Control method for texturing process, training method for control network, and related apparatuses

Abstract

The present disclosure relates to the technical field of data processing, and provides a control method for a texturing process, a training method for a control network, and related apparatuses. The specific implementation solution comprises: sequentially collecting preset control parameters at a plurality of moments in a texturing process flow to obtain a first control parameter sequence; inputting the first control parameter sequence into a parameter prediction model to predict predicted parameters at a plurality of future moments; selecting a target parameter at a target moment from among the predicted parameters at the plurality of moments; on the basis of the control parameters at the plurality of moments before and after the target moment, constructing a second control parameter sequence containing the target parameter; processing the second control parameter sequence on the basis of an index prediction model to obtain an index prediction result; and when the index prediction result does not satisfy an expected value, inputting the index prediction result into a data generation model to obtain an expected control parameter for the target moment. In the present disclosure, a problem occurring at a future moment can be evaluated, and a control parameter in texturing process optimization is given accordingly.

Description

Control method of texturing process, training method of control network and related devices Technical Field

[0001] The present disclosure relates to the field of data processing technology, and in particular to a control method for a texturing process, a training method for a control network, and related devices. Background Art

[0002] In the texturing process, POY (Pre-Oriented Yarn), also known as POY raw yarn, is processed into DTY (Draw Textured Yarn), and the DTY yarn is wound on a paper tube to form a DTY yarn ingot for subsequent transportation and management.

[0003] There are many controlled components in the texturing process, and the parameter setting and control of the corresponding components have a significant impact on the final silk spindle quality.

[0004] Summary of the Invention

[0005] The present disclosure provides a texturing process control method, a control network training method and related devices, which are used to improve the texturing process flow.

[0006] According to one aspect of the present disclosure, a method for controlling a texturing process is provided, comprising:

[0007] sequentially collecting preset control parameters at multiple moments in the texturizing process to obtain a first control parameter sequence;

[0008] Inputting the first control parameter sequence into a parameter prediction model to predict the prediction parameters at multiple moments in the future;

[0009] Selecting a target parameter at a target moment from prediction parameters at multiple moments;

[0010] Constructing a second control parameter sequence including the target parameters based on the control parameters at multiple times before and after the target moment;

[0011] Processing the second control parameter sequence based on the indicator prediction model to obtain an indicator prediction result;

[0012] When the indicator prediction results do not meet the expected values, the indicator prediction results are input into the data generation model to obtain the expected control parameters for the target moment.

[0013] According to one aspect of the present disclosure, a method for training a control network for a texturing process is provided, wherein the control network includes a parameter prediction subnetwork, an indicator prediction subnetwork, and a data generation subnetwork. The method includes:

[0014] Obtaining a first training sample; the first training sample includes preset control parameters of the texturing process at multiple moments;

[0015] Inputting the first training sample into the parameter prediction subnetwork to obtain a parameter prediction value at at least one future moment;

[0016] Process the parameter prediction values ​​based on the indicator prediction sub-network to obtain the prediction indicator;

[0017] Determine the training loss based on the gap between the predicted and true metrics, and the gap between the predicted and true parameter values;

[0018] Adjust the parameter prediction subnetwork and indicator prediction subnetwork based on training loss;

[0019] When the training convergence conditions are met, the parameter prediction model corresponding to the parameter prediction subnetwork and the indicator prediction model corresponding to the indicator prediction subnetwork are obtained;

[0020] Based on the parameter prediction model and the index prediction model, the data generation subnetwork is trained to obtain the data generation model. The data generation model is used to generate the expected control parameters of the texturing machine according to the index prediction results output by the index prediction model.

[0021] According to another aspect of the present disclosure, a control device for a texturing process is provided, comprising:

[0022] a collecting unit, configured to sequentially collect preset control parameters at multiple moments in the texturizing process to obtain a first control parameter sequence;

[0023] A parameter prediction unit, configured to input the first control parameter sequence into a parameter prediction model to predict the prediction parameters at multiple moments in the future;

[0024] A selection unit, configured to select a target parameter at a target moment from prediction parameters at multiple moments;

[0025] A construction unit, configured to construct a second control parameter sequence including the target parameters based on the control parameters at multiple moments before and after the target moment;

[0026] An indicator prediction unit, configured to process the second control parameter sequence based on the indicator prediction model to obtain an indicator prediction result;

[0027] The data generation unit is used to input the indicator prediction result into the data generation model to obtain the expected control parameters for the target moment when the indicator prediction result does not meet the expected value.

[0028] According to one aspect of the present disclosure, a training device for a control network of a texturing process is provided, wherein the control network includes a parameter prediction subnetwork, an indicator prediction subnetwork, and a data generation subnetwork. The device includes:

[0029] An acquisition unit, configured to acquire a first training sample; the first training sample includes preset control parameters of the texturing process at multiple moments;

[0030] An input unit, configured to input the first training sample into the parameter prediction subnetwork to obtain a parameter prediction value at at least one future moment;

[0031] a processing unit, configured to process parameter prediction values ​​based on the indicator prediction subnetwork to obtain a prediction indicator;

[0032] A training loss unit, which is used to determine the training loss based on the gap between the predicted and true metrics, and the gap between the predicted and true parameter values;

[0033] An adjustment unit, used to adjust the parameter prediction subnetwork and the indicator prediction subnetwork based on the training loss;

[0034] A first generating unit is configured to obtain a parameter prediction model corresponding to the parameter prediction subnetwork and an indicator prediction model corresponding to the indicator prediction subnetwork when a training convergence condition is met;

[0035] The second generation unit is used to train the data generation subnetwork based on the parameter prediction model and the index prediction model to obtain a data generation model. The data generation model is used to generate the expected control parameters of the texturing machine according to the index prediction results output by the index prediction model.

[0036] According to another aspect of the present disclosure, there is provided an electronic device, comprising:

[0037] at least one processor; and

[0038] a memory communicatively connected to the at least one processor; wherein,

[0039] The memory stores instructions that can be executed by the at least one processor. The instructions are executed by the at least one processor to enable the at least one processor to perform the method of any embodiment of the present disclosure.

[0040] According to another aspect of the present disclosure, a non-transitory computer-readable storage medium storing computer instructions is provided, wherein the computer instructions are used to enable the computer to execute the method according to any embodiment of the present disclosure.

[0041] According to another aspect of the present disclosure, a computer program product is provided, comprising a computer program, which implements the method according to any embodiment of the present disclosure when executed by a processor.

[0042] In this disclosed embodiment, preset control parameters are collected, future parameters are predicted using a parameter prediction model, target parameters are selected, and a second control parameter sequence is constructed. The second control parameter sequence is processed using an indicator prediction model to obtain indicator prediction results. If the expected values ​​are not met, the data generation model is used to derive the desired control parameters, which are then used to adjust the parameters in the texturing process.

[0043] It should be understood that the contents described in this section are not intended to identify the key or important features of the embodiments of the present disclosure, nor are they intended to limit the scope of the present disclosure. Other features of the present disclosure will become readily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] The accompanying drawings are used to better understand the present solution and do not constitute a limitation of the present disclosure.

[0045] FIG1 is a schematic diagram of a texturing machine according to an embodiment of the present disclosure.

[0046] FIG2 is a flow chart of a control method for a texturing process according to an embodiment of the present disclosure.

[0047] FIG3 is a flow chart of a method for training a control network for a texturing process according to an embodiment of the present disclosure.

[0048] FIG4 is a schematic diagram of the structure of a training parameter prediction subnetwork according to an embodiment of the present disclosure.

[0049] FIG5 is a schematic diagram of a process flow for obtaining a texturing process architecture diagram according to an embodiment of the present disclosure.

[0050] FIG6 is a schematic diagram of a texturing process architecture diagram obtained by a second control parameter sequence according to an embodiment of the present disclosure.

[0051] FIG7 is a flow chart of a feature-proposing module according to an embodiment of the present disclosure.

[0052] FIG8 is a schematic structural diagram of a control device for a texturizing process according to an embodiment of the present disclosure.

[0053] FIG9 is a schematic structural diagram of a control network training device for a texturing process according to an embodiment of the present disclosure.

[0054] FIG10 is a block diagram of an electronic device for implementing the texturing process control method / control network training method according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0055] The following description of exemplary embodiments of the present disclosure is made in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be appreciated by those skilled in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope of the present disclosure. Similarly, for the sake of clarity and conciseness, descriptions of well-known functions and structures are omitted in the following description.

[0056] In addition, it should be noted that the terms "first," "second," and the like in the specification and claims of this application and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or precedence. It should be understood that the numbers used in this manner are interchangeable where appropriate so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. The implementations described in the following exemplary embodiments do not represent all implementations consistent with the present application. Instead, they are merely examples of devices and methods consistent with certain aspects of the present application as detailed in the appended claims.

[0057] A schematic diagram of a possible texturing machine is shown in Figure 1. The key components of the texturing machine include: a raw yarn rack 101, a slitter 102, a first roller 103, a first heat box 104, a cooling plate 105, a false twist component 106, a nozzle 107, a second roller 108, a second heat box 109, a third roller 110, a broken end detection device 111, and a winding component 112.

[0058] The first roller 103, the second roller 108, and the third roller 110 are used to ensure that the silk thread is processed along a predetermined path. During the texturing process, the speeds of the first roller 103, the second roller 108, and the third roller 110 are matched to each other to ensure that the silk fabric is not broken or stacked.

[0059] According to the requirements of product form, the nozzle 107 can be used to process the silk thread processed by the false twist component into a network-shaped silk thread to ensure the touch and form required by different silk threads.

[0060] When producing high-stretch yarn, the second heat box 109 is not used. When producing medium-stretch yarn, the temperature of the second heat box 109 can be adjusted to a first preset temperature, for example, the first preset temperature can be around 140°C. When producing low-stretch yarn, the temperature of the second heat box 109 can be adjusted to a second preset temperature, for example, the second preset temperature can be between 165°C and 195°C.

[0061] When the broken end detection device 111 detects a broken end, it will trigger the wire cutter 102 to cut the wire, so as to avoid the accumulation of POY raw yarn in the subsequent process.

[0062] In the texturing process, at least one of the following core control parameters may be included, such as:

[0063] 1) Vehicle speed, that is, the speed of the second roller.

[0064] 2) Draft ratio, which is the ratio of the speed of the second roller to the speed of the first roller.

[0065] 3) Second overfeed, which is the ratio of the speed difference between the third roller and the second roller to the speed of the second roller.

[0066] 4) False twist ratio, that is, the ratio of the disc speed in the false twist component to the vehicle speed.

[0067] 5) Oil tanker speed, that is, the oil tanker rotation speed. The oil tanker is located between the third roller and the winding component and is used to oil the silk thread.

[0068] 6) Full winding time, that is, the time required for the winding component to fully wind a ingot of silk.

[0069] 7) The temperature of the hot box on H1, that is, the temperature of the deformation hot box (that is, the first hot box).

[0070] 8) The temperature of the hot box under H2, that is, the temperature of the shaping hot box (that is, the second hot box).

[0071] 9) Air pressure value, that is, the air pressure value required for the nozzle to print the network.

[0072] Of course, the above parameters are only examples. During implementation, the corresponding key parameters can be determined according to specific circumstances.

[0073] The impact of the texturing process on DTY yarn quality is a complex, systemic issue. Improper texturing parameters can lead to defects in the final product. Therefore, a method is needed to optimize the texturing process to promptly identify and address issues during production, thereby improving product quality and production efficiency.

[0074] In view of this, an embodiment of the present disclosure provides a control method for a texturing process to solve the above-mentioned problem.

[0075] As shown in FIG2 , the control method for the texturizing process provided in the embodiment of the present disclosure includes the following contents:

[0076] S201 , sequentially collecting preset control parameters at multiple moments in a texturing process flow to obtain a first control parameter sequence.

[0077] In the embodiment of the present disclosure, the preset control parameters include vehicle speed, draft ratio, second overfeed, false twist ratio, cruise ship speed, full roll time, H1 upper hot box temperature, H2 lower hot box temperature, air pressure value, etc.

[0078] S202: Input the first control parameter sequence into a parameter prediction model to predict the prediction parameters at multiple moments in the future.

[0079] S203, selecting target parameters at a target moment from the prediction parameters at multiple moments.

[0080] S204: Construct a second control parameter sequence including the target parameters based on the control parameters at multiple times before and after the target time.

[0081] S205: Process the second control parameter sequence based on the indicator prediction model to obtain an indicator prediction result.

[0082] S206: When the indicator prediction result does not meet the expected value, the indicator prediction result is input into the data generation model to obtain the expected control parameters for the target time.

[0083] In the embodiment of the present disclosure, preset control parameters at multiple moments in the texturing process are collected and input into the parameter prediction model as a first control parameter sequence to predict the prediction parameters for multiple moments in the future. Then, the target parameters for the target moment are selected from these prediction parameters. Based on the control parameters of multiple moments before and after the target moment, a second control parameter sequence containing the target parameters is constructed. By processing the second control parameter sequence using the indicator prediction model, the indicator prediction results at the future target moment can be predicted. If the indicator prediction results do not meet the expected values, problems with possible parameters in the future can be discovered in a timely manner, and then the indicator prediction results are input into the data generation model to obtain the expected control parameters for the future target moment. By predicting and adjusting the control parameters, it helps to optimize the texturing process and ensure that the quality of the final product meets expectations as much as possible.

[0084] For a better understanding, the following is an introduction from two aspects: model training and model use:

[0085] 1. Model Training

[0086] The parameter prediction model, index prediction model, and data generation model described above in the embodiments of the present disclosure are obtained by training a control network. The control network includes a parameter prediction subnetwork (used to train and generate the corresponding parameter prediction model), an index prediction subnetwork (used to generate the corresponding index prediction model after training), and a data generation subnetwork (used to train and generate the corresponding data generation model).

[0087] Accordingly, the present disclosure provides a method for training a control network for a texturing process, as shown in FIG3 , which may include the following steps:

[0088] S301, obtaining a first training sample; the first training sample includes preset control parameters of the texturing process at multiple moments.

[0089] For example, data can be collected on key parameters in the actual texturing process, and multiple preset control parameters in the texturing process can be obtained at the same time. These preset control parameters are the key parameters in the texturing process.

[0090] A first training sample corresponding to time t can be constructed with time t as the center. For example, preset control parameters corresponding to L times are selected to form the first training sample. For example, preset control parameters corresponding to times t-2, t-1, t, t+1, and t+2 are selected to form the first training sample.

[0091] Therefore, a corresponding first training sample can be constructed at each sampling moment.

[0092] S302: Input the first training sample into the parameter prediction subnetwork to obtain a parameter prediction value for at least one future moment.

[0093] In the embodiment of the present disclosure, the Koopman time series model is selected as the parameter prediction subnetwork. The specific training steps are:

[0094] S3021: Detangle the input first training sample and decompose it into a time-varying component and a time-invariant component.

[0095] For example, if the training sample length is T, then perform Fourier transform on each sequence to obtain the amplitude corresponding to each frequency in the spectrum set S = {0, 1, ..., {T / 2}}. For each frequency, average the amplitude corresponding to the frequency to obtain the main frequency size of the sample population (the frequency with the highest average amplitude can be regarded as the main frequency). The frequencies with the highest average amplitude in the first α proportion are selected from the set is the time-invariant component X inv , then the remaining is the time-varying component X var The specific steps are shown in formula (1):

[0096] In formula (1), represents FFT (fast Fourier transform), It is its inverse. Filter() only passes the spectrum corresponding to the input. X represents the parameters of each sample in the training sample.

[0097] S3022: For the time-invariant component, use a time-invariant Koopman predictor to extract the embedded features of the time-invariant component as the first parameter prediction value.

[0098] As shown in Figure 4a, the time-invariant component is input into the time-invariant Koopman predictor, and the time-invariant component is mapped to R through the Encoder encoder. D The latent space of Z is as shown in expression (2): back =Encoder(X inv ) (2)

[0099] Then use the size R in the latent space D×D The learnable matrix K inv Make a transformation, that is, Z fore =K inv Z back (3)

[0100] Finally, use the decoder to map to R H×C , that is, Y inv =Decoder(Z fore ) (4)

[0101] Y inv That is to obtain the predicted value of the first parameter at the future moment.

[0102] S3023: For the time-varying component, use a time-varying Koopman predictor to extract the embedded features and residuals of the time-varying component as the second parameter prediction value.

[0103] As shown in Figure 4b, the time-varying component is input into the time-varying Koopman predictor. In order to improve the processing efficiency, the input time-varying component will be divided into multiple segments of length S (3 segments are used as an example in the figure), and each segment is marked as X. j , X j Taking any element from all segments, we can get the value as shown in expression (5): j =[X (j-1) S+1,…,X jS ] T ∈R SⅹC ,j=1,2,…,T / S (5)

[0104] In formula (5), X j Represents each time-varying component that is segmented.

[0105] And use a pair of encoder / decoder to learn the transformation to the high-dimensional linear system, that is,

[0106] Z j =Encoder(X j ),

[0107] In formula (6), Z jIndicates that the corresponding segments each obtain corresponding features through the corresponding encoder, Express The new variable obtained by decoding.

[0108] According to the corresponding features obtained, the time-varying operator is obtained using the data-driven EDMD (modal decomposition algorithm) algorithm, as shown in expression (7): back =[Z1,Z2,……,Z T / S-1 ],Z fore =[Z2,Z3,…..Z T / S ],K var =Z fore Z back (7)

[0109] Based on the obtained operator, the fitting result of the historically observed dynamic system can be calculated, as shown in expression (8):

[0110] Further extrapolation to obtain the future prediction result Z T / S+t , as shown in expression (9) Z T / S+t =(K var ) T Z T / S ,t=1,2,….,H / S (9)

[0111] Finally, these segments are recombined to obtain the module's fitting output and sequence prediction, as shown in expression (10):

[0112] In formula (10), Y is the input data of the next KoopaBlock layer, var That is to obtain the predicted value of the second parameter at the future moment.

[0113] The above S3022 and S3023 are not performed in any particular order and can be performed simultaneously in some cases.

[0114] S303: Process the parameter prediction values ​​based on the indicator prediction sub-network to obtain the prediction indicator. The implementation steps are as follows:

[0115] S3031, when the parameter prediction value is a third parameter sequence constructed by parameters at multiple moments, select a specified parameter at a specified moment from the third parameter sequence; based on the obtained specified parameter, construct control parameter samples at multiple moments with the moment of the specified parameter as a reference to obtain a fourth control parameter sequence.

[0116] S3032: Process the fourth control parameter sequence based on the indicator prediction subnetwork to obtain a prediction indicator.

[0117] In some embodiments, a corresponding texturing process architecture diagram can be generated based on the control parameters at each moment of the fourth control parameter sequence to obtain multiple texturing process architecture diagrams; the multiple texturing process architecture diagrams are input into the indicator prediction subnetwork to obtain prediction indicators.

[0118] For each control parameter at each moment in the obtained fourth control parameter sequence, the following operations (including steps A1 and A2) are performed to generate a corresponding texturing process architecture diagram:

[0119] In step A1, a plurality of sub-parameters in the control parameter are respectively standardized to map the sub-parameters to a preset value range to obtain standard values ​​of the sub-parameters.

[0120] Step A2: scaling the processing path of the POY yarn in the texturing process to the initialization image in equal proportion to obtain a texturing process architecture diagram corresponding to the moment.

[0121] Among them, the pixel points outside the processing path in the texturing process architecture diagram are set to the default value, the points corresponding to the sub-parameters in the control parameters on the processing path are set to the standard value of the sub-parameters, and the points outside the sub-parameters in the control parameters on the processing path are set to target values ​​different from the default value.

[0122] The aforementioned S3032 can be implemented as the following steps B1 to B4:

[0123] Step B1, based on the short-term feature extraction module constructed based on the attention mechanism of the indicator prediction subnetwork, feature extraction is performed within the third feature extraction range of multiple texturing process architecture diagrams to obtain short-term features; wherein, the third feature extraction range includes the texturing process architecture diagrams at n moments centered on the texturing process architecture diagram at the target moment; n is a positive integer; wherein, the short-term feature extraction module is used to use the texturing process architecture diagram at the target moment as a query feature, and use the texturing process architecture diagrams at n moments other than the target moment as key features and value features to obtain short-term features.

[0124] Step B2, based on the indicator prediction subnetwork, a long-term feature extraction module constructed based on the attention mechanism performs feature extraction within the fourth feature extraction range in multiple texturing process architecture diagrams to obtain long-term features; wherein the fourth feature extraction range includes the texturing process architecture diagrams at m moments centered on the texturing process architecture diagram at the target moment; m is a positive integer greater than n; the long-term feature extraction module is used to split the texturing process architecture diagrams at m moments using a sliding window mechanism, and determine the pooling information within each sliding window; and based on the pooling information of multiple sliding windows, obtain the key features and value features required by the attention mechanism; based on the key features, value features, and query features, obtain the long-term features.

[0125] In step B3, a fusion module is used to fuse the long-term features and the short-term features to obtain fused features.

[0126] In step B4, the prediction module based on the indicator prediction subnetwork processes the fused features to obtain the prediction indicator.

[0127] In order to make the prediction index results obtained by the indicator prediction sub-network processing the parameter prediction values ​​more accurate, the indicator prediction sub-network also includes a supplementary feature extraction module. Based on the obtained parameter prediction values, the supplementary feature extraction module performs the following operations to obtain supplementary features:

[0128] Step C1: Perform data analysis on the obtained fourth control parameter sequence to obtain the cumulative difference of parameters at each moment in the fourth control parameter sequence relative to the first moment in the fourth control parameter sequence, thereby obtaining a cumulative difference sequence. The parameter at the first moment is the first control parameter in the fourth control parameter sequence.

[0129] Step C2: Based on the cumulative difference sequence, determine the duration of the moment when the cumulative difference is greater than the preset threshold relative to the first moment in the fourth control parameter sequence to obtain a duration sequence.

[0130] Step C3: extracting features from the obtained fourth control parameter sequence, cumulative difference sequence, and duration sequence to obtain supplementary features.

[0131] In step C4, a fusion module is used to fuse the long-term features, short-term features, and supplementary features to obtain fused features.

[0132] S304 : Determine the training loss based on the gap between the predicted indicator and the true indicator, and the gap between the parameter predicted value and the true parameter value.

[0133] S305 , adjusting the parameter prediction subnetwork and the indicator prediction subnetwork based on the training loss.

[0134] S306: When the training convergence conditions are met, a parameter prediction model corresponding to the parameter prediction subnetwork and an indicator prediction model corresponding to the indicator prediction subnetwork are obtained.

[0135] In the disclosed embodiments, mean squared error can be used as a loss function to measure the gap between the predicted indicator and the true indicator, as well as the gap between the parameter prediction value and the true parameter value, to adjust the parameter prediction subnetwork and the indicator prediction subnetwork. Finally, the model is iteratively trained and the model performance is evaluated using a validation set or a test set. When the required conditions are met, the tuned parameter prediction subnetwork is determined to be the parameter prediction model, and the tuned indicator prediction subnetwork is determined to be the indicator prediction model.

[0136] S307: Based on the parameter prediction model and the indicator prediction model, the data generation subnetwork is trained to obtain a data generation model. The data generation model is used to generate desired control parameters for the texturing machine based on the indicator prediction results output by the indicator prediction model. The generation subnetwork includes a generator to be trained and a discriminator to be trained. Based on the parameter prediction model and the indicator prediction model, the data generation subnetwork is trained to obtain the data generation model.

[0137] In order to ensure the accuracy of the resulting data generation model, the data generation subnetwork needs to be trained for multiple rounds, where each round of training performs the following operations:

[0138] Step D1, using the real control parameters to input the discriminator to be trained to obtain the first discrimination result of the discriminator on the real control parameters;

[0139] Step D2, using the generator to be trained to generate false control parameters;

[0140] Step D3, inputting the false control parameters into the discriminator to be trained to obtain a second discrimination result;

[0141] Step D4, adjusting the model parameters of the discriminator to be trained based on the first discrimination result and the second discrimination result;

[0142] For example, a generator to be trained is used to generate fake control parameters. These fake control parameters and real control parameters are combined to form a true and false dataset. Based on this true and false dataset, the parameters of the generator are fixed and the discriminator to be trained is trained. For example, given a random training sample, the discriminator to be trained can be trained to determine whether it is a real control parameter or a fake sample generated by the generator to be trained. The real control parameters are input into the discriminator to be trained, and the resulting discrimination result is used as the real result. The fake control parameters are input into the discriminator to be trained, and the resulting discrimination result is used as the fake result. Using the real and fake results as references, the training of the discriminator to be trained is completed.

[0143] Step D5: fix the model parameters of the discriminator to be trained and train the generator to be trained.

[0144] After the discriminator to be trained is trained, the generator to be trained is connected in series with the previously trained discriminator to be trained, with the parameters of the discriminator to be trained fixed. The generator to be trained is given a prediction indicator output by the indicator prediction model to obtain the control parameters generated by the generator to be trained. This control parameter is then input into the discriminator to be trained for judgment. The recognition result of the discriminator to be trained is used as the processing object of the loss function to generate a loss value, which is then used to update the parameters of the generator to be trained.

[0145] After the generator is trained, its parameters are fixed and the discriminator is trained again. This process is repeated until the generator is able to generate sufficiently realistic data that the discriminator cannot accurately distinguish between real and fake data. This results in a data generation model.

[0146] In the embodiment of the present disclosure, the prediction parameters generated by the parameter prediction model and the indicator prediction results generated by the indicator prediction model are processed by the training generator and the training discriminator to obtain a data generation model, thereby providing efficient optimization efficiency for the parameters in the texturing process.

[0147] In the disclosed embodiment, a parameter prediction subnetwork is trained using a first training sample to obtain parameter prediction values. Based on the obtained parameter prediction values, an indicator prediction subnetwork is trained to obtain indicator parameters. By determining the training loss, the prediction subnetwork and the indicator prediction subnetwork are adjusted to obtain parameter prediction models and indicator prediction models. Based on these models, a data generation subnetwork is trained to obtain a data generation model. The training of the subnetworks yields a final prediction model, thereby improving the efficiency of parameter optimization during the springing process.

[0148] 2. Model Usage

[0149] The POY yarn is processed through the yarn processing path to complete the texturing process operation and obtain the desired DTY yarn. The entire silk path is designed according to the texturing process of the texturing machine. During the entire process, in addition to the processing of the corresponding key links, some mechanical properties of the texturing machine, and the processing process of the POY yarn on the silk path will also affect the final yarn quality. In order to be able to explore the impact of potential hidden features on the quality of DTY yarn, in the embodiment of the present disclosure, the second control parameter sequence is processed based on the indicator prediction model to obtain the indicator prediction result, which can be implemented as follows: based on the control parameters of each moment of the second control parameter sequence, a corresponding texturing process architecture diagram is generated to obtain multiple texturing process architecture diagrams, and then the multiple texturing process architecture diagrams are input into the indicator prediction model to obtain the indicator prediction result.

[0150] In the disclosed embodiments, the texturing process architecture diagram can demonstrate the yarn processing path in the texturing machine, as well as key parameters in key steps. By combining predicted control data with the texturing process path of the texturing machine and converting it into a graph, potential features and associations can be discovered to facilitate quality prediction at future times.

[0151] During implementation, the texturing process architecture diagram is generated in the same manner as when training the model. As shown in FIG5 , for each control parameter of the second control parameter sequence at each moment, the following operations as shown in FIG5 are performed to generate the corresponding texturing process architecture diagram:

[0152] S501: performing standardization operations on multiple sub-parameters in the control parameter to map the sub-parameters to preset value ranges to obtain standard values ​​of the sub-parameters.

[0153] For example, for each sub-parameter, the maximum value and the minimum value are calculated to determine the range of the data of the sub-parameter.

[0154] The normalization operation for sub-parameters can be calculated by the following formula: y = (x-min) / (max-min) (11)

[0155] In formula (11), y represents the conversion of each data point in the subparameter from its original range to a new value in the range of [0, 1], x represents each subparameter, min represents the minimum value in the subparameter, and max represents the maximum value in the subparameter.

[0156] In order to reduce the amount of calculation in the subsequent training process, the calculated new value is mapped to a preset value range. In this embodiment, the preset value range is 0 to 255. Therefore, the normalized value can be multiplied by 255 and then added with 0 to obtain a value within the range of 0-255.

[0157] S502, scaling the processing path of the POY yarn in the texturing process to the initialization image in proportion to obtain the texturing process architecture diagram corresponding to the moment, wherein the pixel points outside the processing path in the texturing process architecture diagram are set to default values, the points corresponding to the sub-parameters in the control parameters on the processing path are set to standard values ​​of the sub-parameters, and the points outside the sub-parameters in the control parameters on the processing path are set to target values ​​different from the default values.

[0158] In the disclosed embodiment, the default value can be set to 255 as the value for pixels outside the processing path in the initialization image. The target value can be set to 0 as the value for points outside the sub-parameters in the control parameters on the processing path. The remaining values ​​are used as the values ​​for points corresponding to the sub-parameters in the control parameters on the processing path, resulting in the texturing process architecture diagram shown in Figure 6. This can be understood as scaling the entire texturing process and corresponding components to the initial image, and the values ​​of the key links in the diagram can represent the process parameters of that link.

[0159] In the embodiment of the present disclosure, a standard value is obtained by performing a standardization operation on multiple sub-parameters in the second control parameter sequence and mapping them to a preset value range. Then, it scales the processing path of the POY yarn in the texturing process to the initialization image in proportion. The pixel points outside the processing path in the initialization image take the default value, and the points corresponding to the sub-parameters in the control parameters on the processing path are set to the standard value of the sub-parameter, and the points outside the sub-parameters are set to target values ​​different from the default value. In this way, the texturing process architecture diagram is obtained. By converting the real-time control parameters into a graphical representation, a good data foundation is provided for the subsequent indicator prediction model to mine potential correlation relationships and features.

[0160] In some embodiments, in order to further explore features and improve the accuracy of index prediction, multiple texturing process architecture diagrams are input into the index prediction model to obtain index prediction results, which can be specifically implemented as shown in FIG7 :

[0161] S701, a short-term feature extraction module constructed based on an attention mechanism of an indicator prediction model performs feature extraction within a first feature extraction range in multiple texturing process architecture diagrams to obtain short-term features; wherein the first feature extraction range includes texturing process architecture diagrams at n moments centered on the texturing process architecture diagram at a target moment; n is a positive integer; wherein the short-term feature extraction module is used to use the texturing process architecture diagram at the target moment as a query feature, and use the texturing process architecture diagrams at n moments except the target moment as key features and value features to obtain short-term features.

[0162] For example, if there is a series of texturing process architecture diagrams, each diagram represents the process status at a specific moment. In the disclosed embodiment, it is desired to predict the process status at a certain point in the future and extract short-term features based on the attention mechanism in historical data.

[0163] Therefore, we first determine the first feature extraction range, which is the texturing process architecture diagram for n moments centered on the target moment. If five moments are selected as the first feature extraction range, two moments forward and two moments backward from the current moment (the target moment) will be selected.

[0164] Next, the short-term feature extraction module performs feature extraction. This module uses the texturing process architecture diagram at the target moment as the query feature, and the texturing process architecture diagrams at the other four moments as the key and value features, respectively. By calculating the similarity between these features, the model can capture correlations and trends between different moments.

[0165] S702, a long-term feature extraction module constructed based on the attention mechanism of the indicator prediction model performs feature extraction within the second feature extraction range of multiple texturing process architecture diagrams to obtain long-term features; wherein the second feature extraction range includes the texturing process architecture diagrams at m moments centered on the texturing process architecture diagram at the target moment; m is a positive integer greater than n; the long-term feature extraction module is used to split the texturing process architecture diagrams at m moments using a sliding window mechanism, and determine the pooling information within each sliding window; and construct the key features and value features required for the attention mechanism based on the pooling information of multiple sliding windows; and obtain long-term features based on the key features, value features, and query features.

[0166] For example, if there is a texturing process architecture sequence containing 100 moments, each moment represents a different process state.

[0167] Therefore, we first determine the second feature extraction range, which is the texturing process architecture diagram for m moments centered on the target moment. If 20 moments are selected as the second feature extraction range, 10 moments forward and 10 moments backward from the current moment (the target moment) will be selected.

[0168] Next, the long-term feature extraction module performs feature extraction. This module uses a sliding window mechanism to split the elastic process architecture diagram at these 20 moments and determine the pooled information within each sliding window. For example, a window size of 5 can be set and then a sliding window can be applied over these 20 moments. Assuming that the windows may or may not overlap, pooled information for at least four windows can be obtained. Within each window, the model constructs key and value features using an attention mechanism.

[0169] S703: After obtaining the short-term features and the long-term features, a fusion module is used to fuse the long-term features and the short-term features to obtain fused features.

[0170] S704: Process the fusion features based on the prediction module to obtain an indicator prediction result.

[0171] In this disclosed embodiment, a short-term feature extraction module extracts short-term features from multiple texturing process structure diagrams, while a long-term feature extraction module extracts long-term features from multiple single-piece processes. These short-term and long-term features are then fused as input parameters for a prediction module to generate indicator prediction results. By fusion of short-term and long-term feature extraction, data parameters with varying time scales can be processed to more comprehensively capture data characteristics. This improves the performance and effectiveness of the prediction model in texturing process flows.

[0172] Furthermore, in some embodiments, in order to make the indicator prediction results obtained by the prediction model more accurate, the indicator prediction model further includes a supplementary feature extraction module, which performs the following operations based on the supplementary feature extraction module to obtain supplementary features, specifically:

[0173] The supplementary feature extraction module performs data analysis on the second control parameter sequence to obtain cumulative parameter differences at each moment in the second control parameter sequence relative to a first moment in the second control parameter sequence, thereby obtaining a cumulative difference sequence; and, based on the cumulative difference sequence, determines the duration of a moment at which the cumulative difference is greater than a preset threshold relative to the first moment, thereby obtaining a duration sequence; and performs feature extraction on the second control parameter sequence, the cumulative difference sequence, and the duration sequence to obtain supplementary features;

[0174] For example, if there is a second control parameter sequence including 10 moments, each moment represents a different control parameter value.

[0175] Therefore, it is first necessary to determine the first moment as the reference moment. In the embodiment of the present disclosure, the first moment is selected as the first moment.

[0176] Next, perform data analysis on the second control parameter sequence and calculate the cumulative difference between each moment and the first moment. For example, if the control parameter value at the first moment is 10 and the control parameter value at the second moment is 12, then the cumulative difference at that moment is 2. Similarly, a new cumulative difference sequence is obtained.

[0177] Then, based on the cumulative difference sequence, the duration of the moments where the cumulative difference is greater than a preset threshold relative to the first moment is determined. For example, if the preset threshold is 3, then all moments where the cumulative difference is greater than 3 are found and the duration between them and the first moment is calculated. This results in a new duration sequence.

[0178] Finally, the second control parameter sequence, the accumulated difference sequence, and the duration sequence are input into a feature extraction module to obtain supplementary features.

[0179] On the basis of supplementary features, a fusion module is used to fuse long-term features, short-term features and supplementary features to obtain fused features.

[0180] In this disclosed embodiment, by extracting supplementary features, more information about the second control parameter sequence is obtained, helping the model better understand and predict future texturing process architectures. By incorporating the extraction module, the prediction model is further optimized, resulting in more accurate prediction results.

[0181] In the embodiment of the present disclosure, for the texturing process, the indicator prediction result includes at least one of the following: spindle grade, qualified rate, full roll rate and dyeing uniformity.

[0182] In the disclosed embodiment, by predicting key indicators such as spindle grade, qualified rate, full roll rate and dyeing uniformity, problems can be discovered in a timely manner and measures can be taken to improve them, thereby reducing the defective rate and thus reducing production costs.

[0183] Based on the same technical concept, the present disclosure proposes a control device for a texturizing process, as shown in FIG8 , comprising:

[0184] The collecting unit 801 is used to sequentially collect preset control parameters at multiple moments in the texturizing process to obtain a first control parameter sequence;

[0185] The parameter prediction unit 802 is configured to input the first control parameter sequence into a parameter prediction model to predict the prediction parameters at multiple moments in the future;

[0186] A selection unit 803 is configured to select a target parameter at a target moment from prediction parameters at multiple moments;

[0187] A construction unit 804 is configured to construct a second control parameter sequence including target parameters based on control parameters at multiple moments before and after the target moment;

[0188] An indicator prediction unit 805 is configured to process the second control parameter sequence based on the indicator prediction model to obtain an indicator prediction result;

[0189] The data generation unit 806 is used to input the indicator prediction result into the data generation model to obtain the expected control parameters for the target time when the indicator prediction result does not meet the expected value.

[0190] In some embodiments, the indicator prediction unit includes:

[0191] a generating subunit, configured to generate a corresponding texturing process architecture diagram based on the control parameters at each moment of the second control parameter sequence, to obtain a plurality of texturing process architecture diagrams;

[0192] The prediction subunit is used to input multiple texturing process architecture diagrams into the index prediction model to obtain index prediction results.

[0193] In some embodiments, the generating subunit is specifically configured to:

[0194] For the control parameters at each moment of the second control parameter sequence, perform the following operations to generate the corresponding texturing process architecture diagram:

[0195] Performing standardization operations on multiple sub-parameters in the control parameters to map the sub-parameters to preset value ranges to obtain standard values ​​of the sub-parameters;

[0196] The processing path of the POY yarn in the texturing process is scaled to the initialization image to obtain the corresponding texturing process architecture diagram at that moment;

[0197] Among them, the pixel points outside the processing path in the texturing process architecture diagram are set to the default value, the points corresponding to the sub-parameters in the control parameters on the processing path are set to the standard value of the sub-parameters, and the points outside the sub-parameters in the control parameters on the processing path are set to target values ​​different from the default value.

[0198] In some embodiments, the prediction subunit is specifically configured to:

[0199] A short-term feature extraction module constructed based on an attention mechanism and an indicator prediction model is configured to extract features within a first feature extraction range of multiple texturing process architecture diagrams to obtain short-term features; wherein the first feature extraction range includes texturing process architecture diagrams at n moments centered on the texturing process architecture diagram at a target moment, where n is a positive integer; wherein the short-term feature extraction module is configured to use the texturing process architecture diagram at the target moment as a query feature and use the texturing process architecture diagrams at n moments other than the target moment as key features and value features to obtain short-term features;

[0200] A long-term feature extraction module based on an attention mechanism and an indicator prediction model is configured to extract features within a second feature extraction range in multiple texturing process architecture diagrams to obtain long-term features. The second feature extraction range includes texturing process architecture diagrams at m moments centered on the texturing process architecture diagram at a target moment, where m is a positive integer greater than n. The long-term feature extraction module is configured to split the texturing process architecture diagrams at the m moments using a sliding window mechanism and determine pooling information within each sliding window. The module then constructs key features and value features required for the attention mechanism based on the pooling information of the multiple sliding windows. The module then obtains long-term features based on the key features, value features, and query features.

[0201] The fusion module is used to fuse long-term features and short-term features to obtain fused features;

[0202] The prediction module based on the indicator prediction model processes the fusion features to obtain the indicator prediction results.

[0203] In some embodiments, the indicator prediction model further includes a supplementary feature extraction module, and the prediction subunit is further configured to:

[0204] The following operations are performed based on the supplementary feature extraction module to obtain supplementary features:

[0205] Performing data analysis on the second control parameter sequence to obtain cumulative parameter differences at each moment in the second control parameter sequence relative to the first moment in the second control parameter sequence, thereby obtaining a cumulative difference sequence; and

[0206] Based on the cumulative difference sequence, determine the duration of the moment when the cumulative difference is greater than a preset threshold relative to the first moment to obtain a duration sequence;

[0207] Extracting features from the second control parameter sequence, the cumulative difference sequence, and the duration sequence to obtain supplementary features;

[0208] The prediction subunit is further configured to:

[0209] The fusion module is used to fuse long-term features, short-term features and supplementary features to obtain fused features.

[0210] In some embodiments, the indicator prediction result includes at least one of the following:

[0211] Silk spindle grade, qualified rate, full roll rate and dyeing uniformity.

[0212] In an embodiment of the present disclosure, a training device for a control network of a texturing process is proposed. The control network includes a parameter prediction subnetwork, an indicator prediction subnetwork, and a data generation subnetwork, as shown in FIG9 , including:

[0213] An acquisition unit 901 is configured to acquire a first training sample; the first training sample includes preset control parameters of the texturing process at multiple moments;

[0214] An input unit 902 is configured to input a first training sample into a parameter prediction subnetwork to obtain a parameter prediction value at at least one future moment;

[0215] A processing unit 903 is configured to process the parameter prediction value based on the indicator prediction subnetwork to obtain a prediction indicator;

[0216] A training loss unit 904 is configured to determine a training loss based on the gap between the predicted indicator and the true indicator, and the gap between the parameter predicted value and the true parameter value;

[0217] An adjustment unit 905 is configured to adjust the parameter prediction subnetwork and the indicator prediction subnetwork based on the training loss;

[0218] The first generating unit 906 is configured to obtain a parameter prediction model corresponding to the parameter prediction subnetwork and an indicator prediction model corresponding to the indicator prediction subnetwork when a training convergence condition is met;

[0219] The second generating unit 907 is used to train the data generation subnetwork based on the parameter prediction model and the index prediction model to obtain a data generation model. The data generation model is used to generate the desired control parameters of the texturing machine according to the index prediction results output by the index prediction model.

[0220] In some embodiments, the data generation subnetwork includes a generator to be trained and a discriminator to be trained, and the second generation unit is specifically configured to perform multiple rounds of training on the data generation network, wherein each round of training performs the following operations:

[0221] Using the real control parameters to input the discriminator to be trained to obtain a first discrimination result of the discriminator on the real control parameters;

[0222] Generate fake control parameters using the generator to be trained;

[0223] Inputting the false control parameters into the discriminator to be trained to obtain a second discrimination result;

[0224] Adjusting the model parameters of the discriminator to be trained based on the first discrimination result and the second discrimination result;

[0225] Fix the model parameters of the discriminator to be trained and train the generator to be trained.

[0226] For the description of specific functions and examples of each module, sub-module\unit of the device in the embodiment of the present disclosure, please refer to the relevant description of the corresponding steps in the above method embodiment, which will not be repeated here.

[0227] In the technical solutions disclosed herein, the acquisition, storage, and application of user personal information involved comply with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0228] Figure 10 is a block diagram of the structure of an electronic device according to an embodiment of the present disclosure. As shown in Figure 10, the electronic device includes: a memory 1010 and a processor 1020, and the memory 1010 stores a computer program that can be run on the processor 1020. The number of memories 1010 and processors 1020 can be one or more. The memory 1010 can store one or more computer programs. When the one or more computer programs are executed by the electronic device, the electronic device executes the method provided by the above method embodiment. The electronic device may also include: a communication interface 1030 for communicating with external devices and performing data exchange and transmission.

[0229] If the memory 1010, processor 1020, and communication interface 1030 are implemented independently, the memory 1010, processor 1020, and communication interface 1030 can be interconnected via a bus and communicate with each other. This bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus. This bus can be divided into an address bus, a data bus, a control bus, etc. For ease of illustration, FIG10 shows only one thick line, but this does not mean that there is only one bus or only one type of bus.

[0230] Optionally, in a specific implementation, if the memory 1010, the processor 1020 and the communication interface 1030 are integrated on a chip, the memory 1010, the processor 1020 and the communication interface 1030 can communicate with each other through an internal interface.

[0231] It should be understood that the processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field programmable gate arrays (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor, etc. It is worth noting that the processor may be a processor that supports the Advanced RISC Machines (ARM) architecture.

[0232] Furthermore, optionally, the above-mentioned memory may include a read-only memory and a random access memory, and may also include a non-volatile random access memory. The memory may be a volatile memory or a non-volatile memory, or may include both volatile and non-volatile memories. Among them, the non-volatile memory may include a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory. The volatile memory may include a random access memory (RAM), which is used as an external cache. By way of example but not limitation, many forms of RAM are available. For example, static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link dynamic random access memory (SLDRAM) and direct memory bus random access memory (DR RAM).

[0233] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer instructions are loaded and executed on a computer, the process or function according to the embodiment of the present disclosure is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network or other programmable device. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website, computer, server or data center to another website, computer, server or data center by wired (for example: coaxial cable, optical fiber, data subscriber line (Digital Subscriber Line, DSL)) or wireless (for example: infrared, Bluetooth, microwave, etc.) means. The computer-readable storage medium can be any available medium that can be accessed by a computer, or a data storage device such as a server or data center that includes one or more available media integrated. Available media may be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., digital versatile discs (DVDs)), or semiconductor media (e.g., solid-state drives (SSDs)). It is worth noting that the computer-readable storage media mentioned in the present disclosure may be non-volatile storage media, in other words, non-transitory storage media.

[0234] Those skilled in the art will understand that all or part of the steps of implementing the above embodiments may be accomplished by hardware, or by programs instructing related hardware to accomplish the steps. The programs may be stored in a computer-readable storage medium, and the above-mentioned storage medium may be a read-only memory, a disk, or an optical disk, etc.

[0235] In the description of the embodiments of the present disclosure, the reference terms "one embodiment," "some embodiments," "example," "specific example," or "some examples" mean that the specific features, structures, materials, or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present disclosure. Moreover, the specific features, structures, materials, or characteristics described may be combined in any appropriate manner in any one or more embodiments or examples. In addition, those skilled in the art may combine and combine different embodiments or examples described in this specification, as well as features of different embodiments or examples, unless they are mutually inconsistent.

[0236] In the description of the embodiments of the present disclosure, unless otherwise specified, " / " means or. For example, A / B can mean A or B. "And / or" in this document is only a description of the association relationship between related objects, indicating that three relationships can exist. For example, A and / or B can mean: A exists alone, A and B exist at the same time, and B exists alone.

[0237] In the description of the embodiments of the present disclosure, the terms "first" and "second" are used for descriptive purposes only and should not be understood to indicate or imply relative importance or implicitly specify the number of the technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features. In the description of the embodiments of the present disclosure, unless otherwise specified, "plurality" means two or more.

[0238] The above are merely exemplary embodiments of the present disclosure and are not intended to limit the present disclosure. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present disclosure shall be included in the scope of protection of the present disclosure.

Claims

1. A control method for a texturing process, comprising: Sequentially collecting preset control parameters at multiple moments in the texturing process flow to obtain a first control parameter sequence; Inputting the first control parameter sequence into a parameter prediction model to predict prediction parameters at multiple future moments; Selecting a target parameter at a target moment from the prediction parameters at the multiple moments; Based on the control parameters at multiple moments before and after the target moment, constructing a second control parameter sequence including the target parameter; Processing the second control parameter sequence based on an index prediction model to obtain an index prediction result; In the case where the index prediction result does not meet the expected value, inputting the index prediction result into a data generation model to obtain an expected control parameter for the target moment.

2. The method according to claim 1, wherein The processing the second control parameter sequence based on an index prediction model to obtain an index prediction result includes: Generating corresponding texturing process architecture diagrams based on the control parameters at each moment of the second control parameter sequence to obtain multiple texturing process architecture diagrams; Inputting the multiple texturing process architecture diagrams into the index prediction model to obtain the index prediction result.

3. The method according to claim 2, wherein The generating corresponding texturing process architecture diagrams based on the control parameters at each moment of the second control parameter sequence to obtain multiple texturing process architecture diagrams includes: For the control parameters at each moment of the second control parameter sequence, respectively perform the following operations to generate corresponding texturing process architecture diagrams: Performing standardization operations on multiple sub-parameters in the control parameters to map the sub-parameters to a preset value range to obtain standard values of the sub-parameters; Scaling the processing path of the POY thread in the texturing process proportionally into an initialization image to obtain the texturing process architecture diagram corresponding to the moment; Wherein, the pixel values outside the processing path in the texturing process architecture diagram are default values, the points corresponding to the sub-parameters in the control parameters on the processing path are set to the standard values of the sub-parameters, and the points other than the sub-parameters in the control parameters on the processing path are set to target values different from the default values.

4. The method according to claim 2, wherein The inputting the multiple texturing process architecture diagrams into the index prediction model to obtain the index prediction result includes: Based on a short-term feature extraction module constructed based on an attention mechanism in the index prediction model, performing feature extraction within a first feature extraction range in the multiple texturing process architecture diagrams to obtain short-term features; wherein, the first feature extraction range includes the texturing process architecture diagrams at n moments centered on the texturing process architecture diagram at the target moment; n is a positive integer; wherein, the short-term feature extraction module is used to use the texturing process architecture diagram at the target moment as a query feature, and use the texturing process architecture diagrams at the n moments other than the target moment as key features and value features to obtain the short-term features; The long-term feature extraction module constructed based on the attention mechanism of the index prediction model extracts features within the second feature extraction range in the multiple texturing process architecture diagrams to obtain long-term features; wherein, the second feature extraction range includes the texturing process architecture diagrams of m moments centered on the texturing process architecture diagram at the target moment; m is a positive integer greater than n; the long-term feature extraction module is used to split the texturing process architecture diagrams of the m moments by using a sliding window mechanism, and determine the pooling information within each sliding window; and construct the key features and value features required by the attention mechanism based on the pooling information of multiple sliding windows; obtain the long-term features based on the key features, the value features, and the query features; use a fusion module to perform a fusion operation on the long-term features and the short-term features to obtain fusion features; process the fusion features based on the prediction module of the index prediction model to obtain the index prediction result.

5. The method according to claim 4, wherein the index prediction model further includes a supplementary feature extraction module, and the method further includes: performing the following operations based on the supplementary feature extraction module to obtain supplementary features: performing data analysis on the second control parameter sequence to obtain the parameter cumulative difference of each moment in the second control parameter sequence relative to the first moment in the second control parameter sequence, and obtaining a cumulative difference sequence; and, based on the cumulative difference sequence, determining the duration of the moment with the cumulative difference greater than the preset threshold relative to the first moment to obtain a duration sequence; performing feature extraction on the second control parameter sequence, the cumulative difference sequence, and the duration sequence to obtain the supplementary features; The step of using a fusion module to perform a fusion operation on the long-term features and the short-term features to obtain fusion features includes: using a fusion module to perform a fusion operation on the long-term features, the short-term features, and the supplementary features to obtain the fusion features.

6. The method according to any one of claims 1-5, wherein the index prediction result includes at least one of the following: spindle grade, qualification rate, full bobbin rate, and dyeing uniformity.

7. A training method for a control network of a texturing process, the control network includes a parameter prediction sub-network, an index prediction sub-network, and a data generation sub-network, and the method includes: obtaining a first training sample; the first training sample includes preset control parameters of the texturing process at multiple moments; inputting the first training sample into the parameter prediction sub-network to obtain parameter prediction values for at least one future moment; processing the parameter prediction values based on the index prediction sub-network to obtain predicted indexes; determining a training loss based on the gap between the predicted index and the true index, and the gap between the parameter prediction value and the true parameter value; adjusting the parameter prediction sub-network and the index prediction sub-network based on the training loss; obtaining the parameter prediction model corresponding to the parameter prediction sub-network and the index prediction model corresponding to the index prediction sub-network when the training convergence condition is satisfied; Based on the parameter prediction model and the index prediction model, train the data generation sub-network to obtain a data generation model, which is used to generate the expected control parameters of the texturing machine according to the index prediction results output by the index prediction model.

8. The method according to claim 7, wherein The data generation sub-network includes a generator to be trained and a discriminator to be trained. Training the data generation sub-network based on the parameter prediction model and the index prediction model to obtain a data generation model includes: Perform multiple rounds of training on the data generation network, and the following operations are performed in each round of training: Input the real control parameters into the discriminator to be trained to obtain the first discrimination result of the discriminator to be trained for the real control parameters; Use the generator to be trained to generate fake control parameters; Input the fake control parameters into the discriminator to be trained to obtain a second discrimination result; Adjust the model parameters of the discriminator to be trained based on the first discrimination result and the second discrimination result; Fix the model parameters of the discriminator to be trained and train the generator to be trained.

9. A control device for a texturing process, comprising: An acquisition unit for sequentially acquiring preset control parameters at multiple moments in the texturing process flow to obtain a first control parameter sequence; A parameter prediction unit for inputting the first control parameter sequence into a parameter prediction model to predict the predicted parameters at multiple future moments; A selection unit for selecting the target parameter at the target moment from the predicted parameters at the multiple moments; A construction unit for constructing a second control parameter sequence including the target parameter based on the control parameters at multiple moments before and after the target moment; An index prediction unit for processing the second control parameter sequence based on an index prediction model to obtain an index prediction result; A data generation unit for, when the index prediction result does not meet the expected value, inputting the index prediction result into the data generation model to obtain the expected control parameter for the target moment.

10. The device according to claim 9, wherein, The index prediction unit includes: A generation sub-unit for generating corresponding texturing process architecture diagrams based on the control parameters at each moment of the second control parameter sequence to obtain multiple texturing process architecture diagrams; A prediction sub-unit for inputting the multiple texturing process architecture diagrams into the index prediction model to obtain the index prediction result.

11. The apparatus according to claim 10, wherein, The generation sub-unit is specifically used for: For the control parameters at each moment of the second control parameter sequence, perform the following operations respectively to generate the corresponding texturing process architecture diagram: Perform standardization operations on multiple sub-parameters in the control parameters to map the sub-parameters to a preset value range to obtain the standard values of the sub-parameters; Scale the processing path of the POY wire in the texturing process proportionally into the initialization image to obtain the texturing process architecture diagram corresponding to the moment; Among them, the pixel values of the points outside the processing path in the texturing process architecture diagram are default values, the points corresponding to the sub-parameters in the control parameters on the processing path are set to the standard values of the sub-parameters, and the points other than the sub-parameters in the control parameters on the processing path are set to target values different from the default values.

12. The apparatus according to claim 10, wherein The prediction subunit is specifically configured to: Based on the short-term feature extraction module constructed based on the attention mechanism of the index prediction model, perform feature extraction within the first feature extraction range in the multiple texturing process architecture diagrams to obtain short-term features; wherein, the first feature extraction range includes the texturing process architecture diagrams at n moments centered on the texturing process architecture diagram at the target moment; n is a positive integer; wherein, the short-term feature extraction module is used to use the texturing process architecture diagram at the target moment as the query feature, and use the texturing process architecture diagrams at the n moments other than the target moment as the key feature and the value feature to obtain the short-term features; Based on the long-term feature extraction module constructed based on the attention mechanism of the index prediction model, perform feature extraction within the second feature extraction range in the multiple texturing process architecture diagrams to obtain long-term features; wherein, the second feature extraction range includes the texturing process architecture diagrams at m moments centered on the texturing process architecture diagram at the target moment; m is a positive integer greater than n; the long-term feature extraction module is used to split the texturing process architecture diagrams at the m moments by using a sliding window mechanism, and determine the pooling information within each sliding window; and construct the key feature and the value feature required by the attention mechanism based on the pooling information of multiple sliding windows; obtain the long-term features based on the key feature, the value feature, and the query feature; Use a fusion module to perform a fusion operation on the long-term features and the short-term features to obtain fusion features; Process the fusion features based on the prediction module of the index prediction model to obtain the index prediction result.

13. According to the device described in claim 12, the index prediction model further includes a supplementary feature extraction module, and the prediction subunit is further used to: Perform the following operations based on the supplementary feature extraction module to obtain supplementary features: Perform data analysis on the second control parameter sequence to obtain the parameter cumulative difference of each moment in the second control parameter sequence relative to the first moment in the second control parameter sequence, and obtain a cumulative difference sequence; and, Based on the cumulative difference sequence, determine the duration of the moment when the cumulative difference is greater than the preset threshold relative to the first moment to obtain a duration sequence; Perform feature extraction on the second control parameter sequence, the cumulative difference sequence, and the duration sequence to obtain the supplementary features; The prediction subunit is further used to: Use a fusion module to perform a fusion operation on the long-term features, the short-term features, and the supplementary features to obtain the fusion features.

14. According to the device described in any one of claims 9-13, the index prediction result includes at least one of the following: Spindle grade, qualification rate, full bobbin rate, and dyeing uniformity.

15. A training device for a control network of a texturing process, the control network includes a parameter prediction sub-network, an index prediction sub-network, and a data generation sub-network, including: An acquisition unit for acquiring a first training sample; The first training sample includes preset control parameters of the texturing process at multiple moments; An input unit for inputting the first training sample into the parameter prediction sub-network to obtain parameter prediction values at at least one future moment; A processing unit for processing the parameter prediction values based on the metric prediction sub-network to obtain predicted metrics; A training loss unit for determining a training loss based on the gap between the predicted metric and the true metric, and the gap between the parameter prediction value and the true parameter value; An adjustment unit for adjusting the parameter prediction sub-network and the metric prediction sub-network based on the training loss; A first generation unit for obtaining a parameter prediction model corresponding to the parameter prediction sub-network and a metric prediction model corresponding to the metric prediction sub-network when the training convergence condition is satisfied; A second generation unit for training the data generation sub-network based on the parameter prediction model and the metric prediction model to obtain a data generation model, and the data generation model is used to generate expected control parameters of the texturing machine according to the metric prediction results output by the metric prediction model.

16. The apparatus according to claim 15, wherein, The data generation sub-network includes a generator to be trained and a discriminator to be trained. The second generation unit is specifically configured to perform multiple rounds of training on the data generation network, and each round of training performs the following operations: Inputting real control parameters into the discriminator to be trained to obtain a first discrimination result of the discriminator to be trained on the real control parameters; Generating fake control parameters using the generator to be trained; Inputting the fake control parameters into the discriminator to be trained to obtain a second discrimination result; Adjusting the model parameters of the discriminator to be trained based on the first discrimination result and the second discrimination result; Fixing the model parameters of the discriminator to be trained and training the generator to be trained.

17. An electronic device, comprising: At least one processor; And A memory communicatively connected to the at least one processor; wherein, The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the method according to any one of claims 1-8.

18. A non-transitory computer-readable storage medium storing computer instructions, wherein, Computer instructions are used to cause a computer to execute the method according to any one of claims 1-8.

Images