips acquired through continuous image acquisition. The aim is to extract visual features such as edge variations and texture smoothness of the deformed areas. Image sequence analysis can utilize computer vision algorithms, such as edge detection, feature point matching, and optical flow, to quantify the deformation state of the prepreg strips. Temporal smoothing is then applied to the extracted visual features to reduce random noise in the visual signal and ensure feature stability. Temporal smoothing can employ methods such as moving average or exponential smoothing. Based on the visual features after temporal smoothing, the actual degree and rate of deformation are identified.

[0085] Therefore, the actual movement direction and amplitude of the tiling head are compared with the expected effect of the preset correction command, and the actual deformation degree and rate of the prepreg strip are compared with the expected effect of the preset tiling speed adjustment, thereby calculating the deviation of the correction effect. The comparison process can employ methods such as difference calculation, correlation analysis, or error function evaluation. Based on the deviation of the correction effect, the preset feedback sensitivity threshold, and the adjustment step size, the relevant parameters in the nonlinear conversion rule are corrected in a small-amplitude, gradual manner. The feedback sensitivity threshold is used to control the response speed of the correction, while the adjustment step size ensures the stability and convergence of the correction process.

[0086] When revising rules, priority is given to adjusting parameters that directly affect the tiling head's motion response and prepreg strip deformation behavior, and the revised rules are locally verified. For example, for tiling head motion deviation, the attenuation coefficient of the physical correction command amplitude is adjusted first; for prepreg strip deformation, the reduction ratio of the tiling speed is adjusted first. Furthermore, in multi-layer tiling operations, the results of fine-tuning the nonlinear transformation rules after each layer is completed are historically recorded, and trend analysis is performed on rule changes between adjacent layers. When the rule change trend exceeds the preset stability range, an alarm is triggered to prompt operators or the system to conduct further inspection or intervention.

[0087] This application's solution constructs a closed-loop adaptive feedback system by introducing high-frequency, high-precision sensor data acquisition and a refined data processing and analysis mechanism. Specifically, a high-speed sensor array and a high-frequency vision sensor can capture, in real time and accurately, the actual motion response of the laying head after executing correction commands and the actual deformation behavior of the prepreg strip after adjusting the laying speed. These raw data undergo time-domain filtering and smoothing to effectively remove noise, ensuring the accuracy of subsequent motion trend analysis and image sequence analysis. By precisely comparing the actually observed motion direction, amplitude, deformation degree, and rate with the preset expected effect, the system can quantify the deviation of the correction effect. Based on this precise deviation, combined with a preset feedback sensitivity threshold and adjustment step size, the relevant parameters in the nonlinear conversion rule can be corrected in a small, gradual manner. This correction strategy avoids system oscillations that may result from over-adjustment while ensuring continuous optimization of the rule. Furthermore, by analyzing historical records and trends of the results of rule fine-tuning in multi-layer tiling operations, this application can identify long-term system drift or material variation, thereby triggering alarms before problems worsen, achieving predictive maintenance and deeper quality control of the tiling process.

[0088] Through the aforementioned technical solutions, this application achieves more accurate and timely correction of nonlinear transformation rules by employing high-frequency, multi-dimensional data acquisition and refined data processing, effectively avoiding under-correction or over-correction caused by inaccurate data or rough processing. Especially in complex curved surface tiling and multi-layer tiling scenarios, this application can more effectively handle the transient response of the tiling head movement and the dynamic deformation of the prepreg strip, thereby ensuring the accuracy of the tiling position and the stability of the tiling quality. Furthermore, the introduction of historical record and trend analysis mechanisms enables the system to possess self-learning and long-term adaptability capabilities, allowing for timely detection and early warning of potential systemic problems or batch-to-batch material variations.

[0089] However, in the actual carbon fiber prepreg laying process, factors such as laying speed, local curvature of the mold, and material properties of the prepreg strip are dynamically changing. Using static or preset parameter filtering methods may not effectively adapt to these changes, leading to incomplete noise removal in some conditions or over-filtering in others, resulting in the loss of true details of the motion trajectory. This, in turn, affects the accuracy of subsequent motion trend analysis and may ultimately reduce the precision of correction commands.

[0090] In response, this application further proposes the following steps for performing time-domain filtering on the high-frequency sampled instantaneous motion trajectory data of the tiling head to remove high-frequency noise: Continuously collect instantaneous motion trajectory data of the tiling head; Real-time acquisition of current laying speed, local curvature of mold, and material property parameters of prepreg strips; The center frequency and bandwidth of a notch filter are dynamically adjusted based on the laying speed, the local curvature of the mold, and the material properties of the prepreg strip. The adjusted notch filter is applied to the instantaneous motion trajectory data of the tiling head to remove periodic noise; The cutoff frequency of a low-pass filter is dynamically adjusted based on the laying speed, the local curvature of the mold, and the material properties of the prepreg strip. The adjusted low-pass filter is applied to the motion trajectory data after notch filtering. Motion trend analysis is performed on the motion trajectory data after double filtering.

[0091] Specifically, continuously acquiring the instantaneous motion trajectory data of the tiling head refers to continuously and densely sampling the tiling head's corrected motion trajectory using a high-frequency sensor array to obtain its precise position and attitude information in the time dimension. Real-time acquisition of the current tiling speed, local curvature of the mold, and material property parameters of the prepreg strip is to provide the context information required for dynamic filtering. The tiling speed can be understood as the instantaneous rate at which the tiling head moves on the mold surface; the local curvature of the mold refers to the degree of geometric curvature of the area where the tiling head is currently located, which may affect the smoothness of the tiling head's movement and the introduced vibration frequency; the material property parameters of the prepreg strip, such as its stiffness and damping characteristics, may affect the natural vibration frequency generated during the tiling process.

[0092] Based on this, the center frequency and bandwidth of a notch filter are dynamically adjusted according to real-time acquired parameters such as the laying speed, local curvature of the mold, and material properties of the prepreg strip. A notch filter is a filter capable of attenuating signals within a specific frequency range, aiming to precisely remove periodic noise caused by mechanical vibrations, motor harmonics, etc. By dynamically adjusting its center frequency and bandwidth, the filter can accurately match the frequency of periodic noise generated under the current operating conditions, avoiding false filtering or missed filtering. For example, when the laying speed increases, some mechanical vibration frequencies may rise, and the center frequency of the notch filter should be adjusted accordingly.

[0093] Furthermore, the adjusted notch filter is applied to the instantaneous motion trajectory data of the tiling head to effectively remove periodic noise. Subsequently, the cutoff frequency of a low-pass filter is dynamically adjusted based on the tiling speed, the local curvature of the die, and the material properties of the prepreg strip. The low-pass filter is used to filter out all signals above a certain cutoff frequency, aiming to remove random high-frequency noise and smooth the motion trajectory data. Dynamically adjusting its cutoff frequency allows for an optimal balance between smoothness and detail retention based on actual needs. For example, a lower cutoff frequency may be needed in high-speed tiling or areas with complex curvature to obtain a smoother trajectory, while a higher cutoff frequency can be appropriately increased in low-speed or flat areas to retain more detail.

[0094] Finally, the adjusted low-pass filter is applied to the motion trajectory data after notch filtering, thus completing the dual filtering process. Motion trend analysis of the dual-filtered motion trajectory data involves conducting in-depth analysis of the tiling head's movement direction, amplitude, and speed changes based on this purer and more accurate data. This provides a reliable basis for subsequent evaluation of the correction effect and fine-tuning of the nonlinear conversion rules.

[0095] Based on the same inventive concept, this application also discloses a real-time positioning and calibration system for carbon fiber prepreg laying, such as... Figure 2 As shown, the system includes: Parameter acquisition module 1 is used to acquire laying process parameters that affect the optical properties of the prepreg strip surface; The visual information acquisition module 2 is used to acquire visual information of the edge of the prepreg strip and extract visual features reflecting the optical properties of the prepreg strip surface from the visual information. The association preset module 3 is used to preset the association between tiling process parameters and visual features; Measurement result receiving module 4 is used to receive visual measurement results indicating that there is a deviation in the laying position of the prepreg strip; Deviation source judgment module 5 is used to determine the source of the paving position deviation based on paving process parameters, visual characteristics and correlation relationships; The calibration or compensation execution module 6 is used to perform calibration or compensation operations based on the determination of the source of deviation. Specifically, the calibration or compensation execution module includes: The measurement drift processing unit 61 is used to compensate for the visual measurement results or adjust the visual information processing method when measurement drift is determined. The physical correction instruction generation unit 62 is used to generate a physical correction instruction to adjust the position of the tiling head when it is determined that there is a real physical deviation.

[0096] This application can effectively avoid the accumulation of errors caused by misjudgment in traditional methods, and ensure the manufacturing quality of composite material parts.

[0097] The above are merely embodiments of this application and are not intended to limit the scope of protection of this application.

Claims

1. A method for real-time positioning and calibration of carbon fiber prepreg laying, characterized in that, Includes the following steps: Obtain the laying process parameters that affect the optical properties of the prepreg strip surface; Visual information of the edge of the prepreg strip is collected, and visual features reflecting the optical properties of the prepreg strip surface are extracted from the visual information; The relationship between preset tiling process parameters and visual features; Receive visual measurement results indicating deviations in the placement of prepreg strips; Based on the tiling process parameters, visual characteristics, and correlations, determine the source of tiling position deviation; Based on the determination of the source of the deviation, perform calibration or compensation operations, specifically including: When measurement drift is detected, the visual measurement results are compensated or the visual information processing method is adjusted. When a physical deviation is determined to be real, a physical correction command is generated to adjust the position of the tiling head.

2. The real-time positioning and calibration method for carbon fiber prepreg laying according to claim 1, characterized in that, The step of determining the source of the tiling position deviation based on tiling process parameters, visual characteristics, and correlations includes: Obtain information on the micromechanical response of prepreg strips; Obtain transient thermal response information of prepreg strips; Preset anomaly patterns that include the effects of material variations on visual features, micromechanical response information, and transient thermal response information; Based on the current tiling process parameters and relationships, determine the expected changes in visual features; Obtain the actual changes in current visual features; Compare the actual changes in visual features with the expected changes; Based on the laying process parameters, visual characteristics, micromechanical response information, transient thermal response information, correlation, anomaly patterns, and comparison results, determine whether the laying position deviation is caused by measurement drift due to changes in the optical properties of the prepreg strip surface, the actual physical deviation of the laying head, or the optical response caused by material variation.

3. The method for real-time positioning and calibration of carbon fiber prepreg laying according to claim 2, characterized in that, The step of determining whether the installation position deviation is mainly caused by measurement drift due to changes in the optical properties of the prepreg strip surface, the actual physical deviation of the installation head, or optical response caused by material variation, based on installation process parameters, visual characteristics, micromechanical response information, transient thermal response information, correlation, anomaly patterns, and comparison results, further includes: Obtain the numerical deviation between the actual and expected changes in visual features; Obtain the numerical difference between the micromechanical response information and the normal baseline; Obtain the numerical difference between transient thermal response information and the normal baseline; The comprehensive uncertainty index is calculated based on the deviation value, the difference value, and the preset weighting factors. When the overall uncertainty index exceeds a preset threshold, a risk aversion strategy is implemented. This risk aversion strategy includes: Reduce the magnitude of the physical correction command; Adjust the tiling speed; Increase the frequency of data acquisition; and The manual review alert has been triggered.

4. The real-time positioning and calibration method for carbon fiber prepreg laying according to claim 3, characterized in that, The steps for implementing a risk avoidance strategy when the overall uncertainty index exceeds a preset threshold include: Obtain the geometric shape information of the mold area where the current tiling head is located; Assess the degree of material abnormality in the current prepreg strip; Obtain the production cycle time requirement for the current tiling task; Based on geometric information, the degree of material anomaly, and production cycle requirements, dynamically adjust the combination of risk avoidance strategies; The intensity of risk avoidance strategies is dynamically adjusted based on geometric information, the degree of material anomalies, and production cycle requirements.

5. The real-time positioning and calibration method for carbon fiber prepreg laying according to claim 4, characterized in that, The steps of dynamically adjusting the combination of risk avoidance strategies based on geometric information, the degree of material anomaly, and production cycle requirements include: Continuously monitor the stability of the tiling head's movement; Evaluate the rate of change of the surface optical properties of the prepreg strip; When the tiling head moves with jitter or oscillation, and the rate of change of optical characteristics exceeds a preset threshold, a combination of strategies is selected to reduce the tiling speed, increase the data acquisition frequency, and limit the amplitude of physical correction commands. The strength of each strategy in the selected strategy combination is adjusted differently based on the complexity of the mold geometry.

6. The method for real-time positioning and calibration of carbon fiber prepreg laying according to claim 5, characterized in that, The step of differentially adjusting the strength of each strategy in the selected strategy combination based on the complexity of the mold geometry includes: Obtain real-time strain information of the prepreg strip along its length and width directions; Obtain the principal direction of local curvature of the mold area where the current tiling head is located; Obtain the current laying direction of the prepreg strip; Calculate the angle between the prepreg strip laying direction and the principal direction of mold curvature; Based on the included angle, real-time strain information, and preset material anisotropic deformation characteristics, the risk of local stress concentration of prepreg strip in the current laying area is predicted; When the predicted risk of local stress concentration exceeds a preset threshold, the strength of each strategy in the selected strategy combination is adjusted differentially, specifically as follows: The magnitude of the reduction in physical correction instructions will be further reduced based on the degree of stress concentration risk. To adjust the laying speed, further reduce the laying speed based on the degree of stress concentration risk; To enhance the data acquisition frequency, the frequency should be further increased based on the degree of stress concentration risk.

7. The method for real-time positioning and calibration of carbon fiber prepreg laying according to claim 6, characterized in that, The step of differentially adjusting the strength of each strategy in the selected strategy combination based on the complexity of the mold geometry includes: Obtain the material thickness information for the current paving area; Obtain mold surface roughness information; Obtain the resin flowability parameters of the prepreg strip; Based on material thickness information, mold surface roughness information, resin flow parameters, and the degree of stress concentration risk, the attenuation coefficient of the physical correction command amplitude is calculated through a preset nonlinear conversion rule. Based on material thickness information, mold surface roughness information, resin flow parameters, and the degree of stress concentration risk, the reduction ratio of laying speed is calculated through a preset nonlinear transformation rule. Based on material thickness information, mold surface roughness information, resin flow parameters, and the degree of stress concentration risk, the data acquisition frequency increase factor is calculated through preset nonlinear transformation rules. The calculated attenuation coefficient is applied to a strategy to reduce the amplitude of the physical correction command, thereby further reducing its amplitude; The calculated reduction ratio is applied to adjust the tiling speed strategy to further reduce the tiling speed; The calculated improvement factor is then applied to a strategy to enhance the data acquisition frequency, thereby further increasing the data acquisition frequency. Monitor the actual motion response of the tiling head after executing the calibration command; Monitor the actual deformation behavior of the prepreg strip at the adjusted laying speed; Based on the actual motion response of the laying head after executing the correction command and the actual deformation behavior of the prepreg strip at the adjusted laying speed, the nonlinear conversion rule is fine-tuned. Based on the fine-tuned nonlinear conversion rules, the attenuation coefficient of the physical correction command amplitude, the reduction ratio of the laying speed, and the increase factor of the data acquisition frequency are recalculated.

8. The method for real-time positioning and calibration of carbon fiber prepreg laying according to claim 7, characterized in that, The step of fine-tuning the nonlinear conversion rule based on the actual motion response of the laying head after executing the correction command and the actual deformation behavior of the prepreg strip at the adjusted laying speed includes: After the tiling head executes the correction command, the high-speed sensor array is activated to perform high-frequency sampling of the instantaneous motion trajectory of the tiling head after correction. Using a high-frequency vision sensor, continuous image acquisition is performed on the local deformation area of ​​the prepreg strip after the laying speed is adjusted; Time-domain filtering is performed on the instantaneous motion trajectory data of the paving head sampled at high frequencies to remove high-frequency noise; Based on the data after time-domain filtering, motion trend analysis is performed to identify the actual motion direction and amplitude after the correction command is executed; Image sequence analysis was performed on the local deformation area data of prepreg strips acquired through continuous image acquisition to extract visual features of edge changes and texture smoothness of the deformation area. Temporal smoothing is performed on the extracted visual features to reduce random noise in the visual signal; Based on the visual features after temporal smoothing, the degree and rate of actual deformation are identified; The actual movement direction and amplitude of the laying head are compared with the expected effect of the preset correction command, and the actual deformation degree and rate of the prepreg strip are compared with the expected effect of the preset laying speed adjustment, so as to calculate the deviation of the correction effect. Based on the deviation of the correction effect, the preset feedback sensitivity threshold, and the adjustment step size, the relevant parameters in the nonlinear conversion rule are corrected in a small and gradual manner. When revising the rules, priority is given to adjusting parameters that directly affect the motion response of the laying head and the deformation behavior of the prepreg strip, and the revised rules are locally verified. In multi-layer tiling operations, the results of fine-tuning of non-linear transformation rules after each layer is completed are recorded in history, and trend analysis is performed on the rule changes between adjacent layers. An alarm is triggered when the trend of rule changes exceeds the preset stable range.

9. The real-time positioning and calibration method for carbon fiber prepreg laying according to claim 8, characterized in that, The step of performing time-domain filtering on the high-frequency sampled instantaneous motion trajectory data of the tiling head to remove high-frequency noise includes: Continuously collect instantaneous motion trajectory data of the tiling head; Real-time acquisition of current laying speed, local curvature of mold, and material property parameters of prepreg strips; The center frequency and bandwidth of a notch filter are dynamically adjusted based on the laying speed, the local curvature of the mold, and the material properties of the prepreg strip. The adjusted notch filter is applied to the instantaneous motion trajectory data of the tiling head to remove periodic noise; The cutoff frequency of a low-pass filter is dynamically adjusted based on the laying speed, the local curvature of the mold, and the material properties of the prepreg strip. The adjusted low-pass filter is applied to the motion trajectory data after notch filtering. Motion trend analysis is performed on the motion trajectory data after double filtering.

10. A real-time positioning and calibration system for carbon fiber prepreg laying, characterized in that, The system includes: The parameter acquisition module is used to acquire laying process parameters, which affect the surface optical properties of the prepreg strip. The visual information acquisition module is used to acquire visual information of the edge of the prepreg strip and extract visual features reflecting the optical properties of the prepreg strip surface from the visual information. The association preset module is used to preset the association between tiling process parameters and visual features; The measurement result receiving module is used to receive visual measurement results indicating that there is a deviation in the laying position of the prepreg strip; The deviation source judgment module is used to determine the source of the tiling position deviation based on tiling process parameters, visual features, and correlation relationships. The calibration or compensation execution module is used to perform calibration or compensation operations based on the determination of the source of deviation. The calibration or compensation execution module specifically includes: The measurement drift processing unit is used to compensate for visual measurement results or adjust the visual information processing method when measurement drift is detected. The physical correction instruction generation unit is used to generate physical correction instructions to adjust the position of the tiling head when a real physical deviation is determined.

Images