Composite material sharing intelligent manufacturing management system based on industrial interconnection

Abstract

The invention relates to the technical field of industrial management, and discloses a composite material sharing intelligent manufacturing management system based on an industrial internet, which comprises a cross-layer state modeling and management and control strategy mapping module, the method comprises the following steps: firstly, based on multi-source process data, respectively quantifying the local cumulative strength of microdefects, the association degree of a macroscopic equipment operation sequence and the defect connectivity risk degree driven by space geometry, and aggregating the indexes into cross-layer self-organizing critical coupling state parameters; and then generating a resource scheduling weight vector and a geometric complexity load scaling coefficient through deterministic calculation, and dynamically adjusting a task allocation probability and an equipment bearing limit according to the resource scheduling weight vector and the geometric complexity load scaling coefficient. According to the method, active avoidance and decriticalization control over the critical risk of the manufacturing system are achieved, and high-quality and efficient collaborative production of complex components is guaranteed.

Description

Technical Field

[0001] This invention relates to the field of industrial management technology, and more specifically, to a shared intelligent manufacturing management system for composite materials based on the Industrial Internet. Background Technology

[0002] With the rapid development of aerospace and high-end equipment, the demand for carbon fiber reinforced composite material components is increasing. Automated Placement (AFP), as a core method in composite material manufacturing, is gradually transforming into a shared intelligent manufacturing model based on the Industrial Internet. In this model, multiple distributed automated placement devices are interconnected through a network to form a resource-sharing pool, handling manufacturing orders from different sources and of different types. The aim is to improve the utilization rate of expensive manufacturing resources and production response speed through networked scheduling, achieving collaborative production across regions and enterprises. Existing shared intelligent manufacturing management solutions for composite materials typically employ optimization scheduling strategies based on Quality of Service (QoS). These solutions focus on macro-level indicators at the order level, such as minimizing completion time, reducing production costs, and maintaining historical yield rates. Such management systems often assign tasks based on equipment idle periods and rated parameters, attempting to find the optimal solution for resource allocation at a macro-statistical level, treating manufacturing tasks as standard discrete data packets for flow and allocation.

[0003] However, the actual AFP manufacturing process exhibits the following characteristics: First, the generation of defects at the microscopic level does not follow a common normal distribution, but rather displays a heavy-tailed distribution and extreme clustering: that is, defects are not uniformly and sparsely scattered on the component surface, but tend to be concentrated in complex geometric areas such as high curvature or large corners of the component, in an explosive manner; once a process deviation occurs, it is often not a single minor flaw, but rather a high probability of continuously inducing the accumulation of serious damage such as pores, wrinkles, or broken wires in that local area. Second, the equipment operating status at the macroscopic level exhibits long-range correlation and a self-organized critical state: that is, the equipment failure downtime behavior has a deep time memory effect, and the current downtime is often the result of long-term high-load operation in history, rather than an independent random event; this causes the entire shared manufacturing system to operate spontaneously on a near-saturation critical edge for a long time, and small disturbances (such as a complex geometric command) can easily break the fragile balance, inducing an avalanche-like chain of downtime or large-area quality fluctuations. Summary of the Invention

[0004] This invention provides a shared intelligent manufacturing management system for composite materials based on the Industrial Internet, which solves the technical problems mentioned in the background art.

[0005] This invention provides a composite material sharing intelligent manufacturing management system based on the Industrial Internet, which solves the technical problems mentioned in the background art.

[0006] This invention provides a shared intelligent manufacturing management system for composite materials based on the Industrial Internet, comprising: The layer manufacturing state feature extraction module is configured to quantify and characterize the local cumulative intensity of micro-layout defects, the temporal correlation degree of macro-equipment operation failures, and the defect connectivity risk degree of spatial geometric feature correlation based on real-time process data and historical quality inspection data of automated laying equipment, and further aggregate them into cross-layer critical coupling state parameters of the manufacturing system. The adaptive collaborative management and control decision module is configured to generate equipment task allocation priority weights and component geometric complexity allowable load limits based on the cross-layer critical coupling state parameters of the manufacturing system; adjust the probability of different automatic placement equipment undertaking manufacturing tasks according to the equipment task allocation priority weights; and dynamically constrain the geometric task intensity that the equipment is allowed to process in the current scheduling cycle according to the component geometric complexity allowable load limits, so as to control the evolution of the cross-layer critical state of the manufacturing system.

[0007] Furthermore, the quantification characterizes the local cumulative intensity of micro-layout defects, including: Within a set time window, the total number of defects generated by all laying tasks of each automated laying device is accumulated to construct a defect total sample sequence; several samples with the largest values ​​in the defect total sample sequence are selected and sorted; the average difference between the logarithm of the samples with the largest values ​​and the logarithm of the smallest sample is calculated, and the defect total tail exponent is obtained using the Hill estimator principle; the reciprocal of the defect total tail exponent is used as the local cumulative intensity of micro-layout defects to characterize the probability density of extreme defect events.

[0008] Furthermore, the quantitative characterization of the timing correlation of failures in macroscopic equipment operation includes: The total defect sequence of the automated deployment equipment is normalized to obtain a normalized defect intensity sequence. Rescaled range analysis is then performed on the normalized defect intensity sequence: the average value at different time scales is calculated, a cumulative deviation sequence is constructed, and the ratio of the range to the standard deviation of this cumulative deviation sequence is calculated. The logarithm of this ratio at multiple time scales is linearly fitted to the logarithm of the time scale, and the slope obtained from the fitting is used as the macroscopic correlation degree of equipment operation failure time series to characterize the long-range dependence of equipment operating status on the time series.

[0009] Furthermore, the defect connectivity risk degree associated with spatial geometric features is quantified, including: A geometric complexity field is constructed based on the curvature, fiber rotation angle variation, and thickness gradient of the component surface. This geometric complexity field is then combined with the actual detected defect metric values ​​to construct a defect weight field. The component surface is discretized into a graph structure, where nodes are discrete sampling points and edges connect adjacent sampling points. The edge weight connecting adjacent nodes is calculated as the average of the defect weight field values ​​of the two endpoints, thereby constructing a weighted adjacency matrix. The maximum eigenvalue of the weighted adjacency matrix is ​​calculated and normalized by comparing it with a preset reference eigenvalue. The normalized maximum eigenvalue is used as the defect connectivity risk degree associated with spatial geometric features to characterize the tendency of defects to form spatially penetrating damage under specific geometric driving.

[0010] Furthermore, the local cumulative intensity, fault timing correlation, and defect connectivity risk are aggregated into cross-layer critical coupling state parameters of the manufacturing system, including: A cross-layer state vector is constructed, comprising the local cumulative intensity of micro-layout defects, the temporal correlation degree of macro-equipment operation failures, the defect connectivity risk degree of spatial geometric feature correlation, and the average geometric complexity of product types. Each component in the cross-layer state vector is linearly normalized to map its value to a unit interval. A cross-layer coupling matrix is ​​defined to describe the mutual influence strength between indicators of each layer, and a weight vector is defined. The product of the weight vector, the cross-layer coupling matrix, and the normalized cross-layer state vector is calculated, and the product result is transformed by applying a sigmoid activation function. The scalar value obtained after transformation is used as the cross-layer critical coupling state parameter of the manufacturing system.

[0011] Furthermore, based on the cross-layer critical coupling state parameters of the manufacturing system, the priority weights for equipment task allocation and the allowable load limits for component geometric complexity are generated, including: For a given product type, a negative exponential weighted value based on the cross-layer critical coupling state parameters of the manufacturing system is calculated for all candidate automated placement equipment. This weighted value is then divided by the sum of the weighted values ​​of all candidate equipment to obtain a normalized equipment task allocation priority weight. The average value of the cross-layer critical coupling state parameters of the manufacturing system for all product types historically processed by each automated placement equipment is calculated. Based on this average value, a load scaling factor that is inversely proportional to the average value is calculated. The load scaling factor is multiplied by a preset baseline geometric complexity limit to obtain the allowable load limit of component geometric complexity for the equipment in the current scheduling cycle.

[0012] Furthermore, adjusting the probability of different automated placement devices undertaking manufacturing tasks based on the priority weight of the equipment task allocation includes: The priority weights of equipment task allocation calculated for a specific product type are used to construct a resource scheduling weight matrix for that product type across all available automated deployment equipment. When dispatching tasks, the probability of each automated deployment equipment being selected to execute that type of task is determined based on the resource scheduling weight matrix. The adjustment logic for the probability is as follows: the automated deployment equipment with the higher cross-layer critical coupling state parameter of the manufacturing system has, the lower its corresponding equipment task allocation priority weight.

[0013] Furthermore, based on the allowable load limit of the component's geometric complexity, the intensity of the geometric tasks that the equipment is allowed to process within the current scheduling cycle is dynamically constrained to control the evolution of the critical state across layers of the manufacturing system, including: The allowable load limit of component geometric complexity for each automated placement device is summarized into a device geometric complexity load upper limit vector. During the production scheduling stage of the industrial internet platform, the sum of the geometric complexity of all tasks to be processed assigned to any automated placement device is constrained so that it does not exceed the allowable load limit of component geometric complexity for that device.

[0014] The beneficial effects of this invention are as follows: by establishing a quantitative control system for the cross-layer self-organized critical coupling mechanism in composite material manufacturing; by compressing the heavy-tailed distribution attributes of micro-defects, the long-range correlation attributes of macro-equipment operation, and the seepage potential of spatial geometry into unified critical state parameters, the system can automatically reduce the probability of high-risk equipment undertaking complex tasks and dynamically tighten its load boundary without human intervention, thereby blocking the nonlinear evolution of cross-layer risks at the source, effectively preventing avalanche-style shutdowns or batch quality accidents induced by complex geometric tasks, and significantly improving the robustness and safety of the shared intelligent manufacturing system under extreme working conditions. Attached Figure Description

[0015] Figure 1 This invention provides a flowchart of a shared intelligent manufacturing management process for composite materials based on the Industrial Internet. Figure 2 This is a schematic diagram illustrating a specific implementation of the present invention. Detailed Implementation

[0016] The subject matter described herein will now be discussed with reference to exemplary embodiments. It should be understood that these embodiments are discussed only to enable those skilled in the art to better understand and implement the subject matter described herein, and changes may be made to the function and arrangement of the elements discussed without departing from the scope of this specification. Various processes or components may be omitted, substituted, or added as needed in the examples. Furthermore, features described in some examples may be combined in other examples.

[0017] like Figure 1As shown, a composite material sharing and intelligent manufacturing management system based on the Industrial Internet includes: The layer manufacturing state feature extraction module is configured to quantify and characterize the local cumulative intensity of micro-layout defects, the temporal correlation degree of macro-equipment operation failures, and the defect connectivity risk degree of spatial geometric feature correlation based on real-time process data and historical quality inspection data of automated laying equipment, and further aggregate them into cross-layer critical coupling state parameters of the manufacturing system. The adaptive collaborative management and control decision module is configured to generate equipment task allocation priority weights and component geometric complexity allowable load limits based on the cross-layer critical coupling state parameters of the manufacturing system; adjust the probability of different automatic placement equipment undertaking manufacturing tasks according to the equipment task allocation priority weights; and dynamically constrain the geometric task intensity that the equipment is allowed to process in the current scheduling cycle according to the component geometric complexity allowable load limits, so as to control the evolution of the cross-layer critical state of the manufacturing system.

[0018] Preferably, quantifying the local cumulative intensity of micro-layout defects includes: Configure the cross-layer manufacturing state feature extraction module to accumulate the total number of defects generated by all layup tasks on each automated layup machine r within a time window index k. : ; Where p is the workpiece identifier. Let k be the set of sampling points processed by device r on workpiece p within window k. This is the defect measurement value at sampling point s; Select the total number of defect sample sequences The largest m values ​​in a given set, after sorting, are denoted as: Calculate the tail index of total defects : ; Calculate the local cumulative strength of the micro-layout defects. : ; Where m is the maximum number of samples involved in the estimation, and ln is the natural logarithm.

[0019] It should be noted that the automated deployment equipment identifier is a unique identifier used to distinguish multiple automated deployment devices deployed in a distributed manner within the shared intelligent manufacturing system. A combination of letters and numbers is preferred, facilitating the system's individual statistical analysis and management of process data and operating status for different devices.

[0020] The time window index is used to divide the statistical period for total defects, avoiding trend distortion caused by data mixing across time periods. It is preferably a positive integer, and the value is determined based on the production rhythm; for example, a time window may be defined as 2 hours, with the index starting from 1 and incrementing sequentially.

[0021] Workpiece identification is used to distinguish different composite material components processed by automated placement equipment. The preferred value is a combination of numbers or alphanumeric symbols arranged in the order of production orders. The selection of the value is based on the ease of associating information such as the component's processing parameters and defect data, thereby achieving full-process traceability.

[0022] The defect measurement value at the sampling point is raw data that directly reflects the severity of local defects in a workpiece. It can be collected using visual inspection equipment, ultrasonic inspection equipment, or infrared inspection equipment. When collecting the data, quantification standards need to be set for different defect types. For example, porosity defects are quantified by area ratio, and wrinkle defects are quantified by height difference.

[0023] The sampling point set is the collection of all discrete points used for defect detection during the processing of a specified workpiece by a specified device within a time window. The sampling path can be preset using the 3D model of the component and determined synchronously with the device's processing trajectory. The density of sampling points needs to be adjusted according to the geometric complexity of the component, with higher density in complex areas than in simple areas.

[0024] The maximum number of samples used for estimation is the number of the top m largest samples selected from the total defect sample sequence. The preferred value is an integer between 20 and 50, chosen to balance estimation accuracy and data representativeness. When the total defect sample sequence length is 100, 30 can be selected as the m value to ensure sufficient capture of the heavy-tailed characteristics of the total defect count.

[0025] Total defects is the sum of defect data generated by all deployment tasks of a specified equipment within a specified time window. It is a comprehensive indicator that integrates defect data from different workpieces and different sampling points within that time period.

[0026] The sorted maximum sample is the first m largest sample obtained after sorting the total defect sample sequence in descending order, where the first one is the largest sample and the mth one is the minimum value among the first m samples.

[0027] The defect total tail index is a parameter calculated based on the Hill estimator principle. It is used to reflect the heavy-tailed distribution characteristics of the total defect amount, and its value is inversely proportional to the thickness of the distribution tail.

[0028] The local cumulative intensity of micro-layout defects is the reciprocal of the tail exponent of the total defect amount, and is used to characterize the probability density of extreme defect events.

[0029] It's important to note the quantification logic behind setting time windows and accumulating the total defects of all tasks deployed by the equipment within those windows to construct a sample sequence of total defects: The time window must match the equipment's processing rhythm. For example, in batch production scenarios, the processing time of a production batch can be set as a time window to ensure that the defect data within each window reflects the defect occurrence under stable operating conditions. For instance, if the equipment completes processing a batch of workpieces every 3 hours, then 3 hours is set as a time window, and the total defects of all workpieces within that window are accumulated to avoid data randomness caused by a small number of workpieces in a single batch.

[0030] It should be noted that the method of selecting the largest m samples in the total defect sample sequence, calculating the average difference between their logarithmic values ​​and the logarithmic value of the smallest sample, and combining this with the Hill estimator principle to obtain the tail exponent of the total defect is as follows: the Hill estimator has good adaptability to the tail characteristics of heavy-tailed distributions. The largest m samples are selected because the core feature of heavy-tailed distributions is reflected in the extreme values ​​at the tail, and these extreme values ​​are directly related to the occurrence pattern of extreme defect events.

[0031] It should be noted that defining the reciprocal of the tail exponent of the total defect count as the local cumulative intensity of micro-layout defects is a way to characterize the probability density of extreme defect events. The smaller the tail exponent, the more pronounced the heavy-tailed characteristic of the total defect count, and the higher the probability of extreme defect events. Its reciprocal transforms this inverse relationship into a positive one, facilitating a more intuitive assessment of risk levels in the subsequent system. For example, a tail exponent of 2 corresponds to a local cumulative intensity of 0.5; a tail exponent of 1.5 corresponds to a local cumulative intensity of 0.67, with the latter indicating a higher probability of extreme defects.

[0032] It should be noted that the duration of the time window can be determined based on the production rhythm and equipment processing efficiency, with a preference for 1 to 4 hours. The sliding step size is consistent with the window duration, and the trigger condition is that the first time window is triggered when the equipment starts continuous processing tasks. For example, if the equipment starts processing at 8:00 a.m. every day, and the window duration is set to 2 hours, then the first window is from 8:00 to 10:00 a.m., the second is from 10:00 to 12:00 p.m., and so on, to ensure that the processing conditions within each window are relatively stable.

[0033] It should be noted that the value of m must satisfy the condition that the sample sequence length is greater than or equal to 2m, and the preferred value is 20 to 50. When the sample sequence length is n, m can be selected from 1 / 5 to 1 / 2 of n. For example, when the sample sequence length is 100, m is 20 to 50; when the sample sequence length is 80, m is 16 to 40, to ensure that there are enough extreme samples to support the tail exponent estimation.

[0034] It should be noted that pore defects are quantified as a proportion of the defect area to the area of ​​the sampling point, with a value ranging from 0 to 1. For example, if the area of ​​the sampling point is 10 square millimeters and the area of ​​the pore defect is 2 square millimeters, the quantification value is 0.2. Wrinkle defects are quantified as a proportion of the wrinkle height to the design thickness of the component, with a value ranging from 0 to 1. Broken fiber defects are quantified as a proportion of the number of broken fibers to the total number of fibers in the sampling point area, with a value ranging from 0 to 1.

[0035] It should be noted that during the automated layup of composite materials, the generation of defects is affected by various factors such as geometric complexity and fluctuations in process parameters. Their distribution does not follow a normal distribution but exhibits heavy-tailed characteristics, meaning that extreme defect events with low probability but high impact occur frequently. The tail exponent estimated by the Hill estimator has a high correlation with the actual frequency of extreme defect occurrences, accurately capturing the heavy-tailed characteristics of the defect distribution in this scenario, and is therefore suitable for this application.

[0036] It should be noted that, considering the characteristics of heavy-tailed distribution and extreme clustering of micro-defects during the automated layup of composite materials, a method is used to quantify the probability density of extreme defect events by accumulating the total number of defects within a set time window, screening extreme samples, calculating the tail exponent using a Hill estimator, and then using its reciprocal as the local cumulative intensity. This design can accurately capture the accumulation pattern of defects at the microscopic level, avoiding the risk of misjudgment caused by traditional statistical methods neglecting the heavy-tailed characteristic.

[0037] Preferably, the quantitative characterization of the fault timing correlation of macroscopic equipment operation includes: Configure the cross-layer manufacturing state feature extraction module to extract the total number of defects. Normalized to a continuous defect intensity sequence And perform rescaled range analysis on the sequence length n; Calculate the cumulative deviation sequence : ; in, The average value of the sequence; Calculate the range with standard deviation : ; ; For multiple scales and Performing a linear fit yields the slope of the Hearst exponent. ; Define the timing correlation of the macroscopic equipment operation failure. for: ; in, The normalized sequence value.

[0038] The normalized defect intensity sequence is a continuous sequence with values ​​ranging from 0 to 1 obtained by normalizing the total defect sequence of automated deployment equipment. It is used to eliminate the magnitude differences between different equipment or tasks.

[0039] The sequence length is the number of data points in the normalized defect intensity sequence, consistent with the length of the total defect sample sequence, corresponding to the total number of time windows. It is preferably a positive integer greater than or equal to 30, chosen to ensure sufficient data volume to support recalibrated range analysis and avoid distortion of analysis results due to insufficient data.

[0040] The sequence index is used to identify each data point in the normalized defect intensity sequence. It is preferably a positive integer starting from 1, chosen to facilitate sequential association of all elements in the sequence over time, ensuring the ordered calculation of cumulative deviation, range, etc.

[0041] The sequence mean is the arithmetic mean of all data points in the normalized defect intensity sequence. It serves as a benchmark for calculating the cumulative bias and is used to eliminate the influence of the overall sequence offset on the bias analysis.

[0042] The cumulative deviation sequence is a sequence obtained by summing the differences between each data point in the normalized defect intensity sequence and the sequence mean. It is used to reflect the cumulative deviation effect of the defect intensity at different index positions relative to the mean.

[0043] The range is the difference between the maximum and minimum values ​​in the cumulative deviation sequence. It is used to characterize the fluctuation range of the cumulative deviation and reflects the overall fluctuation amplitude of the sequence in the time dimension.

[0044] The standard deviation is the square root of the average of the squares of the differences between each data point in the normalized defect intensity sequence and the sequence mean. It is used to characterize the dispersion of the sequence itself and eliminate the influence of magnitude.

[0045] The Hearst exponent is a slope obtained by linearly fitting the logarithm of the ratio of the range to the standard deviation at multiple time scales to the logarithm of the time scale. It is used to characterize the long-range dependence of a sequence.

[0046] The macroscopic equipment operation failure time-series correlation degree is an indicator directly equivalent to the Hearst exponent. It is used to quantify the long-range dependence of the operating status of automated deployment equipment in the time series and reflect the time memory effect of equipment failure downtime behavior.

[0047] It should be noted that the overall scheme of normalizing the total defect sequence of the automated placement equipment into a defect intensity sequence and quantifying the long-term dependence of the equipment's operating status through rescaled range analysis is as follows: the total defect sequence may have differences in magnitude due to differences in equipment model and processing tasks, and direct analysis will be subject to scale interference. After normalization, the sequences under different scenarios can be made uniform and comparable.

[0048] It should be noted that the method of constructing a cumulative deviation sequence, calculating the ratio of its range to standard deviation, and then linearly fitting the ratio to the logarithm of the time scale under multiple time scales, defining the slope as the fault time sequence correlation degree, is as follows: the cumulative deviation sequence can intuitively reflect the cumulative deviation of defect intensity in each period, the range reflects the fluctuation range, the standard deviation reflects the dispersion, and the ratio of the two can normalize the fluctuation range.

[0049] It's important to clarify that directly equating the Hearst exponent with the temporal correlation of macroscopic equipment failures to characterize the time memory effect of equipment downtime behavior relies on the following mapping logic: the numerical range of the Hearst exponent is directly related to long-range dependency characteristics. When the value is greater than 0.5, the sequence exhibits persistence, meaning that the equipment's past high-defect operating state is highly likely to continue. For example, a Hearst exponent of 0.7 indicates strong persistence in the equipment's operating state; the impact of historical high-load operation will continue to affect the present and future, corresponding to a failure temporal correlation of 0.7, which accurately quantifies this time memory effect.

[0050] It should be noted that the minimum-maximum normalization method is preferred. The calculation formula is: the normalized value equals the difference between the original value and the minimum value, divided by the difference between the maximum value and the minimum value. This method is chosen because it can strictly map the sequence to the interval between 0 and 1, is computationally simple, and has a clear physical meaning, making it suitable for non-negative numerical sequences such as the total defect count. For example, if the minimum value of the total defect count sequence is 10, the maximum value is 50, and an original value is 30, then the normalized value will be 0.5.

[0051] It should be noted that the number of time scales should preferably be 5 to 10, with a span ranging from 1 / 10 to 1 / 2 of the sequence length. The selection rule is to choose them in a proportionally increasing manner. For example, when the sequence length is 100, the time scales can be 10, 20, 30, 40, and 50 to ensure coverage of the fluctuation characteristics of different short, medium, and long time periods and to fully capture the long-range dependence of the sequence.

[0052] It should be noted that the criteria for determining the sequence length are as follows: the sequence length must be greater than or equal to 30, and at least twice the maximum time scale. This criterion is related to the time window index. For example, if each time window is 2 hours long, and 30 data points are needed, then a total of 60 hours of defect data must be accumulated to ensure that the sequence length meets the minimum effective requirement for rescaled range analysis.

[0053] It should be noted that the accuracy requirements for linear fitting and the handling of outliers are as follows: The goodness of fit for linear fitting should ideally be greater than or equal to 0.85 to ensure that the fitting result effectively reflects the linear relationship between the two data points. Outliers are handled using the three-standard-deviation method. When the ratio at a certain time scale exceeds the range of the sequence mean plus or minus three standard deviations, the data point is removed, and a new adjacent time scale is selected to supplement it, thus avoiding outliers affecting the fitting accuracy.

[0054] It should be noted that, regarding the long-range correlation and self-organized critical state exhibited by the automated composite material placement equipment, the total defect sequence is normalized to eliminate magnitude interference. Then, through rescaled range analysis, a cumulative deviation sequence is constructed, and the ratio of range to standard deviation is calculated. Combined with multi-timescale linear fitting, the Hearst exponent is obtained and used as the fault time-series correlation degree, thus quantifying the time memory effect of equipment operating status. This design can accurately capture the cumulative characteristics of equipment faults at the macro level, avoiding the limitation of traditional scheduling ignoring the impact of historical equipment operation. It provides reliable macro-indicators for the aggregation of cross-layer critical state parameters, enabling the system to predict the risk of equipment failure due to historical high loads in advance, and providing data support for the dynamic management and control of the manufacturing system.

[0055] Preferably, the quantification of the defect connectivity risk degree associated with spatial geometric features includes: Configure the cross-layer manufacturing state feature extraction module, first constructing a geometric complexity field. : ; in, For point curvature at that point For fiber rotation angle changes, For thickness gradient, These are weighting coefficients; Combined with defect detection data Constructing a defect weight field : ; in, To prevent small positive numbers with a denominator of zero; Construct a weighted adjacency matrix Its elements To connect adjacent points and Boundary weight: ; Calculate the defect connectivity risk associated with the spatial geometric features. : ; in, The largest eigenvalue of the adjacency matrix. Used as a reference eigenvalue.

[0056] Product type identification is used to distinguish composite material component types with different geometric properties and processing requirements. The preferred value is a combination of letters and numbers, chosen to facilitate the association of product attributes with equipment status and risk calculations, enabling differentiated management of different product types.

[0057] The geometric complexity field is a comprehensive index constructed by weighted summation based on the curvature, fiber rotation angle change, and thickness gradient of the component surface sampling points. It is used to quantify the geometric complexity of local areas of the component.

[0058] The curvature weighting coefficient is a coefficient that assigns a degree of influence to the geometric complexity field of a component's surface curvature. A value of 0.4 is preferred, based on experience with composite material layup processes, where curvature has a higher weighting on the impact of curvature on defect formation than other geometric factors.

[0059] The fiber angle variation weighting coefficient assigns a degree of influence to the geometric complexity field based on the fiber angle variation. A value of 0.3 is preferred, chosen to balance the influence of various geometric factors; the influence of fiber angle variation is second only to curvature.

[0060] The thickness gradient weighting coefficient is a coefficient that assigns a degree of influence of the thickness gradient on the geometric complexity field. A value of 0.3 is preferred, based on the premise that the influence of the thickness gradient on defects is relatively mild, consistent with the weighting of fiber rotation changes, and the sum of the three is 1.

[0061] Curvature is a geometric parameter describing the degree of bending at sampling points on the surface of a component. It can be positive or negative, corresponding to different bending directions. It can be extracted from the component's 3D model file (such as STL format) and its specific value can be obtained using the curvature analysis function of 3D modeling software.

[0062] Fiber turn angle variation is the difference in fiber layup angle between a sampling point on the component surface and its adjacent sampling points, characterizing the degree of abrupt change in fiber layup direction. It can be extracted from the layup path planning file of an automated layup device or obtained by analyzing the fiber texture on the component surface using image recognition technology.

[0063] Thickness gradient is the rate of change of thickness along the sampling direction at a sampling point on the surface of a component, characterizing the degree of abrupt change in the local thickness of the component. It can be obtained by measuring the thickness of each sampling point and adjacent points using an ultrasonic thickness gauge, and calculating the ratio of the thickness difference to the distance.

[0064] Discrete sampling points are discrete points on the surface of a component used for defect detection and geometric parameter acquisition. They can be obtained by pre-setting a sampling grid based on the component's 3D model and then laying them out according to the principles of equal spacing or densification in complex areas.

[0065] Adjacent discrete sampling points are two sampling points on the surface of a component that are within a preset threshold distance. This can be determined by calculating the Euclidean distance between all sampling points and filtering out pairs of sampling points with a distance less than 5 millimeters.

[0066] The defect weight field is an index constructed by combining the geometric complexity field with the actual defect measurement value, and is used to quantify the severity of defects under specific geometric conditions.

[0067] The small positive number used to prevent the denominator from being zero is a constant set to avoid the formula becoming meaningless when the value of the geometric complexity field is zero. It is preferably chosen to be between 10 to the power of -6 and 10 to the power of -4, based on the principle that it does not affect the accuracy of the calculation result while effectively avoiding the case where the denominator is zero.

[0068] The weighted adjacency matrix is ​​a matrix constructed by discretizing the surface of a component into a graph structure and using the edge weights of adjacent sampling points as elements. It is used to characterize the spatial distribution of defects.

[0069] Edge weights are the elements in the weighted adjacency matrix that connect adjacent sampling points. They are equal to the average of the defect weight field values ​​of the two endpoints and are used to quantify the correlation strength of defects between adjacent sampling points.

[0070] The largest eigenvalue is the largest eigenvalue in the weighted adjacency matrix, used to characterize the connectivity strength of the graph structure.

[0071] The preset reference eigenvalue is a benchmark value used to normalize the largest eigenvalue of the weighted adjacency matrix. The preferred value is the statistical average of the largest eigenvalues ​​of the weighted adjacency matrix of similar components, chosen to ensure that the normalization results have a unified comparison benchmark.

[0072] The defect connectivity risk degree associated with spatial geometric features is the normalized result of the maximum eigenvalue of the weighted adjacency matrix and the preset reference eigenvalue. It is used to characterize the tendency of defects to form spatially penetrating damage under specific geometric driving conditions.

[0073] It should be noted that the geometric complexity field is constructed by weighted summation based on three dimensions: surface curvature, fiber rotation angle change, and thickness gradient. During the composite material laying process, curvature, fiber rotation angle change, and thickness gradient all affect the generation and diffusion of defects, and the three together determine the processing difficulty of the region.

[0074] It should be noted that the quantitative logic for constructing the defect weight field by combining the geometric complexity field with the actual defect measurement value is as follows: the higher the geometric complexity of a region, the more likely the defects are to be concentrated and the higher the risk of connectivity. Therefore, the geometric characteristics need to be correlated with the actual severity of the defects.

[0075] It should be noted that the scheme of discretizing the component surface into a graph structure and using the average value of the defect weight field of adjacent sampling points as the edge weight to construct a weighted adjacency matrix is ​​as follows: it is difficult to directly quantify the defect connectivity on the continuous component surface. After discretizing into a graph structure, nodes represent sampling points, edges represent adjacency relationships, and edge weights quantify the defect association strength.

[0076] It should be noted that the normalized result of the maximum eigenvalue and the preset reference eigenvalue of the weighted adjacency matrix is ​​used as the indicator definition of the defect connectivity risk: the maximum eigenvalue can effectively characterize the connectivity strength of the matrix structure, and the larger the value, the easier it is for the defect to form a connection through the adjacent regions.

[0077] It should be noted that the weighting coefficients were determined through orthogonal experimental calibration. Different types of composite material components were selected, and layup tests were conducted under the same process parameters. The correlation between the geometric complexity field and the actual amount of defects generated under different combinations of weighting coefficients was statistically analyzed. Finally, the combination of curvature weighting coefficient 0.4, fiber rotation change weighting coefficient 0.3, and thickness gradient weighting coefficient 0.3 was determined.

[0078] It should be noted that the reference eigenvalues ​​are obtained based on historical data statistics of a large number of similar components. More than 100 composite material components of the same type and specification are selected, weighted adjacency matrices are constructed and the maximum eigenvalue is calculated. The arithmetic mean of these eigenvalues ​​is taken as the preset reference eigenvalue to ensure that the normalization result has universality.

[0079] It should be noted that the sampling point density is determined according to the component accuracy requirements. One sampling point is placed per square centimeter in simple geometric areas and three sampling points are placed per square centimeter in complex geometric areas. The threshold for determining adjacent sampling points is 5 millimeters, that is, when the Euclidean distance between two sampling points is less than 5 millimeters, they are determined to be adjacent.

[0080] It should be noted that small positive numbers are preferred to be 10 to the power of -5, with a range of 10 to the power of -6 to 10 to the power of -4. The reason for this choice is that the values ​​in this range are small enough that they will not affect the calculation accuracy of the defect weight field, and at the same time, they can completely avoid the situation where the denominator is zero when the value of the geometric complexity field is zero.

[0081] It should be noted that the combination of the defect weight field and the geometric complexity field means that defects in geometrically complex regions have a greater impact on structural safety, and the risk weight needs to be amplified by the weight field. Through comparative experiments, it was found that under the same defect metric, the matching degree between regions with high geometric complexity and the actual defect connectivity probability is improved after the weight field is corrected.

[0082] To address the issue of through-damage caused by defects driven by spatial geometric characteristics during composite material layup, this paper constructs a geometric complexity field by integrating three major geometric factors: component surface curvature, fiber rotation angle variation, and thickness gradient. This is combined with actual defect metrics to form a defect weight field. Furthermore, the component surface is discretized into a graph structure, and the defect connectivity risk is quantified using the maximum eigenvalue of the weighted adjacency matrix. This design can accurately capture the potential tendency of defect connectivity in the spatial dimension, overcoming the shortcomings of traditional control methods that neglect geometric driving factors. It provides reliable spatial index support for the critical coupling state parameters across layers of the manufacturing system, enabling the system to predict defect connectivity risks under specific geometric conditions in advance.

[0083] Preferably, the local cumulative intensity, fault timing correlation, and defect connectivity risk are aggregated into cross-layer critical coupling state parameters of the manufacturing system, including: Configure the cross-layer manufacturing state feature extraction module to construct a cross-layer state vector for equipment r and product type c. And by linear normalizing it, we obtain : ; ; in, , , These are the normalized local cumulative intensity of microscopic placement defects, the temporal correlation degree of macroscopic equipment operation failures, and the defect connectivity risk degree associated with spatial geometric features. This represents the normalized average geometric complexity of the product type. Calculate the cross-layer critical coupling state parameters of the manufacturing system. : ; Where M is the cross-layer coupling matrix, and w is the weight vector. For the Sigmoid function: ; Results This is a scalar used to characterize the critical hazard level of a system.

[0084] The cross-layer state vector is a four-dimensional vector that integrates the local cumulative intensity of micro-layout defects, the temporal correlation degree of macro-equipment operation failures, the connectivity risk degree of spatial geometric feature-related defects, and the average geometric complexity of product types. It is used to centrally carry the core indicators of cross-layer multi-dimensionality.

[0085] The average geometric complexity of a product type is the comprehensive average of the geometric complexities of all components under a specific product type, used to characterize the overall geometric processing difficulty of that type of product.

[0086] The normalized cross-layer state vector is a vector obtained by linearly normalizing each component of the cross-layer state vector, ensuring that the values ​​of each index are mapped to the interval between 0 and 1.

[0087] The normalized local cumulative intensity of micro-layout defects is the result of linear normalization of the local cumulative intensity of micro-layout defects, eliminating the difference in magnitude of the original index.

[0088] The normalized macroscopic equipment operation failure time-series correlation degree is the result of linear normalization of the macroscopic equipment operation failure time-series correlation degree, achieving comparability with other indicators.

[0089] The normalized spatial geometric feature associated defect connectivity risk degree is the result of linear normalization of the spatial geometric feature associated defect connectivity risk degree, with a unified value range.

[0090] The normalized average geometric complexity of product types is the result of linearly normalizing the average geometric complexity of product types, which is suitable for aggregate calculation requirements.

[0091] A cross-layer coupling matrix describes the strength of mutual influence between indicators at the micro, macro, spatial, and product attribute levels. It is preferably a 4x4 matrix with elements ranging from 0 to 1. The values ​​are determined by calibrating the influence strength between indicators through orthogonal experiments; for example, the influence coefficient of a micro-defect indicator on a macro-failure indicator is set to 0.3.

[0092] The weight vector is a one-dimensional vector that reflects the contribution of each component of the cross-level state vector to the critical state of the system. It is preferably a 1-row, 4-column vector with elements ranging from 0 to 1 and a sum of 1. The values ​​are determined by combining the analytic hierarchy process (AHP) with process experience. For example, the weight of the local cumulative intensity of micro-defects is set to 0.35.

[0093] The sigmoid activation function is a nonlinear function that maps any real number to the interval between 0 and 1, and is used to transform aggregated calculation results into scalars with a uniform range.

[0094] The critical coupling state parameters of a manufacturing system across layers are scalars obtained after weight vector, cross-layer coupling matrix operations, and sigmoid activation function transformation, and are used to characterize the critical risk of the manufacturing system.

[0095] The natural constant is approximately 2.71828, which is used to calculate the sigmoid activation function.

[0096] The input variable of the S-shaped function is the product of the weight vector, the cross-layer coupling matrix, and the normalized cross-layer state vector, which is an intermediate input variable of the activation function.

[0097] It should be noted that the selection logic for constructing a cross-layer state vector containing four-dimensional indicators is as follows: the critical risk in composite material manufacturing stems from the combined effects of microscopic defects, macroscopic equipment condition, spatial geometric characteristics, and the inherent properties of the product itself; none of these four factors can be omitted. For example, in carbon fiber components used in aerospace, the accumulation of microscopic defects may lead to localized failures, macroscopic equipment malfunctions can cause production interruptions, complex spatial geometry makes defect connections more likely, and the product type determines the overall processing difficulty. These four factors together constitute the core influencing factors of the system's critical risk.

[0098] It should be noted that after normalizing the cross-layer state vector, the calculation framework for the critical coupled state parameters is formed by combining the product of the weight vector, the cross-layer coupling matrix, and the normalized vector with the S-shaped activation function: normalization ensures that each indicator participates in the calculation fairly, the coupling matrix quantifies the interaction between indicators, the weight vector highlights the contribution of key indicators, and the activation function constrains the output range.

[0099] It should be noted that the aggregation method, which compresses multi-dimensional, cross-layer indicators into a single scalar parameter, is effective because multi-dimensional indicators are difficult to use directly for system decision-making, while a single scalar can intuitively reflect the overall critical state of the system. For example, when the values ​​of indicators from different dimensions differ significantly, direct superposition can lead to distorted results, while the aggregation method described above can integrate scattered risk information into a unified indicator.

[0100] It should be noted that the matrix has a dimension of 4 rows and 4 columns, corresponding to the four components of the cross-layer state vector. The matrix elements were determined through a large number of orthogonal experiments. Historical data of 100 different types of products were selected, and the influence of changes in each indicator on other indicators was statistically analyzed. The influence was quantified into values ​​from 0 to 1 as matrix elements. For example, the influence coefficient of the local cumulative intensity of micro-defects on the time-series correlation of macro-equipment failures was determined by statistically analyzing the correlation coefficient of 0.3 between the two.

[0101] It should be noted that the hierarchical structure is determined by using the analytic hierarchy process (AHP) to construct a target layer (aggregation of critical state parameters), a criterion layer (four indicators), and a scheme layer (different weight combinations). Five to eight experts in the field of composite material manufacturing are invited to compare the importance of each indicator pairwise to construct a judgment matrix. After passing the consistency test, the weight vector is calculated. For example, if the experts determine that the local cumulative intensity of micro-defects is more important than the time-series correlation of macro-equipment failures, the corresponding element of the judgment matrix is ​​set to 2. Finally, the weight vector is calculated.

[0102] It should be noted that the minimum-maximum normalization formula is used. The normalized value is equal to the difference between the original value and the minimum value of the indicator, divided by the difference between the maximum value and the minimum value of the indicator.

[0103] It should be noted that this function can stably map any real number to the interval between 0 and 1, which is consistent with the value requirements of the system's critical risk level. It also has smooth gradient change characteristics, which can accurately reflect the subtle changes after the aggregation of cross-layer indicators and avoid sudden changes in the judgment of critical state.

[0104] It should be noted that the weighted average method is used, which calculates the weighted sum of the geometric complexity of all components by using the production batch size of each component under this product type as the weight, and then divides it by the total batch size.

[0105] It should be noted that, in order to address the problem of the dispersed multi-dimensional indicators of microscopic, macroscopic, spatial and product attributes in composite material manufacturing systems, which are difficult to integrate into a unified decision-making basis, a cross-layer state vector is constructed to integrate core indicators. After linear normalization to eliminate the difference in magnitude, the cross-layer coupling matrix is ​​used to quantify the mutual influence between indicators. Combined with weight vectors to highlight the contribution of key indicators, the aggregation result is finally mapped to a critical coupling state parameter within a unified range through an S-shaped activation function. This allows the dispersed cross-layer risk information to be compressed into a single scalar.

[0106] Preferably, based on the cross-layer critical coupling state parameters of the manufacturing system, the priority weights for equipment task allocation and the allowable load limits for component geometric complexity are generated, including: Configure the adaptive collaborative management and control decision module to calculate the priority weight of equipment task allocation for a given product type c and candidate automatic deployment equipment r. : ; in, To determine the critical coupling state parameters across layers of the manufacturing system. For the set of candidate devices, For sensitivity parameters; Calculate the historical average critical state parameters of the equipment : ; in, The set of product types that equipment r has historically processed; Calculate the geometric load scaling factor and allowable load limits for component geometric complexity : ; ; in, To scale the sensitivity parameters, The baseline geometric complexity limit preset for the system.

[0107] The priority weight for equipment task allocation is the ratio of the negative exponential weight of the candidate automated deployment equipment to the sum of the weights of all candidate equipment for a specified product type. It is used to characterize the priority of the equipment in undertaking this type of task.

[0108] The sensitivity parameter is a constant that adjusts the sensitivity of the device task allocation priority weight to the response of the critical coupling state parameter across layers. The preferred value is a real number between 0.5 and 2, determined through experimental calibration to balance the suppression of high-risk devices with the flexibility of task allocation; for example, 1.2 is used in mass production scenarios, and 0.8 is used in small-batch customization scenarios.

[0109] The candidate equipment set is a collection of automated placement equipment capable of processing specified product types. Priority is given to equipment groups selected based on their processing range and rated accuracy, with the selection criteria being that the equipment parameters must meet the geometric complexity and placement accuracy requirements for that product type.

[0110] The candidate device index is an identifier used to traverse all devices in the candidate device set. It is preferably a positive integer starting from 1, chosen to facilitate the calculation of the negative exponential weighting value for each candidate device, ensuring no weight calculation is missed.

[0111] The historical average critical state parameter of the equipment is the arithmetic mean of the cross-layer critical coupling state parameters of all product types processed by the equipment in the past, and is used to reflect the average risk level of the equipment in long-term operation.

[0112] The historical product type set is a collection of all product types processed since the equipment was put into use [collected parameters]. It can be extracted through the equipment's production record system, which stores information such as processing orders and product type identifiers for each piece of equipment.

[0113] The historical product type index is an identifier used to traverse the set of historically processed product types. It is preferably a positive integer starting from 1, chosen to facilitate the association of each product type in the set and ensure comprehensive calculation of the average parameter.

[0114] The number of historical processed product types is the number of product types contained in the historical processed product type set, and is used for mean-averaging in the calculation of average parameters.

[0115] The geometric load scaling factor is a coefficient that is inversely proportional to the historical average critical state parameters of the equipment, and is used to dynamically adjust the load-bearing capacity of the equipment.

[0116] The scaling sensitivity parameter is a constant that adjusts the sensitivity of the geometric load scaling factor to the historical average critical state parameter. It is preferably a real number between 0.3 and 1.5, chosen based on a combination of equipment durability and production efficiency; 1.2 is used for high-precision equipment, and 0.6 for general-purpose equipment.

[0117] The allowable load limit for component geometric complexity is the product of the geometric load scaling factor and the baseline geometric complexity limit, used to limit the total intensity of geometric tasks that the equipment is allowed to process within the current scheduling cycle.

[0118] The baseline geometric complexity limit is a system-preset reference value for the basic load of the equipment. The preferred value is a value determined based on the rated processing capacity of the equipment. The value is determined by the total geometric complexity that the equipment can stably process under standard operating conditions. For example, the baseline limit corresponding to the rated capacity of a certain piece of equipment is 5.0.

[0119] It should be noted that the method of calculating the normalized task allocation priority weight based on the negative exponential weighting value of the cross-layer critical coupling state parameter is as follows: the negative exponential function has a monotonically decreasing characteristic, which makes the weighting value of devices with higher critical parameters smaller, and the weighting value after normalization is lower, which fits the scheduling logic of high risk and low priority.

[0120] It should be noted that the logic for constructing a load scaling factor that is inversely proportional to the average critical parameters of all product types processed by the equipment in the past is as follows: the higher the historical average critical parameter of the equipment, the higher its long-term operating risk and the weaker its ability to bear complex tasks, and the load limit needs to be reduced by scaling factor.

[0121] It should be noted that the calculation scheme for obtaining the load limit by multiplying the load scaling factor by the baseline geometric complexity limit is as follows: the baseline limit is the basic load-bearing capacity of the equipment, and the scaling factor is dynamically adjusted according to historical risks to ensure that the load limit matches the actual capacity of the equipment while avoiding critical risks.

[0122] It should be noted that the sensitivity parameter ranges from 0.5 to 2. When production tasks are urgent and equipment redundancy is low, a value of 0.5 to 0.8 is used to reduce sensitivity and ensure timely task completion. When product quality requirements are high and equipment redundancy is high, a value of 1.5 to 2.0 is used to increase sensitivity and strengthen risk management. For example, 1.8 is used for aerospace component production, and 0.7 is used for mass production of civilian components.

[0123] It should be noted that the scaling sensitivity parameter ranges from 0.3 to 1.5. When the equipment has a short service life and stable performance, a value of 0.3 to 0.8 is used to appropriately relax the load constraint; when the equipment has been used for more than 5 years and its performance has degraded, a value of 1.0 to 1.5 is used to tighten the load constraint. For example, 0.5 is used for new equipment and 1.3 is used for old equipment.

[0124] It should be noted that the baseline geometric complexity limit is determined based on the equipment's rated processing capacity and the maximum load during historical stable operation, and is preferably set at 70% to 80% of the equipment's maximum tolerable geometric complexity. For example, if the equipment's maximum tolerable geometric complexity is 7.0, then the baseline limit is set at 5.0 (7.0 × 71.4%), which both preserves a safety margin and makes full use of the equipment's capacity.

[0125] It should be noted that the selection criteria for the candidate equipment set include the equipment's laying accuracy, the range of processed component sizes, and the historical pass rate for processing this type of product. For example, if a product type requires a laying accuracy of ±0.1 mm, then the candidate equipment set will only include equipment that meets the laying accuracy requirement and has a historical pass rate higher than 95%.

[0126] It should be noted that when the number of product types processed by the equipment in the past is less than 3, the historical average critical state parameter adopts a weighted average, with the weight of the most recently processed product type being 0.6 and the weight of the earlier ones being 0.4; when the number is not less than 3, the historical average critical state parameter adopts an arithmetic average to ensure that the average parameter can reflect the recent operating status of the equipment.

[0127] It should be noted that the cross-layer critical coupling state parameters of the manufacturing system are transformed into executable equipment task allocation priority weights and component geometric complexity allowable load limits. High-risk equipment is prioritized and downgraded through negative exponential weighting. A load scaling coefficient is constructed based on historical average critical parameters to dynamically match equipment capacity with risk levels. This design provides a quantitative decision-making basis for adaptive collaborative management, avoiding the limitations of traditional scheduling that relies solely on equipment idle states. It makes task allocation more closely aligned with actual equipment operating risks, and load limits prevent equipment from entering a critical state due to overload. This provides solid support for the dynamic optimization and risk management of the manufacturing system, ensuring the stability and continuity of production under the shared intelligent manufacturing model.

[0128] Preferably, adjusting the probability of different automated placement devices undertaking manufacturing tasks based on the priority weight of the equipment task allocation includes: Configure the adaptive collaborative management and control decision module to target product type The calculated set of device task allocation priority weights Construct as a resource scheduling weight matrix : ; The resource scheduling weight matrix elements in Directly as equipment Product types accepted Task allocation probability; pass The numerical distribution characteristics enable probability regulation: ; That is, when the critical coupling state parameters of the manufacturing system across layers. The probability of the device undertaking the task increases as the number of users increases. It decays exponentially, thereby reducing the risk of the system entering a critical state.

[0129] The resource scheduling weight matrix is ​​a matrix that integrates the task allocation priority weights of all available automated deployment devices for a specific product type. The priority values ​​are a matrix with rows representing the number of devices and columns representing the number of product types. The values ​​are chosen to cover the compatibility between all available devices and product types, ensuring that scheduling decisions are comprehensive.

[0130] The set of all available equipment is the overall collection of automated deployment equipment in the shared intelligent manufacturing system that is in normal operating condition and capable of undertaking manufacturing tasks. The priority value is the equipment group that has been screened by equipment status detection (such as operating parameters and fault records), and the selection is based on troubleshooting, maintenance, or equipment that does not meet performance standards to ensure the feasibility of task execution.

[0131] The "All Product Types" set is the collection of all composite material component types that the system has the capability to process. The preferred value is a set containing all registered product type identifiers, determined by covering the product types corresponding to all orders handled by the system, supporting full-category task scheduling.

[0132] The partial derivative of the priority weight with respect to the critical parameter is the rate of change of the priority weight of the equipment task allocation relative to the critical coupling state parameter of the manufacturing system across layers, and is used to verify the negative correlation between the two.

[0133] The priority weights for assigning tasks to devices of specific product types are constructed into a resource scheduling weight matrix. The matrix elements directly serve as a mapping method for the probability of a device undertaking that type of task. The weight matrix can integrate the adaptation information of all devices and product types in the system and can be directly used for task dispatching decisions without additional conversion.

[0134] By verifying the negative correlation between priority weight and cross-layer critical coupling state parameters through partial derivatives, the adjustment logic for the exponential decay of the acceptance probability of high-risk equipment is realized: the partial derivative result is always less than 0, proving that the higher the critical parameter, the smaller the weight, and the decay rate changes nonlinearly with the weight.

[0135] The resource scheduling weight matrix is ​​invoked in the task dispatch system in real time through real-time calculation and updating. When a device status (such as fault repair or new device access) or a new product type is added, the corresponding weight is immediately recalculated and the matrix is ​​updated. For example, after a new device is added to the system, the system automatically calculates its priority weight for each product type and adds it to the corresponding row of the matrix to ensure that scheduling decisions are synchronized with the actual resource status.

[0136] The specific execution method for task assignment is random sampling. The system generates a random number between 0 and 1, and determines the receiving device based on whether the random number falls within the interval corresponding to the weight of each device. For example, if the device weights are 0.3, 0.5, and 0.2, corresponding to intervals of 0-0.3, 0.3-0.8, and 0.8-1.0, and the random number is 0.45, then it will be assigned to the device with a weight of 0.5.

[0137] The minimum threshold for equipment acceptance probability is preferentially set to 0.05. This value is chosen to avoid resource idleness due to excessively low equipment acceptance probability, while also preventing high-risk equipment from accepting tasks. When the priority weight of an equipment is lower than 0.05, the system automatically removes it from the candidate equipment set for that product type, and reinstates it after the equipment's critical condition improves.

[0138] The priority weights for task allocation to equipment of specific product types are transformed into a resource scheduling weight matrix. The matrix elements directly represent the probability of equipment accepting tasks. Partial derivative verification ensures a negative correlation between critical parameters and acceptance probabilities, achieving an exponential decay in the probability of high-risk equipment accepting tasks. This transforms abstract weighting indicators into executable task assignment probabilities, avoiding subjective intervention in traditional scheduling. Task allocation becomes more aligned with the actual risk status of equipment, providing direct execution basis for dynamic management of the manufacturing system. This supports the system in autonomously avoiding critical risks and ensures the rationality of task allocation and the stability of the production process under the shared intelligent manufacturing model.

[0139] Preferably, the geometric task intensity that the equipment is allowed to process within the current scheduling cycle is dynamically constrained based on the allowable load limit of the component's geometric complexity, in order to control the evolution of the critical state across layers of the manufacturing system, including: Configure the adaptive collaborative management and control decision module to output a vector containing the allowable load limits of component geometric complexity for all available devices in the current scheduling cycle. : ; Where r represents the device index. For equipment collection, The limit for device r; The following constraints are applied during the scheduling phase: ; in, The set of tasks assigned to device r. Let p be the geometric complexity of component p; This constraint enables decritical control of high-risk equipment.

[0140] The equipment geometric complexity load limit vector is a vector formed by summing the allowable load limits of the component geometric complexity of all available automated placement equipment. It is used for unified constraint management in the production scheduling stage of the industrial internet platform.

[0141] The device index is an identifier used to distinguish different automated deployment devices in the device set. It is preferred to use a positive integer starting from 1, chosen to facilitate the association of each device with its corresponding load limit, ensuring no scheduling constraints are missed.

[0142] The equipment set is the overall collection of all automatically deployed equipment in the shared intelligent manufacturing system that is in normal operating condition and capable of undertaking tasks. The priority value is the equipment group selected after being screened by equipment status monitoring. The selection criteria are to eliminate faulty, under-maintained, or substandard equipment to ensure the feasibility of constraint execution.

[0143] The equipment load limit is the sum of geometric task intensity that a single automated placement device is allowed to process within the current scheduling cycle, and is obtained by multiplying the geometric load scaling factor and the baseline geometric complexity limit.

[0144] The set of tasks assigned to a device is the set of all component tasks assigned to a specific automated placement device for processing within the current scheduling cycle, and is determined by the production scheduling system based on load limits and task priorities.

[0145] Component geometric complexity is an indicator that quantifies the difficulty of geometric processing of a component. It can be obtained by extracting geometric features such as curvature, fiber rotation angle changes, and thickness gradients from the component's 3D model and then calculating them using a weighted summation formula.

[0146] The load limits of all available devices are aggregated into a device geometric complexity load upper limit vector, which is used for the system design of production scheduling constraints: the load upper limit vector can centrally present the bearing boundaries of all devices, which facilitates quick calling and unified verification by the scheduling system.

[0147] In the scheduling phase, a method for decritical control of the manufacturing system is achieved by constraining the sum of the geometric complexities of all tasks to be processed on a single machine to not exceed its load limit: the accumulation of geometric complexity of multiple tasks will add up the operating pressure of the equipment, and when the accumulated value exceeds the limit, the risk of cross-layer criticality of the system will increase.

[0148] The geometric complexity of the component is calculated using a weighted summation formula, with weighting coefficients of curvature 0.4, fiber rotation change 0.3, and thickness gradient 0.3.

[0149] The strategy for handling constraint conflicts during the production scheduling phase includes: when the sum of the geometric complexities of multiple tasks exceeds the equipment load limit, a task priority ranking strategy is adopted to prioritize the allocation of high-priority tasks and transfer low-priority tasks to other equipment with sufficient load. For example, if the load limit of equipment B is 3.0, and the priorities of the tasks to be allocated are from high to low as follows: Task 1 (complexity 1.2), Task 2 (complexity 1.0), Task 3 (complexity 0.9), and Task 4 (complexity 0.8), the sum of the first three tasks is 3.1, which slightly exceeds the limit. Therefore, the lowest priority task 4 is eliminated and transferred to equipment C.

[0150] The update cycle and triggering conditions for the load limit vector include: the update cycle is preferably one scheduling cycle (e.g., 2 hours), and the triggering conditions include changes in device status (e.g., fault repair, new device access), task completion, or addition. For example, if device D completes all tasks in the current scheduling cycle, the system immediately recalculates its load limit for the next scheduling cycle and updates it to the load limit vector.

[0151] The accuracy requirements for calculating the sum of geometric complexity include: the calculation result is rounded to three decimal places, and the absolute value of the difference between the sum and the load limit is less than 0.001, which is considered to meet the constraints.

[0152] It should be noted that the allowable load limits for the geometric complexity of each device's components are aggregated into a load upper limit vector. During the production scheduling phase of the industrial internet platform, the operating load of the equipment is controlled by constraining the sum of the geometric complexities of all tasks to be processed on a single device to not exceed the corresponding limit. This strictly limits the equipment load within a safe range, preventing the equipment from entering a critical state due to task overload. This provides execution assurance for the de-criticality control of the manufacturing system. Combined with the task allocation priority mechanism mentioned above, a dual control system of priority regulation and hard load constraints is formed, ensuring the rationality of the production scheduling and operational stability of the shared intelligent manufacturing system.

[0153] like Figure 2 As shown, Figure 2 The demonstration showcased a closed-loop management and control architecture for intelligent manufacturing of composite materials, with an industrial internet platform as its hub. The platform connects upwards to production management terminals and quality inspection systems to achieve global monitoring configuration and interaction of quality inspection data. Downwards, it aggregates real-time process data and historical quality inspection data from multiple automated placement devices (A, B, and C) at the bottom layer. Through comprehensive analysis of this multi-source data, the platform can calculate and feed back precise scheduling instructions and load constraints based on risk assessment to the equipment, thereby forming an automated collaborative mechanism that runs through data acquisition, analysis and decision-making, and execution control, enabling intelligent allocation of manufacturing tasks and dynamic optimization of equipment operating status.

[0154] The embodiments of this example have been described above. However, this example is not limited to the specific implementation methods described above. The specific implementation methods described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms based on the guidance of this example, and all of them are within the protection scope of this example.

Claims

1. A shared intelligent manufacturing management system for composite materials based on the Industrial Internet, characterized in that, include: The layer manufacturing state feature extraction module is configured to quantify and characterize the local cumulative intensity of micro-layout defects, the temporal correlation degree of macro-equipment operation failures, and the defect connectivity risk degree of spatial geometric feature correlation based on real-time process data and historical quality inspection data of automated laying equipment, and further aggregate them into cross-layer critical coupling state parameters of the manufacturing system. The adaptive collaborative management and control decision module is configured to generate equipment task allocation priority weights and component geometric complexity allowable load limits based on the cross-layer critical coupling state parameters of the manufacturing system; adjust the probability of different automatic placement equipment undertaking manufacturing tasks according to the equipment task allocation priority weights; and dynamically constrain the geometric task intensity that the equipment is allowed to process in the current scheduling cycle according to the component geometric complexity allowable load limits, so as to control the evolution of the cross-layer critical state of the manufacturing system.

2. The composite material sharing and intelligent manufacturing management system based on the Industrial Internet according to claim 1, characterized in that, Quantitative characterization of the local cumulative intensity of micro-layout defects includes: Within a set time window, the total number of defects generated by all laying tasks of each automated laying device is accumulated to construct a defect total sample sequence; several samples with the largest values ​​in the defect total sample sequence are selected and sorted; the average difference between the logarithm of the samples with the largest values ​​and the logarithm of the smallest sample is calculated, and the defect total tail exponent is obtained using the Hill estimator principle; the reciprocal of the defect total tail exponent is used as the local cumulative intensity of micro-layout defects to characterize the probability density of extreme defect events.

3. The composite material sharing and intelligent manufacturing management system based on the Industrial Internet according to claim 2, characterized in that, Quantitative characterization of the timing correlation of failures in macroscopic equipment operation includes: The total defect sequence of the automated deployment equipment is normalized to obtain a normalized defect intensity sequence. Rescaled range analysis is then performed on the normalized defect intensity sequence: the average value at different time scales is calculated, a cumulative deviation sequence is constructed, and the ratio of the range to the standard deviation of this cumulative deviation sequence is calculated. The logarithm of this ratio at multiple time scales is linearly fitted to the logarithm of the time scale, and the slope obtained from the fitting is used as the macroscopic correlation degree of equipment operation failure time series to characterize the long-range dependence of equipment operating status on the time series.

4. The composite material sharing and intelligent manufacturing management system based on the Industrial Internet according to claim 3, characterized in that, The degree of defect connectivity risk associated with spatial geometric features is quantitatively characterized, including: A geometric complexity field is constructed based on the curvature, fiber rotation angle variation, and thickness gradient of the component surface. This geometric complexity field is then combined with the actual detected defect metric values ​​to construct a defect weight field. The component surface is discretized into a graph structure, where nodes are discrete sampling points and edges connect adjacent sampling points. The edge weight connecting adjacent nodes is calculated as the average of the defect weight field values ​​of the two endpoints, thereby constructing a weighted adjacency matrix. The maximum eigenvalue of the weighted adjacency matrix is ​​calculated and normalized by comparing it with a preset reference eigenvalue. The normalized maximum eigenvalue is used as the defect connectivity risk degree associated with spatial geometric features to characterize the tendency of defects to form spatially penetrating damage under specific geometric driving.

5. The composite material sharing and intelligent manufacturing management system based on the Industrial Internet according to claim 4, characterized in that, The local cumulative intensity, failure timing correlation, and defect connectivity risk are aggregated into cross-layer critical coupling state parameters for the manufacturing system, including: A cross-layer state vector is constructed, comprising the local cumulative intensity of micro-layout defects, the temporal correlation degree of macro-equipment operation failures, the defect connectivity risk degree of spatial geometric feature correlation, and the average geometric complexity of product types. Each component in the cross-layer state vector is linearly normalized to map its value to a unit interval. A cross-layer coupling matrix is ​​defined to describe the mutual influence strength between indicators of each layer, and a weight vector is defined. The product of the weight vector, the cross-layer coupling matrix, and the normalized cross-layer state vector is calculated, and the product result is transformed by applying a sigmoid activation function. The scalar value obtained after transformation is used as the cross-layer critical coupling state parameter of the manufacturing system.

6. The composite material sharing and intelligent manufacturing management system based on the Industrial Internet according to claim 5, characterized in that, Based on the cross-layer critical coupling state parameters of the manufacturing system, the priority weights for equipment task allocation and the allowable load limits for component geometric complexity are generated, including: For a given product type, a negative exponential weighted value based on the cross-layer critical coupling state parameters of the manufacturing system is calculated for all candidate automated placement equipment. This weighted value is then divided by the sum of the weighted values ​​of all candidate equipment to obtain a normalized equipment task allocation priority weight. The average value of the cross-layer critical coupling state parameters of the manufacturing system for all product types historically processed by each automated placement equipment is calculated. Based on this average value, a load scaling factor that is inversely proportional to the average value is calculated. The load scaling factor is multiplied by a preset baseline geometric complexity limit to obtain the allowable load limit of component geometric complexity for the equipment in the current scheduling cycle.

7. A composite material sharing and intelligent manufacturing management system based on the Industrial Internet according to claim 6, characterized in that, Adjusting the probability of different automated placement devices undertaking manufacturing tasks based on the priority weight of the equipment task allocation includes: The priority weights of equipment task allocation calculated for a specific product type are used to construct a resource scheduling weight matrix for that product type across all available automated deployment equipment. When dispatching tasks, the probability of each automated deployment equipment being selected to execute that type of task is determined based on the resource scheduling weight matrix. The adjustment logic for the probability is as follows: the automated deployment equipment with the higher cross-layer critical coupling state parameter of the manufacturing system has, the lower its corresponding equipment task allocation priority weight.

8. The composite material sharing and intelligent manufacturing management system based on the Industrial Internet according to claim 6, characterized in that, Based on the allowable load limit of the component's geometric complexity, dynamic constraints are applied to the geometric task intensity that the equipment is allowed to process within the current scheduling cycle, in order to control the evolution of the critical state across layers of the manufacturing system, including: The allowable load limit of component geometric complexity for each automated placement device is summarized into a device geometric complexity load upper limit vector. During the production scheduling stage of the industrial internet platform, the sum of the geometric complexity of all tasks to be processed assigned to any automated placement device is constrained so that it does not exceed the allowable load limit of component geometric complexity for that device.

Images