Fire-resistant material performance analysis method and system based on steel plant working condition digital twin data

Abstract

The application discloses a refractory performance analysis method and system based on steel plant working condition digital twin data, and belongs to the technical field of refractory performance analysis; the method comprises the following steps: obtaining historical ore component data, smelting duration data and actual detection data of the same smelting furnace; constructing a material wear shaft, calibrating a plurality of detection wear parameter nodes with the actual detection data, and mapping the historical ore component and the smelting duration in time sequence between the nodes; setting an initial wear coefficient for each historical ore component, constructing an optimization model, optimizing the coefficient through multiple iterations, verifying the node data, and realizing closed-loop optimization of the coefficient; and constructing a judgment model based on the mapping of the optimized coefficient on the wear shaft to predict the real-time wear state of the refractory material. Through structured data fusion and component-specific coefficient optimization, the application effectively solves the problems of low prediction accuracy and poor component adaptability of traditional methods, and improves the accuracy and engineering practicability of refractory wear prediction.

Description

Technical Field

[0001] This invention relates to the field of refractory material performance analysis technology, and in particular to a method and system for refractory material performance analysis based on digital twin data of steel plant operating conditions. Background Technology

[0002] In steel smelting, refractory materials serve as crucial linings for high-temperature molten metal contact equipment such as converters, ladles, and electric furnaces. Their service life directly impacts production safety, equipment utilization, and refractory material consumption costs. However, due to the complex and variable operating parameters during smelting, including ore composition (SiO2, CaO, MgO, Al2O3, FeO, etc.), smelting time, temperature, and slag system, refractory material wear exhibits significant non-uniformity and composition-dependent characteristics. Existing technologies primarily rely on periodic laser or ultrasonic thickness measurements for manual or semi-automatic inspection. This method requires furnace shutdown or cooling, has low inspection frequency, high cost, and struggles to provide forward-looking predictions. Empirical models based on historical average furnace age or simple calorific value statistics cannot effectively distinguish the different impacts of ore composition on wear rates, resulting in significant prediction biases under actual operating conditions with large compositional fluctuations. Summary of the Invention

[0003] The purpose of this invention is to provide a method and system for analyzing the performance of refractory materials that can relatively accurately determine the degree of wear of refractory materials.

[0004] This invention discloses a method for analyzing the performance of refractory materials based on digital twin data of steel plant operating conditions, including: Step S100: Obtain historical smelting data, including historical ore composition data, historical smelting duration data, and historical actual test data; Step S200: Construct a material wear axis based on historical smelting data of the same smelting furnace. Based on each historical actual detection data, calibrate the corresponding wear parameter position on the material wear axis to form several detection wear parameter nodes. Based on the relative occurrence sequence of historical ore composition data and historical smelting time data relative to the detection wear parameter nodes, map the historical ore composition data and historical smelting time data between the detection wear parameter nodes. Step S300: Set an initial material wear coefficient for each historical ore composition data, and build a wear coefficient optimization model. Based on the wear coefficient optimization model, optimize the initial material wear coefficient several times, and verify the wear parameter nodes on the material wear axis to gradually optimize the material wear coefficient. Step S400: Based on the optimized material wear coefficient mapping performance on the material wear axis, a wear coefficient judgment model is constructed, and the wear coefficient judgment model is used to predict the real-time wear state of the refractory material.

[0005] In some embodiments disclosed in this invention, the method for setting an initial material wear coefficient for each historical ore composition data includes: Step S301: Set the initial material wear coefficient based on historical experience. This includes retrieving furnace batches with similar or identical ore composition from the historical smelting data of the digital twin platform, calculating the ratio of their historical actual wear amount to the corresponding historical smelting time, and obtaining the average wear degree per unit time as the initial material wear coefficient.

[0006] In some embodiments disclosed in this invention, the method for constructing a wear coefficient determination model includes: Step S302: Based on a first difference scale above the relative coefficient baseline for each initial material wear coefficient, or a second difference scale below the relative coefficient baseline, determine the step size for adjusting the initial material wear coefficient. Step S303: Analyze the historical smelting duration data between the detection wear parameter nodes to form several historical single smelting durations. Each historical single smelting duration corresponds to a historical ore composition, and the historical ore composition corresponds to an initial wear coefficient. Construct an optimization sub-model for step-by-step optimization and comparison. Determine the number of steps for the change in each initial wear coefficient. Then, uniformly optimize all initial material wear coefficients. Multiply the optimized material wear coefficients and the corresponding historical single smelting durations and sum them to obtain the predicted wear parameter. Compare the predicted wear parameter with the detected wear parameter and provide feedback on the comparison result to the optimization sub-model. This provides data reference for determining the number of steps in the next round until the difference between the predicted wear parameter and the detected wear parameter is less than or equal to a preset value.

[0007] In some embodiments disclosed in this invention, a method for constructing a tuning sub-model for progressively optimizing the alignment includes: Step S3031: Construct a component configuration component representation template for historical ore components, determine the historical ore components of each smelting record between adjacent detection wear parameter nodes to form a historical ore component sequence, and adjust the component representation template based on each historical ore component in the historical ore component sequence to form a component representation template sequence. Step S3032: Randomly determine the number of upward steps or downward steps for adjusting the initial material wear coefficient corresponding to each component representation template in the component representation model sequence, and combine the number of steps corresponding to each component representation template to obtain a sequence of step counts. Step S3033: Based on the step count sequence, calculate the predicted wear parameters, and calculate the wear parameter difference between the predicted wear parameters and the detected wear parameters. The combination of the corresponding wear difference, step count sequence, and component performance model sequence is recorded as the sub-model optimization data. Step S3034: Based on the equivalence between component performance model sequences, the sub-model optimization data is classified and a sub-model optimization data group is constructed. The wear difference in the sub-model optimization data group is sorted, and the step number sequence corresponding to the smallest wear difference is taken as the target step number sequence. Each step number in the target step number sequence is taken as the target direction, and the remaining step number sequences are corrected. After each correction, the wear parameter difference is calculated. If a wear difference smaller than the previous minimum wear difference appears, the corresponding corrected step number sequence is identified as the new target step number sequence. The operation is repeated until the minimum wear difference is less than or equal to the preset value. Then, it is retained together with the corresponding step number sequence and the component performance template sequence to obtain the sub-model optimized data.

[0008] In some embodiments disclosed in this invention, the component representation template includes a component reference point. Several component parameter lines are emitted from the component reference point as the center. Each component parameter line corresponds to a mineral component. Based on historical mineral components, component parameter mapping points are mapped on the component parameter lines. The component parameter mapping points are connected to each other to obtain several mapping point lines.

[0009] In some embodiments disclosed in this invention, the method for determining the equivalence between component representation template sequences includes: Step S30341: Compare the component representation templates of the same order. If the overlap of the corresponding mapping point lines between the templates is within the preset range, then the component representation templates are considered to be equivalent. The overlap of the mapping point lines includes the angle between the mapping point lines and the length difference. Step S30342: If all component representation template sequences are identical, then the component representation template sequences are considered to be identical.

[0010] In some embodiments disclosed in this invention, the method for predicting the real-time wear state of refractory materials using a wear coefficient judgment model includes: Step S401: Determine the position of the previous wear parameter of the refractory material and map it on the material wear axis. Then map the subsequent real-time ore composition and real-time smelting time to obtain the real-time material wear axis. Step S402: Compare the material wear axis in the database with the real-time material wear axis. If the front part of the material wear axis matches the real-time generated part of the real-time material wear axis, then the optimized material wear coefficient mapped on the material wear axis is taken as the reference material wear coefficient. Based on the reference material wear coefficient, the real-time wear parameters of the refractory material are calculated.

[0011] In some embodiments disclosed in this invention, the method for determining whether the front portion of the material wear shaft matches the real-time generation portion of the real-time material wear shaft includes: Step S4021: The difference between the previous wear parameter positions is less than or equal to a preset value, and the difference in ore composition for each smelting is less than or equal to a preset value, and the difference in smelting time for each smelting is less than or equal to a preset value.

[0012] In some embodiments disclosed in this invention, a refractory material performance analysis system based on digital twin data of steel plant operating conditions is also disclosed, including: The first module is used to acquire historical smelting data, including historical ore composition data, historical smelting duration data, and historical actual testing data. The second module is used to construct a material wear axis based on the historical smelting data of the same smelting furnace. Based on each historical actual detection data, the corresponding wear parameter position on the material wear axis is calibrated to form several detection wear parameter nodes. Based on the relative occurrence time of historical ore composition data and historical smelting time data relative to the detection wear parameter nodes, the historical ore composition data and historical smelting time data are mapped to the detection wear parameter nodes. The third module is used to set an initial material wear coefficient for each historical ore composition data, build a wear coefficient optimization model, optimize the initial material wear coefficient several times based on the wear coefficient optimization model, and verify the wear parameter nodes on the material wear axis to gradually optimize the material wear coefficient. The fourth module is used to construct a wear coefficient judgment model based on the mapping performance of the optimized material wear coefficient on the material wear axis, and to use the wear coefficient judgment model to predict the real-time wear state of refractory materials.

[0013] This invention discloses a method and system for analyzing the performance of refractory materials based on digital twin data of steel plant operating conditions, belonging to the field of refractory material performance analysis technology. The method includes acquiring historical ore composition data, smelting time data, and actual test data from the same smelting furnace; constructing a material wear axis, calibrating several wear parameter nodes using actual test data, and mapping historical ore composition and smelting time sequentially between nodes; setting initial wear coefficients for each historical ore composition, constructing an optimization model, and achieving closed-loop coefficient optimization through multiple iterations and verification of node data; and constructing a judgment model based on the mapping of the optimized coefficients on the wear axis to predict the real-time wear state of the refractory material. This invention effectively solves the problems of low prediction accuracy and poor composition adaptability of traditional methods through structured data fusion and composition-specific coefficient optimization, improving the accuracy and engineering practicality of refractory material wear prediction.

[0014] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0015] Figure 1 This is a flowchart illustrating the method steps of the refractory material performance analysis method based on digital twin data of steel plant operating conditions disclosed in this embodiment of the invention. Figure 2 This is a flowchart illustrating the material wear coefficient optimization process disclosed in an embodiment of the present invention. Detailed Implementation

[0016] The technical solution of the present invention will be further described below with reference to the accompanying drawings and embodiments.

[0017] The technical solution of the present invention will be clearly and completely described below with reference to the accompanying drawings and specific embodiments. It should be understood that the preferred embodiments described herein are only for illustration and explanation of the present invention and should not be construed as limiting the scope of protection of the present invention. Those skilled in the art can make some non-essential improvements and adjustments based on the following content of the present invention. In the present invention, unless otherwise expressly specified and limited, the technical terms used in the present invention should have the ordinary meaning understood by those skilled in the art.

[0018] Example: This invention discloses a method for analyzing the performance of refractory materials based on digital twin data of steel plant operating conditions, with reference to... Figure 1 ,include: Step S100: Obtain historical smelting data, including historical ore composition data, historical smelting duration data, and historical actual testing data.

[0019] Step S100 forms the data foundation of the entire method. Its core principle lies in utilizing the multi-source heterogeneous historical data accumulated by the digital twin platform to establish a supervised learning sample set corresponding to the "input-output" relationship. Historical ore composition data and smelting duration data constitute the main process drivers affecting refractory material wear, while historical actual detection data provides scarce but reliable wear reality labels. By acquiring the complete historical records of the same smelting furnace, the continuity and traceability of the data in both time and operating condition dimensions are ensured, providing necessary raw materials and validation benchmarks for subsequent construction of structured models and coefficient optimization. Without complete historical detection data as anchor points, subsequent wear axis construction and coefficient tuning will lack a true reference, making closed-loop optimization difficult to achieve.

[0020] Step S200: Construct a material wear axis based on historical smelting data of the same smelting furnace. Based on each historical actual detection data, calibrate the corresponding wear parameter position on the material wear axis to form several detection wear parameter nodes. Based on the relative occurrence sequence of historical ore composition data and historical smelting duration data relative to the detection wear parameter nodes, map the historical ore composition data and historical smelting duration data between the detection wear parameter nodes.

[0021] The core principle of step S200 is to structurally integrate and time-align discrete actual detection data with continuous process data. By constructing a material wear axis and calibrating each actual detection data point as a node, the wear measurement results, originally scattered across the time axis, are transformed into a cumulative wear parameter axis with clear physical meaning, thus solving the problem of sparse and discontinuous detection data. Simultaneously, all historical ore composition and smelting duration data between two detection nodes are mapped completely to the node interval according to their relative occurrence sequence, achieving a precise correlation between the "process data band" and the "detection anchor point." This design allows subsequent wear coefficient optimization to clearly correspond to specific wear increment intervals, avoiding data confusion and time-series misalignment.

[0022] Step S300: Set an initial material wear coefficient for each historical ore composition data, and build a wear coefficient optimization model. Based on the wear coefficient optimization model, optimize the initial material wear coefficient several times, and verify the wear parameter nodes detected on the material wear axis to gradually optimize the material wear coefficient.

[0023] The principle behind step S300 lies in achieving accurate identification and adaptive optimization of multi-component wear coefficients through a data-driven closed-loop iterative approach. First, initial wear coefficients are statistically analyzed based on historical similar furnace runs, reflecting the idea of ​​transfer learning and quickly obtaining a reasonable starting point using existing experience. Then, an optimization model is constructed, using the measured wear values ​​at detection nodes on the material wear axis as the true values. The wear coefficients of each component are adjusted multiple times, and the difference between the predicted wear and the actual detected values ​​is calculated for feedback correction. This "prediction-comparison-correction" closed-loop mechanism effectively solves the challenge of simultaneously optimizing parameters for multiple component variables when there are limited detection nodes (few samples). The iterative optimization process ensures that the final wear coefficient accurately reflects the erosion contribution of different mineral components under actual working conditions, exhibiting strong physical interpretability.

[0024] Step S400: Based on the optimized material wear coefficient mapping performance on the material wear axis, a wear coefficient judgment model is constructed, and the wear coefficient judgment model is used to predict the real-time wear state of the refractory material.

[0025] The principle of step S400 is to solidify and transfer historical optimization results to real-time operating conditions, realizing the transformation from offline learning to online prediction. By mapping the optimized material wear coefficients onto the wear axis, a judgment model is constructed, linking the coefficients to specific component combinations and wear locations. During real-time prediction, the system matches the similarity between the current real-time ore composition, smelting time, and the optimized intervals on the historical wear axis, reusing the corresponding optimized coefficients for incremental calculation. This prediction mechanism based on reusing similar operating conditions ensures prediction accuracy while avoiding the computational overhead of re-optimizing for each real-time calculation. It enables efficient online application of the model in a digital twin environment, demonstrating the effective transfer of historical experience to real-time decision-making.

[0026] In some embodiments disclosed in this invention, the method for setting an initial material wear coefficient for each historical ore composition data includes: Step S301: Set the initial material wear coefficient based on historical experience. This includes retrieving furnace batches with similar or identical ore composition from the historical smelting data of the digital twin platform, calculating the ratio of their historical actual wear amount to the corresponding historical smelting time, and obtaining the average wear degree per unit time as the initial material wear coefficient.

[0027] The core principle of step S301 is to provide a physically meaningful and data-supported initial wear coefficient for each ore component through experience transfer from similar historical operating conditions, thus laying a good foundation for subsequent closed-loop iterative optimization. This method does not use fixed empirical values ​​or random initialization. Instead, it retrieves furnaces with similar or identical ore components from the massive historical data of the digital twin platform, and calculates the ratio of the actual detected wear amount to the corresponding smelting time for these similar furnaces, obtaining the average wear degree per unit time as the initial coefficient. This historical data statistical method based on component similarity essentially realizes the transfer learning idea of ​​"reusing similar operating conditions," effectively capturing the inherent differences in the erosion rate of refractory materials by different ore components. This makes the initial coefficient closer to the real physical laws, avoiding the slow convergence and easy getting trapped in local optima that may occur with optimization from scratch. At the same time, it significantly improves the iterative efficiency of the subsequent optimization model and the reliability of the final optimization results.

[0028] In some embodiments disclosed in this invention, reference is made to Figure 2 Methods for constructing wear coefficient judgment models include: Step S302: Based on a first difference scale above the relative coefficient baseline for each initial material wear coefficient, or a second difference scale below the relative coefficient baseline, determine the step size for adjusting the initial material wear coefficient.

[0029] The core principle of step S302 is to adaptively determine the adjustment step size based on the deviation of the initial wear coefficient from the baseline, thereby achieving a balance between rapid approximation in the early stages of optimization and fine-tuning in the later stages. This method first calculates the baseline (usually the average or median) for all initial material wear coefficients, then determines whether each coefficient is too high or too low relative to the baseline, and determines the corresponding change step size based on the scale of deviation (first or second scale). Coefficients with larger deviations are assigned larger step sizes to accelerate their convergence to a reasonable range; coefficients with smaller deviations are assigned smaller step sizes to avoid over-adjustment leading to oscillations. This adaptive step size design based on deviation magnitude is essentially a variable step size strategy in heuristic optimization, which improves iteration efficiency and enhances the stability of the optimization process, providing a reasonable search granularity for subsequent unified multi-coefficient tuning.

[0030] Step S303: Analyze the historical smelting duration data between the detection wear parameter nodes to form several historical single smelting durations. Each historical single smelting duration corresponds to a historical ore composition, and the historical ore composition corresponds to an initial wear coefficient. Construct an optimization sub-model for step-by-step optimization and comparison. Determine the number of steps for the change in each initial wear coefficient. Then, uniformly optimize all initial material wear coefficients. Multiply the optimized material wear coefficients and the corresponding historical single smelting durations and sum them to obtain the predicted wear parameter. Compare the predicted wear parameter with the detected wear parameter and provide feedback on the comparison result to the optimization sub-model. This provides data reference for determining the number of steps in the next round until the difference between the predicted wear parameter and the detected wear parameter is less than or equal to a preset value.

[0031] Step S303 transforms the wear coefficient optimization into a closed-loop iterative search and feedback correction process driven by real detection data, addressing the challenge of parameter identification for multi-component coefficients at limited observation points. This step first decomposes the historical smelting duration data between detection nodes into several single smelting records, and associates the ore composition corresponding to each record with its initial wear coefficient, constructing an optimization sub-model. In each iteration, the system determines the number of steps for the coefficient change of each component, uniformly adjusts all initial coefficients, calculates the predicted wear parameter (the cumulative sum of the optimized coefficient and the corresponding duration), and compares it with the actual wear parameter at the detection node. Based on the magnitude of the difference generated by the comparison, the number of steps for each component in the next round is adjusted, gradually reducing the deviation between the predicted and measured values ​​until the difference is less than or equal to a preset threshold. This closed-loop mechanism of "prediction-comparison-feedback-correction" fully utilizes sparse detection nodes as ground truth anchors, achieving efficient convergence in a multi-variable, high-dimensional search space, ensuring that the optimized coefficients accurately reflect the wear contribution of different components under actual working conditions.

[0032] In some embodiments disclosed in this invention, a method for constructing a tuning sub-model for progressively optimizing the alignment includes: Step S3031: Construct a component configuration component representation template for historical ore components, determine the historical ore components of each smelting record between adjacent detection wear parameter nodes to form a historical ore component sequence, and adjust the component representation template based on each historical ore component in the historical ore component sequence to form a component representation template sequence.

[0033] Step S3031 works by using a geometric representation method to transform high-dimensional, variable mineral composition data into comparable and quantifiable "composition representation templates," thus providing efficient structured features for subsequent similarity judgment and grouping optimization. This step first constructs a composition representation template for historical mineral compositions, radiating multiple radial composition parameter lines centered on a composition baseline point. Parameter points are mapped onto each line based on the composition content, and these points are then connected to form a closed geometric shape. For each smelting record between adjacent detection nodes, its historical mineral composition sequence is determined, and the template is adjusted accordingly to generate a corresponding composition representation template sequence. This method of abstracting composition combinations into geometric signatures essentially achieves a dimensionality reduction mapping from a high-dimensional composition space to a low-dimensional geometric feature space. This allows for rapid quantification and comparison of the similarity between different composition sequences through geometric features (connection lengths and angles), laying the foundation for subsequent equivalence classification and search space compression.

[0034] Step S3032: Randomly determine the number of upward steps or downward steps for adjusting the initial material wear coefficient corresponding to each component representation template in the component representation model sequence, and combine the number of steps corresponding to each component representation template to obtain a sequence of step counts.

[0035] Step S3032 employs a random sampling strategy to generate diverse candidate solutions, enabling a broad exploration of the adjustment space for multi-component wear coefficients. This step randomly determines the number of upward or downward steps for the initial material wear coefficient corresponding to each template in the component representation template sequence, and combines the step counts of all templates to form multiple different step count sequences. This random combination generation method avoids local optima that may result from searching in a single fixed direction, and increases the probability of finding a globally optimal solution by generating multiple candidate schemes. Each step count sequence represents a possible scheme for jointly adjusting all component coefficients within the current historical interval, providing rich search samples for subsequent prediction calculations and error assessments.

[0036] Step S3033: Based on the step count sequence, calculate the predicted wear parameters, and calculate the wear parameter difference between the predicted wear parameters and the detected wear parameters. The combination of the corresponding wear difference, the step count sequence, and the component performance model sequence is recorded as the sub-model optimization data.

[0037] Step S3033 transforms each candidate stepping scheme into a quantifiable prediction result and generates an optimized sample with error labels by comparing it with real detection data, thus achieving a mapping from scheme to performance. Based on the generated stepping sequence, this step adjusts the initial wear coefficient of the corresponding component, calculates the predicted wear parameter (the cumulative sum of the adjusted coefficient and the corresponding smelting time), and compares it with the detected wear parameter to obtain the wear difference. Subsequently, the wear difference, the stepping sequence, and the corresponding component performance template sequence are combined and recorded as data for sub-model optimization. This integrated "scheme-prediction-error" recording method essentially constructs an evaluable dataset for the heuristic search process, enabling subsequent sorting and selection of different schemes based on error magnitude.

[0038] Step S3034: Based on the equivalence between component performance model sequences, the sub-model optimization data is classified and a sub-model optimization data group is constructed. The wear difference in the sub-model optimization data group is sorted, and the step number sequence corresponding to the smallest wear difference is taken as the target step number sequence. Each step number in the target step number sequence is taken as the target direction, and the remaining step number sequences are corrected. After each correction, the wear parameter difference is calculated. If a wear difference smaller than the previous minimum wear difference appears, the corresponding corrected step number sequence is identified as the new target step number sequence. The operation is repeated until the minimum wear difference is less than or equal to the preset value. Then, it is retained together with the corresponding step number sequence and the component performance template sequence to obtain the sub-model optimized data.

[0039] Step S3034 utilizes the equivalence of component templates to group the search space and achieves efficient iterative convergence through an "optimal guided correction" strategy. This step first categorizes the sub-model optimization data based on the equivalence between component representation template sequences (geometric feature similarity within a threshold), constructing several sub-model optimization data groups. Within each group, the wear difference is sorted, and the step sequence with the smallest error is selected as the target sequence. Then, using the step count of the target sequence as the direction, other sequences within the group are progressively corrected, and the error is recalculated after each correction. If a smaller error appears, the target sequence is updated and iteration continues until the error is less than or equal to a preset value. This "grouping + elite guidance + local correction" mechanism reduces redundant searching through template equivalence and significantly improves the efficiency of finding the optimal or near-optimal step combination in the multivariable space by continuously approaching the current optimal solution. Ultimately, the optimal step count sequence and its corresponding template sequence are retained as the sub-model optimization result.

[0040] In some embodiments disclosed in this invention, the component representation template includes a component reference point. Several component parameter lines are emitted from the component reference point as the center. Each component parameter line corresponds to a mineral component. Based on historical mineral components, component parameter mapping points are mapped on the component parameter lines. The component parameter mapping points are connected to each other to obtain several mapping point lines.

[0041] In some embodiments disclosed in this invention, the method for determining the equivalence between component representation template sequences includes: Step S30341: Compare component representation templates of the same order. If the overlap of the corresponding mapping point lines between the templates is within the preset range, then the component representation templates are considered to be equivalent. The overlap of the mapping point lines includes the angle between the mapping point lines and the length difference.

[0042] Step S30342: If all component representation template sequences are identical, then the component representation template sequences are considered to be identical.

[0043] In some embodiments disclosed in this invention, the method for predicting the real-time wear state of refractory materials using a wear coefficient judgment model includes: Step S401: Determine the position of the previous wear parameter of the refractory material and map it on the material wear axis. Then map the subsequent real-time ore composition and real-time smelting time to obtain the real-time material wear axis.

[0044] Step S402: Compare the material wear axis in the database with the real-time material wear axis. If the front part of the material wear axis matches the real-time generated part of the real-time material wear axis, then the optimized material wear coefficient mapped on the material wear axis is taken as the reference material wear coefficient. Based on the reference material wear coefficient, the real-time wear parameters of the refractory material are calculated.

[0045] In some embodiments disclosed in this invention, the method for determining whether the front portion of the material wear shaft matches the real-time generation portion of the real-time material wear shaft includes: Step S4021: The difference between the previous wear parameter positions is less than or equal to a preset value, and the difference in ore composition for each smelting is less than or equal to a preset value, and the difference in smelting time for each smelting is less than or equal to a preset value.

[0046] In some embodiments disclosed in this invention, a refractory material performance analysis system based on digital twin data of steel plant operating conditions is also disclosed, including: The first module is used to acquire historical smelting data, including historical ore composition data, historical smelting duration data, and historical actual testing data. The second module is used to construct a material wear axis based on the historical smelting data of the same smelting furnace. Based on each historical actual detection data, the corresponding wear parameter position on the material wear axis is calibrated to form several detection wear parameter nodes. Based on the relative occurrence time of historical ore composition data and historical smelting time data relative to the detection wear parameter nodes, the historical ore composition data and historical smelting time data are mapped to the detection wear parameter nodes. The third module is used to set an initial material wear coefficient for each historical ore composition data, build a wear coefficient optimization model, optimize the initial material wear coefficient several times based on the wear coefficient optimization model, and verify the wear parameter nodes on the material wear axis to gradually optimize the material wear coefficient. The fourth module is used to construct a wear coefficient judgment model based on the mapping performance of the optimized material wear coefficient on the material wear axis, and to use the wear coefficient judgment model to predict the real-time wear state of refractory materials.

[0047] This invention discloses a method and system for analyzing the performance of refractory materials based on digital twin data of steel plant operating conditions, belonging to the field of refractory material performance analysis technology. The method includes acquiring historical ore composition data, smelting time data, and actual test data from the same smelting furnace; constructing a material wear axis, calibrating several wear parameter nodes using actual test data, and mapping historical ore composition and smelting time sequentially between nodes; setting initial wear coefficients for each historical ore composition, constructing an optimization model, and achieving closed-loop coefficient optimization through multiple iterations and verification of node data; and constructing a judgment model based on the mapping of the optimized coefficients on the wear axis to predict the real-time wear state of the refractory material. This invention effectively solves the problems of low prediction accuracy and poor composition adaptability of traditional methods through structured data fusion and composition-specific coefficient optimization, improving the accuracy and engineering practicality of refractory material wear prediction.

[0048] Through the above description of the embodiments, those skilled in the art can clearly understand that the present invention can be implemented in hardware or by means of software plus necessary general-purpose hardware platforms. Based on this understanding, the technical solution of the present invention can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (such as a CD-ROM, USB flash drive, external hard drive, etc.) and includes several instructions to cause a computer device (such as a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0049] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for analyzing the performance of refractory materials based on digital twin data of steel plant operating conditions, characterized in that, include: Step S100: Obtain historical smelting data, including historical ore composition data, historical smelting duration data, and historical actual test data; Step S200: Construct a material wear axis based on historical smelting data of the same smelting furnace. Based on each historical actual detection data, calibrate the corresponding wear parameter position on the material wear axis to form several detection wear parameter nodes. Based on the relative occurrence sequence of historical ore composition data and historical smelting time data relative to the detection wear parameter nodes, map the historical ore composition data and historical smelting time data between the detection wear parameter nodes. Step S300: Set an initial material wear coefficient for each historical ore composition data, and build a wear coefficient optimization model. Based on the wear coefficient optimization model, optimize the initial material wear coefficient several times, and verify the wear parameter nodes on the material wear axis to gradually optimize the material wear coefficient. Step S400: Based on the optimized material wear coefficient mapping performance on the material wear axis, a wear coefficient judgment model is constructed, and the wear coefficient judgment model is used to predict the real-time wear state of the refractory material.

2. The method for analyzing the performance of refractory materials based on digital twin data of steel plant operating conditions according to claim 1, characterized in that, Methods for setting initial material wear coefficients for each historical ore composition data include: Step S301: Set the initial material wear coefficient based on historical experience. This includes retrieving furnace batches with similar or identical ore composition from the historical smelting data of the digital twin platform, calculating the ratio of their historical actual wear amount to the corresponding historical smelting time, and obtaining the average wear degree per unit time as the initial material wear coefficient.

3. The method for analyzing the performance of refractory materials based on digital twin data of steel plant operating conditions according to claim 1, characterized in that, Methods for constructing wear coefficient judgment models include: Step S302: Based on a first difference scale above the relative coefficient baseline for each initial material wear coefficient, or a second difference scale below the relative coefficient baseline, determine the step size for adjusting the initial material wear coefficient. Step S303: Analyze the historical smelting duration data between the detection wear parameter nodes to form several historical single smelting durations. Each historical single smelting duration corresponds to a historical ore composition, and the historical ore composition corresponds to an initial wear coefficient. Construct an optimization sub-model for step-by-step optimization and comparison. Determine the number of steps for the change in each initial wear coefficient. Then, uniformly optimize all initial material wear coefficients. Multiply the optimized material wear coefficients and the corresponding historical single smelting durations and sum them to obtain the predicted wear parameter. Compare the predicted wear parameter with the detected wear parameter and provide feedback on the comparison result to the optimization sub-model. This provides data reference for determining the number of steps in the next round until the difference between the predicted wear parameter and the detected wear parameter is less than or equal to a preset value.

4. The method for analyzing the performance of refractory materials based on digital twin data of steel plant operating conditions according to claim 3, characterized in that, Methods for constructing tuned sub-models for progressively optimizing alignments include: Step S3031: Construct a component configuration component representation template for historical ore components, determine the historical ore components of each smelting record between adjacent detection wear parameter nodes to form a historical ore component sequence, and adjust the component representation template based on each historical ore component in the historical ore component sequence to form a component representation template sequence. Step S3032: Randomly determine the number of upward steps or downward steps for adjusting the initial material wear coefficient corresponding to each component representation template in the component representation model sequence, and combine the number of steps corresponding to each component representation template to obtain a sequence of step counts. Step S3033: Based on the step count sequence, calculate the predicted wear parameters, and calculate the wear parameter difference between the predicted wear parameters and the detected wear parameters. The combination of the corresponding wear difference, step count sequence, and component performance model sequence is recorded as the sub-model optimization data. Step S3034: Based on the equivalence between component performance model sequences, the sub-model optimization data is classified to construct a sub-model optimization data group. The wear difference in the sub-model optimization data group is sorted, and the step number sequence corresponding to the minimum wear difference is taken as the target step number sequence. Each step number in the target step number sequence is taken as the target direction, and the remaining step number sequences are corrected. After each correction, the wear parameter difference is calculated. If a wear difference smaller than the previous minimum wear difference appears, the corresponding corrected step number sequence is identified as the new target step number sequence. The operation is repeated until the minimum wear difference is less than or equal to a preset value. Then, it is retained together with the corresponding step number sequence and the component performance template sequence to obtain the sub-model optimized data.

5. The method for analyzing the performance of refractory materials based on digital twin data of steel plant operating conditions according to claim 4, characterized in that, The composition representation template includes a composition reference point. Centered on the composition reference point, several composition parameter lines are emitted. Each composition parameter line corresponds to a mineral composition. Based on historical mineral compositions, composition parameter mapping points are mapped onto the composition parameter lines. The composition parameter mapping points are connected to each other to obtain several mapping point lines.

6. The method for analyzing the performance of refractory materials based on digital twin data of steel plant operating conditions according to claim 5, characterized in that, Methods for determining the equivalence between component representation template sequences include: Step S30341: Compare the component representation templates of the same order. If the overlap of the corresponding mapping point lines between the templates is within the preset range, then the component representation templates are considered to be equivalent. The overlap of the mapping point lines includes the angle between the mapping point lines and the length difference. Step S30342: If all component representation template sequences are identical, then the component representation template sequences are considered to be identical.

7. The method for analyzing the performance of refractory materials based on digital twin data of steel plant operating conditions according to claim 1, characterized in that, Methods for predicting the real-time wear state of refractory materials using wear coefficient judgment models include: Step S401: Determine the position of the previous wear parameter of the refractory material and map it on the material wear axis. Then map the subsequent real-time ore composition and real-time smelting time to obtain the real-time material wear axis. Step S402: Compare the material wear axis in the database with the real-time material wear axis. If the front part of the material wear axis matches the real-time generated part of the real-time material wear axis, then the optimized material wear coefficient mapped on the material wear axis is taken as the reference material wear coefficient. Based on the reference material wear coefficient, the real-time wear parameters of the refractory material are calculated.

8. The method for analyzing the performance of refractory materials based on digital twin data of steel plant operating conditions according to claim 7, characterized in that, Methods for determining the match between the initial portion of the material wear axis and the real-time generated portion of the real-time material wear axis include: Step S4021: The difference between the previous wear parameter positions is less than or equal to a preset value, and the difference in ore composition for each smelting is less than or equal to a preset value, and the difference in smelting time for each smelting is less than or equal to a preset value.

9. A refractory material performance analysis system based on digital twin data of steel plant operating conditions, characterized in that, include: The first module is used to acquire historical smelting data, including historical ore composition data, historical smelting duration data, and historical actual testing data. The second module is used to construct a material wear axis based on the historical smelting data of the same smelting furnace. Based on each historical actual detection data, the corresponding wear parameter position on the material wear axis is calibrated to form several detection wear parameter nodes. Based on the relative occurrence time of historical ore composition data and historical smelting time data relative to the detection wear parameter nodes, the historical ore composition data and historical smelting time data are mapped to the detection wear parameter nodes. The third module is used to set an initial material wear coefficient for each historical ore composition data, build a wear coefficient optimization model, optimize the initial material wear coefficient several times based on the wear coefficient optimization model, and verify the wear parameter nodes on the material wear axis to gradually optimize the material wear coefficient. The fourth module is used to construct a wear coefficient judgment model based on the mapping performance of the optimized material wear coefficient on the material wear axis, and to use the wear coefficient judgment model to predict the real-time wear state of refractory materials.

Images