Intelligent knapsack fabric simulation and optimization system based on composite high polymer material

Abstract

The invention discloses an intelligent simulation and optimization system for a knapsack fabric based on a composite high polymer material, relates to the technical field of high polymer compounding, and solves the problems that the performance of a traditional composite high polymer knapsack fabric is difficult to regulate and control uniformly under multiple working conditions, technological parameters depend on experience, the optimization efficiency is low and the like. Comprising the steps of multi-scale data acquisition, attention mechanism fusion modeling, adaptive PID control of fuzzy parameter adjustment and constraint perception hybrid evolutionary algorithm optimization. The method has the technical effects that high-precision prediction of the fatigue life and the wear resistance of the material and real-time compensation control of process parameters are realized, a multi-target optimal solution is obtained under complex constraints, and the product consistency and the production intelligence level are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of polymer composites, and more specifically, to an intelligent simulation and optimization system for backpack fabrics based on composite polymer materials. Background Art

[0002] With the rapid development of outdoor sports equipment and functional textiles, backpack fabrics based on composite polymer materials are facing increasingly demanding performance requirements. Traditional backpack fabrics mainly use single materials or simple laminated structures, making it difficult to simultaneously meet comprehensive requirements such as lightweight, abrasion resistance, and environmental adaptability. In recent years, the multi-layer co-extrusion composite process represented by TPU (thermoplastic polyurethane) and PET (polyester) has become the industry mainstream, achieving an optimized combination of material properties through temperature gradient control and interlayer pressure regulation.

[0003] In the prior art, mainly two solutions are adopted: First, a convolutional neural network (CNN) is trained with laboratory standard test data for fatigue life prediction. This method performs well under constant load conditions, but when simulating dynamic complex loads in actual use (such as multi-angle impacts, temperature and humidity alternation), due to the lack of a cross-scale data fusion mechanism, there are significant deviations between the prediction results and the actual working conditions. Second, the NSGA-I multi-objective optimization algorithm is used to balance material cost and performance parameters. Although it can theoretically optimize tensile strength and weight, when dealing with a high-dimensional parameter space (including extrusion temperature, layer thickness ratio, and filler ratio, etc.), it is both prone to falling into local optimal solutions and difficult to take into account material physical property constraints (such as the TPU melt index threshold), resulting in the optimization results being often infeasible in actual production. In-depth analysis shows that there are three key defects in the prior art: First, small fluctuations in process parameters in the multi-layer co-extrusion process will cause uneven layer thickness and interfacial defects, and traditional control methods lack a real-time compensation mechanism; second, the material performance prediction model fails to effectively integrate the correlation between microscopic interfacial defects (infrared data) and macroscopic working condition loads (dynamic stress); finally, the optimization algorithm considers process parameters and material performance separately, forming a "black box optimization" mode that not only affects the convergence speed but also makes it difficult to ensure process feasibility. These problems seriously restrict the performance improvement and production efficiency of composite polymer backpack fabrics, and there is an urgent need to establish an intelligent system that integrates process control, performance prediction, and multi-objective optimization.

[0004] In view of the above problems, the present invention proposes a solution. Summary of the Invention

[0005] In order to overcome the above-mentioned defects of the prior art, an embodiment of the present invention provides an intelligent simulation and optimization system for backpack fabrics based on composite polymer materials to solve the problems raised in the above background art.

[0006] To achieve the above object, the present invention provides the following technical solutions: An intelligent simulation and optimization system for backpack fabrics based on composite polymer materials, comprising: A data acquisition layer, which includes a microscopic data acquisition module, a mesoscopic data acquisition module, and a macroscopic data acquisition module, respectively used for acquiring interface chemical bond characteristics, interface morphology parameters, layer thickness distribution data, interlayer defect data, dynamic load data, and environmental parameters; A cross-scale fusion layer, which uses a multi-modal attention neural network model or a hierarchical attention fusion network model to dynamically allocate weights to features at different scales, fuse microscopic, mesoscopic, and macroscopic features, and output material property prediction values; A process control layer, which includes an adaptive PID controller for dynamically adjusting process parameters according to real-time layer thickness deviation and deviation change rate, and generating control instructions to drive the actuator; An optimization decision layer, which uses a multi-objective optimization algorithm to generate an optimal process parameter combination that meets multi-objective performance through parameter coding, population initialization, fitness evaluation, and constraint handling.

[0007] In a preferred embodiment, the data acquisition layer includes: A microscopic data acquisition module: configured with a spectral analysis device and a microscopic detection device for periodically acquiring the intensity of interface chemical bond characteristic peaks and interface roughness; A mesoscopic data acquisition module: configured with a thickness measurement device and a defect detection device for real-time monitoring of layer thickness distribution and interlayer bubble defect density; A macroscopic data acquisition module: configured with a dynamic load simulation device and an environmental parameter acquisition device for simulating actual dynamic impact loads and recording environmental alternating data.

[0008] In a preferred embodiment, the cross-scale fusion layer includes: An input layer, receiving microscopic-scale interface chemical bond characteristics and roughness, mesoscopic-scale layer thickness deviation and defect density, macroscopic-scale dynamic load and environmental parameters; An attention mechanism module for dynamically allocating weights to features at different scales, including in-scale feature weight allocation and cross-scale weight allocation; A feature fusion module that uses a gating mechanism or a splicing mechanism to fuse multi-scale features; An output layer that outputs material fatigue life and wear resistance prediction values.

[0009] In a preferred embodiment, the process control layer includes: A fuzzy inference module for dynamically adjusting PID control parameters based on a preset rule base, and adjusting proportional, integral, and differential coefficients according to layer thickness deviation and its change rate; The actuator module, including a temperature control system and a pressure regulation system, responds to control instructions to adjust process parameters and ensure that the layer thickness deviation is controlled within a preset range.

[0010] In a preferred embodiment, the optimization decision-making layer includes: A constraint handling module that uses a dynamic penalty coefficient and a feasible solution retention strategy to handle process constraints; A hybrid evolution module that combines a differential evolution strategy and a simulated crossover strategy for parameter optimization; An adaptive adjustment module that dynamically adjusts the crossover probability and mutation probability according to population diversity and the number of generations of evolution.

[0011] In a preferred embodiment, the hierarchical attention fusion network further includes: A first-level attention module that calculates the weights of features within the same scale; A second-level attention module that dynamically assigns fusion weights to features of different scales; A gated fusion module that controls the information flow of multi-scale features through a sigmoid function.

[0012] In a preferred embodiment, the hybrid evolution module specifically includes: Using a differential evolution strategy for continuous variables to generate new individuals; Using a simulated binary crossover strategy for discrete variables; Adapting the mutation probability according to the population diversity index.

[0013] In a preferred embodiment, the system further includes: A data preprocessing module for denoising, normalizing, and detecting outliers in the original data; A model training module that trains a prediction model using cross-validation and regularization methods; A real-time prediction module that periodically updates the prediction results and provides a confidence interval.

[0014] In a preferred embodiment, the system achieves dynamic optimization in the following ways: In the data acquisition stage, multi-scale data is periodically collected at different frequencies; In the process control stage, the control parameters are adjusted at a fixed period and the control effect is verified; In the optimization decision-making stage, multiple sets of Pareto optimal solutions are generated based on preset convergence conditions.

[0015] An intelligent simulation and optimization system for backpack fabrics based on composite polymer materials, the hardware configuration of the system includes: A graphics processing unit for accelerating model calculations; A data communication module that meets the real-time control cycle; A high-bandwidth data storage device.

[0016] Technical effects and advantages of the intelligent simulation and optimization system for backpack fabrics based on composite polymer materials of the present invention: The present invention provides an intelligent simulation and optimization system for backpack fabrics based on composite polymer materials, which constructs a full-process architecture for multi-level process parameter acquisition, fusion, and feedback control from micro, meso to macro levels. By introducing a multi-modal attention neural network (MMANN) and an improved non-dominated sorting genetic algorithm (NSGA-III), this system can predict the fatigue life and wear resistance of materials in real time and achieve high-precision process parameter optimization. Further, in Embodiment 2, a hierarchical attention fusion network (HAFN) and a constraint-aware hybrid evolutionary algorithm (CAHEA) are adopted, significantly improving the feature expression ability and optimization efficiency of the model in a dynamic environment. The system integrates an adaptive PID control strategy based on fuzzy inference to achieve precise adjustment of layer thickness deviation and response closed-loop compensation. Overall, the present invention has advantages such as efficient cross-scale feature modeling, intelligent prediction of multi-objective performance, global optimal search under complex constraint conditions, and high dynamic response ability, significantly improving the performance consistency and production automation level of backpack fabrics in complex environments, and is applicable to various high-end functional composite material manufacturing scenarios. Description of the Drawings

[0017] Figure 1 It is a schematic structural diagram of the intelligent simulation and optimization system for backpack fabrics based on composite polymer materials of the present invention; Figure 2 It is a structural diagram of the multi-modal attention neural network model adopted by the cross-scale fusion layer of the present invention; Figure 3 It is a flow chart of the improved NSGAIII algorithm adopted by the optimization decision layer of the present invention; Figure 4 It is a timing diagram of the overall system of the present invention. Detailed Embodiments

[0018] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0019] The present invention relates to an intelligent simulation and optimization system for backpack fabric based on composite polymer materials. The system includes three layers of data acquisition modules at the microscale, mesoscale, and macroscale, a multimodal fusion modeling module, an adaptive process control module, and an intelligent optimization decision-making module. Multiscale feature data is obtained through means such as online Raman spectroscopy, laser thickness measurement, X-ray imaging, and dynamic fatigue simulation. The accurate prediction of material properties is achieved by using a multimodal attention neural network (MMANN) or a hierarchical attention fusion network (HAFN). Combining an adaptive PID control strategy with fuzzy parameter tuning and a constraint-aware hybrid evolutionary algorithm (CAHEA), real-time adjustment of key process parameters and multi-objective global optimization during the production process are realized. This system has high prediction accuracy, strong dynamic adaptability, and excellent optimization convergence, and is applicable to the intelligent manufacturing field of high-performance composite materials.

[0020] Embodiment 1 provides a specific implementation solution of the intelligent simulation and optimization system for backpack fabric based on composite polymer materials. The system realizes the intelligent optimization and control of the production process of composite polymer backpack fabric by integrating process data at the microscale, mesoscale, and macroscale.

[0021] The intelligent simulation and optimization system for backpack fabric based on composite polymer materials of the present invention, as Figure 1 and Figure 4 shown, mainly includes the following components: Data acquisition layer; The data acquisition layer consists of three sub-modules: Microscale data acquisition module: An online Raman spectrometer (wavelength 532 nm, resolution 1 cm -1 ) is used to collect the characteristic peak intensity of chemical bonds at the TPU / PET interface, and the interface roughness (Ra ≤ 0.5 μm is the qualified standard) is detected by an atomic force microscope (AFM, scanning range 10 μm × 10 μm); Mesoscale data acquisition module: A laser thickness gauge (accuracy ±1 μm) is configured to monitor the thickness distribution of the TPU layer (target thickness 0.2 mm) and the PET layer (target thickness 0.4 mm) in real time, and an X-ray transmission instrument is used to detect interlayer bubble defects (defects are determined when the diameter ≥ 0.1 mm); Macroscale data acquisition module: A multi-axis fatigue testing machine (load range 0 - 500 N, frequency 0 - 10 Hz) is used to simulate the dynamic impact load in actual use, and an environmental chamber (temperature range 20°C - 60°C, humidity range 30% - 90%) is used to collect environmental alternating data.

[0022] Cross-scale fusion layer; The cross-scale fusion layer uses a multimodal attention neural network (MMANN) model, and its specific structure is as Figure 2 shown: Input layer: Receives feature vectors at three scales: Microscale features: 1700 cm-1 The peak area A1 of the C=O bond and the surface roughness Ra at the location; Mesoscopic features: TPU layer thickness deviation Δh_TPU, PET layer thickness deviation Δh_PET, bubble density D; Macroscopic features: impact load F, ambient temperature T, and ambient humidity RH.

[0023] Attention layer: Dynamically allocate the weight coefficients of the three-scale features (macroscopic feature weight 0.4 - 0.6, mesoscopic 0.3 - 0.4, microscopic 0.1 - 0.2), which can also be adjusted according to the actual situation; Hidden layer: Contains 2 fully connected layers, with 32 neurons in each layer, and the ReLU activation function is used; Output layer: Output the predicted values of fatigue life L (unit: thousand times) and wear resistance W (unit: g / m²).

[0024] Process control layer; The process control layer uses an adaptive PID controller to achieve real-time compensation, and its working process is as follows: Input module: Receive the layer thickness data detected by the laser thickness gauge in real time; Calculation module: Calculate the current layer thickness deviation e(t) and the deviation change rate de(t) / dt; Fuzzy inference module: Adjust the PID parameters according to the preset rule base. When |e(t)| > 5% and de(t) / dt > 0, ΔKp = +0.2. When |e(t)| < 2% and de(t) / dt ≈ 0, ΔKi = +0.1; Output module: Generate the control quantity u(t) and send it to the actuator.

[0025] Optimization decision-making layer; The optimization decision-making layer uses an improved NSGAIII algorithm, and its optimization process is as Figure 3 shown: Parameter encoding: Take the TPU melt index x1 (15 - 25 g / 10 min), TPU layer thickness x2 (0.195 - 0.205 mm), and PET layer thickness x3 (0.395 - 0.405 mm) as the optimization variables; Population initialization: Generate 100 groups of initial parameter combinations that meet the constraints; Fitness evaluation: Predict the L and W values of each group of parameters through the MMANN model; Evolution operation: Use simulated binary crossover (η_c = 20) and polynomial mutation (η_m = 20); Constraint handling: Apply a penalty coefficient of 0.5 to individuals that violate the constraints.

[0026] More specifically, the working process of the system of the present invention is as follows: In the data acquisition stage, the specific acquisition method is as follows: Microscopic data: Raman spectroscopy data are collected every 5 seconds, and the AFM scans the interface morphology every 10 minutes. Mesoscopic data: The laser thickness gauge collects layer thickness data at a frequency of 10 Hz, and the X-ray performs a full-frame scan every 30 seconds. Macroscopic data: The fatigue testing machine is loaded according to the preset working condition program, and the temperature and humidity chamber records the environmental parameters in real time.

[0027] In the model training and prediction stage, the work process is as follows: Data preprocessing: The original data is normalized (Zscore normalization). Model training: The MMANN model is trained using 1000 groups of historical data, of which 800 groups are used for training and 200 groups are used for validation. Performance prediction: Three-scale features are input in real time, and the predicted values of material performance are output.

[0028] In the process control implementation stage, the work process is as follows: Control cycle: The PID parameters are adjusted every 1 second. Response of the actuator: Extruder temperature control system: Temperature adjustment accuracy ±0.5°C, Pressure pump: Pressure adjustment accuracy ±0.05 MPa. Effect verification: The layer thickness deviation is controlled within ±5 μm.

[0029] The optimization decision-making process is as follows: Optimization cycle: Global optimization is performed before each batch of production. Convergence condition: The change in the Pareto front for 10 consecutive generations < 1%. Result output: 35 groups of optimal parameter combinations are provided for the operator to choose.

[0030] In an optional example, the specific implementation parameters of the present invention can be as follows: Material parameters: TPU melt index: 1822 g / 10 min, PET intrinsic viscosity: 0.65 - 0.70 dL / g, Interface bonding strength: ≥15 N / cm.

[0031] In an optional example, the specific process parameters of the present invention can be as follows: Extrusion temperature: TPU 180 - 200°C, PET 260 - 280°C; Interlayer pressure: 3.5 - 4.5 MPa, Traction speed: 5 - 8 m / min.

[0032] In an optional example, the specific equipment parameters of the present invention can be as follows: Extruder: Screw diameter 65 mm; Thickness gauge: Scanning width 500 mm, Resolution 1 μm; Control system: Response time < 100 ms.

[0033] Through the above specific implementation methods, this embodiment realizes the effective fusion of cross-scale data, the real-time and accurate control of process parameters, and the intelligent multi-objective optimization decision-making in the production process of composite polymer backpack fabrics, significantly improving product quality and production efficiency.

[0034] Furthermore, the present invention also includes Embodiment 2. On the basis of Embodiment 1, Embodiment 2 further optimizes the intelligent simulation and optimization system for backpack fabrics based on composite polymer materials, focusing on improving the model architecture of the cross-scale fusion layer and the algorithm implementation of the optimization decision layer to enhance the prediction accuracy and optimization efficiency of the system under dynamic working conditions.

[0035] Specifically, the system of the present invention is composed as follows: Data acquisition layer; The data acquisition layer is composed of three sub-modules: Microscopic data acquisition module: An online Raman spectrometer (wavelength 532nm, resolution 1cm-1) is used to collect the characteristic peak intensity of chemical bonds at the TPU / PET interface, and the interface roughness (Ra≤0.5μm is the qualified standard) is detected by an atomic force microscope (AFM, scanning range 10μm×10μm); Mesoscopic data acquisition module: A laser thickness gauge (accuracy ±1μm) is configured to monitor the thickness distribution of the TPU layer (target thickness 0.2mm) and the PET layer (target thickness 0.4mm) in real time, and an X-ray transmission instrument is used to detect interlayer bubble defects (defects are determined when the diameter ≥0.1mm); Macroscopic data acquisition module: A multi-axis fatigue testing machine (load range 0500N, frequency 010Hz) is used to simulate the dynamic impact load in actual use, and an environmental chamber (temperature range 20°C~60°C, humidity range 30%90%) is used to collect environmental alternating data.

[0036] Cross-scale fusion layer; In this Embodiment 2, an improved hierarchical attention fusion network (HAFN) model is adopted, and its innovative structure is as Figure 2 shown: Input layer: Receive feature vectors of three scales: Microscopic features: The peak area A1 of the C=O bond at 1700cm-1 and the interface roughness Ra; Mesoscopic features: TPU layer thickness deviation Δh_TPU, PET layer thickness deviation Δh_PET, bubble density D; Macroscopic features: Impact load F, environmental temperature T, and environmental humidity RH.

[0037] Hierarchical attention mechanism: First-level attention: Calculate the internal feature weights of each scale ; where is the i-th feature of the m-th scale, is a trainable parameter matrix; Second-level attention: Dynamically allocate weights among three scales ; where is the aggregated feature of the m-th scale, and are trainable parameters; Feature fusion layer: Adopt a gating mechanism to control the information flow: ; ; where is the sigmoid function, represents element-wise multiplication; represents feature concatenation; Output layer: Output the predicted values of fatigue life L (unit: thousand times) and wear resistance W (unit: g / m²).

[0038] Process control layer; The process control layer uses an adaptive PID controller to achieve real-time compensation, and its working process is as follows: Input module: Receive the layer thickness data detected by the laser thickness gauge in real time; Calculation module: Calculate the current layer thickness deviation e(t) and the deviation change rate de(t) / dt; Fuzzy inference module: Adjust the PID parameters according to the preset rule base. When |e(t)| > 5% and de(t) / dt > 0, ΔKp = +0.2; when |e(t)| < 2% and de(t) / dt ≈ 0, ΔKi = +0.1; Output module: Generate the PID control quantity u(t) and send it to the actuator.

[0039] Optimization decision layer; In this second embodiment, a constraint-aware hybrid evolutionary algorithm (CAHEA) is adopted, and its innovative improvements include: Constraint handling mechanism: Adopt a dynamic penalty coefficient: ; where t is the current generation number and T is the total generation number, Introduce a feasible solution retention strategy: Forcefully retain 20% of the individuals that satisfy all constraints in each generation; Hybrid evolutionary operation: Differential evolution: For continuous variables (x1, x2, x3), where x1 is the TPU melt index and x2 is the layer thickness constraint, adopt the DE / rand / 1 strategy ; where F = 0.5, and r1, r2, r3 are random individual indices; Simulated binary crossover: Adopt the SBX operator for discrete variables (such as process route selection); Adaptive parameter adjustment: Crossover probability adapts to change with the evolutionary generation number: ; Mutation probability is associated with the population diversity; ; where is the standard deviation of the distance of the current generation population, is the initial standard deviation.

[0040] The working process of the system of the present invention is as follows: In the data acquisition stage, the specific acquisition method is as follows: Microscopic data: Raman spectrum data is collected every 5 seconds, and the interface topography is scanned by AFM every 10 minutes; Mesoscopic data: The layer thickness data is collected by a laser thickness gauge at a frequency of 10 Hz, and a full-frame scan is performed by X-ray every 30 seconds; Macroscopic data: The fatigue testing machine is loaded according to the preset working condition program, and the environmental parameters are recorded in real time by the temperature and humidity chamber.

[0041] In an optional example, in the second embodiment, the model training and prediction specifically include: Data preprocessing: Wavelet denoising (db4 wavelet, 5-layer decomposition) is performed on the Raman spectrum data; Outlier detection (3σ criterion) is performed on the layer thickness data; Model training: Hierarchical k-fold cross-validation (k = 5) is adopted; Optimization objective function: ; where is the mean square error, is the L2 regularization term of the weight matrix.

[0042] Online prediction: The prediction result is updated every 30 seconds; Provide a prediction confidence interval (95% confidence level).

[0043] In the process control implementation stage, the working process is as follows: Control period: The PID parameters are adjusted every 1 second; Actuator response: Extruder temperature control system: Temperature adjustment accuracy ±0.5°C Pressure pump: Pressure adjustment accuracy ±0.05 MPa; Effect verification: The layer thickness deviation is controlled within ±5 μm.

[0044] In an optional example, the optimization decision-making process of the second embodiment of the present invention includes the following contents: Initialization stage: Latin hypercube sampling is used to generate the initial population, and individuals satisfying XXX are pre-screened; Evolution stage: 100 individuals are evaluated in each generation, and the elite retention strategy (retaining the top 10% of the optimal solutions) is adopted; Termination condition: The maximum number of generations is 100 generations or the Pareto front improvement rate is continuously less than 0.5% for 5 generations.

[0045] In the second embodiment, the specific implementation parameters can be set as follows: HAFN model parameters: Hidden layer dimension: 64; Number of attention heads: 4; Dropout rate: 0.2; Batch size: 32; Learning rate: 0.001.

[0046] CAHEA algorithm parameters: Population size: 120; Differential evolution weight F: 0.5; Minimum mutation probability: 0.05; Number of reference points: 15.

[0047] Hardware configuration: GPU acceleration: NVIDIA Tesla T4; Real-time control cycle: 50 ms; Data storage bandwidth: 1 Gbps.

[0048] Through the innovative design of the hierarchical attention fusion network and the constraint-aware hybrid evolution algorithm, this embodiment significantly improves the adaptability of the system to dynamic working conditions and the optimization efficiency on the basis of maintaining all the functions of Embodiment 1, and is particularly suitable for the production scenario of special backpack fabrics with high-precision requirements.

[0049] The above formulas are all dimensionless and take their numerical values for calculation. The formulas are obtained by collecting a large amount of data for software simulation to get a formula closest to the real situation. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0050] The above embodiments can be implemented in whole or in part by software, hardware, firmware, or any other combination. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product.

[0051] Those of ordinary skill in the art can realize that the modules and algorithm steps of each example described in combination with the embodiments disclosed in this article can be implemented by electronic hardware, or by a combination of computer software and electronic hardware. Whether these functions are executed in hardware or software depends on the specific application of the technical solution and the invention constraints. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of this application.

[0052] In addition, in each embodiment of this application, the functional modules can be integrated into one processing module, or each module can exist physically alone, or two or more modules can be integrated into one module.

[0053] The above is only the specific implementation manner of this application, but the protection scope of this application is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed in this application, and all should be covered by the protection scope of this application. Therefore, the protection scope of this application should be subject to the protection scope of the claims.

[0054] Finally: The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included within the protection scope of the present invention.

Claims

1. An intelligent simulation and optimization system for backpack fabrics based on composite polymer materials, characterized in that Including: A data acquisition layer, which includes a microscopic data acquisition module, a mesoscopic data acquisition module, and a macroscopic data acquisition module, respectively used for collecting interface chemical bond characteristics, interface morphology parameters, layer thickness distribution data, interlayer defect data, dynamic load data, and environmental parameters; A cross-scale fusion layer, which uses a multi-modal attention neural network model or a hierarchical attention fusion network model to dynamically allocate weights to features of different scales, fuse microscopic, mesoscopic, and macroscopic features, and output material property prediction values; A process control layer, which includes an adaptive PID controller for dynamically adjusting process parameters according to real-time layer thickness deviation and deviation change rate, and generating control instructions to drive the actuator; An optimization decision layer, which uses a multi-objective optimization algorithm to generate an optimal process parameter combination that meets multi-objective performance through parameter coding, population initialization, fitness evaluation, and constraint handling.

2. The intelligent simulation and optimization system for backpack fabrics based on composite polymer materials according to claim 1, wherein The data acquisition layer includes: A microscopic data acquisition module: configured with a spectroscopic analysis device and a microscopic detection device for periodically collecting the intensity of interface chemical bond characteristic peaks and interface roughness; A mesoscopic data acquisition module: configured with a thickness measurement device and a defect detection device for real-time monitoring of layer thickness distribution and interlayer bubble defect density; A macroscopic data acquisition module: configured with a dynamic load simulation device and an environmental parameter acquisition device for simulating actual dynamic impact loads and recording environmental alternating data.

3. The intelligent simulation and optimization system for backpack fabrics based on composite polymer materials according to claim 2, wherein: The cross-scale fusion layer includes: An input layer, which receives microscopic-scale interface chemical bond characteristics and roughness, mesoscopic-scale layer thickness deviation and defect density, macroscopic-scale dynamic load, and environmental parameters; An attention mechanism module, which dynamically allocates weights to features of different scales, including in-scale feature weight allocation and cross-scale weight allocation; A feature fusion module, which uses a gating mechanism or a splicing mechanism to fuse multi-scale features; An output layer, which outputs material fatigue life and wear resistance prediction values.

4. The intelligent simulation and optimization system for backpack fabrics based on composite polymer materials according to claim 3, characterized in that; The process control layer includes: A fuzzy inference module, which dynamically adjusts PID control parameters based on a preset rule base, and adjusts proportional, integral, and differential coefficients according to layer thickness deviation and its change rate; An actuator module, which includes a temperature control system and a pressure regulation system, and responds to control instructions to adjust process parameters to ensure that the layer thickness deviation is controlled within a preset range.

5. The intelligent simulation and optimization system for backpack fabrics based on composite polymer materials according to claim 4, characterized in that: The optimization decision layer includes: A constraint handling module, which uses a dynamic penalty coefficient and a feasible solution retention strategy to handle process constraints; A hybrid evolution module, which combines a differential evolution strategy and a simulated crossover strategy for parameter optimization; An adaptive adjustment module, which dynamically adjusts the crossover probability and mutation probability according to population diversity and the number of generations of evolution.

6. The intelligent simulation and optimization system for backpack fabrics based on composite polymer materials according to claim 3, characterized in that: The hierarchical attention fusion network further includes: A first-level attention module, which calculates the weights of features within the same scale; A second-level attention module, which dynamically allocates the fusion weights of features of different scales; A gating fusion module, which controls the information flow of multi-scale features through a sigmoid function.

7. The intelligent simulation and optimization system for backpack fabrics based on composite polymer materials according to claim 5, characterized in that: The hybrid evolution module specifically includes: Using a differential evolution strategy to generate new individuals for continuous variables; Using a simulated binary crossover strategy for discrete variables; Adaptive adjustment of the mutation probability according to the population diversity index.

8. The intelligent simulation and optimization system for backpack fabrics based on composite polymer materials according to claim 1, wherein: The system further includes: A data preprocessing module for denoising, normalizing, and detecting outliers in the original data; A model training module that trains a prediction model using cross-validation and regularization methods; A real-time prediction module that periodically updates the prediction results and provides confidence intervals.

9. The intelligent simulation and optimization system for backpack fabrics based on composite polymer materials according to claim 1, wherein: The system achieves dynamic optimization in the following ways: In the data acquisition stage, multi-scale data is periodically acquired at different frequencies; In the process control stage, the control parameters are adjusted at a fixed period and the control effect is verified; In the optimization decision-making stage, multiple sets of Pareto optimal solutions are generated based on preset convergence conditions.

10. The intelligent simulation and optimization system for backpack fabrics based on composite polymer materials according to any one of claims 1-9, characterized in that, The hardware configuration of the system includes: A graphics processing unit for accelerating model calculations; A data communication module that meets the real-time control cycle; A high-bandwidth data storage device.

Images