Strip shape prediction method and device based on Bayesian data fusion

Abstract

The invention discloses a plate shape prediction method and device based on Bayesian data fusion, and relates to the technical field of rolling control. The method comprises the steps that finite element working condition data of to-be-predicted equipment are determined based on a pre-constructed finite element model, and the finite element model is a simulation model of the to-be-predicted equipment; acquiring actual working condition data of to-be-predicted equipment, and performing Bayesian data fusion on the actual working condition data and the finite element working condition data to obtain fused working condition data; taking the fused working condition data as training data, training a random forest model until the random forest model meets a preset training standard, and obtaining a trained plate shape prediction model; and determining a plate shape coefficient corresponding to the to-be-predicted equipment based on the plate shape prediction model and the to-be-predicted working condition data. By adopting the Bayesian data fusion method, the finite element working condition data and the actual working condition data can be effectively integrated, so that the working condition prediction precision and robustness are remarkably improved, and the plate shape quality can be accurately evaluated.

Description

Technical Field

[0001] The present invention relates to the technical field of rolling control, and in particular to a plate shape prediction method and device based on Bayesian data fusion. Background Art

[0002] The quality of flatness has become a key concern for manufacturers and users. Strip steel with noticeable wavy shapes can severely impact the user experience and even the market share of strip steel. Flatness control requires that flatness information be processed by a flatness control system. The flatness deviation is calculated based on the target and actual flatness, and the appropriate actuator is selected and the adjustment amount is calculated to minimize the flatness deviation. Different flatness defects require different actuators, so accurately identifying the current flatness defect or predicting the future flatness defect based on parameter changes, and then implementing appropriate control strategies in advance, are key areas of flatness control research.

[0003] Currently, non-mechanistic modeling based on data-driven data is often used to predict plate shape. However, in the complex process of steel rolling, the accuracy and real-time nature of process data directly impacts production efficiency and product quality. Existing process data processing methods typically rely on data from a single source, such as finite element simulation data or actual production data. While finite element data can provide a relatively accurate theoretical model, it often fails to fully reflect the variability and complexity of the actual production process. Actual production data can be affected by noise, missing values, and measurement errors, resulting in an inaccurate estimate of the process conditions. Summary of the Invention

[0004] In order to solve at least one technical problem in the background technology section, the present application provides a plate shape prediction method and device based on Bayesian data fusion, which can effectively integrate finite element working condition data and actual working condition data, thereby significantly improving the accuracy and robustness of working condition prediction, so as to accurately evaluate the plate shape quality.

[0005] In a first aspect, an embodiment of the present invention provides a method for predicting flatness based on Bayesian data fusion, the method comprising:

[0006] Determining finite element operating condition data of the device to be predicted based on a pre-built finite element model, wherein the finite element model is a simulation model of the device to be predicted;

[0007] Acquiring actual operating condition data of the equipment to be predicted, and performing Bayesian data fusion on the actual operating condition data and the finite element operating condition data to obtain fused operating condition data;

[0008] Using the fused working condition data as training data, training a random forest model until the random forest model meets a preset training standard, thereby obtaining a trained flatness prediction model;

[0009] Based on the shape prediction model and the operating condition data to be predicted, a shape coefficient corresponding to the equipment to be predicted is determined.

[0010] In some optional aspects of this embodiment, the step of constructing the finite element model includes:

[0011] Obtaining geometric parameters and material parameters of the device to be predicted;

[0012] Invoking a solid simulation tool, and constructing the finite element model based on the solid simulation tool, geometric parameters, and material parameters;

[0013] Among them, the geometric parameters include the working roll diameter, the working roll body length, the spacing between the working roll bending force concentration points, the support roll diameter, the support roll body length and the spacing between the support roll support reaction force concentration points; the material parameters include the density of the rolled product and the roll, the elastic modulus of the rolled product, the Poisson's ratio of the rolled product, the elastic modulus of the roll and the Poisson's ratio of the roll.

[0014] In some optional aspects of this embodiment, determining the finite element operating condition data of the equipment to be predicted based on the pre-built finite element model includes:

[0015] Acquiring process parameters and inlet plate shape defect parameters of the equipment to be predicted, and determining an initial operating condition parameter combination based on the process parameters and the inlet plate shape defect parameters;

[0016] Determining, based on the finite element model and the physical simulation tool, a first influence weight of each of the process parameters on the output plate shape and a second influence weight of each of the inlet plate shape defect parameters on the output plate shape;

[0017] Remove, from the initial operating condition parameter combination, process parameters whose first influence weight is less than a preset weight threshold and inlet flatness defect parameters whose second influence weight is less than the preset weight threshold, to obtain a candidate operating condition parameter combination;

[0018] Zero mean processing is performed on each candidate operating condition parameter in the candidate operating condition parameter combination, and outliers in each candidate operating condition parameter are removed according to the 3 sigma principle to obtain the finite element operating condition data.

[0019] In some optional aspects of this embodiment, performing Bayesian data fusion on the actual operating condition data and the finite element operating condition data to obtain fused operating condition data includes:

[0020] Determining a prior distribution based on the mean and variance of the finite element working condition data;

[0021] Determining a likelihood function based on the mean and variance of the actual operating condition data;

[0022] Determining a marginal likelihood based on the likelihood function and the prior distribution;

[0023] determining a posterior distribution based on the likelihood function, the prior distribution, and the marginal likelihood;

[0024] Based on the posterior distribution, Bayesian data fusion is performed on the actual working condition data and the finite element working condition data to obtain the fused working condition data.

[0025] In some optional aspects of this embodiment, the method further includes: determining an output shape coefficient corresponding to the finite element working condition data;

[0026] The method of training a random forest model using the fused operating condition data as training data includes:

[0027] Training a random forest model using the fused working condition data as training input data and the output shape coefficient as training output data;

[0028] In the random forest model, the number of decision trees is 50, the maximum number of features used when each decision tree is split is 4, the maximum depth of the decision tree is 4, and the minimum number of samples required to split a node is 30.

[0029] In a second aspect, an embodiment of the present invention further provides a flatness prediction device based on Bayesian data fusion, comprising:

[0030] a finite element operating condition data determination module, configured to determine finite element operating condition data of the device to be predicted based on a pre-built finite element model, wherein the finite element model is a simulation model of the device to be predicted;

[0031] a fused operating condition data determination module, configured to obtain actual operating condition data of the equipment to be predicted, and perform Bayesian data fusion on the actual operating condition data and the finite element operating condition data to obtain fused operating condition data;

[0032] a flatness prediction model training module configured to train a random forest model using the fused operating condition data as training data until the random forest model meets a preset training standard, thereby obtaining a trained flatness prediction model;

[0033] The shape coefficient prediction module is configured to determine the shape coefficient corresponding to the equipment to be predicted based on the shape prediction model and the working condition data to be predicted.

[0034] In some optional aspects of this embodiment, the fusion operating condition data determination module includes:

[0035] a prior distribution determining unit, configured to determine a prior distribution based on the mean and variance of the finite element working condition data;

[0036] a likelihood function determining unit, configured to determine a likelihood function based on a mean and a variance of the actual operating condition data;

[0037] a marginal likelihood determination unit configured to determine a marginal likelihood based on the likelihood function and the prior distribution;

[0038] a posterior distribution determining unit configured to determine a posterior distribution based on the likelihood function, the prior distribution, and the marginal likelihood;

[0039] The fused operating condition characteristic data determining unit is configured to perform Bayesian data fusion on the actual operating condition data and the finite element operating condition data based on the posterior distribution to obtain the fused operating condition characteristic data.

[0040] In a third aspect, an embodiment of the present invention further provides an electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned plate shape prediction method based on Bayesian data fusion when executing the computer program.

[0041] An embodiment of the present invention further provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, the computer program implements the above-mentioned plate shape prediction method based on Bayesian data fusion.

[0042] An embodiment of the present invention further provides a computer program product, which includes a computer program. When the computer program is executed by a processor, it implements the above-mentioned plate shape prediction method based on Bayesian data fusion.

[0043] An embodiment of the present invention provides a plate shape prediction method and device based on Bayesian data fusion, which fuses finite element working condition data and actual production working condition data through Bayesian fusion. It can comprehensively consider the theoretical accuracy of the finite element model and the actual situation of the production data, dynamically update the estimation of the working condition parameters, establish a data-driven plate shape prediction model, and predict the plate shape, which is convenient for the subsequent calculation of the control efficiency of the actuator. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative work. In the drawings:

[0045] Figure 1 is an exemplary system architecture diagram in which an embodiment of the present application may be applied;

[0046] Figure 2 This is a flow chart of a method for predicting flatness based on Bayesian data fusion according to an embodiment of the present invention;

[0047] Figure 3 This is one of the structural diagrams of the finite element model in an embodiment of the present invention;

[0048] Figure 4 This is the second structural diagram of the finite element model in an embodiment of the present invention;

[0049] Figure 5 This is a second flow chart of a method for predicting flatness based on Bayesian data fusion in an embodiment of the present invention;

[0050] Figure 6 This is a third flow chart of a method for predicting flatness based on Bayesian data fusion in an embodiment of the present invention;

[0051] Figure 7 Schematic diagram of the training process of the flatness prediction model in an embodiment of the present invention;

[0052] Figure 8 Schematic diagram of the structure of a flatness prediction device based on Bayesian data fusion in an embodiment of the present invention;

[0053] Figure 9 This is a schematic structural diagram of a finite element working condition data determination module in an embodiment of the present invention;

[0054] Figure 10 This is a structural diagram of a fusion working condition data determination module in an embodiment of the present invention;

[0055] Figure 11 A schematic diagram of the physical structure of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0056] 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. These details should be considered as merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit 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.

[0057] The information collected in the technical solution of this application is information and data authorized by the user or fully authorized by all parties, and the collection, storage, use, processing, transmission, provision, disclosure and application of relevant data comply with the relevant laws, regulations and standards of relevant countries and regions, take necessary confidentiality measures, do not violate public order and good morals, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0058] The acquisition, transmission, storage, use, and processing of data in the technical solution of this application are in compliance with the relevant provisions of national laws and regulations. It should be noted that in the embodiments of this application, certain software, components, models, and other existing solutions in the industry may be mentioned. These should be considered as exemplary. Their purpose is only to illustrate the feasibility of implementing the technical solution of this application, but it does not mean that the applicant has or will necessarily use such solutions.

[0059] It should be noted that, in the absence of conflict, the embodiments and features of the embodiments in this application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0060] Figure 1 An exemplary system architecture is shown to which an embodiment of the flatness prediction method or flatness prediction device of the present application can be applied.

[0061] like Figure 1 As shown, system architecture 100 may include terminal devices 101, 102, 103, a network 104, and a server 105. Network 104 is a medium for providing communication links between terminal devices 101, 102, 103 and server 105. Network 104 may include various connection types, such as wired or wireless communication links or fiber optic cables.

[0062] Users can use terminal devices 101, 102, 103 to interact with server 105 via network 104 to receive shape prediction results or send geometric parameters and material parameters of the equipment to be predicted. Various physical simulation tools can be installed on terminal devices 101, 102, 103.

[0063] Terminal devices 101, 102, and 103 can be hardware or software. When terminal devices 101, 102, and 103 are hardware, they can be various electronic devices, including but not limited to smartphones, tablet computers, e-book readers, laptop computers, and desktop computers. When terminal devices 101, 102, and 103 are software, they can be installed in the electronic devices listed above. They can be implemented as multiple software or software modules (for example, to provide distributed services), or as a single software or software module. No specific limitations are given here.

[0064] The server 105 may be a server that provides various services, such as a backend server that reviews texts generated by users through the terminal devices 101, 102, and 103. The backend server may review the obtained texts and feed back the processing results (e.g., flatness prediction results, flatness coefficients) to the terminal devices 101, 102, and 103.

[0065] It should be noted that the server 105 can be hardware or software. When the server 105 is hardware, it can be implemented as a distributed server cluster consisting of multiple servers, or it can be implemented as a single server. When the server 105 is software, it can be implemented as multiple software or software modules (for example, to provide distributed services), or it can be implemented as a single software or software module. No specific limitations are given here.

[0066] It should be understood that Figure 1 The number of terminal devices, networks and servers in the embodiment is merely illustrative. Any number of terminal devices, networks and servers may be provided as required.

[0067] It should be noted that the flatness prediction method based on Bayesian data fusion provided in the embodiment of the present application is generally executed by the server 105. Accordingly, the flatness prediction device based on Bayesian data fusion is generally provided in the server 105.

[0068] In this application, plate shape refers to the distribution of residual stress inside the strip after rolling along the direction of the bandwidth. If a certain length of strip is naturally placed on a plane, the warping of the strip can often be observed. Among them, warping has various forms, most of which are wavy, and thin strips often produce wrinkles or local bumps; warping can sometimes spread throughout the entire bandwidth, and sometimes only locally; this warping and uneven deformation of the strip are closely related to the uneven distribution of internal stress. At present, there are eight main types of plate shape defects for strips on cold rolling lines: left wave, right wave, middle wave, double-sided wave, left three-point wave, right three-point wave, four-point wave and edge-middle compound wave. The main task of plate shape pattern recognition is to map the plate shape value detected online into a few characteristic parameters through a certain mathematical method, and use this to determine the control amount of the actuator.

[0069] The strip is cut into several longitudinal strips and flattened into several longitudinal strips. The rolled length at different points in the transverse direction is generally measured using the relative length difference of each longitudinal strip along the length of the strip to express the flatness. The relative length difference is also called the flatness index ε, where ε = ΔL / L. Since ε is a very small value, the I unit is often used to express the flatness in actual production. The relationship between the I unit and the flatness index is shown as follows:

[0070]

[0071] Where ΔL is the difference between the length of the longitudinal strip and the reference length in the longitudinal direction of the strip; L is the reference length of the strip, which is generally the average value of the lengths of the longitudinal strips.

[0072] In practical applications, the internal mechanisms of many complex systems are poorly understood, or the subject of study is highly time-varying, making it virtually impossible to describe them using mechanistic models. In such cases, understanding the relationships between variables within the system based solely on mechanisms becomes increasingly difficult. To address these challenges, data-driven analysis and modeling methods, such as data mining, machine learning, and pattern recognition, have proven effective. Furthermore, with the continuous advancement of measuring instruments and sensors, obtaining input and output data for systems is becoming increasingly accessible. Consequently, data-driven modeling approaches, which utilize sample or measurement data from measuring instruments or sensors to analyze the interdependencies between variables within the system, and then use this data to build a mathematical model of the system, are becoming increasingly common. This approach treats the system as a black box, bypassing internal mechanics analysis and instead directly modeling the interdependencies between the input and output data of the system. These models offer strong online calibration capabilities and are applicable to highly nonlinear and highly uncertain systems, providing an effective approach for modeling complex systems.

[0073] However, data-based non-mechanistic modeling can also lead to poor generalization capabilities of the established models due to problems such as data noise pollution. Especially in complex processes such as steel rolling, the accuracy and real-time nature of working condition data directly affect production efficiency and product quality. Existing working condition data processing methods usually rely on data from a single source, such as finite element simulation data or actual production data. However, these methods each have certain limitations. Although finite element data can provide a relatively accurate theoretical model, it often cannot fully reflect the changes and complexity of the actual production process; and actual production data may be affected by noise, missing values ​​and measurement errors, resulting in insufficient accurate estimation of the working conditions. At the same time, most models are "black box" structure models that cannot reflect the true characteristics of the system, which in turn affects the research on the system.

[0074] To this end, this application proposes a flatness prediction method based on Bayesian data fusion. This method fuses finite element operating condition data with actual production operating condition data using a Bayesian approach. This method comprehensively considers the theoretical accuracy of the finite element model and the actual production data, dynamically updating the estimation of the operating condition parameters, thereby improving the accuracy of the prediction and reducing the uncertainty of the system. Furthermore, data fusion not only improves the robustness of the model but also provides reliable operating condition prediction results even when data is scarce or of poor quality. This is of great significance for improving production efficiency, optimizing process control, and improving product quality.

[0075] It should also be noted that combining data-driven non-mechanistic modeling with process knowledge and experience, using prior knowledge to save training samples for data-driven models, and using data-driven models to compensate for characteristics that the original model cannot explain, greatly improves the interpretability and application scope of the model.

[0076] Specifically, such as Figure 2 As shown, the plate shape prediction method based on Bayesian data fusion of the present application includes:

[0077] Step 10: Determine finite element operating condition data of the equipment to be predicted based on a pre-built finite element model, wherein the finite element model is a simulation model of the equipment to be predicted.

[0078] In some optional aspects of this embodiment, the step of constructing the finite element model includes:

[0079] Acquiring geometric parameters and material parameters of the device to be predicted; calling a physical simulation tool, and constructing the finite element model based on the physical simulation tool, the geometric parameters and the material parameters;

[0080] Among them, the geometric parameters include the working roll diameter, the working roll body length, the spacing between the working roll bending force concentration points, the support roll diameter, the support roll body length and the spacing between the support roll support reaction force concentration points; the material parameters include the density of the rolled product and the roll, the elastic modulus of the rolled product, the Poisson's ratio of the rolled product, the elastic modulus of the roll and the Poisson's ratio of the roll.

[0081] In a specific example, the finite element model is modeled as follows:

[0082] This paper demonstrates the use of a solid-state simulation software to establish a coupled model of the rolling stock roll system, using the process parameters of a four-roller skin-pass mill provided by a certain factory. It should be noted that the simulation ignores the effects of factors such as roll wear, temperature, and lubrication. Because the strip temper rolling process exhibits low elongation, minimal plastic deformation, and the primary form of roll deformation is elastic, the anisotropy of the strip and roll materials has little impact on the strip shape simulation. Therefore, the strip and roll materials are isotropic.

[0083] The geometric parameters involved in the modeling process are shown in Table 1, and the material parameters are shown in Table 2.

[0084] Table 1

[0085] Geometric parameters size Working roll diameter Db 450mm Working roll barrel length Lb 1450mm The distance between the concentrated force action points of the work roll bending force Lf 2450mm Support roller diameter Dw 400mm Support roller body length Lw 1400mm Support roller reaction concentrated force action point distance Lp 2450mm

[0086] Table 2

[0087]

[0088]

[0089] Use the entity simulation tool to create Figure 3 The three-dimensional solid finite element model shown in the figure ignores some chamfers, fillets, and undercuts that do not affect the simulation results during the model building process. The dimensions of the three-dimensional solid finite element model after expansion are as follows: Figure 4 As shown in Table 1, no further details are given here.

[0090] In some optional embodiments of this embodiment, such as Figure 5 As shown, the finite element operating condition data of the equipment to be predicted is determined based on the pre-built finite element model, including:

[0091] Step 101A: Obtain process parameters and inlet flatness defect parameters of the equipment to be predicted, and determine an initial operating condition parameter combination based on the process parameters and the inlet flatness defect parameters.

[0092] In this embodiment, all relevant working condition parameters of the equipment to be predicted, such as plate width, inlet plate thickness, material deformation resistance, and bending roll force, are pre-extracted from the database, while ensuring that each parameter has a clear value or range. The process parameters of the equipment to be predicted are shown in Table 3, and the inlet plate shape defect parameters of the equipment to be predicted are shown in Table 4.

[0093] Table 3

[0094] Parameter Type Value range Board width / mm 900:100:1300 Strip entrance thickness / mm 0.5:0.3:2.0 Strip deformation resistance / MPa 100:50:400 Bending roller force / kN -50:25:50 Inlet tension / kN 5:25:55 Outlet tension / kN 15:25:65

[0095] Table 4

[0096]

[0097] It should be noted that when determining the initial operating condition parameter combination, the maximum and minimum values ​​of each wave shape of the entrance flatness defect are set for the operating condition, and the maximum, minimum and intermediate values ​​of the entrance transverse thickness difference are set for the operating condition. According to the above process parameters, entrance plate shape defect parameters and selection methods, different operating conditions are simulated by adjusting the geometric scale parameters and material parameters of the rolled piece in the aforementioned finite element model.

[0098] According to Table 3 and Table 4 above, appropriate operating condition parameters are selected for combination to cover all possible operating conditions. 680,400 operating conditions, i.e., initial operating condition parameter combinations, can be set.

[0099] Step 102A: Based on the finite element model and the physical simulation tool, determine a first influence weight of each process parameter on the output plate shape and a second influence weight of each inlet plate shape defect parameter on the output plate shape.

[0100] In the present application, based on the finite element model and the solid simulation tool, the output shape coefficient corresponding to each initial operating parameter combination can be determined, the output shape coefficients under different initial operating parameter combinations can be sorted and compared, and it can be determined which parameters have a greater impact on the output shape coefficient.

[0101] For example, based on the finite element model and the physical simulation tool, the first influence weight of each process parameter on the output plate shape and the second influence weight of each inlet plate shape defect parameter on the output plate shape are determined. Specifically, the first influence weight and the second influence weight can be determined in the following way.

[0102] (1) Finite element analysis and simulation: In the finite element model, each process parameter and defect parameter is changed (for example, item by item), and the changes in the output plate shape are recorded; and the model under different conditions is analyzed multiple times using a physical simulation tool to evaluate the impact of the process parameters and defect parameters.

[0103] (2) Statistical analysis: Apply statistical regression analysis to establish a mathematical model between the output shape coefficient and various influencing factors, and determine the relative weight of each parameter through the regression coefficient; and use sensitivity analysis to calculate the degree of influence of each parameter change on the output result, and quantify the influence weight through the sensitivity index.

[0104] (3) Weighted synthesis: Based on the obtained sensitivity analysis or regression analysis results, the weights of different influencing factors are calculated comprehensively to identify the main influencing factors.

[0105] Step 103A: Remove the process parameters whose first influence weight is less than a preset weight threshold and the inlet flatness defect parameters whose second influence weight is less than the preset weight threshold from the initial operating condition parameter combination to obtain a candidate operating condition parameter combination.

[0106] Specifically, according to the above steps, the influence weights of process parameters such as plate width, strip entrance thickness, strip deformation resistance, bending roll force and entrance tension can be determined, and the influence weights are compared with the preset weight thresholds respectively, and the process parameters with influence weights less than the preset weight thresholds are removed; similarly, according to the above steps, the influence weights of plate shape defect parameters such as entrance flatness and entrance transverse thickness difference can be determined, and the influence weights are compared with the preset weight thresholds respectively, and the plate shape defect parameters with influence weights less than the preset weight thresholds are removed, and finally a candidate working condition parameter combination is obtained.

[0107] Step 104A: performing zero mean processing on each candidate operating condition parameter in the candidate operating condition parameter combination, and removing abnormal values ​​in each candidate operating condition parameter according to the 3 sigma principle to obtain the finite element operating condition data.

[0108] Specifically, the values ​​of each candidate operating condition parameter in the candidate operating condition parameter combination are subjected to zero mean processing, and according to the 3sigma principle, the data exceeding the 3sigma range are considered as outliers. The outliers in each candidate operating condition parameter are removed to ensure the validity and accuracy of the data, and the finite element operating condition data are obtained.

[0109] Step 20: Acquire actual operating condition data of the equipment to be predicted, and perform Bayesian data fusion on the actual operating condition data and the finite element operating condition data to obtain fused operating condition data.

[0110] In some optional embodiments of this embodiment, such as Figure 6 As shown, the Bayesian data fusion is performed on the actual working condition data and the finite element working condition data to obtain the fused working condition data, including:

[0111] Step 201: Determine a priori distribution based on the mean and variance of the finite element working condition data.

[0112] Step 202: Determine a likelihood function based on the mean and variance of the actual operating condition data.

[0113] Step 203: Determine the marginal likelihood based on the likelihood function and the prior distribution.

[0114] Step 204: Determine a posterior distribution based on the likelihood function, the prior distribution, and the marginal likelihood.

[0115] Step 205: Based on the posterior distribution, perform Bayesian data fusion on the actual working condition data and the finite element working condition data to obtain the fused working condition characteristic data.

[0116] Specifically, the Bayesian fusion method of the present invention utilizes finite element data as a prior distribution. Finite element data, which can be artificially generated and set to conform to a normal distribution, reflects the common characteristics of the steel rolling process and provides a theoretical basis for the model. Actual production data, used as observational data, reflects the individual characteristics of specific production batches. By calculating the posterior distribution, estimates of unknown parameters can be dynamically updated to obtain fused operating condition data. This process not only improves the accuracy of operating condition predictions but also quantifies the uncertainty of the predictions, providing a reliable basis for subsequent control decisions.

[0117] In a specific example, by calculating the mean and variance of the posterior distribution, more reliable working condition input parameters can be extracted for subsequent model training and prediction. n} and the likelihood function of the normal distribution with μ as the mean and δ as the standard deviation. The specific data fusion process is as follows:

[0118] Obtaining a prior distribution: The normal distribution followed by the finite element data is used as the prior distribution. The prior distribution is expressed as follows.

[0119]

[0120] Wherein, μ0 is the mean of the normal distribution of the finite element working condition data; is the normal distribution variance of the finite element working condition data; N represents the normal distribution.

[0121] Calculate the likelihood function of the observed data: The actual production data is used as the observed data. When the actual production data output is sufficient, it also obeys the normal distribution. Its likelihood function can be expressed as shown below:

[0122]

[0123] Where x i is the actual production condition data sample, n is the actual production condition data sample size, s 2 is the normal distribution variance of actual production condition data.

[0124] Calculate marginal likelihood: The marginal likelihood is the integration of the likelihood function for all possible μ, and the formula is:

[0125] P(D)=∫P(D|μ)P(μ)dμ (3)

[0126] Obtain the posterior distribution: According to Bayes' theorem, the posterior distribution is:

[0127]

[0128] In a specific example, according to the above, the finite element working condition data includes multiple process parameters of the equipment to be predicted, namely: strip width, in millimeters (mm); strip thickness, in millimeters (mm); elongation, in percentage (%); deformation resistance, in megapascals (MPa); working roll bending force, in tons (t); intermediate roll bending force, in tons (t); intermediate roll shifting, in millimeters (mm).

[0129] According to the finite element data, the mean μ0 and variance of each process parameter of each specification strip steel are calculated according to the following formulas (5) and (6): Similarly, the mean of each process parameter in the actual production data is calculated as The variance is s 2 , the data volume is n. Data fusion is performed on each process parameter according to the Bayesian method.

[0130]

[0131] Where x0i is the i-th data point of a certain process parameter of the finite element, and the total number of data points is N.

[0132] The normal distribution is obtained from the finite element working condition data as the prior distribution, and the specific form is shown in the following formula:

[0133]

[0134] The specific form of the likelihood function is as follows:

[0135]

[0136] The marginal likelihood is the integration of the likelihood function over all possible μ, and the formula is:

[0137]

[0138] According to Bayes' theorem, the posterior distribution is:

[0139]

[0140] Usually, the posterior distribution obtained by the maximum likelihood method is also a normal distribution, and its mean is calculated as follows:

[0141]

[0142] The updated posterior distribution mean is organized into a data set to complete the working condition data fusion, which serves as the input data set of the model.

[0143] Obtaining fused data: The mean of the posterior distribution, updated based on actual operating conditions, is obtained as the fused data to form the model training dataset. Specific input features include: strip width (in millimeters (mm); strip thickness (in millimeters (mm)); elongation (in percentage (%)); deformation resistance (in megapascals (MPa)); work roll bending force (in tons (t)); intermediate roll bending force (in tons (t)); and intermediate roll shifting (in millimeters (mm)). Output features include: primary, secondary, cubic, and quartic shape coefficients, representing the basic wave patterns of the strip.

[0144] This application provides a working condition data fusion method based on Bayesian fusion. By combining finite element data and actual production data, the deficiencies of the two are compensated, thereby improving the accuracy and reliability of the working condition data. Then, a data-driven plate shape prediction model is established to predict the plate shape, which is convenient for the subsequent calculation of the control efficiency of the actuator. An accurate plate shape prediction model can provide a basis for plate shape control and compensate for the time lag problem of the plate shape control system. Predicting the subsequent plate shape in advance is conducive to improving the accuracy of plate shape control, making full use of the information of industrial big data, and effectively improving the plate shape control capability of the leveling machine.

[0145] Step 30: Using the fusion working condition data as training data, training a random forest model until the random forest model meets a preset training standard, thereby obtaining a trained flatness prediction model.

[0146] In some optional aspects of this embodiment, the method further includes: determining an output shape coefficient corresponding to the finite element working condition data;

[0147] The method of training a random forest model using the fused operating condition data as training data includes:

[0148] Training a random forest model using the fused working condition data as training input data and the output shape coefficient as training output data;

[0149] In the random forest model, the number of decision trees is 50, the maximum number of features used when each decision tree is split is 4, the maximum depth of the decision tree is 4, and the minimum number of samples required to split a node is 30.

[0150] In this application, the random forest algorithm is a random regression method based on a combination of classification trees and multiple basic units of decision trees. The final output category is determined based on the category output by each individual tree, and the mode of the individual results is adopted. Although the classification accuracy of a single tree is not high, the accumulation of small amounts can add up to a large amount. Using a large number of randomly generated decision trees as an experimental strategy, a test sample is passed through many different branching trees layer by layer, and the results obtained are summarized and statistically analyzed to obtain the classification with the highest probability of selection.

[0151] The steps of model building using random forest model can be roughly divided into several stages, such as data preprocessing, model building, training, evaluation, and generating results. The basic process is as follows: Figure 7 As shown:

[0152] Obtain and separate the data set, randomly select 80% of the data, generate a training set, set the random forest parameters, train and generate the random forest, and use the remaining 20% ​​of the data to generate a test set, import the random forest for testing, calculate the accuracy, determine whether the accuracy meets the standard, if so, determine whether the evaluation indicators are good, if so, calculate the importance of each indicator random forest visualization, if not, reset the random forest parameters or try other models.

[0153] In a specific example, the basic process of random forest generation is as follows:

[0154] Step 1: Randomly extract n times in the data set to obtain multiple samples;

[0155] Step 2: Randomly select k features from all features to build a decision tree for the selected sample;

[0156] Step 3: Repeat steps 1 and 2 m times to obtain m decision trees, forming a random forest.

[0157] Step 4: Each decision tree can be regarded as a classifier. Therefore, if there are p trees, then p classifiers will produce corresponding p classification results;

[0158] Step 5. After summarizing all classification results, sort them according to the number of votes and select the classification result with the most votes, which can be regarded as the final result of the random forest algorithm.

[0159] Next, select R 2 and MSE as the evaluation criteria of the flatness prediction model, where R 2 Determine the model's fit to the true value. MSE is used to determine the deviation between the predicted value and the actual value. The calculation formula is as follows:

[0160]

[0161] Where y i is the true value of the data, is the predicted value, and n is the number of data points.

[0162] It should be noted that R 2 The closer the value is to 1, the better the model fit is; the smaller the MSE value is, the higher the model accuracy is. Based on the model evaluation results, it is determined whether the model training meets the requirements. Models that meet the requirements are deployed and implemented for plate shape prediction.

[0163] In a specific example, once the data is prepared, we can begin building a random forest model. Adjusting parameters to regenerate the random forest is a process that requires repeated trade-offs. If the prediction accuracy is extremely high through parameter adjustment, the prediction efficiency may decrease due to the excessive size of the model. Therefore, it is necessary to strike a balance between model accuracy and model efficiency. The key hyperparameters of random forest and their settings are as follows:

[0164] Number of decision trees (n_estimators): The default value is 10. More trees generally lead to better performance, but training time will also increase accordingly. In practical applications, the number of trees is usually set in units of 10, and the parameter adjustment range is between 1 and 201.

[0165] Maximum number of features to use when splitting each tree (max_features): The maximum number of features to use when splitting each tree. By default, all features are considered. By limiting the number of features used per tree, the risk of overfitting can be reduced. Since the model has a small number of features, try adjusting max_features one by one until you find an ideal value.

[0166] Maximum depth of the tree (max_depth): The maximum depth of the tree. The default value is "None" which does not limit the depth. Since the model has a large number of samples, limiting the maximum depth helps prevent the tree from being too deep and causing overfitting.

[0167] Minimum number of samples required to split a node (min_samples_split): The minimum number of samples required to split a node. The default value is 2. Increasing this value can reduce overfitting. In this application, due to the large number of samples, we choose to increase this value.

[0168] By optimizing the parameters, in this application, the final parameters are set as follows: the number of decision trees is 50, the maximum number of features used when splitting each tree is 4, the maximum depth of the tree is 4, and the minimum number of samples required to split a node is 30.

[0169] Select R 2 and MSE as the evaluation criteria of the flatness prediction model, where R 2 Determine the model's fit to the true value, and MSE determines the deviation between the predicted value and the actual value; R 2 The closer the value is to 1, the better the model fit is; the smaller the MSE value is, the higher the accuracy of the model is, and the R 2 The value is 0.96 and the MSE value is 0.035.

[0170] In this application, the advantages of random forests over decision trees are mainly:

[0171] (1) Reduce the impact of outliers: Because random forests select some data to construct many decision trees, even if there are individual outlier decision trees, the prediction results may not be 100% accurate. However, since the prediction results are obtained by referencing the results of multiple decision trees and summarizing them, the minority obeys the majority, and the herd effect reduces the impact of outliers;

[0172] (2) Reduced overfitting: Because all samples and feature values ​​are used by the decision tree, overfitting may occur, that is, the effect is very good when training samples, but the effect is unsatisfactory when running the test set. To reduce this risk, the random forest method adopts the method of selecting some features of some samples to construct many decision trees, reducing the feature values ​​and data of a single decision tree, and reducing the possibility of overfitting.

[0173] The random forest algorithm offers significant advantages in both real-time and interpretability, while ensuring accurate predictions. In terms of real-time performance, random forests generate multiple decision trees in parallel, enabling rapid response to new data and making them suitable for processing large amounts of data. In terms of interpretability, random forests reveal the model's decision logic through feature importance analysis, enabling users to understand the impact of each feature on the results. These characteristics make random forests widely applicable in industrial fields that demand high real-time performance and transparency.

[0174] Step 40: Determine the shape coefficient corresponding to the equipment to be predicted based on the shape prediction model and the operating condition data to be predicted.

[0175] Specifically, the prediction model is deployed. After the feature value is input, the model will make predictions through each decision tree, and finally give the prediction result of the board shape through integration of all decision trees.

[0176] The present invention adopts a Bayesian data fusion method, which can effectively integrate finite element data with actual production data, thereby significantly improving the accuracy and robustness of working condition predictions. Finite element data provides theoretical predictions based on physical models, while actual production data reflects the details and changes in the actual production process. By integrating the advantages of both, the accuracy of working condition predictions can be greatly improved. Especially when faced with complex and dynamically changing production environments, the fused data can more realistically reflect the actual status of the working conditions. In addition, data fusion effectively reduces the uncertainty that may be brought about by a single data source, provides more comprehensive and accurate information, and significantly improves the quality of working condition predictions.

[0177] The proposed model can predict the five Legendre coefficients at a future point in time based on current input information, enabling early assessment of flatness quality. If the model meets predetermined accuracy standards, the predicted Legendre coefficients can be used to determine flatness type. It also serves as an experimental platform, allowing for adjustments to preset values ​​in input features to evaluate the impact of different preset values ​​on flatness curve fitting. This approach can improve production efficiency and reduce economic losses, laying the foundation for optimization and intelligent control of industrial production processes.

[0178] Based on the same inventive concept, the embodiments of the present application also provide a plate shape prediction device based on Bayesian data fusion, which can be used to implement the method described in the above embodiments, as described in the following embodiments. Since the principle of the problem solved by the plate shape prediction device based on Bayesian data fusion is similar to that of a plate shape prediction method based on Bayesian data fusion, the implementation of a plate shape prediction device based on Bayesian data fusion can refer to the implementation of a plate shape prediction method based on Bayesian data fusion, and the repeated parts will not be repeated. As used below, the term "unit" or "module" can be a combination of software and / or hardware that implements a predetermined function. Although the system described in the following embodiments is preferably implemented in software, the implementation of hardware, or a combination of software and hardware, is also possible and conceived.

[0179] like Figure 8 As shown, a flatness prediction device based on Bayesian data fusion includes:

[0180] The finite element operating condition data determination module 801 is configured to determine the finite element operating condition data of the device to be predicted based on a pre-built finite element model, wherein the finite element model is a simulation model of the device to be predicted;

[0181] The fused operating condition data determination module 802 is configured to obtain actual operating condition data of the equipment to be predicted, and perform Bayesian data fusion on the actual operating condition data and the finite element operating condition data to obtain fused operating condition data;

[0182] The flatness prediction model training module 803 is configured to train a random forest model using the fused operating condition data as training data until the random forest model meets a preset training standard, thereby obtaining a trained flatness prediction model;

[0183] The shape coefficient prediction module 804 is configured to determine the shape coefficient corresponding to the equipment to be predicted based on the shape prediction model and the working condition data to be predicted.

[0184] In some optional aspects of this embodiment, the step of constructing the finite element model includes:

[0185] Obtaining geometric parameters and material parameters of the device to be predicted;

[0186] Invoking a solid simulation tool, and constructing the finite element model based on the solid simulation tool, geometric parameters, and material parameters;

[0187] Among them, the geometric parameters include the working roll diameter, the working roll body length, the spacing between the working roll bending force concentration points, the support roll diameter, the support roll body length and the spacing between the support roll support reaction force concentration points; the material parameters include the density of the rolled product and the roll, the elastic modulus of the rolled product, the Poisson's ratio of the rolled product, the elastic modulus of the roll and the Poisson's ratio of the roll.

[0188] In some optional embodiments of this embodiment, such as Figure 9 As shown, the finite element working condition data determination module includes:

[0189] The initial operating condition parameter combination determining unit 8011 is configured to obtain the process parameters and inlet flatness defect parameters of the equipment to be predicted, and determine the initial operating condition parameter combination based on the process parameters and the inlet flatness defect parameters;

[0190] An influence weight determination unit 8012 is configured to determine, based on the finite element model and the physical simulation tool, a first influence weight of each of the process parameters on the output flatness and a second influence weight of each of the inlet flatness defect parameters on the output flatness;

[0191] The candidate operating condition parameter combination determining unit 8013 is configured to remove the process parameters whose first influence weight is less than a preset weight threshold and the inlet flatness defect parameters whose second influence weight is less than the preset weight threshold from the initial operating condition parameter combination to obtain a candidate operating condition parameter combination;

[0192] The finite element operating condition data determination unit 8014 is configured to perform zero mean processing on each candidate operating condition parameter in the candidate operating condition parameter combination, and remove abnormal values ​​in each candidate operating condition parameter according to the 3 sigma principle to obtain the finite element operating condition data.

[0193] In some optional aspects of this embodiment, the finite element operating condition data includes a plurality of finite element operating condition feature data, the actual operating condition data includes a plurality of actual operating condition feature data, and the fused operating condition data includes a plurality of fused operating condition feature data. The Bayesian data fusion of the actual operating condition data and the finite element operating condition data to obtain the fused operating condition data includes:

[0194] Bayesian data fusion is performed on the actual working condition characteristic data and the finite element working condition characteristic data respectively to obtain the fused working condition characteristic data.

[0195] In some optional embodiments of this embodiment, such as Figure 10 As shown, the fusion working condition data determination module includes:

[0196] A prior distribution determining unit 8021 is configured to determine a prior distribution based on the mean and variance of the finite element working condition characteristic data;

[0197] A likelihood function determining unit 8022 is configured to determine a likelihood function based on the mean and variance of the actual operating condition characteristic data;

[0198] A marginal likelihood determination unit 8023 is configured to determine a marginal likelihood based on the likelihood function and the prior distribution;

[0199] a posterior distribution determining unit 8024, configured to determine a posterior distribution based on the likelihood function, the prior distribution, and the marginal likelihood;

[0200] The fused operating condition characteristic data determining unit 8025 is configured to perform Bayesian data fusion on the actual operating condition characteristic data and the finite element operating condition characteristic data based on the posterior distribution to obtain the fused operating condition characteristic data.

[0201] In some optional aspects of this embodiment, the method further includes: determining an output shape coefficient corresponding to the finite element working condition data;

[0202] The method of training a random forest model using the fused operating condition data as training data includes:

[0203] Training a random forest model using the fused working condition data as training input data and the output shape coefficient as training output data;

[0204] In the random forest model, the number of decision trees is 50, the maximum number of features used when each decision tree is split is 4, the maximum depth of the decision tree is 4, and the minimum number of samples required to split a node is 30.

[0205] According to an embodiment of the present disclosure, the present disclosure also provides an electronic device, a readable storage medium, and a computer program product.

[0206] An electronic device comprises: 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 so as to enable the at least one processor to perform the steps of a plate shape prediction method based on Bayesian data fusion of the aforementioned embodiment.

[0207] A non-transitory computer-readable storage medium storing computer instructions, wherein the computer instructions are used to enable a computer to execute the steps of a plate shape prediction method based on Bayesian data fusion in the above embodiment.

[0208] A computer program product includes a computer program / instruction, which, when executed by a processor, implements the steps of a plate shape prediction method based on Bayesian data fusion in the aforementioned embodiment.

[0209] Figure 11 A schematic block diagram of an example electronic device 900 that can be used to implement embodiments of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital assistants, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are provided as examples only and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0210] like Figure 11 As shown, the device 900 includes a computing unit 901, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 902 or a computer program loaded from a storage unit 908 into a random access memory (RAM) 903. Various programs and data required for the operation of the device 900 can also be stored in the RAM 903. The computing unit 901, the ROM 902, and the RAM 903 are connected to each other via a bus 904. An input / output (I / O) interface 905 is also connected to the bus 904.

[0211] Various components in the device 900 are connected to the I / O interface 905, including an input unit 906, such as a keyboard, a mouse, etc.; an output unit 907, such as various types of displays, speakers, etc.; a storage unit 908, such as a magnetic disk, an optical disk, etc.; and a communication unit 909, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 909 allows the device 900 to exchange information / data with other devices via a computer network such as the Internet and / or various telecommunication networks.

[0212] The computing unit 901 can be any general-purpose and / or specialized processing component with processing and computing capabilities. Some examples of the computing unit 901 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various specialized artificial intelligence (AI) computing chips, various computing units that run machine learning model algorithms, a digital signal processor (DSP), and any appropriate processor, controller, microcontroller, etc. The computing unit 901 executes the various methods and processes described above, such as a board shape prediction method based on Bayesian data fusion.

[0213] For example, in some embodiments, a method for flatness prediction based on Bayesian data fusion can be implemented as a computer software program, which is tangibly contained in a machine-readable medium, such as a storage unit 908. In some embodiments, part or all of the computer program can be loaded and / or installed on the device 900 via the ROM 902 and / or the communication unit 909. When the computer program is loaded into the RAM 903 and executed by the computing unit 901, one or more steps of the method for flatness prediction based on Bayesian data fusion described above can be performed. Alternatively, in other embodiments, the computing unit 901 can be configured to perform a method for flatness prediction based on Bayesian data fusion by any other appropriate means (e.g., by means of firmware).

[0214] Various embodiments of the systems and techniques described herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), system-on-chip systems (SOCs), programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments can include being implemented in one or more computer programs that are executable and / or interpreted on a programmable system that includes at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0215] The program code for implementing the method of the present disclosure can be written in any combination of one or more programming languages. These program codes can be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device so that when the program code is executed by the processor or controller, the functions / operations specified in the flow chart and / or block diagram are implemented. The program code can be executed entirely on the machine, partially on the machine, as a stand-alone software package, partially on the machine and partially on a remote machine, or entirely on a remote machine or server.

[0216] In the context of the present disclosure, a machine-readable medium can be a tangible medium that can contain or store a program for use by or in conjunction with an instruction execution system, device or equipment. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium can include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0217] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0218] The systems and techniques described herein can be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer having a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network (LAN), a wide area network (WAN), and the Internet.

[0219] A computer system may include a client and a server. The client and server are generally remote from each other and typically interact through a communication network. The client-server relationship arises through computer programs running on the respective computers and having a client-server relationship with each other. The server may be a cloud server, a server in a distributed system, or a server integrated with a blockchain.

[0220] It should be understood that the various forms of the processes shown above can be used to reorder, add, or delete steps. For example, the steps described in this disclosure can be performed in parallel, sequentially, or in a different order, as long as the desired results of the technical solutions of this disclosure can be achieved. This is not a limitation herein.

[0221] The above specific embodiments do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art will appreciate that various modifications, combinations, sub-combinations, and substitutions may be made based on design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure shall be included within the scope of protection of this disclosure.

Claims

1. A flatness prediction method based on Bayesian data fusion, characterized in that: include: Determining finite element operating condition data of the equipment to be predicted based on a pre-built finite element model, wherein the finite element model is a simulation model of the equipment to be predicted; Acquire actual operating condition data of the equipment to be predicted, and perform Bayesian data fusion on the actual operating condition data and the finite element operating condition data to obtain fused operating condition data; Using the fusion working condition data as training data, training a random forest model until the random forest model meets a preset training standard, thereby obtaining a trained plate shape prediction model; Based on the plate shape prediction model and the operating condition data to be predicted, a plate shape coefficient corresponding to the equipment to be predicted is determined.

2. The plate shape prediction method according to claim 1, characterized in that: The steps of constructing the finite element model include: Obtaining geometric parameters and material parameters of the device to be predicted; Invoking a solid simulation tool, and constructing the finite element model based on the solid simulation tool, geometric parameters and material parameters; Among them, the geometric parameters include the working roll diameter, the working roll body length, the spacing between the working roll bending force concentration points, the support roll diameter, the support roll body length and the spacing between the support roll support reaction force concentration points; the material parameters include the workpiece and roll density, the workpiece elastic modulus, the workpiece Poisson's ratio, the roll elastic modulus and the roll Poisson's ratio.

3. The plate shape prediction method according to claim 2, characterized in that: The step of determining finite element operating condition data of the equipment to be predicted based on the pre-built finite element model includes: Acquiring process parameters and inlet plate shape defect parameters of the equipment to be predicted, and determining an initial operating condition parameter combination based on the process parameters and the inlet plate shape defect parameters; Based on the finite element model and the physical simulation tool, determining a first influence weight of each of the process parameters on the output plate shape and a second influence weight of each of the inlet plate shape defect parameters on the output plate shape; From the initial operating condition parameter combination, remove the process parameters whose first influence weight is less than a preset weight threshold and the inlet plate shape defect parameters whose second influence weight is less than the preset weight threshold to obtain a candidate operating condition parameter combination; Zero mean processing is performed on each candidate operating condition parameter in the candidate operating condition parameter combination, and according to the 3 sigma principle, outliers in each candidate operating condition parameter are removed to obtain the finite element operating condition data.

4. The plate shape prediction method according to claim 1, characterized in that: The performing Bayesian data fusion on the actual working condition data and the finite element working condition data to obtain fused working condition data includes: Determining a prior distribution based on the mean and variance of the finite element working condition data; Determining a likelihood function based on the mean and variance of the actual operating condition data; Determining a marginal likelihood based on the likelihood function and the prior distribution; Determining a posterior distribution based on the likelihood function, the prior distribution, and the marginal likelihood; Based on the posterior distribution, Bayesian data fusion is performed on the actual operating condition data and the finite element operating condition data to obtain the fused operating condition data.

5. The plate shape prediction method according to claim 1, characterized in that: Also includes: Determine the output shape coefficient corresponding to the finite element working condition data; The step of training a random forest model using the fused operating condition data as training data includes: Using the fusion working condition data as training input data and the output plate shape coefficient as training output data, training a random forest model; Among them, in the random forest model: the number of decision trees is 50, the maximum number of features used when each decision tree is split is 4, the maximum depth of the decision tree is 4, and the minimum number of samples required to split a node is 30.

6. A flatness prediction device based on Bayesian data fusion, characterized in that: include: A finite element working condition data determination module is configured to determine the finite element working condition data of the device to be predicted based on a pre-built finite element model, wherein the finite element model is a simulation model of the device to be predicted; a fused operating condition data determination module, configured to obtain actual operating condition data of the equipment to be predicted, and perform Bayesian data fusion on the actual operating condition data and the finite element operating condition data to obtain fused operating condition data; A plate shape prediction model training module is configured to use the fusion working condition data as training data to train a random forest model until the random forest model meets a preset training standard, thereby obtaining a trained plate shape prediction model; The shape coefficient prediction module is configured to determine the shape coefficient corresponding to the equipment to be predicted based on the shape prediction model and the working condition data to be predicted.

7. The plate shape prediction device according to claim 6, characterized in that: The fusion working condition data determination module includes: A prior distribution determination unit is configured to determine a prior distribution based on a mean and a variance of the finite element operating condition data; A likelihood function determination unit is configured to determine a likelihood function based on a mean and a variance of the actual operating condition data; A marginal likelihood determination unit, configured to determine a marginal likelihood based on the likelihood function and the prior distribution; a posterior distribution determination unit, configured to determine a posterior distribution based on the likelihood function, the prior distribution and the marginal likelihood; The fused operating condition characteristic data determination unit is configured to perform Bayesian data fusion on the actual operating condition data and the finite element operating condition data based on the posterior distribution to obtain the fused operating condition characteristic data.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the plate shape prediction method based on Bayesian data fusion described in any one of claims 1 to 5 is implemented.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method for predicting plate shape based on Bayesian data fusion according to any one of claims 1 to 5 is implemented.

10. A computer program product, characterized in that The computer program product comprises a computer program, and when the computer program is executed by a processor, the method for predicting plate shape based on Bayesian data fusion according to any one of claims 1 to 5 is implemented.

Images