A cord steel decarburization process quality optimization method and system based on data mining

Abstract

The application discloses a cord steel decarburization process quality optimization method and system based on data mining, and relates to the technical field of intelligent manufacturing of the steel industry. The method obtains full-process process parameters and decarburization layer depth detection data; performs data cleaning and sample equalization preprocessing based on a metallurgical mechanism model and a steady-state working condition benchmark value; constructs an effective feature set through parallel feature screening and generation branches; trains multiple heterogeneous benchmark prediction models and fuses to generate a decarburization quality prediction model; key parameters are extracted by using an explainability algorithm, and a global optimization algorithm is used to reversely search for an optimal process parameter combination. The application realizes closed-loop optimization from quality prediction to parameter recommendation, greatly shortens the trial production cycle and reduces the defect rate.

Description

Technical Field

[0001] This invention relates to the field of intelligent manufacturing technology in the steel industry, and in particular to a method and system for optimizing the quality of decarburization process of cord steel based on data mining. Background Technology

[0002] Cord steel is a key high-end steel grade used in the manufacture of radial tire skeleton materials. The depth of its surface decarburization layer is one of the core quality indicators affecting the fatigue life of the product, and the requirements are extremely strict. The decarburization process is affected by the complex coupling of multiple processes and dozens of process parameters, including smelting composition, furnace temperature regime, rolling process, and cooling parameters, and exhibits characteristics of high dimensionality, strong nonlinearity, and high noise.

[0003] Currently, the optimization of the process for steel cord mainly relies on engineers' field experience and single-factor adjustment experiments using a "trial and error" approach, which presents significant bottlenecks. (1) Human experience is insufficient to fully analyze the complex interactions between parameters in long processes and multiple procedures. Traditional statistical regression or models based on simplified physical mechanisms are insufficient to accurately characterize the above complex relationships, resulting in insufficient prediction accuracy. (2) A large amount of data has been accumulated during the production process, but there is a lack of systematic data mining methods, which makes it difficult to effectively accumulate and reuse process knowledge; (3) In actual production, there are few defective products that fail to meet decarbonization standards. When constructing the classification model, there is a serious sample imbalance problem, which leads to the poor ability of the model to identify defects.

[0004] With the development of machine learning technology, ensemble learning, automatic feature construction, model interpretability, and intelligent optimization algorithms have provided new approaches to solving the aforementioned problems. Therefore, by combining data mining methods with practical R&D scenarios, a complete methodology integrating data preprocessing, intelligent feature engineering, precise modeling, model interpretation, and process inversion optimization is developed. This is of urgent need and significant importance for improving the quality stability of tire cord steel and achieving intelligent and refined process optimization. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to overcome the shortcomings of the prior art and provide a method and system for optimizing the quality of decarburization process of cord steel based on data mining.

[0006] To solve the above technical problems, the technical solution of the present invention is as follows: In a first aspect, the present invention provides a method for optimizing the quality of decarburization process of tire cord steel based on data mining, comprising: Obtain process parameter data for the entire production process of steel cord and corresponding decarburization layer depth detection data; The process parameter data and decarburized layer depth detection data are preprocessed to obtain a preprocessed dataset. Feature selection and generation are performed on the preprocessed dataset to construct an effective feature set; Using an effective feature set as input, a decarbonization quality prediction model is constructed and trained. The decarbonization quality prediction model is used as the objective function, and a multi-parameter global optimization algorithm is used to solve it under given constraints to determine the target combination of process parameters.

[0007] Secondly, this invention provides a data mining-based system for optimizing the decarburization process quality of steel cord, which includes the following functional modules: The data acquisition module is used to acquire process parameter data for the entire production process of steel cord and the corresponding decarburization layer depth detection data. The data preprocessing module is used to preprocess process parameter data and decarburization layer depth detection data to obtain a preprocessed dataset. The feature generation module is used to perform feature selection and feature generation on the preprocessed dataset to construct an effective feature set. The model building module is used to build and train a decarbonization quality prediction model using an effective feature set as input. The parameter optimization module is used to take the decarbonization quality prediction model as the objective function and solve it using a multi-parameter global optimization algorithm under given constraints to determine the target combination of process parameters.

[0008] Thirdly, the present invention provides a computer device including a memory and a processor, the memory storing a computer program, the processor executing the computer program to implement the steps of the above-described method.

[0009] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the above-described method.

[0010] The beneficial effects of this invention are: (1) This invention sets up parallel feature filtering and generation branches. Not only are redundant features eliminated through coverage and variance inflation factor (VIF), but expert features are also constructed by combining the experience of process experts, and composite features are generated by mathematical combination using evolutionary algorithms. This dual-drive mechanism successfully uncovers hidden information that has not been explicitly recorded but is of great value, such as the implicit impact of the difference between the temperature of a specific heating zone and the surface temperature measurement on decarburization, so that the effective feature set of the final input model has both physical meaning and data discriminability.

[0011] (2) This invention uses a hyperparameter optimization algorithm to construct multiple heterogeneous benchmark prediction models, and integrates them through weighted calculation based on performance weights or stacking fusion strategy based on meta-learners. This heterogeneous fusion strategy effectively overcomes the overfitting problem of a single model under specific working conditions, effectively reduces the data noise interference caused by a single algorithm, and significantly improves the model's AUC index.

[0012] (3) This invention utilizes SHAP values ​​to quantify the degree and direction of the influence of various process parameters on decarbonization quality, extracts high-confidence process rules, and then uses the prediction model as the objective function. A global optimization algorithm based on biological evolution is employed to search for the optimal parameter combination within the physically feasible range. This scheme transforms traditional post-process verification into pre-process parameter recommendation, reducing the average number of trial and optimization rounds of the new process from 7 to 2, resulting in significant economic benefits.

[0013] (4) This invention abandons the blind interpolation of pure data science and adopts a strategy that combines "metallurgical mechanism model" and "steady-state operating condition benchmark value" to complete the missing data, and identifies anomalies based on "process parameter range" and "statistical distribution deviation criterion". At the same time, in response to the pain point of low defect rate in industrial quality inspection, a sample balancing strategy is adopted to reconstruct the dataset. This scheme completely eliminates data noise introduced by sensor failure or operating condition fluctuation and solves the model failure problem caused by imbalance of positive and negative samples.

[0014] (5) This invention addresses the data distribution shift caused by aging steel mill equipment and fluctuations in raw material batches by periodically acquiring newly added qualified process data and manually confirmed defect data, and using a sliding window mechanism to incrementally learn the model. This mechanism maintains the long-term prediction accuracy of the model with extremely low computing power cost, and gives the system the ability to be quickly reused and generalized across steel grades. Attached Figure Description

[0015] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. 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 effort.

[0016] Figure 1 A flowchart illustrating the data mining-based method for optimizing the decarburization process quality of tire cord steel provided by this invention. Figure 2 A schematic diagram of the data mining-based decarburization process quality optimization system for steel cord provided by the present invention; Figure 3 A schematic diagram of the computer device provided by the present invention. Detailed Implementation

[0017] To make the content of this invention easier to understand, the invention will be further described in detail below with reference to specific embodiments and accompanying drawings.

[0018] Example 1: This example provides a data mining-based method for optimizing the decarburization process quality of tire cord steel. Taking the optimization of the decarburization quality of 72, 82, and 92 grade tire cord steel from a steel mill as an example, the composition range is: C: 0.68%~0.96%, Si: 0.12%~0.35%, Mn: 0.35%~0.85%, P≤0.025%, S≤0.020%, Al: 0.001%~0.008%, N≤0.008%, O≤0.0025%, with the balance being Fe and unavoidable impurities. See also... Figure 1 The method specifically includes the following steps: Step S1: Obtain process parameter data for the entire production process of tire cord steel and corresponding decarburization layer depth detection data.

[0019] Specifically, production process data for approximately 16,000 heats (billets) from 2023 to 2024 were collected from the steel plant's Manufacturing Execution System (MES), the secondary process control system for steelmaking and high-speed wire rod rolling, and the quality inspection and testing system. Process parameter data covered: smelting and continuous casting parameters (content of elements such as C, Si, and Mn, superheat, tundish temperature, casting speed, etc.), heating parameters (temperature of heating furnace zones 1-15, billet residence time in each zone), and rolling and cooling parameters (initial rolling temperature, wire drawing temperature, flow rate and opening degree of water tanks in each section, air volume of blowers 1-14, etc.). The corresponding decarburization layer depth detection data were the decarburization depth values ​​(range 20-140 μm) measured by the quality inspection department using a metallographic microscope.

[0020] Step S2: Preprocess the process parameter data and decarburized layer depth detection data to obtain a preprocessed dataset.

[0021] Specifically, this step aims to address the issues of missing, anomalies, and imbalances in industrial data: (1) Handling missing and outlier values: For features with high missing rates, such as "liquidite temperature", theoretical completion is performed based on the metallurgical mechanism model (thermodynamic calculation formula); for sporadic missing values ​​caused by occasional packet loss from other sensors, historical steady-state benchmark values ​​(such as the median under the same pulling speed) are used for filling. At the same time, based on the process parameter range set in the equipment manual (such as the physical upper limit of the heating furnace temperature) and the 3σ statistical distribution deviation criterion, abnormal data is identified and replaced (for example, replacing the garbled code "reducing sizing inlet temperature" of 0 in 9 records with the median).

[0022] (2) Sample Balancing: Based on the company's internal control standards and a preset decarburization layer depth threshold (≥60μm), the data was divided into positive samples (4694 unqualified or critical samples) and negative samples (11173 samples). The number of negative samples was adjusted using a random undersampling strategy to construct a balanced dataset with 4694 positive and negative samples each, thus obtaining the preprocessed dataset.

[0023] Step S3: Perform feature selection and feature generation on the preprocessed dataset to construct an effective feature set; Specifically, this step includes parallel filtering and branch generation steps, as well as subsequent integration steps, as follows: (1) Feature screening branch: Single feature screening removes constant features with coverage <10% and variance of zero, and calculates the population stability index (PSI) to ensure stable feature distribution; multi-feature screening calculates the Pearson correlation coefficient to remove strongly correlated redundant features (such as pre-finishing mill exit temperature and BGV inlet temperature, keeping only one), and iteratively removes features with variance inflation factor (VIF) >10 to eliminate multicollinearity. After screening, the number of features decreased from the initial 152 to 81, resulting in candidate features.

[0024] (2) Simultaneously execute the feature generation branch: Combine the experience of process experts to construct expert features (such as "the difference between the temperature of zone 15 of the heating furnace and the surface temperature measurement" and "the time interval between two automatic temperature measurements in the tundish"); use the genetic programming-based evolutionary algorithm to perform addition, subtraction, multiplication, and division on the candidate features to generate about 200 candidate composite features such as "(rolling temperature × average pulling speed) / three-section water-passing length", and screen out 15 effective composite features through the random forest evaluation model.

[0025] (3) Feature integration step: The candidate features, expert features and effective composite features are integrated to generate a final effective feature set containing about 40 high-value dimensions.

[0026] Step S4: Using the effective feature set as input, construct and train the decarbonization quality prediction model.

[0027] Specifically, using the average ROC_AUC value of 5-fold cross-validation as the optimization objective, a Bayesian hyperparameter optimization algorithm is employed to iteratively optimize within a defined parameter space. Multiple heterogeneous benchmark prediction models (such as XGBoost, logistic regression, single decision tree, and random forest) with optimal parameter configurations are trained in parallel. Subsequently, performance evaluation metrics for each model on independent validation sets are obtained, and corresponding performance weights are calculated. A model fusion strategy (preferably a weighted calculation strategy based on performance weights in this embodiment) is used to integrate the prediction results of the heterogeneous benchmark prediction models to generate the final decarbonization quality prediction model.

[0028] The evaluation metrics for each model on the independent test set (approximately 3200 samples) are shown in Table 1: Table 1

[0029] The fusion model (primarily XGBoost) achieved an AUC of 0.932 on the test set, representing an improvement of 22.5% and 16.1% compared to logistic regression (0.761) and a single decision tree (0.803), respectively.

[0030] Step S5: Using the decarbonization quality prediction model as the objective function, solve it using a multi-parameter global optimization algorithm under given constraints to determine the target process parameter combination.

[0031] Specifically, firstly, the decarbonization quality prediction model is analyzed using an interpretability algorithm based on SHAP values. The influence of each process parameter on the probability of decarbonization qualification is quantified, key process parameters such as the temperature of the key zone of the heating furnace, the water tank flow rate, and the wire-spinning temperature are extracted, and a process rule library with high confidence (>0.85) is generated.

[0032] Subsequently, using the decarbonization pass probability output by the model as the maximization objective and key process parameters as decision variables, an iterative search is performed using a global optimization algorithm (genetic algorithm) based on biological evolution mechanisms, within the physically feasible range allowed by the equipment. The system ultimately outputs the top three combinations of target process parameters for on-site verification. For example, the optimal combination recommended by the system for the current operating conditions is shown in Table 2.

[0033] Table 2

[0034] Under this combination, the model predicts a decarbonization compliance probability of 97.3%, which is 6.1 percentage points higher than the current best historical sample (compliance probability of 91.2%). The top 3 optimized parameter sets are output for on-site process personnel to refer to and verify.

[0035] In addition, to address the concept drift caused by equipment aging and raw material fluctuations, an offline periodic update + incremental learning strategy is implemented. Monthly, newly added qualified process parameter data from the production line is periodically acquired as negative samples, and manually confirmed defect data is acquired as positive samples. These two data are combined, and a sliding window mechanism (extracting data from the past 12 months) is used to incrementally learn and retrain the decarburization quality prediction model. Each update takes approximately 15 minutes, and the model's AUC fluctuation remains within ±0.01. Furthermore, when this method is migrated to the new SWRH82A steel grade, only the classification threshold needs to be adjusted to 50μm and retrained; the defect prediction F1 score reaches 0.89, demonstrating extremely strong cross-steel grade generalization ability.

[0036] After a steel plant's cord steel production line operated using the above optimization method for 6 months, statistical data showed that: (1) Quality improvement: The defect rate of decarburization depth of cord steel exceeding the tolerance (≥60μm) decreased from 3.2% before the system went online to 1.7%, a relative decrease of 46.9%. Based on an annual output of 200,000 tons, the annual reduction of unqualified products is about 3,000 tons, resulting in significant direct economic benefits.

[0037] (2) Efficiency improvement: The time for locating the cause of process abnormalities was reduced from an average of 4.5 hours to 1.2 hours (a reduction of 73%); when the new process was trial-produced, parameter recommendations replaced the traditional trial-and-error method, and the number of optimization rounds was reduced from an average of 7 rounds to 2 rounds (a reduction of 71.4%), which greatly shortened the process development cycle.

[0038] (3) Knowledge Accumulation: Through SHAP analysis, corresponding high-confidence process rules were generated. Hidden knowledge that was never explicitly recorded by field engineers but has been verified to be effective (such as the key influence of the temperature in Zone 15 of the final heating furnace) has been incorporated into operating standards and experience transfer.

[0039] Example 2: This example provides a data mining-based system for optimizing the decarburization process quality of tire cord steel. See [link to example]. Figure 2 The system includes the following functional modules: data acquisition module, data preprocessing module, feature generation module, model building module, and parameter optimization module.

[0040] Specifically, the data acquisition module is used to acquire process parameter data for the entire production process of steel cord and the corresponding decarburization layer depth detection data. This module extracts process parameters such as smelting composition (C, Si, Mn, etc.), furnace zone temperatures (zones 1-15), water tank flow rates, and wire drawing temperatures from the steelmaking secondary system, high-speed wire rolling secondary system, and LIMS system in real time or in batches via the industrial OPC / UA protocol or direct database connection. It then matches these parameters with decarburization layer depth detection data (e.g., measured values ​​of 20-140 μm) obtained by the quality inspection department using a metallographic microscope to construct a full-process data mapping table.

[0041] The data preprocessing module is used to preprocess process parameter data and decarburization layer depth detection data to obtain a preprocessed dataset. This module has a built-in industrial data cleaning unit, which first calls metallurgical mechanism models (such as thermodynamic formulas to calculate liquidus temperature) or extracts historical steady-state benchmark values ​​(such as the median at the same pulling speed) to intelligently fill in missing values ​​caused by sensor packet loss; secondly, it identifies and replaces abnormal data according to the process parameter range (physical limit) set in the equipment manual and the 3σ statistical distribution deviation criterion; finally, it divides positive and negative samples by a preset decarburization layer depth threshold (such as 60μm) and runs the RandomUnderSampler algorithm to adjust the sample ratio, outputting a balanced preprocessed dataset with a 1:1 ratio of positive and negative samples.

[0042] The feature generation module is used to filter and generate features from the preprocessed dataset to construct an effective feature set. This module employs a parallel computing architecture: in the filtering branch, it automatically calculates the Population Stability Index (PSI), Pearson correlation coefficient, and Variance Inflation Factor (VIF), eliminating redundant features with low coverage, strong correlation, and multicollinearity; in the generation branch, it calls upon a process expert knowledge base to construct expert features such as "the difference between the temperature of zone 15 of the heating furnace and the surface temperature measurement," and runs an evolutionary algorithm based on genetic programming to combine existing features through addition, subtraction, multiplication, and division to generate candidate composite features such as "(initial rolling temperature × average drawing speed) / three-segment water penetration length." After evaluation by a random forest, the retained original features, expert features, and effective composite features are fused to output a high-value effective feature set of approximately 40 dimensions.

[0043] The model building module is used to construct and train a decarbonization quality prediction model using an effective feature set as input. Specifically, this module optimizes the average ROC_AUC value of 5-fold cross-validation using a Bayesian hyperparameter optimization algorithm within a defined parameter space, and trains multiple heterogeneous benchmark prediction models such as XGBoost and logistic regression in parallel. Subsequently, performance evaluation metrics of each model on independent validation sets are obtained to calculate performance weights, and model fusion strategies (such as weighted calculation based on performance weights or stacked fusion based on meta-learners) are used to integrate the prediction results of each model to generate the final decarbonization quality prediction model (with an AUC of up to 0.932 as tested).

[0044] The parameter optimization module uses the decarbonization quality prediction model as the objective function and solves it using a multi-parameter global optimization algorithm under given constraints to determine the target combination of process parameters. Specifically, this module includes an interpretation engine and a reverse optimization engine: the interpretation engine uses an interpretability algorithm based on SHAP values ​​to quantify the influence direction of each process parameter on the decarbonization qualification probability, extracts key process parameters such as "heater critical zone temperature, water tank flow rate, and wire feeding temperature," and generates high-confidence process rules; the reverse optimization engine encapsulates the prediction model as a fitness function, uses key process parameters as decision variables, and within the physically feasible range allowed by the equipment, uses a global optimization algorithm based on biological evolution mechanisms (such as a genetic algorithm) to perform crossover, mutation, and iterative search, finally outputting the top 3 sets of target process parameter combinations that maximize the decarbonization qualification probability (e.g., up to 97.3%), which are directly pushed to the secondary system for on-site process personnel to refer to and verify.

[0045] Example 3: See Figure 3 This embodiment also provides a computer device, the components of which may include, but are not limited to: one or more processors or processing units, system memory, and buses connecting different system components (including system memory and processing units).

[0046] A bus refers to one or more of several bus architectures, including a memory bus or memory controller, a peripheral bus, a graphics acceleration port, a processor, or a local bus using any of the various bus architectures. Examples of these architectures include, but are not limited to, the Industry Standard Architecture (ISA) bus, the Micro Channel Architecture (MAC) bus, the Enhanced ISA bus, the Video Electronics Standards Association (VESA) local bus, and the Peripheral Component Interconnect (PCI) bus.

[0047] Computer systems / servers typically include a variety of computer system-readable media. These media can be any available media that can be accessed by the computer system / server, including volatile and non-volatile media, and removable and non-removable media.

[0048] System memory may include computer system readable media in the form of volatile memory, such as random access memory (RAM) and / or cache memory. The computer device may further include other removable / non-removable, volatile / non-volatile computer system storage media. By way of example only, the storage system may be used to read and write non-removable, non-volatile magnetic media. Disk drives for reading and writing to removable non-volatile disks (e.g., "floppy disks") and optical disc drives for reading and writing to removable non-volatile optical discs (e.g., CD-ROMs, DVD-ROMs, or other optical media) may be provided. In these cases, each drive may be connected to a bus via one or more data media interfaces. The memory may include at least one program product having a set (e.g., at least one) of program modules configured to perform the functions of the embodiments of the present invention.

[0049] A program / utility having a set (at least one) of program modules can be stored, for example, in memory. Such program modules include—but are not limited to—an operating system, one or more application programs, other program modules, and program data. Each or some combination of these examples may include an implementation of a network environment. The program modules typically perform the functions and / or methods described in the embodiments of this invention.

[0050] Computer devices can also communicate with one or more external devices (such as keyboards, pointing devices, monitors, etc.). This communication can be done through input / output (I / O) interfaces. Furthermore, computer devices can communicate with one or more networks (such as local area networks (LANs), wide area networks (WANs), and / or public networks, such as the Internet) via network adapters.

[0051] The processing unit executes the functions and / or methods described in the embodiments of the present invention by running programs stored in the system memory.

[0052] The aforementioned computer program can be stored in a computer storage medium, that is, the computer storage medium is encoded with a computer program, which, when executed by one or more computers, causes one or more computers to perform the method flow and / or device operation shown in the above embodiments of the present invention.

[0053] With the development of time and technology, the meaning of "medium" has become increasingly broad. The dissemination of computer programs is no longer limited to tangible media; they can also be downloaded directly from the network. Any combination of one or more computer-readable media can be used. A computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. Computer-readable storage media can be, for example,—but not limited to—electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatuses, or devices, or any combination thereof. More specific examples of computer-readable storage media (a non-exhaustive list) include: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this document, a computer-readable storage medium can be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0054] Computer-readable signal media may include data signals propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals may take various forms, including—but not limited to—electromagnetic signals, optical signals, or any suitable combination thereof. Computer-readable signal media may also be any computer-readable medium other than computer-readable storage media, capable of transmitting, propagating, or transmitting programs for use by or in connection with an instruction execution system, apparatus, or device.

[0055] The program code contained on a computer-readable medium may be transmitted using any suitable medium, including—but not limited to—wireless, wire, optical fiber, RF, etc., or any suitable combination thereof.

[0056] Computer program code for performing the operations of this invention can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, as well as conventional procedural programming languages ​​such as "C" or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0057] In addition to the above embodiments, the present invention may have other implementation methods; all technical solutions formed by equivalent substitution or equivalent transformation fall within the protection scope claimed by the present invention.

Claims

1. A method for optimizing the quality of decarburization process of tire cord steel based on data mining, characterized in that, include: Obtain process parameter data for the entire production process of steel cord and corresponding decarburization layer depth detection data; The process parameter data and the decarburized layer depth detection data are preprocessed to obtain a preprocessed dataset; The preprocessed dataset is subjected to feature filtering and feature generation to construct an effective feature set; Using the effective feature set as input, a decarbonization quality prediction model is constructed and trained; The decarbonization quality prediction model is used as the objective function, and a multi-parameter global optimization algorithm is used to solve it under given constraints to determine the target combination of process parameters.

2. The method for optimizing the quality of decarburization process of tire cord steel based on data mining according to claim 1, characterized in that, The process parameter data and the decarburized layer depth detection data are preprocessed to obtain a preprocessed dataset, which includes: Missing values ​​in the process parameter data and the decarburization layer depth detection data are filled based on the metallurgical mechanism model or steady-state operating condition benchmark value, and outliers are identified and replaced based on the preset process parameter range and statistical distribution deviation criteria. Based on a preset decarburization layer depth threshold, the decarburization layer depth detection data is divided into positive samples and negative samples, and the ratio of the number of positive samples to the number of negative samples is balanced to obtain the preprocessed dataset.

3. The method for optimizing the quality of decarburization process of tire cord steel based on data mining according to claim 1, characterized in that, The preprocessed dataset is subjected to feature selection and feature generation to construct an effective feature set, including: The preprocessed dataset is subjected to single-feature screening based on feature coverage, data distribution stability, and variance, and multi-feature screening based on feature correlation and multicollinearity to obtain candidate features. Construct expert characteristics by combining process expert knowledge; The candidate features are combined using an evolutionary algorithm to generate candidate composite features, and the candidate composite features are then screened using an evaluation model to obtain effective composite features. The candidate features, the expert features, and the effective composite features are fused to generate the effective feature set.

4. The method for optimizing the quality of decarburization process of tire cord steel based on data mining according to claim 1, characterized in that, Using the aforementioned effective feature set as input, constructing and training a decarbonization quality prediction model includes: Using the effective feature set as input, a hyperparameter optimization algorithm is used to iteratively optimize within a set parameter space, thereby training and obtaining multiple heterogeneous benchmark prediction models with optimal parameter combinations. Obtain the performance evaluation index of each heterogeneous benchmark prediction model on the validation set, and calculate the performance weight of the corresponding model based on the performance evaluation index; A model fusion strategy is adopted to combine multiple heterogeneous benchmark prediction models to construct a final decarbonization quality prediction model by integrating the prediction results of each heterogeneous benchmark prediction model.

5. The method for optimizing the quality of decarburization process of tire cord steel based on data mining according to claim 1, characterized in that, Using the aforementioned decarbonization quality prediction model as the objective function, a multi-parameter global optimization algorithm is employed under given constraints to determine the target process parameter combination, including: The decarbonization quality prediction model is analyzed using an interpretable algorithm to quantify the degree and direction of influence of each process parameter in the effective feature set on the decarbonization quality prediction results, and to extract key process parameters. Using the decarbonization qualification probability output by the decarbonization quality prediction model as the maximization objective, and the key process parameters as decision variables, an iterative search is performed using a global optimization algorithm based on biological evolution within a preset physical feasible range to output the target process parameter combination.

6. The method for optimizing the quality of decarburization process of tire cord steel based on data mining according to claim 5, characterized in that, The key process parameters include the temperature of the critical zone of the heating furnace, the water tank flow rate, and the wire-spinning temperature.

7. The method for optimizing the quality of decarburization process of tire cord steel based on data mining according to claim 1, characterized in that, After determining the target combination of process parameters, the process further includes: Periodically acquire newly added qualified process parameter data and manually confirmed defect data, and use the newly added qualified process parameter data as negative samples and the manually confirmed defect data as positive samples. The negative samples are merged with the positive samples, and a sliding window mechanism is used to extract historical data within a preset time period for incremental learning and retraining of the decarbonization quality prediction model.

8. A data mining-based quality optimization system for decarburization process of steel cord, characterized in that, include: The data acquisition module is used to acquire process parameter data for the entire production process of steel cord and decarburization layer depth detection data corresponding to the process parameter data. The data preprocessing module is used to preprocess the process parameter data and the decarburization layer depth detection data to obtain a preprocessed dataset; The feature generation module is used to perform feature filtering and feature generation on the preprocessed dataset to construct an effective feature set. The model building module is used to build and train a decarbonization quality prediction model using the effective feature set as input. The parameter optimization module is used to take the decarbonization quality prediction model as the objective function, and solve it using a multi-parameter global optimization algorithm under given constraints to determine the target combination of process parameters.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.

Images