A steam bath high multiple drafting parameter dynamic prediction system and method
By acquiring, preprocessing, and utilizing finite element analysis combined with parameter interaction influence models and multi-source heterogeneous prediction models, dynamic prediction of high-stretch parameters in steam baths is achieved. This solves the problems of parameter correlation and working condition adaptability, and improves the accuracy and efficiency of prediction.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- CHANGSHU XIANGYING SPECIAL FIBER
- Filing Date
- 2025-05-21
- Publication Date
- 2026-06-02
Smart Images
Figure CN120524756B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of high-ratio drawing parameter prediction technology in steam baths, specifically to a dynamic prediction system and method for high-ratio drawing parameters in steam baths. Background Technology
[0002] Predicting high-ratio drafting parameters using steam baths can improve yarn quality and production efficiency. Traditional methods for predicting high-ratio drafting parameters rely on empirical and experimental methods, which are time-consuming and labor-intensive. With the development of machine learning, using trained models to predict high-ratio drafting parameters can effectively improve efficiency.
[0003] Currently, existing technologies for predicting high-ratio drawing parameters in steam baths have shortcomings. On the one hand, existing technologies only predict single parameters and do not combine the influence and correlation between parameters to achieve coordinated dynamic prediction of steam bath parameters and high-ratio drawing parameters. On the other hand, the prediction parameters obtained by existing technologies using machine learning do not have the ability to adaptively adjust to changes in operating conditions and production deviations.
[0004] To address this, a dynamic prediction system and method for high-ratio stretching parameters in steam baths are proposed. Summary of the Invention
[0005] The purpose of this invention is to provide a dynamic prediction system and method for high-ratio drawing parameters in steam baths. First, a data acquisition unit acquires parameter data, equipment status data, and environmental data. Then, a data processing unit preprocesses the acquired data and uses finite element analysis to predict the drawing parameters, obtaining initial drawing prediction parameters. Next, a parameter interaction influence unit inputs the preprocessed data into a parameter interaction influence prediction model to obtain parameter adjustment data and parameter interaction influence coefficients. Finally, a drawing parameter prediction unit inputs the parameter adjustment data and parameter interaction influence coefficients into a multi-source heterogeneous drawing parameter prediction model, and uses an optimization algorithm to optimize and adjust the model's drawing parameters based on the initial drawing prediction parameters, obtaining the final drawing parameters, which are then output through an output unit. This invention can improve the accuracy of high-ratio drawing parameter prediction in steam baths.
[0006] To achieve the above objectives, the present invention provides the following technical solution:
[0007] A dynamic prediction system for high-ratio drawing parameters in a steam bath includes:
[0008] The data acquisition unit is used to acquire parameter data, device status data, and environmental data.
[0009] The data processing unit is used to preprocess the parameter data, the equipment status data and the environmental data to obtain preprocessed data, and to use finite element analysis to predict the stretching parameters to obtain the initial stretching prediction parameters.
[0010] The parameter interaction effect unit is used to input the preprocessed data into the parameter interaction effect prediction model to obtain parameter adjustment data and parameter interaction effect coefficients;
[0011] The stretching parameter prediction unit is used to input the parameter adjustment data and the parameter interaction coefficient into the multi-source heterogeneous stretching parameter prediction model, and use the optimization algorithm and the initial stretching prediction parameters to adjust the stretching parameters predicted by the model to obtain the final stretching parameters.
[0012] The output unit is used to output the final stretching parameters.
[0013] Furthermore, the parameter data includes steam bath parameters and stretching parameters; the equipment status data includes: equipment temperature, equipment pressure, equipment speed, equipment power, and equipment operating time; the environmental data includes: ambient temperature, ambient humidity, ambient air pressure, and dust concentration.
[0014] Furthermore, the data processing unit preprocesses the parameter data, the equipment status data, and the environmental data to obtain preprocessed data, and then uses finite element analysis to predict the stretching parameters to obtain the initial stretching prediction parameters. The process includes:
[0015] The parameter data, the device status data, and the environmental data are cleaned, aligned, and transformed to obtain the preprocessed data.
[0016] The initial stretching prediction parameters are obtained by performing finite element analysis on the preprocessed data.
[0017] Furthermore, the process by which the parameter interaction influence unit inputs the preprocessed data into the parameter interaction influence prediction model to obtain parameter adjustment data and parameter interaction influence coefficients includes:
[0018] The input layer of the prediction model utilizes the parameter interaction to receive preprocessed parameter data, device status data, and environmental data.
[0019] The parameter data, the equipment status data, and the environmental data are input into the multi-source feature extraction layer of the parameter interaction influence prediction model to obtain parameter feature vectors, equipment status feature vectors, and environmental feature vectors.
[0020] The device state feature vector and the environment feature vector are input into the external influence prediction layer of the parameter interaction influence prediction model, and the output external influence coefficients are used to adjust the parameter feature vector to obtain the parameter adjustment feature vector.
[0021] The parameter adjustment feature vector is input into the internal interaction influence prediction layer of the parameter interaction influence prediction model to obtain the internal interaction influence coefficient feature.
[0022] The output layer of the parameter interaction influence prediction model receives the parameter adjustment feature vector and the internal interaction influence coefficient feature, and outputs the parameter adjustment data and the parameter interaction influence coefficient.
[0023] Furthermore, the process by which the stretching parameter prediction unit inputs the parameter adjustment data and the parameter interaction coefficient into the multi-source heterogeneous stretching parameter prediction model includes:
[0024] The parameter adjustment data and the parameter interaction coefficient are input into the data embedding layer of the multi-source heterogeneous stretching parameter prediction model to obtain the parameter adjustment embedding vector and the interaction coefficient embedding vector.
[0025] The parameter adjustment embedding vector and the interaction influence coefficient embedding vector are input into the stretching parameter prediction layer of the multi-source heterogeneous stretching parameter prediction model, and the output of the stretching parameter prediction layer is adjusted using an optimization algorithm and the initial stretching prediction parameters to obtain the final stretching parameter feature vector.
[0026] The final stretching parameter feature vector is input into the output layer of the multi-source heterogeneous stretching parameter prediction model to obtain the final stretching parameter.
[0027] Furthermore, the process by which the stretching parameter prediction unit adjusts the output of the stretching parameter prediction layer using an optimization algorithm and the initial stretching prediction parameters to obtain the final stretching parameter features includes:
[0028] The initial stretching prediction parameters are input into the data embedding layer of the multi-source heterogeneous stretching parameter prediction model to obtain the initial stretching prediction parameter embedding vector.
[0029] The objective function of the optimization algorithm is constructed, the optimization algorithm is initialized using the initial stretching prediction parameter embedding vector, the output of the stretching parameter prediction layer is used as the input of the optimization algorithm, and the final stretching parameter feature under the minimized objective function is obtained using the optimization algorithm.
[0030] A method for dynamic prediction of high-ratio drawing parameters in a steam bath includes:
[0031] Acquire parameter data, device status data, and environmental data;
[0032] The parameter data, the equipment status data, and the environmental data are preprocessed to obtain preprocessed data, and the stretching parameters are predicted using finite element analysis to obtain the initial stretching prediction parameters.
[0033] The preprocessed data is input into the parameter interaction influence prediction model to obtain parameter adjustment data and parameter interaction influence coefficients.
[0034] The parameter adjustment data and the parameter interaction coefficient are input into the multi-source heterogeneous stretching parameter prediction model, and the stretching parameters predicted by the model are adjusted using the optimization algorithm and the initial stretching prediction parameters to obtain the final stretching parameters and output them.
[0035] Compared with the prior art, the beneficial effects of the present invention are as follows:
[0036] 1. This invention proposes a parameter interaction influence prediction method to obtain parameter adjustment data and parameter interaction influence coefficients. This method learns the influence relationship between equipment status and external environment on parameters, as well as the influence relationship between parameters, through a parameter interaction influence prediction model. This allows for parameter adjustment to avoid external interference. At the same time, the output parameter interaction influence coefficients can improve the prediction accuracy of subsequent stretching parameters, thereby improving the accuracy of dynamic prediction of high-stretching parameters in steam baths.
[0037] 2. This invention proposes a multi-source heterogeneous drawing parameter prediction method for predicting high-ratio drawing parameters in steam baths. This method inputs parameter adjustment data and parameter interaction coefficients into a multi-source heterogeneous drawing parameter prediction model for prediction. By utilizing the parameters after interference adjustment and the influence coefficients between parameters, the model can predict high-ratio drawing parameters in steam baths that conform to actual working conditions, thereby improving the accuracy of dynamic prediction of high-ratio drawing parameters in steam baths.
[0038] 3. This invention proposes a dynamic adjustment method for stretching prediction to achieve dynamic stretching prediction. This method introduces an optimization algorithm and initial stretching prediction parameters obtained through finite element analysis during the stretching parameter prediction process. The prediction results of the multi-source heterogeneous stretching parameter prediction model are dynamically adjusted to obtain the optimal stretching parameter prediction results, thereby effectively improving the accuracy of dynamic prediction of high-magnification stretching parameters in steam baths. Attached Figure Description
[0039] Figure 1 This is a schematic diagram of the structure of a steam bath high-ratio drawing parameter dynamic prediction system according to the present invention;
[0040] Figure 2 This is a flowchart illustrating the parameter interaction influence unit processing procedure of the present invention;
[0041] Figure 3 This is a schematic flowchart of a method for dynamically predicting high-stretch parameters in a steam bath according to the present invention. Detailed Implementation
[0042] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0043] Please see Figures 1 to 3 This invention provides a dynamic prediction system and method for high-ratio drawing parameters in a steam bath, the technical solution of which is as follows:
[0044] Example 1:
[0045] To achieve dynamic prediction of high-ratio drafting parameters in steam baths and improve the accuracy of drafting parameter prediction, a textile company used a dynamic prediction system for high-ratio drafting parameters in steam baths proposed in this invention. The structure of this system can be referenced. Figure 1 ,include:
[0046] The data acquisition unit is used to acquire parameter data, device status data, and environmental data.
[0047] Furthermore, parameter data, equipment status data, and environmental data are all obtained through sensor devices;
[0048] Furthermore, the parameter data includes steam bath parameters and stretching parameters; the equipment status data includes: equipment temperature, equipment pressure, equipment speed, equipment power, and equipment operating time; the environmental data includes: ambient temperature, ambient humidity, ambient air pressure, and dust concentration.
[0049] Furthermore, the steam bath parameters include: steam temperature, steam pressure, steam humidity, steam treatment time, etc.; the drawing parameters include: draw ratio, drawing temperature, drawing speed, tension, etc.
[0050] The following table provides reference values for obtaining the parameter data.
[0051] Table 1 Reference values for obtaining parameter data
[0052]
[0053] Furthermore, parameter data, equipment status data, and environmental data are interconnected; equipment status data and environmental data can affect changes in parameter data; and in actual operation, steam bath parameters and stretching parameters can also influence each other.
[0054] By introducing multi-source parameter data, equipment status data, and environmental data, a data foundation can be provided for subsequent prediction of parameter interaction effects and stretching parameters, thereby improving the accuracy of dynamic prediction of high-strength stretching parameters in steam baths.
[0055] The data processing unit is used to preprocess parameter data, equipment status data and environmental data to obtain preprocessed data, and to use finite element analysis to predict the stretching parameters to obtain the initial stretching prediction parameters.
[0056] Furthermore, the data processing unit preprocesses the parameter data, equipment status data, and environmental data to obtain preprocessed data, and then uses finite element analysis to predict the stretching parameters, obtaining the initial stretching prediction parameters. The process includes:
[0057] The parameter data, equipment status data, and environmental data are cleaned, aligned, and transformed to obtain preprocessed data.
[0058] Finite element analysis was used to obtain the initial stretching prediction parameters from the preprocessed data;
[0059] Furthermore, data cleaning includes: handling missing values, outliers, and duplicate values; data alignment mainly involves time alignment to ensure that the data are on the same time reference; data transformation utilizes standardization, nonlinear transformation, and feature encoding to convert the data into a form suitable for modeling and analysis, thereby improving the performance and interpretability of the model.
[0060] Furthermore, the process of obtaining the initial draw prediction parameters by using finite element analysis on the preprocessed data is as follows:
[0061] A three-dimensional geometric model was established based on the structural parameters of the drawing equipment;
[0062] Import the completed model and material properties into the finite element analysis software;
[0063] Load the parameter data from the preprocessed data and set boundary conditions based on the device status data and environmental data;
[0064] Select an appropriate solver based on the model and start the solution program of the finite element analysis software to obtain the initial stretching prediction parameters.
[0065] By preprocessing the data received by the data acquisition unit, the reliability of subsequent model inputs can be improved. By using finite element analysis to process the data and obtain initial stretching prediction parameters, initial solutions can be provided for subsequent dynamic prediction of stretching parameters, accelerating the process of stretching parameter adjustment and optimization, thereby improving the accuracy of dynamic prediction of high-magnification stretching parameters in steam baths.
[0066] The parameter interaction effect unit is used to input preprocessed data into the parameter interaction effect prediction model to obtain parameter adjustment data and parameter interaction effect coefficients;
[0067] Furthermore, the structure of the parameter interaction influence prediction model includes: an input layer, a multi-source feature extraction layer, an external influence prediction layer, an internal interaction influence prediction layer, and an output layer;
[0068] Furthermore, the flowchart illustrating the parameter interaction affecting the unit processing is as follows: Figure 2 As shown, the details are as follows:
[0069] The input layer of the prediction model utilizes parameter interaction to receive preprocessed parameter data, equipment status data, and environmental data.
[0070] The parameter data, equipment status data, and environmental data are input into the multi-source feature extraction layer of the parameter interaction influence prediction model to obtain parameter feature vectors, equipment status feature vectors, and environmental feature vectors.
[0071] The equipment state feature vector and environmental feature vector are input into the external influence prediction layer of the parameter interaction influence prediction model, and the output external influence coefficients are used to adjust the parameter feature vector to obtain the parameter adjustment feature vector.
[0072] The parameter adjustment feature vector is input into the internal interaction effect prediction layer of the parameter interaction effect prediction model to obtain the internal interaction effect coefficient feature.
[0073] The output layer of the prediction model utilizing parameter interaction influence receives parameter adjustment feature vectors and internal interaction influence coefficients, and outputs parameter adjustment data and parameter interaction influence coefficients.
[0074] Furthermore, the multi-source feature extraction layer utilizes three independent autoencoders to handle feature extraction tasks for different data types respectively;
[0075] Furthermore, the external impact prediction layer is used to predict the degree of influence of equipment status and environmental data on different parameter data. The external impact prediction layer uses a hybrid causal neural network for prediction processing. The hybrid causal neural network includes a causal processing module, which incorporates causal graph information into the neural network through graph convolution operations to model the causal relationship between equipment status and environmental data and parameters. The causal graph information can be obtained by using causal discovery algorithms, such as the PC algorithm and the GES algorithm.
[0076] Furthermore, the internal interaction effect prediction layer uses a combination of temporal convolutional networks and multi-scale convolutions for processing; it uses convolutional kernels of different sizes to capture interaction relationships at different scales; and the internal interaction effect coefficient features obtained through the internal interaction effect prediction layer reflect the degree of influence of steam bath parameters on stretching parameters.
[0077] The following are three sets of parameter adjustment data and parameter interaction coefficients from test data at different time periods; the test data are denoted as Data 1, Data 2, and Data 3, respectively; due to the large number of parameter types and parameter interaction types, this embodiment selects steam temperature as the adjustment reference and steam temperature-drawing speed as the interaction reference to obtain the parameter interaction effect test results for each set, as shown in Table 2; where, the parameter adjustment data represents the steam temperature adjusted after being affected by equipment status and environmental factors; the parameter interaction coefficient represents the degree of influence of steam temperature on drawing speed.
[0078] Table 2 Test Results of Parameter Interaction Effects
[0079]
[0080] By using a parameter interaction influence prediction model, the influence relationship between equipment status and external environment on parameters, as well as the influence relationship between parameters, can be learned. This allows for parameter adjustments to avoid external interference. At the same time, the output parameter interaction influence coefficient can improve the prediction accuracy of subsequent drawing parameters, thereby improving the accuracy of dynamic prediction of high-magnification drawing parameters in steam baths.
[0081] The stretching parameter prediction unit is used to input parameter adjustment data and parameter interaction coefficients into the multi-source heterogeneous stretching parameter prediction model, and to adjust the stretching parameters predicted by the model using optimization algorithms and initial stretching prediction parameters to obtain the final stretching parameters.
[0082] Furthermore, the process by which the stretching parameter prediction unit inputs parameter adjustment data and parameter interaction coefficients into the multi-source heterogeneous stretching parameter prediction model includes:
[0083] The parameter adjustment data and parameter interaction coefficients are input into the data embedding layer of the multi-source heterogeneous stretching parameter prediction model to obtain the parameter adjustment embedding vector and the interaction coefficient embedding vector.
[0084] The parameter adjustment embedding vector and the interaction influence coefficient embedding vector are input into the stretching parameter prediction layer of the multi-source heterogeneous stretching parameter prediction model, and the output of the stretching parameter prediction layer is adjusted using the optimization algorithm and the initial stretching prediction parameters to obtain the final stretching parameter feature vector.
[0085] The final stretching parameter feature vector is input into the output layer of the multi-source heterogeneous stretching parameter prediction model to obtain the final stretching parameters;
[0086] Furthermore, the data embedding layer uses a linear layer to convert the parameter adjustment data and parameter interaction coefficients into continuous, low-dimensional embedding vectors, which facilitates subsequent calculations.
[0087] Furthermore, the stretching parameter prediction layer uses a cross-modal Transformer to perform the prediction task. The parameter adjustment embedding vector and the interaction influence coefficient embedding vector are fed as a whole into the cross-modal Transformer encoder for cross-modal attention mechanism processing to learn the interaction relationship between different parameter types. Then, the output of the cross-modal Transformer encoder is fused to obtain the stretching parameter feature vector.
[0088] Furthermore, the output layer uses a fully connected layer to map the stretching parameter feature vector output by the encoder to the predicted stretching parameter values.
[0089] By inputting parameter adjustment data and parameter interaction coefficients into a multi-source heterogeneous drawing parameter prediction model, the model can predict steam bath high-ratio drawing parameters that conform to actual working conditions, thereby improving the accuracy of dynamic prediction of steam bath high-ratio drawing parameters.
[0090] Furthermore, the process by which the stretching parameter prediction unit adjusts the output of the stretching parameter prediction layer using an optimization algorithm and initial stretching prediction parameters to obtain the final stretching parameter features includes:
[0091] The initial stretching prediction parameters are input into the data embedding layer of the multi-source heterogeneous stretching parameter prediction model to obtain the initial stretching prediction parameter embedding vector.
[0092] The objective function of the optimization algorithm is constructed, the optimization algorithm is initialized using the initial stretching prediction parameter embedding vector, the output of the stretching parameter prediction layer is used as the input of the optimization algorithm, and the final stretching parameter features under the minimized objective function are obtained using the optimization algorithm.
[0093] Furthermore, the objective function includes performance indicators, cost indicators, and constraint penalty terms. Performance indicators are obtained by comparing the predicted drawing parameters with the standard drawing parameters corresponding to the performance requirements. Cost indicators are production cost-related indicators calculated from equipment status data and environmental data, such as energy consumption and material consumption. Constraint penalty terms are used to penalize solutions that violate constraints, i.e., when the value of the drawing parameters obtained through optimization exceeds the maximum value allowed by the equipment, this situation needs to be suppressed. At the same time, the penalty value of each parameter in the constraint penalty term is different and needs to be adjusted according to the actual situation.
[0094] Furthermore, optimization algorithms can employ genetic algorithms, particle swarm optimization, ant colony optimization, etc. Before optimization, the initial stretching prediction parameter embedding vector needs to be set as the initial solution, which can improve optimization efficiency.
[0095] By introducing optimization algorithms and initial drawing prediction parameters obtained from finite element analysis into the drawing parameter prediction process, the prediction results of the multi-source heterogeneous drawing parameter prediction model are dynamically adjusted to obtain the optimal drawing parameter prediction results, thereby effectively improving the accuracy of dynamic prediction of high-magnification drawing parameters in steam baths.
[0096] The output unit is used to output the final stretching parameters.
[0097] This embodiment proposes a dynamic prediction system for high-ratio drawing parameters in a steam bath. First, a data acquisition unit acquires parameter data, equipment status data, and environmental data. Then, a data processing unit preprocesses the acquired data and uses finite element analysis to predict the drawing parameters, obtaining initial drawing prediction parameters. Next, a parameter interaction influence unit inputs the preprocessed data into a parameter interaction influence prediction model to obtain parameter adjustment data and parameter interaction influence coefficients. Finally, a drawing parameter prediction unit inputs the parameter adjustment data and parameter interaction influence coefficients into a multi-source heterogeneous drawing parameter prediction model, and uses an optimization algorithm to optimize and adjust the model's drawing parameters based on the initial drawing prediction parameters, obtaining the final drawing parameters, which are then output through an output unit. This invention can improve the accuracy of high-ratio drawing parameter prediction in a steam bath.
[0098] Example 2:
[0099] This invention also proposes a dynamic prediction method for high-ratio drawing parameters in a steam bath, the flowchart of which is shown below. Figure 3 As shown, it includes:
[0100] Acquire parameter data, device status data, and environmental data;
[0101] The parameter data, equipment status data, and environmental data are preprocessed to obtain preprocessed data, and the initial stretching prediction parameters are obtained by using finite element analysis.
[0102] The preprocessed data is input into the parameter interaction effect prediction model to obtain parameter adjustment data and parameter interaction effect coefficients;
[0103] Furthermore, the process of inputting the preprocessed data into the parameter interaction effect prediction model to obtain the parameter adjustment data and parameter interaction effect coefficients includes:
[0104] The input layer of the prediction model utilizes parameter interaction to receive preprocessed parameter data, equipment status data, and environmental data.
[0105] The parameter data, equipment status data, and environmental data are input into the multi-source feature extraction layer of the parameter interaction influence prediction model to obtain parameter feature vectors, equipment status feature vectors, and environmental feature vectors.
[0106] The equipment state feature vector and environmental feature vector are input into the external influence prediction layer of the parameter interaction influence prediction model, and the output external influence coefficients are used to adjust the parameter feature vector to obtain the parameter adjustment feature vector.
[0107] The parameter adjustment feature vector is input into the internal interaction effect prediction layer of the parameter interaction effect prediction model to obtain the internal interaction effect coefficient feature.
[0108] The output layer of the parameter interaction influence prediction model receives parameter adjustment feature vectors and internal interaction influence coefficients, and outputs parameter adjustment data and parameter interaction influence coefficients.
[0109] The parameter adjustment data and parameter interaction coefficients are input into the multi-source heterogeneous stretching parameter prediction model. The stretching parameters predicted by the model are adjusted using the optimization algorithm and the initial stretching prediction parameters to obtain the final stretching parameters and output them.
[0110] Furthermore, the process of inputting parameter adjustment data and parameter interaction coefficients into the multi-source heterogeneous stretching parameter prediction model to obtain the final stretching parameters includes:
[0111] The parameter adjustment data and parameter interaction coefficients are input into the data embedding layer of the multi-source heterogeneous stretching parameter prediction model to obtain the parameter adjustment embedding vector and the interaction coefficient embedding vector.
[0112] The parameter adjustment embedding vector and the interaction influence coefficient embedding vector are input into the stretching parameter prediction layer of the multi-source heterogeneous stretching parameter prediction model, and the output of the stretching parameter prediction layer is adjusted using the optimization algorithm and the initial stretching prediction parameters to obtain the final stretching parameter feature vector.
[0113] The final stretching parameter feature vector is input into the output layer of the multi-source heterogeneous stretching parameter prediction model to obtain the final stretching parameters.
[0114] To verify the effectiveness of the proposed method for predicting stretching parameters, 600 sets of preprocessed data were randomly selected as test data to test the effectiveness of the stretching parameter prediction method. The test data were applied to three different prediction methods to obtain the stretching parameter prediction results of each method. The proportion of prediction results of each method within a reasonable range was obtained through manual verification.
[0115] The prediction schemes are as follows: the "multi-source heterogeneous stretching parameter prediction + optimization algorithm adjustment" scheme proposed in this invention is referred to as Scheme 1; the parameter interaction coefficient used in the prediction process of this invention is removed, while the rest remain unchanged, which is referred to as Scheme 2; the process of using the optimization algorithm to optimize and adjust the initial stretching prediction parameters in this invention is removed, which is referred to as Scheme 3.
[0116] The results of the effectiveness test of the stretching parameter prediction scheme are shown in Table 3.
[0117] Table 3 Results of the effectiveness test of the stretching parameter prediction scheme
[0118]
[0119] As shown in Table 3, the prediction results obtained by using the drawing parameter prediction scheme (Scheme 1) proposed in this invention are better than those obtained by using other schemes. This indicates that it is necessary to add the parameter interaction coefficient and optimize the algorithm adjustment steps in the drawing parameter prediction process, thereby improving the accuracy of dynamic prediction of high-magnification drawing parameters in steam bath.
[0120] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.
Claims
1. A dynamic prediction system for high-ratio drawing parameters in a steam bath, characterized in that, include: The data acquisition unit is used to acquire parameter data, device status data, and environmental data. The data processing unit is used to preprocess parameter data, equipment status data and environmental data to obtain preprocessed data, and to use finite element analysis to predict the stretching parameters to obtain the initial stretching prediction parameters. The parameter interaction influence unit is used to input preprocessed data into the parameter interaction influence prediction model to obtain parameter feature vectors, equipment state feature vectors, and environmental feature vectors; input the equipment state feature vectors and environmental feature vectors into the external influence prediction layer of the parameter interaction influence prediction model, and use the output external influence coefficients to adjust the parameter feature vectors to obtain parameter adjustment feature vectors; The external influence prediction layer employs a hybrid causal neural network for prediction processing. The causal processing module incorporates causal graph information into the neural network through graph convolution operations to model the causal relationships between equipment status and environmental data and parameters. The causal graph information is obtained using a causal discovery algorithm. The parameter adjustment feature vector is input into the internal interaction effect prediction layer of the parameter interaction effect prediction model to obtain the internal interaction effect coefficient feature. The internal interaction effect prediction layer is processed by a combination of temporal convolutional network and multi-scale convolution; different sizes of convolutional kernels are used to capture interaction relationships at different scales; the internal interaction effect coefficient feature obtained by the internal interaction effect prediction layer reflects the degree of influence of steam bath parameters on stretching parameters. By utilizing the output layer of the parameter interaction influence prediction model, which receives parameter adjustment feature vectors and internal interaction influence coefficients, parameter adjustment data and parameter interaction influence coefficients are obtained. The drawing parameter prediction unit is used to input parameter adjustment data and parameter interaction coefficients into the multi-source heterogeneous drawing parameter prediction model, and adjust the drawing parameters predicted by the model using an optimization algorithm and initial drawing prediction parameters to obtain the final drawing parameters. The objective function of the optimization algorithm is constructed, and the optimization algorithm is initialized using the initial drawing prediction parameter embedding vector. The output of the drawing parameter prediction layer is used as the input of the optimization algorithm. The objective function includes a constraint penalty term. The constraint penalty term suppresses the drawing parameter values obtained through optimization from exceeding the maximum allowable value of the equipment. The optimization algorithm is used to obtain the final drawing parameter characteristics under the minimized objective function. The output unit is used to output the final stretching parameters.
2. The dynamic prediction system for high-ratio drawing parameters in a steam bath according to claim 1, characterized in that, The parameter data includes steam bath parameters and stretching parameters; the equipment status data includes: equipment temperature, equipment pressure, equipment speed, equipment power, and equipment running time; the environmental data includes: ambient temperature, ambient humidity, ambient air pressure, and dust concentration.
3. The dynamic prediction system for high-ratio drawing parameters in a steam bath according to claim 1, characterized in that, The data processing unit preprocesses the parameter data, equipment status data, and environmental data to obtain preprocessed data, and uses finite element analysis to predict the stretching parameters to obtain the initial stretching prediction parameters. The process includes: cleaning, aligning, and transforming the parameter data, equipment status data, and environmental data to obtain the preprocessed data; and using finite element analysis on the preprocessed data to obtain the initial stretching prediction parameters.
4. The dynamic prediction system for high-ratio drawing parameters in a steam bath according to claim 1, characterized in that, The process by which the drawing parameter prediction unit inputs the parameter adjustment data and the parameter interaction coefficients into the multi-source heterogeneous drawing parameter prediction model includes: inputting the parameter adjustment data and the parameter interaction coefficients into the data embedding layer of the multi-source heterogeneous drawing parameter prediction model to obtain parameter adjustment embedding vectors and interaction coefficient embedding vectors; inputting the parameter adjustment embedding vectors and the interaction coefficient embedding vectors into the drawing parameter prediction layer of the multi-source heterogeneous drawing parameter prediction model, and adjusting the output of the drawing parameter prediction layer using an optimization algorithm and the initial drawing prediction parameters to obtain the final drawing parameter feature vector; and inputting the final drawing parameter feature vector into the output layer of the multi-source heterogeneous drawing parameter prediction model to obtain the final drawing parameters.
5. The dynamic prediction system for high-ratio drawing parameters in a steam bath according to claim 4, characterized in that, The process by which the stretching parameter prediction unit adjusts the output of the stretching parameter prediction layer using an optimization algorithm and the initial stretching prediction parameters to obtain the final stretching parameter features includes: inputting the initial stretching prediction parameters into the data embedding layer of the multi-source heterogeneous stretching parameter prediction model to obtain the initial stretching prediction parameter embedding vector.
6. A method for dynamic prediction of high-ratio drawing parameters in a steam bath, characterized in that, include: Acquire parameter data, device status data, and environmental data; The parameter data, equipment status data, and environmental data are preprocessed to obtain preprocessed data, and the initial stretching prediction parameters are obtained by using finite element analysis. The preprocessed data is input into the parameter interaction influence prediction model to obtain parameter feature vectors, equipment state feature vectors, and environmental feature vectors. The equipment state feature vectors and environmental feature vectors are input into the external influence prediction layer of the parameter interaction influence prediction model, and the output external influence coefficients are used to adjust the parameter feature vectors to obtain the parameter adjustment feature vectors. The external influence prediction layer employs a hybrid causal neural network for prediction processing. The causal processing module incorporates causal graph information into the neural network through graph convolution operations to model the causal relationships between equipment status and environmental data and parameters. The causal graph information is obtained using a causal discovery algorithm. The parameter adjustment feature vector is input into the internal interaction effect prediction layer of the parameter interaction effect prediction model to obtain the internal interaction effect coefficient feature. The internal interaction effect prediction layer is processed by a combination of temporal convolutional network and multi-scale convolution; different sizes of convolutional kernels are used to capture interaction relationships at different scales; the internal interaction effect coefficient feature obtained by the internal interaction effect prediction layer reflects the degree of influence of steam bath parameters on stretching parameters. By utilizing the output layer of the parameter interaction influence prediction model, which receives parameter adjustment feature vectors and internal interaction influence coefficients, parameter adjustment data and parameter interaction influence coefficients are obtained. The output layer of the parameter interaction influence prediction model receives the parameter adjustment feature vector and the internal interaction influence coefficient feature to obtain parameter adjustment data and parameter interaction influence coefficient. The parameter adjustment data and parameter interaction coefficients are input into the multi-source heterogeneous drawing parameter prediction model. The model-predicted drawing parameters are adjusted using an optimization algorithm and the initial drawing prediction parameters to obtain the final drawing parameters. The objective function of the optimization algorithm is constructed, and the optimization algorithm is initialized using the embedding vector of the initial drawing prediction parameters. The output of the drawing parameter prediction layer is used as the input of the optimization algorithm. The objective function includes a constraint penalty term. The constraint penalty term suppresses the drawing parameter values obtained through optimization from exceeding the maximum value allowed by the equipment. The final drawing parameter features under the minimized objective function are obtained using the optimization algorithm and output.