Surface antistatic treatment method and system for polyester knitted fabric

Abstract

The invention discloses a polyester knitted fabric surface antistatic treatment method and system, and relates to the technical field of polyester knitted fabrics, the method comprises the following steps: determining a treatment process scheme according to a target application scene, calling a sample process record, and constructing a standard process training set; constructing and training an antistatic performance prediction model based on the standard process training set in combination with a knowledge distillation mechanism; a cost evaluation factor is constructed, a process evaluation function is defined in combination with the antistatic performance prediction model and the cost evaluation factor, and the process evaluation function is associated with a penalty space; and by taking the process evaluation function as an optimization target, carrying out adaptive capacity expansion and shrinkage iterative optimization on the treatment process parameters, determining a target process parameter combination for the polyester knitted fabric according to an iterative optimization result, and carrying out surface antistatic treatment. Therefore, the technical effects of reducing the parameter adjustment cost, improving the adjustment efficiency and enhancing the anti-static treatment flexibility are achieved.

Description

Technical Field

[0001] This invention relates to the field of polyester knitted fabric technology, and in particular to a method and system for surface antistatic treatment of polyester knitted fabric. Background Technology

[0002] Polyester knitted fabrics are widely used in clothing, home textiles, and industrial fabrics due to their excellent properties. However, in practical use, polyester knitted fabrics are prone to static electricity, which not only affects the user experience but may also pose safety hazards in certain scenarios. Therefore, surface antistatic treatment is necessary. Currently, the main methods for surface antistatic treatment of polyester knitted fabrics rely on traditional, experience-based process adjustments. For example, based on fixed process parameters, small adjustments are made manually based on experience. This makes it difficult to flexibly adjust according to actual production data and specific application requirements, resulting in insufficient precision in process parameter optimization and an inability to effectively meet antistatic requirements under different conditions. While some existing methods incorporate optimization algorithms, they struggle to achieve a balance between global exploration and local convergence when dealing with complex process parameter optimization problems, leading to time-consuming and inaccurate optimization processes. Summary of the Invention

[0003] This invention provides a method and system for surface antistatic treatment of polyester knitted fabric, which solves the technical problems of high cost and low efficiency of parameter adjustment in the prior art, and achieves the technical effects of reducing parameter adjustment cost, improving adjustment efficiency, and enhancing the flexibility of antistatic treatment.

[0004] In a first aspect, the present invention provides a method for surface antistatic treatment of polyester knitted fabric, wherein the method for surface antistatic treatment of polyester knitted fabric includes: Based on the processing technology scheme determined for the target application scenario, sample process records are called to construct a standard process training set.

[0005] Based on the standard process training set, and combined with the knowledge distillation mechanism, an antistatic performance prediction model is constructed and trained.

[0006] A cost evaluation factor is constructed, and a process evaluation function is defined by combining the antistatic performance prediction model and the cost evaluation factor, wherein the process evaluation function is associated with a penalty space.

[0007] Using the process evaluation function as the optimization objective, the processing parameters are iteratively optimized by adaptive expansion and contraction, and the target combination of process parameters for polyester knitted fabric is determined based on the iterative optimization results, and surface antistatic treatment is performed.

[0008] In one feasible implementation, a process evaluation function is defined by combining the antistatic performance prediction model and the cost evaluation factor, including: Obtain the antistatic treatment task sheet, parse the antistatic treatment task sheet, and extract the compliance constraints.

[0009] Based on the compliance constraints, the penalty space of the process evaluation function is defined.

[0010] The output of the antistatic performance prediction model is weighted and fused with the cost evaluation factor to obtain the process evaluation function, and the process evaluation function is associated with the penalty space.

[0011] In one feasible implementation, the process parameters are iteratively optimized by adaptive scaling, taking the process evaluation function as the optimization objective, including: Based on the standard process training set, a process parameter particle swarm containing multiple candidate process parameter combinations is initialized.

[0012] Based on the initialized process parameters and particle swarm, an iterative optimization process based on the particle swarm optimization algorithm is executed. In each iteration, the optimization direction of each particle is calculated, and the bidirectional virtual intersection point of the particle in the search space is identified based on the optimization direction.

[0013] Based on the bidirectional virtual intersection identification results, the process evaluation function adaptively expands and shrinks the particle swarm of process parameters.

[0014] The iterative optimization results are output based on the preset iterative constraints.

[0015] In one feasible implementation, the process parameter particle swarm adaptively expands and shrinks in conjunction with the process evaluation function, further comprising: When the optimization directions of the first preset number of particles within the preset range have a common virtual intersection point in the multidimensional parameter space, the process evaluation function value corresponding to the current position of each particle is obtained, and the fusion weight is defined accordingly.

[0016] Based on the fusion weight, a first preset number of particles are weighted and fused, and the first fusion process evaluation function value corresponding to the weighted fusion result is calculated based on the process evaluation function.

[0017] The process evaluation function value at the dummy intersection point is determined based on the process evaluation function value.

[0018] The spatial point corresponding to the larger value between the first fusion process evaluation function value and the virtual intersection process evaluation function value is compared and selected as a new particle and added to the process parameter particle swarm.

[0019] In one feasible implementation, the process parameter particle swarm adaptively expands and shrinks in conjunction with the process evaluation function, further comprising: In the multidimensional parameter space, identify the reverse virtual intersection points formed by the intersection of the reverse extensions of the second preset number of optimization directions.

[0020] When the reverse virtual intersection point is within a preset range, the angular variance of the adjacent included angles formed by the second preset number of optimized directions is calculated.

[0021] The number of new particles to be inserted is determined based on the angular variance, and new particles are inserted on the central axis of the adjacent angles of the second preset number of optimized directions in descending order of adjacent included angles.

[0022] In one feasible implementation, the process parameter particle swarm adaptively expands and shrinks in conjunction with the process evaluation function, further comprising: A dynamic fluctuation range for the particle swarm size is defined, and the particle swarm size is adaptively adjusted within the dynamic fluctuation range based on the initial particle swarm size.

[0023] If the current particle swarm count is lower than the fluctuation lower limit, the adaptive scaling down is paused until the current particle swarm count is higher than the initial particle swarm count.

[0024] If the current particle swarm size is higher than the fluctuation limit, pause adaptive expansion until the current particle swarm size is lower than the initial particle swarm size.

[0025] In one feasible implementation, based on the processing technology scheme determined for the target application scenario, sample process records are invoked to construct a standard process training set, including: Obtain sample process parameter data, including at least the auxiliary agent ratio, treatment temperature, treatment time, and fabric moisture content.

[0026] Obtain the sample antistatic performance data associated with the sample process parameter data, wherein the antistatic performance includes at least static voltage, surface charge density and half-life.

[0027] The sample process parameter data and the sample antistatic performance data are preprocessed and target values ​​are calibrated. Based on the target value calibration results, they are divided into positive samples and negative samples, and the output is the standard process training set.

[0028] In one feasible implementation, based on the aforementioned standard process training set and combined with a knowledge distillation mechanism, an antistatic performance prediction model is constructed and trained, including: A benchmark prediction model based on a deep neural network is constructed and trained using the standard process training set.

[0029] Based on the knowledge distillation mechanism, a simplified prediction model is trained by combining the prediction output of the baseline prediction model with the standard process training set.

[0030] The simplified prediction model is verified for accuracy. If the set performance evaluation threshold is met on both positive and negative samples in the standard process training set, the simplified prediction model is output and deployed as the antistatic performance prediction model.

[0031] In one feasible implementation, the cost evaluation factors include at least the processing cost per unit of fabric, the processing energy consumption, the processing time, and the resource consumption.

[0032] Secondly, the present invention also provides a surface antistatic treatment system for polyester knitted fabrics, wherein the surface antistatic treatment system for polyester knitted fabrics includes: The training set construction module is used to construct a standard process training set by calling sample process records based on the processing technology scheme determined by the target application scenario.

[0033] The prediction model training module is used to construct and train an antistatic performance prediction model based on the standard process training set and in conjunction with a knowledge distillation mechanism.

[0034] The evaluation function definition module is used to construct cost evaluation factors and, in combination with the antistatic performance prediction model and the cost evaluation factors, define a process evaluation function, wherein the process evaluation function is associated with a penalty space.

[0035] The process parameter optimization module is used to perform adaptive expansion and contraction iterative optimization of the processing process parameters with the process evaluation function as the optimization target, and to determine the target process parameter combination for polyester knitted fabric based on the iterative optimization results for surface antistatic treatment.

[0036] This invention discloses a method and system for surface antistatic treatment of polyester knitted fabric, comprising: constructing a standardized process training dataset by calling historical sample process records based on a treatment process scheme determined by the target application scenario; establishing and training a machine learning model for predicting antistatic performance based on the training dataset and combining it with a knowledge distillation mechanism; constructing a cost-related evaluation factor and combining the antistatic performance prediction model with the cost evaluation factor to set a process evaluation function including a penalty term to quantify the comprehensive effect of the treatment process; using the process evaluation function as the optimization objective, iteratively optimizing the treatment process parameters using an adaptive scaling strategy, and finally determining the target process parameter combination for surface antistatic treatment of polyester knitted fabric. This invention solves the technical problems of high cost and low efficiency in adjusting treatment parameters, achieving the technical effects of reducing parameter adjustment costs, improving adjustment efficiency, and enhancing the flexibility of antistatic treatment. Attached Figure Description

[0037] Figure 1This is a schematic flowchart of a method for surface antistatic treatment of polyester knitted fabric according to the present invention.

[0038] Figure 2 This is a schematic diagram of the surface antistatic treatment system for polyester knitted fabric according to the present invention.

[0039] Figure labeling: Training set construction module 11, prediction model training module 12, evaluation function definition module 13, process parameter optimization module 14. Detailed Implementation

[0040] The above technical solutions will now be described in detail with reference to the accompanying drawings and specific embodiments to provide a better understanding of them. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. It should be understood that the present invention is not limited to the exemplary embodiments used only to explain the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention. Furthermore, it should be noted that, for ease of description, only the parts related to the present invention are shown in the drawings, not all of them.

[0041] Example 1, as Figure 1 This is a schematic flowchart of a method for surface antistatic treatment of polyester knitted fabric according to the present invention, wherein the method for surface antistatic treatment of polyester knitted fabric includes: S100: Based on the processing technology scheme determined for the target application scenario, call the sample process records to construct a standard process training set.

[0042] Specifically, the target application scenario refers to the end use of polyester knitted fabric (such as clothing, home textiles, industrial fabrics, etc.). Based on the target application scenario, specific requirements for properties such as antistatic properties, hand feel, and color fastness can be determined.

[0043] Specifically, different target application scenarios correspond to different types of processes, such as surface coating, chemical modification, and conductive fiber embedding. In addition, different processing solutions involve different numbers and types of process parameters, including, but not limited to, the type of antistatic agent, the addition ratio, padding process parameters, drying temperature and time, and other process elements.

[0044] Specifically, the sample process records are the process parameters that have been verified and effective in historical production processes, along with their corresponding performance test data. For example, we can obtain the processing records of all polyester knitted fabrics produced in the past year, including different combinations of parameters such as temperature, time, type and amount of auxiliaries, as well as the corresponding surface resistivity (an indicator of antistatic performance). From these, we can select samples that meet the requirements of the target application scenario, such as surface coating treatment records in the clothing industry, and construct a standard process training set.

[0045] Through the above process, an accurate and targeted data foundation can be provided for the subsequent antistatic performance prediction model. In other words, the constructed standard process training set closely revolves around the target application scenario, ensuring that the rules learned by the model match the actual application requirements.

[0046] In some embodiments, based on the processing technology scheme determined according to the target application scenario, sample process records are invoked to construct a standard process training set, including: Obtain sample process parameter data, wherein the process parameters include at least the auxiliary agent ratio, treatment temperature, treatment time, and fabric moisture content; obtain sample antistatic performance data associated with the sample process parameter data, wherein the antistatic performance includes at least static voltage, surface charge density, and half-life; preprocess the sample process parameter data and the sample antistatic performance data and calibrate the target values, and divide them into positive samples and negative samples according to the target value calibration results, and output them as the standard process training set.

[0047] Specifically, the sample process parameter data refers to the actual or historically recorded process operation parameters during the antistatic treatment of polyester knitted fabric surfaces. These mainly include auxiliary agent ratios (such as the ratio of antistatic agent to water), treatment temperature, treatment time, and fabric moisture content. The sample antistatic performance data includes static voltage, surface charge density, and half-life, used to measure the antistatic effect of polyester knitted fabrics.

[0048] Specifically, first, the target antistatic performance value is determined based on the target application scenario (e.g., clothing, home textiles). Then, historical sample process records are retrieved to extract relevant process parameter data and corresponding antistatic performance data, and the data is standardized (e.g., outlier removal, normalization). Next, the samples are calibrated according to preset performance thresholds. Positive samples (achieving target performance) are defined as those with static voltage not exceeding a certain set value, charge surface density below a specified upper limit, and half-life less than a specified upper limit; otherwise, they are considered negative samples (not achieving target performance). Finally, a dataset containing both positive and negative samples is output as a standard process training set.

[0049] For example, if the static voltage of the fabric is 350V (lower than the target 400V) and the surface charge density is 0.7μC / m under a certain set of process parameters... 2 (1.0 μC / m below the target) 2 If the half-life is 0.510s (less than the target 2s), then the sample is labeled as a positive sample.

[0050] Through the above process, a standard process training set that meets the target application requirements can be systematically constructed, improving the scientific nature and accuracy of process development, providing strong data support for the parameter optimization of the antistatic treatment process for polyester knitted fabrics, thereby improving product consistency, stability and production efficiency.

[0051] S200: Based on the standard process training set, and combined with the knowledge distillation mechanism, construct and train the antistatic performance prediction model.

[0052] In some embodiments, based on the standard process training set and combined with a knowledge distillation mechanism, an antistatic performance prediction model is constructed and trained, including: A benchmark prediction model based on a deep neural network is constructed and trained using the standard process training set. A simplified prediction model is trained by combining the prediction output of the benchmark prediction model with the standard process training set based on the knowledge distillation mechanism. The simplified prediction model is then verified for accuracy. If the set performance evaluation threshold is met on both positive and negative samples in the standard process training set, the simplified prediction model is output and deployed as the antistatic performance prediction model.

[0053] Specifically, knowledge distillation is used to transfer knowledge from complex and high-performance baseline prediction models (typically deep neural network models) to simplified prediction models that are structurally simpler and computationally more efficient. The baseline prediction model refers to a deep neural network model trained on a complete standard process training set, possessing high accuracy in predicting antistatic performance. The simplified prediction model refers to a smaller, faster-inference model obtained through knowledge distillation, facilitating rapid deployment in real-world applications.

[0054] Specifically, firstly, using the process parameters and antistatic performance data in the standard process training set, a benchmark prediction model based on deep neural networks (including multilayer perceptrons, convolutional neural networks, or long short-term memory networks) is trained so that it can accurately predict the antistatic performance under given process parameters.

[0055] Furthermore, after training is completed, knowledge distillation is adopted, using the benchmark prediction model as the teacher model. Its prediction outputs on the training set samples (such as probability distributions, feature representations, and other soft labels) are used together with the real labels of the standard process training set to train a more simplified student model. By designing a distillation loss function, the simplified prediction model can not only fit the real labels, but also approximate the output performance of the teacher model as closely as possible.

[0056] For example, the distillation loss function can adopt the following composite loss: ; in, The cross-entropy loss is the value of the true label. The Kullback-Leibler divergence is the output of the teacher model (baseline prediction model) and the student model (simplified prediction model), where α is the weighting coefficient (e.g., 0.5). y represents the output of the teacher model and the student model at temperature T, respectively; y represents the true label, such as the true antistatic performance category or actual performance value of the sample. Then the predicted output of the student model.

[0057] Furthermore, the simplified prediction model is evaluated on positive and negative samples in the standard process training set. If its prediction accuracy, recall and other indicators all reach the preset threshold, the performance of the simplified prediction model is considered to meet the requirements, and the simplified prediction model can be used as the final antistatic performance prediction model for output and actual deployment.

[0058] Through the above process, while ensuring the accuracy of antistatic performance prediction, the inference efficiency and deployment flexibility of the model can be significantly improved, which helps to achieve real-time process parameter optimization in actual production environments and further enhances the automation and intelligence level of the antistatic treatment process on the surface of polyester knitted fabrics.

[0059] S300: Construct a cost evaluation factor, and combine the antistatic performance prediction model and the cost evaluation factor to define a process evaluation function, wherein the process evaluation function is associated with a penalty space.

[0060] Specifically, cost evaluation factors are indicators used to quantify the economic viability of antistatic treatment processes for polyester knitted fabrics, facilitating quantitative comparisons of the costs of different process solutions. For example, cost evaluation factors may include, but are not limited to: auxiliary agent usage costs, such as the market price of the auxiliary agent required per square meter of fabric multiplied by the actual usage; energy consumption costs, such as the electricity consumption costs corresponding to the treatment temperature and time; water costs, such as water resource consumption during the treatment process; and equipment depreciation and labor costs, such as the equipment depreciation and labor costs allocated to each batch of production.

[0061] For example, assuming a single batch of auxiliary agent usage is 0.8%, the unit price of the auxiliary agent is 100 yuan / kg, and 8g of auxiliary agent is used per square meter of fabric, then the auxiliary agent cost is 0.8 yuan / square meter; the processing temperature is 140℃, electricity consumption is 0.5 kWh, and the electricity price is 0.8 yuan / kWh, then the energy cost is 0.4 yuan / square meter; water consumption is 0.2 tons, and the water price is 3 yuan / ton, then the water cost is 0.6 yuan / square meter; equipment depreciation and labor costs total 0.5 yuan / square meter. Therefore, the total cost is 2.3 yuan / square meter.

[0062] By combining cost-based process evaluation functions, a comprehensive trade-off between antistatic performance and economic cost can be achieved in process solutions. This ensures that the recommended process not only has excellent antistatic effect but also takes into account production cost control. Furthermore, the penalty space mechanism can further avoid inefficient or unusable combinations of process parameters that do not meet performance requirements, thereby improving the practicality and reliability of process optimization and intelligent recommendation.

[0063] In some embodiments, the cost evaluation factors include at least the unit fabric processing cost, processing energy consumption, processing time, and resource consumption.

[0064] Preferably, the cost evaluation factors include at least the following sub-items: The unit fabric treatment cost item refers to the direct economic input required to complete antistatic treatment of a unit area (or unit weight) of fabric, including auxiliary agent costs, labor costs, etc. The treatment energy consumption item refers to the energy costs consumed in the process, such as electricity and steam. The treatment time item refers to the actual time required to complete one batch of antistatic treatment, reflecting process efficiency and directly impacting production capacity. The resource consumption item refers to the cost of water, chemicals, and other resources consumed in the process, reflecting the green and environmentally friendly characteristics of the process.

[0065] In some embodiments, a process evaluation function is defined by combining the antistatic performance prediction model and the cost evaluation factor, including: Obtain the antistatic treatment task book and parse it to extract the compliance constraints; define the penalty space of the process evaluation function based on the compliance constraints; weight and fuse the output of the antistatic performance prediction model with the cost evaluation factor to obtain the process evaluation function, and associate the process evaluation function with the penalty space.

[0066] Specifically, firstly, the antistatic treatment task specification is obtained and analyzed to extract compliance constraints. For example, a task specification might stipulate: static voltage ≤ 500V, half-life ≥ 20s, and treatment cost per unit area ≤ 3 yuan / m². 2 Then, the above constraints can be transformed into a structured set of compliance constraint parameters. Subsequently, based on the compliance constraints, a penalty space for the process evaluation function is defined to negatively correct process parameter combinations that do not meet the above constraints. For example, the expression for the penalty space is: ; in, To predict static voltage, To predict half-life, To predict unit cost, λ is a penalty coefficient (e.g., all are taken as 1000), used to amplify the negative impact of exceeding the limit.

[0067] Furthermore, a weighted summation method is used to integrate the output of the antistatic performance prediction model, the cost evaluation factor, and the expression of the aforementioned penalty space to reduce the overall score of unqualified solutions.

[0068] Through the above process, it is possible to achieve automated, multi-objective, and constrained optimization evaluation of different combinations of antistatic treatment process parameters, ensuring that the recommended solution obtains the optimal comprehensive score while meeting both performance and cost requirements.

[0069] S400: Using the process evaluation function as the optimization target, the processing parameters are iteratively optimized by adaptive expansion and contraction, and the target process parameter combination for polyester knitted fabric is determined based on the iterative optimization results, and surface antistatic treatment is performed.

[0070] Specifically, adaptive scaling-up and scaling-down iterative optimization can efficiently find the optimal combination of process parameters by dynamically adjusting the particle swarm size, ensuring that the optimization algorithm achieves a balance between global exploration and local convergence, avoiding getting trapped in local optima, and improving the accuracy and efficiency of optimization.

[0071] In some embodiments, the process parameters are adaptively scaled up and down iteratively optimized using the process evaluation function as the optimization objective, including: Based on the standard process training set, a process parameter particle swarm containing multiple candidate process parameter combinations is initialized; according to the initialized process parameter particle swarm, an iterative optimization process based on the particle swarm optimization algorithm is executed, calculating the optimization direction of each particle in each iteration, and identifying the bidirectional virtual intersection points of the particles in the search space based on the optimization direction; based on the bidirectional virtual intersection point identification results, the process evaluation function adaptively expands and shrinks the process parameter particle swarm; according to the preset iterative constraints, the iterative optimization results are output.

[0072] Specifically, the process parameter particle swarm refers to a group of candidate process parameters combined during the optimization process, where each particle represents a set of possible process parameters. Bidirectional virtual intersection points refer to virtual points in the search space that particles may intersect during particle swarm optimization, identified based on their movement trends and optimization directions. These points are used to determine whether the particle aggregation degree and optimization direction are consistent.

[0073] Specifically, firstly, based on the standard process training set, an initial range of process parameter values ​​is set, and multiple candidate process parameter combinations are randomly generated within this range as the initial particle swarm. The optimization direction and velocity of the particles are then initialized and defined. For example, a particle swarm containing 50 particles is initialized, where each particle represents a set of process parameters, such as the ratio of additives, treatment temperature, treatment time, and fabric moisture content, and its current fitness (i.e., the process evaluation function value) is recorded.

[0074] Then, based on the initialized process parameters, an iterative optimization process based on the particle swarm optimization algorithm is executed: In each iteration, the fitness value of each particle is first calculated, that is, the merits of the process parameter combination corresponding to the particle are evaluated according to the process evaluation function. Then, the velocity and position of each particle are updated according to the particle velocity and the update rules of the particle swarm optimization algorithm.

[0075] Next, after updating the particle positions, the movement trends of the particles in the historical optimal direction or the global optimal direction are analyzed to calculate the theoretical bidirectional virtual intersection points. Based on the distribution of virtual intersection points, it is determined whether the current particle swarm is tending towards convergence or dispersion, providing a basis for subsequent expansion and contraction adjustments. The bidirectional virtual intersection points include virtual intersection points on the extension line of the optimization direction and reverse virtual intersection points on the reverse extension line of the optimization direction.

[0076] Furthermore, based on the bidirectional virtual intersection identification results and the process evaluation function, the particle swarm size for process parameters is adaptively expanded or reduced. For example, if the particle swarm has high diversity and the bidirectional virtual intersections indicate that the particles are dispersed in a large search space, the size of the particle swarm can be appropriately increased to enhance the global search capability; conversely, if the particle swarm has low diversity and the bidirectional virtual intersections indicate that the particles are concentrated in a small region, the size of the particle swarm can be appropriately reduced to improve search efficiency.

[0077] Finally, based on the preset iterative constraints (such as the maximum number of iterations, the convergence threshold of the evaluation function, etc.), when the termination condition is met, the optimal combination of process parameters in the current particle swarm is output as the target process parameters for the antistatic treatment of the polyester knitted fabric surface.

[0078] The above process improves the efficiency of process development and the reliability of the optimal solution. The adaptive scaling mechanism effectively avoids local optima and low search efficiency, ensuring that the output process parameter combinations achieve optimal performance, cost, and other objectives.

[0079] In some implementations, the process parameter particle swarm optimization is adaptively scaled up or down in conjunction with the process evaluation function, including: When the optimization directions of a first preset number of particles within a preset range have a common virtual intersection point in the multidimensional parameter space, the process evaluation function value corresponding to the current position of each particle is obtained, and a fusion weight is defined accordingly; based on the fusion weight, the first preset number of particles are weighted and fused, and the first fused process evaluation function value corresponding to the weighted fusion result is calculated based on the process evaluation function; the virtual intersection point process evaluation function value at the common virtual intersection point is determined based on the process evaluation function value; the spatial point corresponding to the larger value between the first fused process evaluation function value and the virtual intersection point process evaluation function value is compared and selected as a new particle and added to the process parameter particle group.

[0080] Specifically, a common virtual intersection point refers to the point where the straight lines extending from the optimization directions (i.e., velocity vectors) of several particles intersect in the multidimensional process parameter space. This common virtual intersection point can be considered as a potential convergence target for multiple particles. Fusion weight refers to the proportion assigned to each particle in the fusion process based on its current position and process evaluation function value. The first fusion process evaluation function value is the process evaluation function value corresponding to the new process parameter combination obtained by weighted fusion of multiple particles. The virtual intersection point process evaluation function value refers to the process evaluation function value calculated at the common virtual intersection point, assuming a corresponding process parameter combination exists.

[0081] Specifically, firstly, in each iteration, it is determined whether there are intersections in the optimization directions of a first preset number of particles (e.g., 2-3) within a preset range (e.g., a region of a specific radius in the parameter space), and the set of particles that meet the condition is recorded as the candidate particle group participating in the fusion. Then, the process evaluation function value of the current position of each candidate particle is obtained, and fusion weights are assigned according to their magnitude (e.g., weight allocation based on normalization). Next, the parameter combinations of each particle are weighted and fused based on the aforementioned fusion weights to obtain a new parameter combination (i.e., the first fusion point), and the process evaluation function value of this weighted fusion point (i.e., the first fusion point) is calculated using the process evaluation function and recorded as the first fused process evaluation function value.

[0082] Furthermore, using the parameter combination of the common virtual intersection point as input, the process evaluation function value is calculated through the process evaluation function and recorded as the virtual intersection point process evaluation function value. Then, the virtual intersection point process evaluation function value is compared with the first fused process evaluation function value, and the parameter space point corresponding to the larger value (i.e., the point with better performance) is taken as a new particle and added to the process parameter particle swarm for subsequent iterations. At the same time, the aforementioned first preset number of particles are removed from the particle swarm.

[0083] The above process fully utilizes information from multiple high-quality parameter combinations in the particle swarm. Through weighted fusion and virtual intersection analysis, it dynamically generates new particles with superior performance, enhancing the exploration capability of the search space and the probability of finding the global optimum. Simultaneously, based on feedback from the process evaluation function, the particle swarm structure is adjusted in real time to achieve adaptive scaling, further accelerating the optimization convergence speed and improving the intelligence and efficiency of process parameter optimization.

[0084] In some implementations, the process parameter particle swarm optimization is adaptively scaled up and down in conjunction with the process evaluation function, further including: In the multidimensional parameter space, the reverse virtual intersection points formed by the intersection of the reverse extensions of the second preset number of optimization directions are identified; when the reverse virtual intersection points are within a preset range, the angular variance of the adjacent included angles formed by the second preset number of optimization directions is calculated; based on the angular variance, the number of new particles to be inserted is determined, and new particles are inserted on the central axis of the adjacent included angles of the second preset number of optimization directions in descending order of the adjacent included angles.

[0085] Specifically, the reverse virtual intersection point refers to the theoretical point formed when the backward extensions of the optimization directions of several particles (i.e., the backward extensions of the velocity vectors) intersect in the multidimensional parameter space, reflecting the potential divergence trend of the particle swarm or an unexplored spatial region. The angular variance refers to the variance calculated from the angles between each pair of these optimization directions, reflecting the degree of dispersion in the particle distribution. The midline of adjacent angles refers to the direction of the bisector of the angle between two optimization directions in the parameter space.

[0086] Specifically, firstly, based on the same principle as the aforementioned common virtual intersection point, within the process parameter space, the optimization directions of a second preset number (e.g., 3-5) of particles are selected, and their velocity vectors are extended in the opposite direction. The intersection points of these reverse extensions are then calculated. If an intersection point exists and its coordinates fall within the preset parameter range (i.e., meet the preset range requirements), the intersection point of the reverse extensions is recorded as a reverse virtual intersection point. Then, based on the aforementioned second preset number of particle optimization directions, the angles between adjacent pairwise optimization directions (i.e., velocity vectors) are calculated to obtain a set of angle values ​​(and the aforementioned adjacent angles, including multiple angles), and the variance of these angles is calculated.

[0087] Furthermore, based on the magnitude of the variance of the included angle, the number of new particles to be inserted is determined. A larger variance indicates a more dispersed particle distribution, requiring more new particles to be inserted to enhance the search in that region. Optionally, the variance of the included angle can be mapped to the number of new particles to be inserted using a preset mapping rule or function.

[0088] Furthermore, according to all adjacent included angles sorted from largest to smallest, the direction pairs with larger included angles (i.e., velocity vector pairs) are selected first, and new particles are inserted on the central axis of adjacent included angles (i.e., the direction of the bisector of the two optimized directions) with the reverse virtual intersection point as the reference.

[0089] Optionally, the parameter combination of the newly inserted particle can be set to the intermediate value between two adjacent original particles in the direction of the central axis, or directly set to the position on the central axis at a certain step length away from the reverse virtual intersection point.

[0090] Through the above process, the particle swarm structure can be dynamically adjusted based on the discreteness of the particle swarm distribution and the spatial characteristics of the reverse virtual intersections. This actively fills potential exploration blind spots in the parameter space, enhances global search capabilities, prevents getting trapped in local optima, and further improves the efficiency and effectiveness of process optimization.

[0091] In some implementations, the process parameter particle swarm optimization is adaptively scaled up and down in conjunction with the process evaluation function, further including: A dynamic fluctuation range for the particle swarm size is set, and the particle swarm size is adaptively adjusted within the dynamic fluctuation range based on the initial particle swarm size. If the current particle swarm size is lower than the lower limit of fluctuation, adaptive scaling down is paused until the current particle swarm size is higher than the initial particle swarm size. If the current particle swarm size is higher than the upper limit of fluctuation, adaptive scaling up is paused until the current particle swarm size is lower than the initial particle swarm size.

[0092] Specifically, the dynamic fluctuation range of the particle swarm size refers to the upper and lower limits of the allowable particle number variation, based on the initial particle swarm size. For example, if the initial particle swarm size is N0, the dynamic fluctuation range can be defined as [N...]. min N max ], where N min =N0−ΔN,N max =N0+ΔN, where ΔN is the allowable fluctuation value (e.g., 20% of N0).

[0093] Specifically, firstly, a dynamic fluctuation range for the particle swarm size is set, allowing the particle swarm size to flexibly expand or shrink according to search needs during the optimization process. For example, the initial particle swarm size is 50, with a dynamic fluctuation range of 40~60.

[0094] Then, using the initial particle swarm size as a baseline, the current particle swarm size is monitored in real time. When the particle swarm size falls below the lower limit (e.g., 40) due to adaptive shrinkage operations, the shrinkage operation is automatically paused. Shrinkage can only resume after the particle swarm size recovers (e.g., through the insertion of new particles or natural growth during iteration) to a level higher than the initial particle swarm size (e.g., 50). Similarly, when the particle swarm size exceeds the upper limit (e.g., 60) due to adaptive expansion operations, expansion operations are automatically paused until the particle swarm size falls back below the initial particle swarm size (e.g., 50) before expansion can resume.

[0095] Through the above process, it is possible to effectively prevent the particle swarm size from shrinking excessively, resulting in insufficient search capability, or from expanding excessively, causing a waste of computing resources. This achieves a dynamic balance between the global exploration capability and computational efficiency of the optimization process, and improves the stability and efficiency of process parameter optimization.

[0096] In summary, the surface antistatic treatment method for polyester knitted fabric provided by this invention has the following technical effects: By determining the processing scheme based on the target application scenario, and calling historical sample process records, a standardized process training dataset is constructed. Based on the training dataset, a machine learning model for predicting antistatic performance is established and trained using a knowledge distillation mechanism. A cost-related evaluation factor is constructed, and the antistatic performance prediction model is combined with the cost evaluation factor to set a process evaluation function containing a penalty term to quantify the comprehensive effect of the processing process. Using the process evaluation function as the optimization objective, an adaptive scaling strategy is used to iteratively optimize the processing parameters, and finally, the target process parameter combination for antistatic treatment of polyester knitted fabric surfaces is determined, thereby achieving the technical effects of reducing parameter adjustment costs, improving adjustment efficiency, and enhancing the flexibility of antistatic treatment.

[0097] Example 2, as Figure 2 This is a schematic diagram of the surface antistatic treatment system for polyester knitted fabric according to the present invention. For example, Figure 1 A schematic flowchart of a surface antistatic treatment method for polyester knitted fabric according to the present invention can be seen as follows: Figure 2 The structure shown is implemented.

[0098] Based on the same concept as the surface antistatic treatment method for polyester knitted fabric described in the above embodiment, the present invention also provides a surface antistatic treatment system for polyester knitted fabric comprising: Training set construction module 11 is used to construct a standard process training set by calling sample process records based on the processing technology scheme determined by the target application scenario.

[0099] The prediction model training module 12 is used to construct and train an antistatic performance prediction model based on the standard process training set and in conjunction with a knowledge distillation mechanism.

[0100] Evaluation function definition module 13 is used to construct cost evaluation factors and, in combination with the antistatic performance prediction model and the cost evaluation factors, define a process evaluation function, wherein the process evaluation function is associated with a penalty space.

[0101] The process parameter optimization module 14 is used to perform adaptive expansion and contraction iterative optimization of the processing process parameters with the process evaluation function as the optimization target, and to determine the target process parameter combination for polyester knitted fabric based on the iterative optimization results for surface antistatic treatment.

[0102] In some embodiments, the training set construction module 11 includes: The sample process parameter data acquisition unit is used to acquire sample process parameter data, wherein the process parameters include at least the auxiliary agent ratio, treatment temperature, treatment time and fabric moisture content.

[0103] The sample antistatic performance data acquisition unit is used to acquire sample antistatic performance data associated with the sample process parameter data, wherein the antistatic performance includes at least static voltage, surface charge density and half-life.

[0104] The data preprocessing and standard process training set generation unit is used to preprocess and target value calibrate the sample process parameter data and the sample antistatic performance data, and divide them into positive samples and negative samples according to the target value calibration results, and output the standard process training set.

[0105] In some embodiments, the prediction model training module 12 includes: The benchmark prediction model construction and training unit is used to construct and train a benchmark prediction model based on a deep neural network using the standard process training set.

[0106] A simplified prediction model training unit is used to train a simplified prediction model based on a knowledge distillation mechanism, combining the prediction output of the baseline prediction model with the standard process training set.

[0107] A simplified prediction model accuracy verification and output unit is used to verify the accuracy of the simplified prediction model. If the set performance evaluation threshold is met on both positive and negative samples in the standard process training set, the simplified prediction model is output and deployed as the antistatic performance prediction model.

[0108] In some embodiments, the cost evaluation factors in the evaluation function definition module 13 include at least the processing cost per unit of fabric, the processing energy consumption, the processing time, and the resource consumption.

[0109] In some embodiments, the evaluation function definition module 13 includes: The antistatic treatment task book acquisition and parsing unit is used to acquire the antistatic treatment task book, parse the antistatic treatment task book, and extract the compliance constraints.

[0110] The penalty space definition unit is used to define the penalty space of the process evaluation function according to the compliance constraints.

[0111] The process evaluation function acquisition and association unit is used to weight and fuse the output of the antistatic performance prediction model with the cost evaluation factor to obtain the process evaluation function, and associate the process evaluation function with the penalty space.

[0112] In some embodiments, the process parameter optimization module 14 includes: The process parameter particle swarm initialization unit is used to initialize a process parameter particle swarm containing multiple candidate process parameter combinations based on the standard process training set.

[0113] The iterative optimization and optimization direction calculation unit is used to perform an iterative optimization process based on the particle swarm optimization algorithm according to the initialized process parameters. In each iteration, it calculates the optimization direction of each particle and identifies the bidirectional virtual intersection point of the particle in the search space according to the optimization direction.

[0114] An adaptive contraction unit is used to adaptively contract the process parameter particle swarm based on the bidirectional virtual intersection identification result and the process evaluation function.

[0115] The iterative optimization result output unit is used to output the iterative optimization result according to the preset iterative constraints.

[0116] In some implementations, the execution steps of the adaptive scaling unit in the process parameter optimization module 14 include: When the optimization directions of the first preset number of particles within the preset range have a common virtual intersection point in the multidimensional parameter space, the process evaluation function value corresponding to the current position of each particle is obtained, and the fusion weight is defined accordingly.

[0117] Based on the fusion weight, a first preset number of particles are weighted and fused, and the first fusion process evaluation function value corresponding to the weighted fusion result is calculated based on the process evaluation function.

[0118] The process evaluation function value at the dummy intersection point is determined based on the process evaluation function value.

[0119] The spatial point corresponding to the larger value between the first fusion process evaluation function value and the virtual intersection process evaluation function value is compared and selected as a new particle and added to the process parameter particle swarm.

[0120] In some implementations, the execution steps of the adaptive scaling unit in the process parameter optimization module 14 also include: In the multidimensional parameter space, identify the reverse virtual intersection points formed by the intersection of the reverse extensions of the second preset number of optimization directions.

[0121] When the reverse virtual intersection point is within a preset range, the angular variance of the adjacent included angles formed by the second preset number of optimized directions is calculated.

[0122] The number of new particles to be inserted is determined based on the angular variance, and new particles are inserted on the central axis of the adjacent angles of the second preset number of optimized directions in descending order of adjacent included angles.

[0123] In some implementations, the adaptive scaling unit in the process parameter optimization module 14 further includes: The scale constraint subunit is used to set the dynamic fluctuation range of the particle swarm size, and adaptively adjust the particle swarm size within the dynamic fluctuation range based on the initial particle swarm size.

[0124] The adaptive scaling-down pause subunit is used to pause adaptive scaling-down if the current particle swarm count is lower than the fluctuation lower limit, until the current particle swarm count is higher than the initial particle swarm count.

[0125] The adaptive expansion pause subunit is used to pause adaptive expansion if the current particle swarm number is higher than the fluctuation limit, until the current particle swarm number is lower than the initial particle swarm number.

[0126] It should be understood that the focus of the embodiments mentioned in this specification is their difference from other embodiments. The specific embodiments in the aforementioned Embodiment 1 are also applicable to the surface antistatic treatment system for polyester knitted fabric described in Embodiment 2. For the sake of brevity, they will not be elaborated further here.

[0127] It should be understood that the embodiments disclosed in this invention and the above description enable those skilled in the art to implement this invention. However, this invention is not limited to the embodiments mentioned above. It should be understood that those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this invention, and should all be included within the protection scope of this invention.

Claims

1. A method for surface antistatic treatment of polyester knitted fabric, characterized in that, include: Based on the processing technology scheme determined for the target application scenario, call the sample process records to construct a standard process training set; Based on the standard process training set, and combined with the knowledge distillation mechanism, an antistatic performance prediction model is constructed and trained. A cost evaluation factor is constructed, and a process evaluation function is defined by combining the antistatic performance prediction model and the cost evaluation factor, wherein the process evaluation function is associated with a penalty space. Using the process evaluation function as the optimization objective, the processing parameters are iteratively optimized by adaptive expansion and contraction, and the target combination of process parameters for polyester knitted fabric is determined based on the iterative optimization results, and surface antistatic treatment is performed.

2. The surface antistatic treatment method for polyester knitted fabric as described in claim 1, characterized in that, Combining the antistatic performance prediction model and the cost evaluation factor, a process evaluation function is defined, including: Obtain the antistatic treatment task sheet, parse the antistatic treatment task sheet, and extract the compliance constraints; Based on the compliance constraints, define the penalty space of the process evaluation function; The output of the antistatic performance prediction model and the cost evaluation factor are weighted and fused to obtain the process evaluation function, and the process evaluation function is associated with the penalty space.

3. The surface antistatic treatment method for polyester knitted fabric as described in claim 2, characterized in that, Using the process evaluation function as the optimization objective, the processing parameters are iteratively optimized through adaptive scaling, including: Based on the standard process training set, initialize a process parameter particle swarm containing multiple candidate process parameter combinations; Based on the initialized process parameters and particle swarm, an iterative optimization process based on the particle swarm optimization algorithm is executed. In each iteration, the optimization direction of each particle is calculated, and the bidirectional virtual intersection point of the particle in the search space is identified based on the optimization direction. Based on the bidirectional virtual intersection identification results, the process evaluation function adaptively expands and shrinks the particle swarm of process parameters. The iterative optimization results are output based on the preset iterative constraints.

4. The surface antistatic treatment method for polyester knitted fabric as described in claim 3, characterized in that, The process parameter particle swarm optimization, combined with the process evaluation function, also includes: When the optimization directions of the first preset number of particles within the preset range have a common virtual intersection point in the multidimensional parameter space, obtain the process evaluation function value corresponding to the current position of each particle, and define the fusion weight accordingly; Based on the fusion weight, a first preset number of particles are weighted and fused, and the first fusion process evaluation function value corresponding to the weighted fusion result is calculated based on the process evaluation function. The process evaluation function value at the dummy intersection point is determined based on the process evaluation function value; The spatial point corresponding to the larger value between the first fusion process evaluation function value and the virtual intersection process evaluation function value is compared and selected as a new particle and added to the process parameter particle swarm.

5. The surface antistatic treatment method for polyester knitted fabric as described in claim 3, characterized in that, The process parameter particle swarm optimization, combined with the process evaluation function, also includes: In the multidimensional parameter space, identify the reverse virtual intersection points formed by the intersection of the reverse extensions of the second preset number of optimization directions; When the reverse virtual intersection point is within a preset range, calculate the angular variance of the adjacent included angles formed by the second preset number of optimized directions; The number of new particles to be inserted is determined based on the angular variance, and new particles are inserted on the central axis of the adjacent angles of the second preset number of optimized directions in descending order of adjacent included angles.

6. The surface antistatic treatment method for polyester knitted fabric as described in claim 3, characterized in that, The process parameter particle swarm optimization, combined with the process evaluation function, also includes: A dynamic fluctuation range for the particle swarm size is set, and the particle swarm size is adaptively adjusted within the dynamic fluctuation range based on the initial particle swarm size. If the current particle swarm count is lower than the fluctuation lower limit, pause adaptive scaling down until the current particle swarm count is higher than the initial particle swarm count. If the current particle swarm size is higher than the fluctuation limit, pause adaptive expansion until the current particle swarm size is lower than the initial particle swarm size.

7. The surface antistatic treatment method for polyester knitted fabric as described in claim 1, characterized in that, Based on the processing technology scheme determined for the target application scenario, sample process records are retrieved to construct a standard process training set, including: Obtain sample process parameter data, including at least the auxiliary agent ratio, treatment temperature, treatment time and fabric moisture content; Obtain the sample antistatic performance data associated with the sample process parameter data, wherein the antistatic performance includes at least static voltage, surface charge density and half-life; The sample process parameter data and the sample antistatic performance data are preprocessed and target values ​​are calibrated. Based on the target value calibration results, they are divided into positive samples and negative samples, and the output is the standard process training set.

8. The surface antistatic treatment method for polyester knitted fabric as described in claim 7, characterized in that, Based on the aforementioned standard process training set, and combined with a knowledge distillation mechanism, an antistatic performance prediction model is constructed and trained, including: Using the standard process training set, a benchmark prediction model based on a deep neural network is constructed and trained. Based on the knowledge distillation mechanism, a simplified prediction model is trained by combining the prediction output of the benchmark prediction model with the standard process training set. The simplified prediction model is verified for accuracy. If the set performance evaluation threshold is met on both positive and negative samples in the standard process training set, the simplified prediction model is output and deployed as the antistatic performance prediction model.

9. The method for surface antistatic treatment of polyester knitted fabric as described in claim 1, characterized in that, The cost evaluation factors include at least the unit fabric processing cost, processing energy consumption, processing time, and resource consumption.

10. A surface antistatic treatment system for polyester knitted fabric, characterized in that, A method for implementing the surface antistatic treatment of polyester knitted fabric according to any one of claims 1 to 9 includes: The training set construction module is used to construct a standard process training set by calling sample process records based on the processing technology scheme determined by the target application scenario. The prediction model training module is used to construct and train an antistatic performance prediction model based on the standard process training set and in conjunction with a knowledge distillation mechanism. The evaluation function definition module is used to construct cost evaluation factors and, in combination with the antistatic performance prediction model and the cost evaluation factors, define a process evaluation function, wherein the process evaluation function is associated with a penalty space. The process parameter optimization module is used to perform adaptive expansion and contraction iterative optimization of the processing process parameters with the process evaluation function as the optimization target, and to determine the target process parameter combination for polyester knitted fabric based on the iterative optimization results for surface antistatic treatment.

Images