Production method of annealing-free cold forging steel hot-rolled wire rod

Abstract

The invention discloses a production method of a non-annealing cold heading steel hot-rolled wire rod, and relates to the technical field of steel production, the method comprises the following steps: selecting a steel billet meeting the production specification requirement of the cold heading steel wire rod, and carrying out coping finishing and preheating treatment on the steel billet to be processed; the steel billet is conveyed to a large-compression-ratio rolling unit, the hot rolling process is monitored in real time, hot rolling segmented regulation and control are conducted, multi-pass temperature gradient rolling is achieved, and a cold heading steel wire rod formed through hot rolling is obtained; and the hot-rolled cold heading steel wire rod is conveyed to a heat preservation tunnel furnace, a cold heading steel wire rod finished product is obtained through segmented controlled cooling and surface treatment, and production of the annealing-free cold heading steel hot-rolled wire rod is achieved. According to the method, the advanced rolling mill unit with the large compression ratio is adopted for producing the cold heading steel wire rods of different specifications, the heat preservation tunnel furnace is adopted, the cooling temperature and the cooling speed of the wire rods are controlled, production of the annealing-free cold heading steel hot-rolled wire rods is achieved, the production process is greatly simplified, the production period is shortened, the production cost is reduced, and the production efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of steel production, and particularly to a production method for hot-rolled wire rods of cold-heading steel without annealing. Background Art

[0002] In the technical field of steel production, hot-rolling technology is an important processing method. It rolls the steel billet at high temperature to reduce its thickness and increase its length, thereby obtaining steel products with the required shape and size. Cold-heading steel is a high-quality carbon structural steel, mainly used for manufacturing standard parts such as bolts, nuts, and bearing rings. Its characteristics are high strength, good toughness, and wear resistance. However, the traditional production process of cold-heading steel requires annealing treatment to improve its organizational structure and performance, which not only increases the production cost but also prolongs the production cycle.

[0003] To solve this problem, some enterprises have adopted the production method of cold-heading steel without annealing. This method mainly changes the composition and organizational structure of the steel billet so that it can achieve the required performance without annealing treatment. In addition, some enterprises have adopted special rolling processes, such as large reduction ratio rolling, to improve the strength and toughness of cold-heading steel.

[0004] However, the existing production methods of cold-heading steel without annealing still have some problems. First, these methods often require special equipment and complex processes, which increase the production cost. Second, these methods have strict requirements on the composition and organizational structure of the steel billet, which limits their application range. Finally, the performance of the cold-heading steel produced by these methods still needs to be improved, especially in terms of strength and toughness.

[0005] Therefore, how to produce high-performance cold-heading steel without annealing through a simple process without changing the composition and organizational structure of the steel billet is still an urgent problem to be solved. Summary of the Invention

[0006] Based on this, it is necessary to provide a production method for hot-rolled wire rods of cold-heading steel without annealing in view of the above technical problems.

[0007] The present invention provides a production method for hot-rolled wire rods of cold-heading steel without annealing, including:

[0008] S1. Select a steel billet that meets the production specification requirements of cold-heading steel wire rods, and perform grinding, finishing, and preheating treatment on the steel billet to be processed;

[0009] S2. Convey the steel billet to a large reduction ratio rolling mill set, and monitor the hot-rolling process in real time to perform hot-rolling segmented regulation, so as to achieve multi-pass temperature gradient rolling and obtain hot-rolled formed cold-heading steel wire rods;

[0010] S3. Convey the cold-heading steel wire rod after hot rolling to an insulation tunnel furnace, and through segmented controlled cooling and surface treatment, obtain the finished cold-heading steel wire rod, realizing the production of hot-rolled wire rods of cold-heading steel without annealing.

[0011] Furthermore, select steel billets that meet the production specification requirements of cold-heading steel wire rods, and perform grinding, finishing and preheating treatment on the steel billets to be processed, including:

[0012] S11. Preset the production specifications of cold-heading steel wire rods in advance, select steel billets of corresponding sizes, and detect the component contents of various elements in the steel billets through a laser spectrometer;

[0013] S12. Adopt a two-round grinding wheel grinding method. In the first round, use a 16-20 mesh grinding wheel to remove the wrinkles at the corners of the steel billet, and in the second round, switch to a 24-26 mesh grinding wheel to finely polish the surface of the steel billet;

[0014] S13. Place the ground steel billet in a preheating environment of 600-800 °C, gradually raise the temperature of the preheating environment to 1090-1170 °C, and then introduce a CO2 / N2 mixed gas to increase the surface hardness of the steel billet.

[0015] Furthermore, convey the steel billets to a high compression ratio rolling mill unit, and monitor the hot rolling process in real time to perform hot rolling segmented regulation, realizing multi-pass temperature gradient rolling, and obtaining hot-rolled cold-heading steel wire rods, including: S21. According to the production specifications of cold-heading steel wire rods and the component contents of the steel billets, divide the rolling process into three stages: rough rolling, medium rolling and finish rolling, and use a physical constraint element learning algorithm to match the optimal rolling parameters;

[0016] S22. Monitor the hot rolling data in the hot rolling process in real time, perform dynamic compensation on the optimal rolling parameters, and dynamically adjust the pass interval time according to the monitoring and compensation results; S23. After rolling, spray a nano-Al2O3-MgO composite coating on the surface of the cold-heading steel wire rod, use the waste heat to generate a dense oxide film, and verify the actual production specifications of the cold-heading steel wire rod.

[0017] Furthermore, according to the production specifications of cold-heading steel wire rods and the component contents of the steel billets, divide the rolling process into three stages: rough rolling, medium rolling and finish rolling, and use a physical constraint element learning algorithm to match the optimal rolling parameters, including:

[0018] S211. Based on the preset production specifications of cold-heading steel wire rods, reverse deduce the cross-sectional dimensions of the steel billets, and calculate the target compression ratio for hot rolling the steel billets to cold-heading steel wire rods; S212. Divide the rolling process of the steel billets into three stages: rough rolling, medium rolling and finish rolling, establish a parameter mapping table using the historical database, and output the initial compression ratio and rolling temperature range for each stage;

[0019] S213. Establish physical constraints for hot rolling of steel billets based on metallurgical theory. The physical constraints include phase transformation kinetics constraints, deformation resistance and rolling force constraints, and precipitation phase pinning constraints.

[0020] S214. Set input variables and output variables, construct and train a parameter recommendation model based on meta-learning, integrate physical metallurgical constraints, and match rolling parameters corresponding to different steel billets.

[0021] S215. Input the size and composition content of the steel billet to be processed into the parameter recommendation model, generate rolling parameters for each stage and real-time adjustment logic, and obtain the optimal rolling parameters.

[0022] Furthermore, setting input variables and output variables, constructing and training a parameter recommendation model based on meta-learning, integrating physical metallurgical constraints, and matching rolling parameters corresponding to different steel billets includes:

[0023] S2141. Corresponding each original task to a steel grade, set the production specifications of cold heading steel wire rods and the content of each component in the steel billet as input variables, and rolling parameters as output variables to construct a support set.

[0024] S2142. Link a physical calculation module after the output of the hidden layer to achieve physical information embedding, which is used to calculate the temperature and reduction ratio for steel billet rolling in each stage.

[0025] S2143. Set two physical loss term functions of phase transformation loss and rolling force loss, and combine the phase transformation loss and rolling force loss to set a comprehensive loss function to construct a parameter recommendation model.

[0026] S2144. Pre-train model parameters on all steel grade tasks, initially form a hot rolling knowledge base, and when facing a new steel grade, achieve fine-tuning of the calculation task through single-step gradient descent.

[0027] Furthermore, inputting the size and composition content of the steel billet to be processed into the parameter recommendation model, generating rolling parameters for each stage and real-time adjustment logic, and obtaining the optimal rolling parameters includes:

[0028] S2151. Obtain the carbon content of the steel billet to be processed, match the rough rolling temperature corresponding to the rough rolling stage according to the carbon content value, and calculate the rough rolling reduction ratio.

[0029] The calculation formula for the rough rolling temperature is:

[0030] ;

[0031] In the formula, T 粗轧 represents the rough rolling temperature in the rough rolling stage; T0 represents the preset initial temperature; [C] represents the carbon content.

[0032] The calculation formula for the rough rolling reduction ratio is:

[0033] ;

[0034] Wherein, r 粗轧 represents the rough rolling reduction rate in the rough rolling stage; r1 represents the initial compression ratio in the rough rolling stage output by the parameter mapping table; R total represents the target compression ratio; S2152. Obtain the titanium content, manganese content of the steel billet to be processed and the grain size after rough rolling, and calculate the intermediate rolling cooling rate and intermediate rolling reduction rate when the steel billet enters the intermediate rolling stage; The calculation formula for the intermediate rolling cooling rate is:

[0035] ; Wherein, v T represents the intermediate rolling cooling rate; Δ T represents the temperature difference between rough rolling and intermediate rolling; Δ t represents the time difference; Mn represents the manganese content; The calculation formula for the intermediate rolling reduction rate is:

[0036] ; Wherein, r 中轧 represents the intermediate rolling reduction rate; r 2 represents the initial compression ratio in the intermediate rolling stage output by the parameter mapping table; Ti represents the titanium content;

[0037] S2153. Calculate the Ar3 phase transformation point of the steel billet to be processed, set the finish rolling temperature in the finish rolling stage according to the Ar3 phase transformation point, and calculate the finish rolling reduction rate in the finish rolling stage according to the carbon content;

[0038] The calculation formula for the finish rolling reduction rate is:

[0039] ;

[0040] Wherein, r 精轧 represents the finish rolling reduction rate; r3 represents the initial compression ratio in the finish rolling stage output by the parameter mapping table.

[0041] Furthermore, the hot rolling data in the hot rolling process is monitored in real time, dynamic compensation is performed on the optimal rolling parameters, and the pass interval time is dynamically adjusted according to the monitoring compensation result, including:

[0042] S221. Use the sensors inside the rolling mill to collect the hot rolling data of the steel billet rolling process in real time;

[0043] S222. Calculate the difference between the actual rolling force and the target rolling force during the rolling process of each stage of the rolling mill, and calculate the reduction rate compensation amount through the proportional-integral coefficient;

[0044] S223. Calculate the temperature difference between the head and the tail of the steel billet after passing through each stage. If the temperature difference between the head and the tail is greater than the preset threshold, trigger the adjustment of extending the pass interval to balance the temperature drop and the rolling rhythm;

[0045] S224. Import the real-time hot rolling data into the historical parameter library, set the update period, and update the comprehensive loss function regularly. If there is a mutation in the hot rolling data, trigger an emergency stop signal.

[0046] Furthermore, convey the cold heading steel wire rod after hot rolling to an insulation tunnel furnace, and through segmented controlled cooling and surface treatment, obtain the finished cold heading steel wire rod. The production of the cold heading steel hot rolled wire rod without annealing includes: S31. Obtain and record the wire rod parameters of the cold heading steel wire rod after hot rolling. The wire rod parameters include the finishing rolling temperature, the grain size after rolling, the pre-judged ferrite ratio, and the rolling force curve;

[0047] S32. Divide the cooling process of the cold heading steel wire rod into three stages: rapid cooling, slow cooling, and final cooling, and build a physical constraint meta-learning framework to generate cooling parameters to control the cooling of the cold heading steel wire rod in segments;

[0048] S33. Use high-pressure water to remove phosphorus from the cooled cold heading steel wire rod, and identify surface defects through on-line laser detection. After the detection is correct, obtain the finished cold heading steel wire rod.

[0049] Furthermore, dividing the cooling process of the cold heading steel wire rod into three stages: rapid cooling, slow cooling, and final cooling, and building a physical constraint meta-learning framework to generate cooling parameters to control the cooling of the cold heading steel wire rod in segments includes:

[0050] S321. Introduce wire rod parameters to expand the input layer, embed the phase transformation kinetics equation and the heat conduction equation to build a physical constraint layer, and build an output layer that outputs the segmented cooling rate value and the final cooling temperature;

[0051] S322. Divide the cooling process of the cold heading steel wire rod into rapid cooling, slow cooling, and final cooling stages, and calculate the cooling rate and cooling time in the rapid cooling stage respectively;

[0052] S323. Based on the preset segmented air cooling strategy, use a fan-water mist combination for cooling in the rapid cooling stage, turn off the fan in the slow cooling stage, and use natural cooling by the roller table. In the final cooling stage, use natural cooling and introduce a nitrogen-carbon dioxide mixed gas to generate a dense oxide film.

[0053] Furthermore, calculating the cooling rate and cooling time in the rapid cooling stage includes:

[0054] Calculation formula for the rapid cooling rate in the rapid cooling stage:

[0055] ; In the formula, v 速冷 represents the rapid cooling rate in the rapid cooling stage; T 终轧 represents the finish rolling temperature; T 速冷终 represents the target temperature in the rapid cooling stage; t 速冷 represents the rapid cooling time;

[0056] Calculation formula for the cooling time in the slow cooling stage:

[0057] ;

[0058] In the formula, t 缓冷 represents the slow cooling time; k0 represents the phase transformation rate constant; represents the activation energy for phase transformation; R represents the gas constant; T 缓冷 represents the slow cooling temperature; X total represents the target ferrite ratio; n represents the Avrami exponent; e represents the natural constant.

[0059] The beneficial effects of the present invention are as follows:

[0060] 1. By using an advanced rolling mill with a large compression ratio to produce cold heading steel wire rods of different specifications, and with the aid of a heat preservation tunnel furnace to control the cooling temperature and cooling speed of the wire rods, the production of hot-rolled cold heading steel wire rods without annealing is realized, greatly simplifying the production process, shortening the production cycle, reducing the production cost, and improving the production efficiency; the cold heading steel wire rods produced do not need to be annealed, avoiding the changes in the organizational structure and performance that may be brought about by the annealing treatment, thus improving the quality stability of the products; at the same time, the present invention has no strict requirements on the composition and organizational structure of the steel billet, so it can be applied to more steel grades, expanding its application range; by precisely controlling the cooling temperature and cooling speed, the strength and toughness of the products are significantly improved, and they have better use performance.

[0061] 2. By precisely controlling the temperature changes in each stage through multi-pass temperature gradient rolling, and combining with the segmented controlled cooling process to achieve the directional phase transformation from austenite to ferrite and the spheroidization of pearlite, breaking through the limitations of traditional processes; in the rolling process, through the coordinated action of differential temperature control and large reduction ratios in the rough, medium, and finish rolling stages, ultrafine grains and uniform structures are directly obtained; in the cooling process, rapid cooling is used to inhibit the coarsening of carbides, slow cooling is used to promote phase transformation equilibrium, supplemented by the final cooling oxide film control technology, completely eliminating the need for annealing; ultimately achieving the integrated performance improvement of high strength, high plasticity, and corrosion resistance of cold heading steel wire rods, while significantly reducing energy consumption and production costs, and promoting the upgrading of the steel hot rolling industry towards green and efficient development. Brief Description of the Drawings

[0062] The drawings described herein are provided to further understand the present invention and form a part of the present invention. The schematic embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings: Figure 1 is a flowchart of a production method for hot-rolled wire rods of non-annealed cold-heading steel according to an embodiment of the present invention; Figure 2 is a typical tissue characterization map of the SCM435 steel grade before optimization according to an embodiment of the present invention; Figure 3 is a typical tissue characterization map of the SCM435 steel grade after optimization according to an embodiment of the present invention; Figure 4 is a mechanical property diagram of different shear turns of the SCM435 steel grade before optimization according to an embodiment of the present invention; Figure 5 is a mechanical property diagram of different shear turns of the SCM435 steel grade after optimization according to an embodiment of the present invention; Figure 6 is a typical tissue diagram of the cold-heading wire of the B7 steel grade before optimization according to an embodiment of the present invention; Figure 7 is a typical tissue diagram of the cold-heading wire of the B7 steel grade after optimization according to an embodiment of the present invention. Detailed Description of the Embodiments

[0063] In order to make the objectives, technical solutions and advantages of the present invention clearer and more understandable, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0064] Please refer to Figure 1 , which provides a production method for hot-rolled wire rods of non-annealed cold-heading steel, including:

[0065] S1. Select a steel billet that meets the production specification requirements of the cold-heading steel wire rod, and perform grinding, finishing and preheating treatments on the steel billet to be processed.

[0066] In the description of the present invention, selecting a steel billet that meets the production specification requirements of the cold-heading steel wire rod and performing grinding, finishing and preheating treatments on the steel billet to be processed includes:

[0067] S11. Preset the production specifications of the cold-heading steel wire rod, select a steel billet of the corresponding size, and detect the component content of each component element in the steel billet by a laser spectrometer.

[0068] Specifically, before the production of cold heading steel wire rods, the billet size needs to be deduced based on the target wire rod specifications (such as a diameter of Φ16mm) to ensure a reduction ratio of ≥80%. For example, to produce a Φ16mm wire rod, a 160mm×160mm square billet is selected, and its reduction ratio is calculated as the ratio of the original cross-sectional area to the target cross-sectional area to ensure sufficient plastic deformation during the rolling process.

[0069] The composition of the billet is detected by a laser spectrometer to analyze the content of key elements such as C, Mn, Cr, and Ti in real time. If the composition deviation exceeds the process requirements, the unqualified billets are automatically removed. For example, when it is detected that the C content of a certain billet is 0.35% and the Mn content is 0.75%, which meets the standard of the SWRCH35K steel grade, it is determined as a qualified billet.

[0070] S12. Adopt a two-round grinding wheel grinding method. In the first round, use a 16-20 mesh grinding wheel to remove the wrinkles at the corners of the billet, and in the second round, switch to a 24-26 mesh grinding wheel to perform fine polishing on the surface of the billet.

[0071] Specifically, adopt a two-round grinding wheel grinding method. In the first round, use a 16-20 mesh grinding wheel (coarser abrasive grain size), with a pressure of 7000-8000N and a feed speed of 0.8-1.2m / s, to mainly remove the wrinkles at the corners of the billet and surface cracks (defects with a depth of ≥0.5mm need to be completely eliminated).

[0072] For example, for a folding defect with a depth of 0.8mm at the corner, after the first-round grinding, the defect depth is reduced to ≤0.2mm. Subsequently, switch to a 24-26 mesh grinding wheel (fine abrasive grain size) and perform secondary polishing with a pressure of 4000-5000N and a feed speed of 1.0-1.5m / s to make the surface roughness Ra≤6.3μm, while avoiding material loss caused by excessive grinding. For example, after grinding a certain billet, there are no visible cracks on the surface, and the roughness Ra = 5.8μm, meeting the surface quality requirements before rolling.

[0073] S13. Place the ground billet in a preheating environment of 600-800°C. After gradually raising the preheating environment temperature to 1090-1170°C, introduce a CO2 / N2 mixed gas to increase the surface hardness of the billet.

[0074] Specifically, the ground billet is first placed in a preheating zone at 600 - 800 °C and gradually soaked through radiation heating (the time is adjusted according to the billet size, for example, a 160 mm square billet requires 30 minutes) to avoid thermal stress cracks caused by direct high-temperature heating. Subsequently, the furnace temperature is raised to the austenitizing temperature range of 1090 - 1170 °C, and a CO2 / N2 mixed gas (with a CO2 proportion of 5% - 8%) is introduced. The active carbon atoms generated by the decomposition of CO2 at high temperatures react with the steel surface to form a carbon concentration gradient (the surface C content increases by 0.05% - 0.10%), enhancing the surface hardness (HRB + 3) and inhibiting the formation of decarburized layers.

[0075] For example, after a certain billet is treated in this way, the surface hardness is increased from HRB 82 to HRB 85, and the decarburized layer thickness is only 0.03 mm, meeting the performance requirements of non-annealed cold-heading steel.

[0076] S2. Transport the billet to a high-reduction rolling mill unit, and monitor the hot rolling process in real time to conduct segmented hot rolling control, achieving multi-pass temperature gradient rolling to obtain a cold-heading steel wire rod in hot-rolled form.

[0077] In the description of the present invention, transporting the billet to a high-reduction rolling mill unit, monitoring the hot rolling process in real time to conduct segmented hot rolling control, and achieving multi-pass temperature gradient rolling to obtain a cold-heading steel wire rod in hot-rolled form includes:

[0078] S21. According to the production specifications of the cold-heading steel wire rod and the component content of the billet, divide the rolling process into three stages: rough rolling, medium rolling, and finish rolling, and use a physical-constrained meta-learning algorithm to match the optimal rolling parameters.

[0079] Among them, physical-constrained meta-learning is a hybrid intelligent algorithm that combines metallurgical physical rules and data-driven meta-learning. Through the "learning to learn" mechanism, it extracts common laws from historical data of multiple steel grades, quickly adapts to new tasks (such as parameter matching for new steel grades), and during the model training and inference processes, it compulsorily adheres to basic metallurgical laws (such as phase transformation kinetics, thermal-mechanical coupling equations) to ensure that the recommended parameters conform to the principles of materials science. Its core lies in embedding key metallurgical equations in cold-heading steel production (such as phase transformation kinetics equations, hot deformation resistance models, oxidation film growth laws, etc.) as hard constraints into the meta-learning framework to ensure that the process parameters generated by the algorithm not only rely on data-driven but also strictly conform to materials science laws. For example, in the optimization of rolling parameters, the algorithm will compulsorily require that the finish rolling temperature must be lower than the Ar3 phase transformation point to avoid grain coarsening caused by austenite recrystallization. At the same time, the meta-learning mechanism enables the model to quickly generalize based on a small amount of data of new steel grades, reducing the cost of traditional trial-and-error experiments.

[0080] The input variables of the physical constraint meta-learning algorithm include steel grade composition (contents of C, Mn, etc.), production specifications (diameter, surface requirements), and blank parameters (size, initial hardness); the output variables include the temperatures at each rolling stage (rough / medium / fine rolling), reduction ratio, and reduction rate.

[0081] In the production method of the present invention, the physical constraint meta-learning algorithm is embodied in the following aspects:

[0082] 1. Dynamic matching of rolling stage parameters During the multi-pass temperature gradient rolling process, the physical constraint meta-learning algorithm calculates the temperature windows and reduction rates at the rough rolling, medium rolling, and fine rolling stages in real time according to the billet composition and the target specifications of the wire rod. For example, for high-carbon steel, the model automatically reduces the initial rolling temperature to 980 - 1030 °C and increases the reduction rate in fine rolling to ≥60% to trigger deformation-induced ferrite transformation (DIFT) and directly generate ultrafine-grained structures. By embedding the rolling force-temperature coupling equation, the physical constraint meta-learning algorithm can dynamically compensate for the load fluctuations of the rolling mill and ensure the uniformity of the properties along the whole bar.

[0083] 2. Optimization of the sectional controlled cooling process During the cooling stage, the physical constraint meta-learning algorithm combines the post-rolling temperature field distribution and the phase transformation kinetics model to generate the cooling rate and time parameters for the three stages of rapid cooling, slow cooling, and final cooling. For example, if an abnormal temperature gradient is detected in a certain section of the wire rod, the model will automatically extend the slow cooling time and adjust the fan power to ensure that the ferrite proportion is ≥76% and the spheroidization rate of pearlite is ≥90%. At the same time, by constraining the final cooling temperature and the atmosphere control parameters through the oxide film thickness diffusion equation, surface decarburization is inhibited.

[0084] 3. Cross-process collaborative control The physical constraint meta-learning algorithm breaks through the data barriers between the rolling and cooling processes and realizes the linkage of parameters in the whole process. For example, when the insufficient ferrite proportion is detected in the cooling section, the feedback signal will trigger the rolling model to increase the reduction rate in fine rolling; conversely, if an abnormal temperature drop occurs during the rolling stage, the cooling model will adjust the fan opening and closing strategy in advance. This closed-loop optimization mechanism significantly reduces the impact of process fluctuations on the final properties.

[0085] In the description of the present invention, according to the production specifications of cold heading steel wire rods and the billet composition content, the rolling process is divided into three stages: rough rolling, medium rolling, and fine rolling, and the physical constraint meta-learning algorithm is used to match the optimal rolling parameters, including: S211. Based on the preset production specifications of cold heading steel wire rods, reverse-derive the billet cross-sectional size and calculate the target reduction ratio for hot-rolling the billet to cold heading steel wire rods.

[0086] Specifically, based on the volume invariance principle of cold heading steel, assuming that the metal volume is basically unchanged during hot rolling (ignoring oxidation and burning loss), that is: ; In the formula, A 0 and A f are the cross-sectional areas of the steel billet and the finished product respectively, L 0 and L f are the lengths before and after rolling.

[0087] The target compression ratio is the ratio of the cross-sectional areas before and after rolling, that is: ; In the formula, D 0 and D f are the diameters of the steel billet and the finished product.

[0088] S212. Divide the rolling process of the steel billet into three stages: rough rolling, medium rolling, and finish rolling. Establish a parameter mapping table using the historical database, and output the initial compression ratio and rolling temperature range for each stage.

[0089] Specifically, based on historical data and steel grade characteristics, divide the rough / medium / finish rolling stages and output the initial parameters. The following division criteria can be referred to:

[0090] Rough rolling stage: large compression ratio (38 - 48%), high temperature zone (1050 - 1100 °C), preferentially break the original austenite grains;

[0091] Medium rolling stage: medium compression ratio (30 - 40%), medium temperature zone (950 - 1000 °C), refine grains and control the precipitation phase;

[0092] Finish rolling stage: supercritical compression ratio (≥60%), low temperature zone (750 - 800 °C), trigger deformation-induced ferrite transformation (DIFT).

[0093] For establishing the parameter mapping table using the historical database, the present invention uses an automated data acquisition system to collect historical production data, including steel billet composition, initial dimensions, rolling temperature at each stage, compression ratio, rolling force, and indicators such as the grain size and mechanical properties of the final product. Clean and standardize the collected data, remove outliers and unify the unit format.

[0094] The present invention uses a clustering algorithm (such as K-means) to automatically group the data according to steel grade composition and billet size, forming different categories of data sets; on this basis, analyze the correlation between the rolling parameters at each stage (such as rough rolling temperature, compression ratio) and the target results (such as grain size, cold heading qualification rate) within each data group through multiple linear regression or decision tree models, and establish a mathematical model, that is, the parameter mapping table.

[0095] When a new steel billet is put into production, the system automatically matches its composition and size to the closest data category, calls the corresponding model to calculate the initial reduction ratios (e.g., rough rolling 5:1, intermediate rolling 3.5:1, finish rolling 2:1) and temperature ranges (e.g., rough rolling 1080 - 1120 °C, intermediate rolling 980 - 1020 °C, finish rolling 880 - 920 °C) for the three stages of rough rolling, intermediate rolling, and finish rolling; at the same time, a real-time feedback mechanism is set up to transmit the rolling effect data in actual production back to the database, and the model is retrained with new data regularly to continuously optimize the accuracy of the parameter mapping table. The entire process is completely data-driven and does not require manual experience intervention.

[0096] It should be noted that in step S212, by converting historical production data into a structured knowledge base, initial recommended values based on statistical laws are provided for rolling parameters. Before the physical constraint meta-learning model intervenes, benchmark parameters that conform to the commonalities of steel grades are quickly generated, avoiding calculation from scratch and shortening the decision-making time. Using the historical optimal parameters as the initial values can effectively reduce the parameter oscillation caused by random initialization of the model. In the subsequent process, the physical constraint meta-learning algorithm can dynamically correct various rolling parameters on the basis of the initial parameters by combining physical constraints and real-time data feedback.

[0097] S213. Establish physical constraints for the hot rolling of steel billets based on metallurgical theory. The physical constraints include phase transformation kinetics constraints, deformation resistance and rolling force constraints, and precipitation phase pinning constraints.

[0098] Specifically, the phase transformation kinetics constraint formula is:

[0099] ;

[0100] In the formula, X(t) represents the volume fraction of ferrite; k(T) represents the temperature-dependent rate constant; t represents time; n represents the Avrami exponent (dimensionless, typical value for cold heading steel is 1.5 - 2.0).

[0101] The deformation resistance and rolling force constraint formula is:

[0102] ;

[0103] In the formula, σ represents the deformation resistance; σ0 represents the reference stress; Q represents the deformation activation energy; R represents the gas constant; T represents the absolute temperature; ε represents the strain rate; ε0 represents the reference strain rate; m represents the strain rate sensitivity coefficient.

[0104] The precipitation phase pinning constraint formula is:

[0105] ;

[0106] In the formula, f TiC represents the volume fraction of TiC precipitation.

[0107] The phase transformation kinetics constraint is based on the thermodynamic and kinetic laws of austenite-to-ferrite transformation, controls the temperature and cooling rate during the rolling process, and ensures that austenite decomposes into ferrite and pearlite within a specific temperature range. Its core lies in the synergistic effect of the temperature window and time parameters to regulate the phase transformation rate and the final tissue proportion, and avoid grain coarsening caused by austenite recrystallization. The phase transformation kinetics constraint forces the finish rolling temperature to be lower than the Ar3 phase transformation point, triggers deformation-induced ferrite transformation (DIFT), directly generates ultra-fine ferrite grains, and replaces the traditional spheroidizing annealing process. By controlling the ferrite volume fraction and the pearlite spheroidization rate, it ensures that the cold heading steel has uniform plasticity and strength, meeting the forming requirements of high-strength fasteners under the condition of annealing-free.

[0108] It should be noted that phase transformation control is the basis of the hot rolling process, especially widely used in the controlled rolling and cooling technology to guide the setting of the finish rolling temperature and cooling rate. In the meta-learning framework, the phase transformation kinetics constraint is transformed into a part of the loss function. For example, if the predicted ferrite proportion deviates from the target value, the recommended rolling temperature value is adjusted through backpropagation to ensure that the tissue performance meets the standards.

[0109] The deformation resistance and rolling force constraint is based on the rheological stress law of materials during high-temperature deformation, combines the billet composition and temperature field distribution, and dynamically calculates the mill load and energy consumption during the rolling process. Its core lies in balancing the rolling efficiency and the equipment load-bearing capacity, and preventing the risk of roll wear or strip breakage caused by too high deformation resistance. The deformation resistance and rolling force constraint is used to limit the rolling force range in the rough rolling and finish rolling stages to avoid equipment overload shutdown. By adjusting the reduction ratio and rolling speed, the energy consumption of ineffective deformation is reduced, and the rolling efficiency is improved.

[0110] It should be noted that the deformation resistance and rolling force control is a standard technology in the industry. The rolling force model and the automatic gauge control (AGC) system have been integrated into modern rolling mills to adjust the process parameters in real time. The meta-learning framework takes the upper limit of the rolling force as a hard boundary condition. If the predicted rolling force exceeds the equipment limit, the model automatically reduces the recommended reduction ratio or adjusts the rolling temperature to ensure safe production.

[0111] The precipitation pinning constraint utilizes the solution and precipitation behavior of microalloying elements (such as Ti, Nb) in austenite, and pins the grain boundaries through the dispersed distribution of carbides and nitrides (TiC, NbC), hindering grain growth. Its core is to control the size, quantity, and distribution of the precipitates through composition design and process parameter matching. The precipitation pinning constraint is used to pin the austenite grain boundaries during rolling, limit grain growth, and refine the final ferrite grains to ≤5μm. By refining the grains and homogenizing the tissue, the cold heading cracking rate is significantly reduced, and the material formability is improved.

[0112] It should be noted that the precipitation pinning principle is mainly used in the production of high-strength steel and micro-alloyed steel. In the meta-learning framework, the precipitation constraint is encoded as a temperature range limit. For example, the model will avoid the temperature range where the precipitates redissolve and preferentially select a process window that can retain the precipitates without affecting plasticity when recommending rolling parameters.

[0113] S214. Set the input variables and output variables, construct and train a parameter recommendation model based on meta-learning, integrate physical metallurgy constraints, and match the rolling parameters corresponding to different steel billets.

[0114] In the description of the present invention, setting the input variables and output variables, constructing and training a parameter recommendation model based on meta-learning, integrating physical metallurgy constraints, and matching the rolling parameters corresponding to different steel billets includes: S2141. Assign each original task to a steel grade, set the production specifications of cold heading steel wire rods and the content of each component in the steel billet as input variables, and the rolling parameters as output variables to construct a support set.

[0115] Specifically, in the meta-learning framework, each meta-task corresponds to a steel grade, and its support set is composed of the historical production data of this steel grade.

[0116] The input variables (input layer) include the production specifications of cold heading steel wire rods and the content of steel billet components (mass percentages of elements such as C, Mn, Cr, Ti, etc.), and the output variables (output layer) are the rolling parameters (temperatures, reduction ratios, rolling forces in the rough / middle / fine rolling stages) corresponding to the steel grade.

[0117] For example, for the SWRCH35K steel grade, the input is C = 0.35%, Mn = 0.75%, target diameter Φ16mm, and the output is rough rolling temperature 1030°C, reduction ratio 45%, fine rolling temperature 780°C, and reduction ratio 65%. The support set needs to cover the typical working conditions of this steel grade (such as component fluctuations and specification changes in different heats), and usually each task contains 50 - 100 groups of data to enable the model to quickly adapt to new variant steel grades.

[0118] S2142. Link a physical calculation module after the output of the hidden layer to achieve physical information embedding for calculating the temperature and reduction ratio during the rolling of the steel billet at each stage.

[0119] Specifically, after the output of the hidden layer of the neural network, link a physical calculation module, combine the data-driven prediction results with metallurgical equations to achieve physical information constraints. For example, the preliminarily predicted finishing temperature output by the hidden layer is 790°C, but according to the Ar3 phase transformation point formula, the Ar3 of this component steel grade is 750°C, and the physical calculation module corrects the finishing temperature to ≤750°C.

[0120] Among them, the hidden layer is the core component of the neural network, located between the input layer and the output layer, and is responsible for performing non-linear transformation and feature extraction on the input data. Through multi-level weighted calculations and activation function mappings, the hidden layer transforms the original input into high-dimensional abstract features, providing an intermediate representation for the final prediction.

[0121] In the cold heading steel rolling parameter recommendation model, the specific implementation of the hidden layer includes the following steps: 1. Input reception: Receive the standardized billet parameters.

[0122] 2. Weighted calculation: Perform a linear transformation through the weight matrix and bias vector. The formula is simplified as: Hidden layer output = activation function (weight × input + bias); 3. Non-linear activation: Introduce non-linearity using the ReLU function or the Sigmoid function. For example: ReLU(x) = max(0, x); Through non-linear activation, the hidden layer can learn complex process rules.

[0123] 4. Feature transfer: Transfer the processed features to the output layer to finally generate a preliminary prediction value.

[0124] It should be noted that the physical calculation module includes a phase transformation kinetics module, a rolling force calculation module, and a precipitation phase control module. The specific content is as follows.

[0125] 1. Phase transformation kinetics module: Input the predicted temperature and time, and calculate the ferrite volume fraction through the Avrami equation. If the predicted value is lower than the target, the rolling temperature and reduction rate are adjusted backward.

[0126] This module predicts the phase transformation ratio of austenite to ferrite based on the steel grade composition and the rolling temperature-time history. The purpose is to ensure that the ferrite ratio in the rolled material reaches the target value by controlling the finish rolling temperature and cooling rate, so as to meet the plasticity and uniformity required for cold heading processing. The implementation methods include: 1.1. Input parameters: The predicted temperature output by the neural network hidden layer and the rolling time at each stage; 1.2. Dynamic correction: If the module calculates that the predicted temperature is higher than the Ar3 phase transformation point (the critical temperature for the transformation of austenite to ferrite) of this steel grade, the finish rolling temperature is forced to be lowered below the Ar3 point; 1.3. Reverse regulation: If the ferrite ratio is lower than the target value, the reduction rate in the finish rolling stage is automatically increased to promote deformation-induced phase transformation.

[0127] 2. Rolling Force Calculation Module: Based on the Sellars deformation resistance model (the Sellars deformation resistance model), substitute the predicted reduction rate and temperature, calculate whether the rolling force exceeds the limit. If it exceeds the limit, reduce the reduction rate by 3% - 5%.

[0128] This module calculates the rolling force for a single pass based on the Sellars deformation resistance model (Sellars Model), combined with roll dimensions and the high-temperature strength characteristics of the material. The purpose is to prevent the rolling mill from being overloaded or the strip from breaking due to excessive reduction rate, and at the same time balance production efficiency and equipment safety. The implementation methods include: 2.1 Input Parameters: The predicted reduction rate output by the hidden layer, rolling temperature; 2.2 Dynamic Verification: The module calculates the actual rolling force. If it exceeds the equipment safety threshold, immediately trigger the reduction rate compensation; 2.3 Temperature Coordination: If the reduction rate can no longer be reduced, increase the rolling temperature to reduce the material deformation resistance.

[0129] 3. Precipitate Phase Control Module: For steel grades containing Ti, according to the TiC precipitate phase volume fraction formula, verify whether the reduction rate in medium rolling meets the requirements of grain refinement. If not, trigger the reduction rate increment compensation.

[0130] This module is for micro-alloyed steel grades containing titanium (Ti), niobium (Nb), etc. This module calculates the pinning strength of the precipitate phase on the austenite grain boundary through the Zener pinning effect. The purpose is to ensure that the combination of rolling temperature and reduction rate can promote the dispersed distribution of the precipitate phase and avoid its re-dissolution or excessive coarsening. The implementation methods include: 3.1 Input Parameters: Steel grade composition, medium rolling temperature; 3.2 Dynamic Regulation: If the module detects that the current temperature is higher than the TiC re-dissolution temperature, reduce the medium rolling temperature to 980 °C to retain the precipitate phase; 3.3 Reduction Rate Compensation: If the number of precipitate phases is insufficient, increase the reduction rate in the medium rolling stage to supplement the precipitate phase through the strain-induced precipitation mechanism.

[0131] S2143. Set two physical loss term functions for phase transformation loss and rolling force loss, and combine the phase transformation loss and rolling force loss to set a comprehensive loss function, and construct a parameter recommendation model.

[0132] Specifically, when training the parameter recommendation model, in addition to the traditional data loss (the mean square error between the predicted parameters and the measured values), physical loss terms need to be added to force the model output to conform to metallurgical laws.

[0133] Among them, the phase transformation loss function is used to ensure the core constraints of material properties. During hot rolling, the austenite of cold heading steel needs to be transformed into ferrite / pearlite structure through phase transformation, and its proportion directly affects the cold heading properties of the material (such as plasticity and crack resistance). The phase transformation loss function quantifies the deviation between the phase transformation result predicted by the model (such as the ferrite proportion) and the target value, forcing the parameters such as rolling temperature and reduction rate recommended by the model to satisfy the phase transformation kinetics law. For example, if the model aims to pursue production efficiency and initially recommends a finishing rolling temperature of 800 °C, but calculates its Ar3 phase transformation point (the critical temperature for austenite to ferrite transformation) to be 750 °C according to the steel grade composition, the phase transformation loss function will determine that the finishing rolling temperature is too high (austenite is not fully transformed) at this time, and force the model to correct the temperature below 750 °C through backpropagation to ensure that the ferrite proportion ≥ 76%.

[0134] The rolling force loss function calculates the theoretical rolling force corresponding to the parameters such as reduction rate and rolling speed recommended by the model, and compares it with the rated load-bearing capacity of the rolling mill to impose penalties on the over-limit parameters. Its core is to prevent accidents such as strip breakage and roll fracture caused by equipment overload. For example, if the parameter recommendation model refines the grain size and recommends a reduction rate of 20% in the medium rolling stage, but calculates according to the material high-temperature deformation resistance model that the rolling force corresponding to this reduction rate is 2800 MPa, exceeding the equipment safety threshold (2600 MPa). At this time, the rolling force loss function will increase significantly, forcing the model to reduce the reduction rate below 18% or increase the rolling temperature to reduce the material resistance.

[0135] The comprehensive loss function combines the phase transformation loss and the rolling force loss according to the distribution weight to form a single optimization objective. During model training, through the gradient descent algorithm, it automatically finds the parameter combination that minimizes the comprehensive loss to achieve the balance between performance and safety. For example, when producing microalloyed steel containing titanium, the model may face a dilemma: on the one hand, increasing the temperature can promote the dissolution of TiC precipitation phase and reduce the rolling force; on the other hand, reducing the temperature can retain TiC particles, refine the grain but increase the rolling force. In this case, the comprehensive loss function will select a compromise solution of a temperature of 960 °C and a reduction rate of 17% according to the weight distribution - retaining some TiC to refine the grain and controlling the rolling force within the safe limit at the same time.

[0136] Specifically, the construction of the comprehensive loss function essentially integrates material property requirements and production safety constraints into a single optimization goal through mathematical means, thereby generating process parameters that not only conform to metallurgical laws but also meet the equipment's load-bearing capacity. The weight allocation between phase transformation loss and rolling force loss needs to be dynamically adjusted in combination with actual production. If the current production line pays more attention to the cold heading qualification rate, the phase transformation loss will occupy a higher weight in the comprehensive loss, enabling the model to prioritize ensuring that the ferrite ratio meets the standard. If the equipment is in a high-load state, the weight ratio of the rolling force loss will be increased to ensure that the parameter recommendation does not trigger the equipment protection mechanism. During the actual training process, the parameter space is continuously explored through gradient descent to automatically find the combination of process parameters that minimizes the comprehensive loss. By coupling physical rules and data patterns, the best balance point is found in the contradiction between phase transformation control and rolling force safety, and finally a parameter scheme that can meet the ferrite ratio requirements and control the rolling force within the safety threshold is generated.

[0137] S2144. Pretrain the model parameters for all steel grade tasks to initially form a hot rolling knowledge base. When facing a new steel grade, perform fine-tuning of the calculation task through single-step gradient descent.

[0138] Specifically, S2144 includes the following aspects: 1. Global pre-training: Pretrain the model on the historical database to learn the common laws across steel grades. For example, high-carbon steel requires a lower final rolling temperature, and Ti-containing steel requires an increased medium rolling reduction ratio. After pre-training, a hot rolling knowledge base is formed, and the model can be generalized to new steel grades with similar compositions.

[0139] 2. Single-step gradient fine-tuning: When facing a new steel grade, only 50 sets of support set data need to be provided to update the model parameters through single-step gradient descent. For example, 10B21 contains boron (B = 0.002%). The model quickly adjusts according to the pre-trained knowledge and generates parameters such as a finishing rolling temperature of 760°C and a reduction ratio of 68% to meet the requirements for improving hardenability.

[0140] S215. Input the size and composition content of the steel billet to be processed into the parameter recommendation model, generate rolling parameters for each stage and the real-time adjustment logic, and obtain the optimal rolling parameters.

[0141] In the description of the present invention, inputting the size and composition content of the steel billet to be processed into the parameter recommendation model, generating rolling parameters for each stage and the real-time adjustment logic, and obtaining the optimal rolling parameters includes:

[0142] S2151. Obtain the carbon content of the steel billet to be processed, match the corresponding rough rolling temperature in the rough rolling stage according to the carbon content value, and calculate the rough rolling reduction ratio.

[0143] The calculation formula for the rough rolling temperature is:

[0144] ;

[0145] Where, T 粗轧 represents the rough rolling temperature in the rough rolling stage; T0 represents the preset initial temperature; [C] represents the carbon content;

[0146] The calculation formula for the rough rolling reduction rate is:

[0147] ; In the formula, r 粗轧 Indicates the rough rolling reduction rate in the rough rolling stage; r 1 represents the initial compression ratio of the rough rolling stage output by the parameter mapping table; R total Indicates the target compression ratio.

[0148] It should be noted that the preset initial temperature T 0 and initial compression ratio r 1 are all obtained from the output of the parameter mapping table established based on the historical database, that is, when the new steel billet is put into production, its composition and size are automatically matched to the closest data category, and the initial compression ratio and initial temperature of the three stages of rough rolling, intermediate rolling and finishing rolling are automatically calculated.

[0149] Specifically, the preset initial temperature T 0 is determined based on the comprehensive setting of steel grade characteristics, equipment capabilities and process objectives: for low carbon steel, T 0 is usually set to 1150-1200℃ to promote dynamic recrystallization; the recrystallization temperature of medium and high carbon steels will decrease due to the increase in carbon content. T 0 needs to be lowered to 1100-1150℃ to avoid grain coarsening; at the same time, combined with the maximum load of the rolling mill, the efficiency of the heating furnace and the final performance requirements of the product, it is finally determined through historical data statistics or experimental verification.

[0150] The rough rolling temperature calculation formula dynamically adjusts the rolling temperature through the carbon content to achieve a precise match between process parameters and material properties. The formula adjusts the heating temperature in the rough rolling stage inversely according to the difference in carbon content in the steel: the higher the carbon content, the lower the rough rolling temperature setting value. On the one hand, by inhibiting the excessive growth of austenite grains at high temperatures and refining the original grain structure of the steel, the uniformity of the organization in the subsequent deformation process is improved; on the other hand, lowering the rolling temperature of high carbon steel can effectively control the amount of surface oxide scale generated and reduce material loss caused by oxidation. At the same time, the negative correlation design between temperature and carbon content also balances the relationship between rolling force and equipment bearing capacity, avoiding insufficient recrystallization of low carbon steel due to insufficient temperature and preventing high carbon steel from exceeding the rolling force limit at high temperatures. The practical significance of the rough rolling temperature calculation formula is to convert the composition difference into controllable process parameters, realize the flexible production of different steel grades on the same production line, and significantly improve the yield rate and product quality stability.

[0151] The calculation formula for rough rolling reduction rate realizes dynamic adaptation of the target compression ratio: If the target compression ratio for a certain batch R total = 85%, r 1 taken from the parameter mapping table is 30% (the best value for similar steel grades in historical data), then the calculated reduction rate is 30% + 0.1×(85 - 80) = 30.5%. This design makes the actual compression ratio approach the target value (error ≤ ±1.5%), avoiding overloading of the rolling mill caused by too high reduction rate in a single pass (e.g., reduction rate exceeding 35% may trigger an alarm) or too low reduction rate resulting in an increase in the number of rolling passes (reducing 1 - 2 passes). At the same time, combined with temperature control, it can refine the initial austenite grains and reduce the risk of cold heading cracking.

[0152] S2152. Obtain the titanium content, manganese content of the steel billet to be processed, and the grain size after rough rolling, and calculate the cooling rate and reduction rate in the intermediate rolling stage when the steel billet enters the intermediate rolling stage.

[0153] The calculation formula for the cooling rate in the intermediate rolling stage is:

[0154] ; In the formula, v T represents the cooling rate in the intermediate rolling stage; Δ T represents the temperature difference between rough rolling and intermediate rolling; Δ t represents the time difference; Mn represents the manganese content.

[0155] The calculation formula for the reduction rate in the intermediate rolling stage is:

[0156] ; In the formula, r 中轧 represents the reduction rate in the intermediate rolling stage; r 2 represents the initial compression ratio in the intermediate rolling stage output by the parameter mapping table; Ti represents the titanium content.

[0157] It should be noted that the cooling rate in the intermediate rolling stage is dynamically adjusted through the manganese content: for every 0.1% increase in the manganese content, the cooling rate increases by 0.05 °C / s, ensuring that the intermediate rolling temperature is stable in the range of 950 - 1000 °C, avoiding coarsening of austenite grains caused by insufficient temperature drop or sudden change in rolling force caused by too fast temperature drop. This control can reduce the proportion of banded structure to ≤8% and improve the uniformity of cold heading steel.

[0158] The reduction rate in the intermediate rolling stage is calculated based on the titanium content to correct the compression ratio: for every 0.01% increase in the titanium content, the reduction rate increases by 1%. Through the precipitation strengthening effect of titanium, it compensates for the impact of the increase in compression on the rolling mill load, and at the same time refines the ferrite grains, increasing the tensile strength after intermediate rolling by 20 - 30 MPa and reducing the surface crack incidence rate.

[0159] S2153. Calculate the Ar3 transformation point of the steel billet to be processed, set the finish rolling temperature in the finish rolling stage according to the Ar3 transformation point, and calculate the finish rolling reduction ratio in the finish rolling stage according to the carbon content.

[0160] The calculation formula for the finish rolling reduction ratio is:

[0161] ;

[0162] In the formula, r 精轧 represents the finish rolling reduction ratio; r3 represents the initial reduction ratio in the finish rolling stage output by the parameter mapping table.

[0163] It should be noted that the finish rolling reduction ratio is calculated to reversely adjust the reduction amount according to the carbon content: when the carbon content drops from 0.35% to 0.25%, the reduction ratio increases by 2.9%. Combining the finish rolling temperature set according to the Ar3 transformation point can not only inhibit the brittleness caused by too high carbon content, but also optimize the cold heading formability through deformation-induced ferrite transformation, effectively improving the dimensional accuracy control effect.

[0164] S22. Real-time monitor the hot rolling data during the hot rolling process, dynamically compensate the optimal rolling parameters, and dynamically adjust the pass interval time according to the monitoring and compensation results.

[0165] In the description of the present invention, real-time monitoring of the hot rolling data during the hot rolling process, dynamically compensating the optimal rolling parameters, and dynamically adjusting the pass interval time according to the monitoring and compensation results include:

[0166] S221. Use the sensors inside the rolling mill unit to collect the hot rolling data of the steel billet rolling process in real time.

[0167] Specifically, deploy high-precision sensors at the key positions of the roll bearing seats, guide devices and roller tables of the rolling mill unit, and the following data need to be collected in real time: 1. Temperature data: Monitor the surface temperature field of the steel billet through an infrared thermal imager, and collect it once every 0.5 seconds to ensure that the temperature gradient is controlled within ±20°C. For example, for a certain Φ16mm wire rod, the head temperature is 1050°C and the tail temperature is 1030°C in the rough rolling stage, with a temperature difference of 20°C, which is judged as normal fluctuation.

[0168] 2. Rolling force data: The rolling force sensor (range 0 - 3000MPa, error ≤ ±1%) records the load of each rolling mill pass in real time, and the data is transmitted to the control center through industrial Ethernet. For example, the measured rolling force in the finish rolling stage is 2450MPa, with a deviation of 2% from the target value of 2500MPa, triggering dynamic compensation.

[0169] 3. Dimension data: The laser diameter gauge detects the diameter and ovality of the rolled piece and gives an immediate feedback when out of tolerance.

[0170] S222. Calculate the difference between the actual rolling force and the target rolling force during the rolling process of each stage of the rolling mill unit, and calculate the reduction rate compensation amount through the proportional-integral coefficient.

[0171] Specifically, the formula for calculating the reduction rate compensation amount by the proportional-integral coefficient is:

[0172] ;

[0173] In the formula, Δr represents the compensation amount; k p , k i respectively represent the proportional and integral coefficients; F 目标 represents the target pressing force; F 实际 represents the actual pressing force.

[0174] Among them, in the production process of cold heading steel hot-rolled wire rods, the determination of the proportional-integral (PI) coefficient is closely related to the physical characteristics of the rolling mill unit and the material deformation behavior. Its essence is to transform the dynamic response characteristics of the rolling mill system into executable control parameters through mathematical means. The proportional coefficient k p and the integral coefficient k i are calibrated based on the mechanical properties of the rolling mill itself, the inertial characteristics of the drive system, and the high-temperature rheological laws of the material.

[0175] The setting of the integral coefficient k i is related to the long-period fluctuation characteristics of the rolling force. For example, the progressive drift of the rolling force caused by roll wear requires a slow correction of the cumulative historical error through a higher k i value. The proportional-integral coefficient is usually obtained by combining off-line experiments and on-line self-tuning: during the equipment commissioning stage, the engineer applies a step reduction command to the rolling mill, and calculates the initial k p and k i based on the measured rolling force response curve; while in actual production, the coefficient is dynamically fine-tuned according to real-time data (such as the rolling force fluctuation frequency, error convergence speed).

[0176] The proportional coefficient k p determines the immediate reaction intensity of the system to the current deviation. When there is a deviation between the measured rolling force and the target value, k p directly adjusts the reduction rate compensation amount according to the deviation ratio. For example, if the rolling force deviation is large, a higher k p will quickly increase the compensation amount and shorten the adjustment time. The integral coefficient ki To eliminate the cumulative effect of historical deviations. When the deviation persists for a long time (e.g., the rolling force remains low continuously due to roll wear), the integral term continuously corrects the compensation amount by accumulating the historical deviation values until the steady-state error is eliminated. For example, if the rolling force is 5% lower than the target for 3 consecutive seconds, the integral term will gradually increase the compensation amount until the deviation returns to zero.

[0177] S223. Calculate the temperature difference between the head and tail of the billet after each stage. If the temperature difference between the head and tail is greater than the preset threshold, trigger the adjustment to extend the pass interval to balance the temperature drop and the rolling rhythm.

[0178] Specifically, after each pass, calculate the temperature difference between the head and tail of the billet. For every 1°C exceeded, extend the interval by 0.5 seconds. For example, if the measured temperature difference is 25°C (exceeding the limit by 5°C), the interval will be extended by 2.5 seconds; for every 1°C exceeded, reduce the reduction ratio by 0.2%. For example, when the temperature difference is 25°C, the reduction ratio is reduced by 1%.

[0179] S224. Import the real-time hot rolling data into the historical parameter library, set the update period, and update the comprehensive loss function regularly. If there is a sudden change in the hot rolling data, trigger an emergency stop signal.

[0180] Specifically, the real-time data (temperature, rolling force, dimensions) are stored classified by steel type, and the model parameter update is triggered every 30 minutes.

[0181] The sudden change detection and emergency response include the following aspects: 1. Sudden change in rolling force: If the change in rolling force between adjacent passes > 20%, it is determined as the risk of strip breakage, and an emergency stop signal is triggered;

[0182] 2. Sudden temperature drop: If the temperature drop rate in a certain area > 10°C / s (normal range 1 - 3°C / s), start the heater to compensate for the temperature drop;

[0183] 3. Model rollback: If the tolerance is still exceeded after 3 consecutive compensations, automatically switch to the historical optimal parameter template.

[0184] S23. After rolling, spray the nano-Al2O3-MgO composite coating on the surface of the cold heading steel wire rod, generate a dense oxide film using the waste heat, and verify the actual production specifications of the cold heading steel wire rod.

[0185] Specifically, after the cold heading steel wire rod is rolled, mix the nano-Al2O3-MgO composite powder with ethanol suspension and spray it onto the surface of the wire rod through a high-pressure spraying device. Using the waste heat of the wire rod (without additional heating), the powder particles melt and diffuse at high temperature to form a dense and uniform composite coating. During this process, Al2O3 provides high hardness and wear resistance, MgO enhances the bonding force between the coating and the steel matrix, and at the same time induces the surface Fe element to oxidize to generate the Fe3O4 oxide film, inhibiting the generation of FeO red rust.

[0186] S3. Convey the cold-heading steel wire rod after hot rolling to a heat preservation tunnel furnace, and through segmented controlled cooling and surface treatment, obtain the finished cold-heading steel wire rod, realizing the production of hot-rolled wire rods of cold-heading steel without annealing.

[0187] In the description of the present invention, conveying the cold-heading steel wire rod after hot rolling to a heat preservation tunnel furnace, and through segmented controlled cooling and surface treatment, obtaining the finished cold-heading steel wire rod, realizing the production of hot-rolled wire rods of cold-heading steel without annealing includes: S31. Obtain and record the wire rod parameters of the cold-heading steel wire rod after hot rolling. The wire rod parameters include the final rolling temperature, the grain size after rolling, the pre-judgment ratio of ferrite, and the rolling force curve.

[0188] S32. Divide the cooling process of the cold-heading steel wire rod into three stages: rapid cooling, slow cooling, and final cooling, and build a physical constraint meta-learning framework to generate cooling parameters to control the cooling of the cold-heading steel wire rod in segments.

[0189] In the description of the present invention, dividing the cooling process of the cold-heading steel wire rod into three stages: rapid cooling, slow cooling, and final cooling, and building a physical constraint meta-learning framework to generate cooling parameters to control the cooling of the cold-heading steel wire rod in segments includes:

[0190] S321. Introduce wire rod parameters to expand the input layer, embed the phase transformation kinetics equation and the heat conduction equation to build a physical constraint layer, and build an output layer that outputs the segmented cooling rate value and the final cooling temperature.

[0191] Specifically, in the input layer of the neural network, in addition to the production specifications and billet composition of the cold-heading steel wire rod, introduce the real-time temperature field distribution and phase transformation progress (ferrite proportion) of the wire rod as extended input variables.

[0192] The physical constraint layer embeds the phase transformation kinetics equation (Avrami equation) and the heat conduction equation. For example, calculate the ferrite volume fraction through the Avrami equation and compare it with the target value (≥76%). If the predicted value deviates, reverse correct the network weights. The output layer is divided into three channels: the cooling rate in the rapid cooling stage (3 - 5 °C / s), the cooling time in the slow cooling stage (300 - 500 seconds), and the final cooling temperature (500 - 600 °C).

[0193] It should be noted that the cooling process of cold heading steel wire rods has the same core logic as the physical constraint meta-learning framework built in the hot rolling stage. Both realize intelligent optimization of process parameters by integrating metallurgical principles and data-driven models. In the hot rolling stage, the physical constraint meta-learning framework dynamically adjusts the rolling temperature and reduction rate by embedding rules such as rolling force models and phase change kinetic equations; in the cooling stage, the framework is extended to the heat conduction equation to control the cooling rate and final cooling temperature. The underlying data closed-loop mechanism of the two (real-time monitoring → dynamic compensation → model update) is exactly the same, and both are based on "physical rule constraints + meta-learning generalization" to ensure that parameter recommendations are both in line with scientific laws and adaptable to multiple working conditions.

[0194] In the physical constraint meta-learning framework of the cooling stage of cold heading steel wire rod, the temperature field evolution law and microstructure evolution mechanism in the cooling process are transformed into mathematical constraints through the dual embedding of heat conduction equation and phase change kinetic equation, driving the neural network model to generate cooling parameters that conform to the laws of metallurgy. Specifically, the physical constraint meta-learning framework first takes the real-time temperature field distribution, steel grade composition, and target microstructure performance of the wire rod as input variables, extracts features through the hidden layer of the neural network, and preliminarily predicts the cooling parameters of the three stages of rapid cooling, slow cooling, and final cooling; then, in the physical constraint layer, the Fourier heat conduction equation is used to calculate the theoretical relationship between temperature gradient and cooling rate, and the Avrami equation is used to infer the progress of ferrite generation under the current cooling rate. If the predicted parameters cause the theoretical ferrite ratio to be lower than the target value or the temperature field distribution violates the law of heat conduction (such as excessive surface cooling causing excessive core thermal stress), the weight parameters of the neural network are forced to be corrected through the back-propagation mechanism so that the output cooling rate, cooling time, and final cooling temperature meet the physical rules. For example, when a cooling rate of 5°C / s was initially recommended for the rapid cooling stage, the physical constraint layer detected that the core heat could not be discharged in time at this rate (the temperature drop rate exceeded the allowable range of the Fourier equation), and immediately triggered parameter correction, reducing the rapid cooling rate to 4.2°C / s and extending the slow cooling time to 380 seconds, thereby balancing tissue performance and thermal stress risks.

[0195] The physical constraint meta-learning framework of the cooling stage adopts a channel output structure - the cooling rate in the rapid cooling stage is achieved by the coordinated control of the fan speed and the water mist flow rate, the slow cooling time is dynamically adjusted by the roller speed, and the final cooling temperature is accurately maintained by the closed-loop temperature control system. The three are linked through real-time verification of the heat conduction equation.

[0196] S322. Divide the cooling process of the cold heading steel wire rod into rapid cooling, slow cooling and final cooling stages, and calculate the cooling rate and cooling time of the rapid cooling stage respectively.

[0197] In the description of the present invention, calculating the cooling rate and cooling time of the rapid cooling stage includes:

[0198] Calculation formula for the rapid cooling rate in the rapid cooling stage:

[0199] ; In the formula, v 速冷 represents the rapid cooling rate in the rapid cooling stage; T 终轧 represents the finishing rolling temperature; T 速冷终 represents the target temperature in the rapid cooling stage; t 速冷 represents the rapid cooling time;

[0200] Calculation formula for the cooling time in the slow cooling stage:

[0201] ;

[0202] In the formula, t 缓冷 represents the slow cooling time; k0 represents the phase transformation rate constant; represents the phase transformation activation energy; R represents the gas constant; T 缓冷 represents the slow cooling temperature; X total represents the target ferrite ratio; n represents the Avrami exponent; e represents the natural constant.

[0203] S323. Based on the preset segmented air cooling strategy, in the rapid cooling stage, a fan - water mist combined cooling is adopted. In the slow cooling stage, the fan is turned off and natural cooling by the roller table is utilized. In the final cooling stage, natural cooling is adopted and a nitrogen - carbon dioxide mixed gas is introduced to generate a dense oxide film.

[0204] Specifically, in the rapid cooling stage, a combined cooling of a fan (power 80% - 100%) and water mist (flow rate 15 - 20 L / min) is adopted. The fan speed is adjusted according to the real - time temperature drop rate. For example, when it is detected that the temperature drop rate of a certain section < 3 °C / s, the fan speed is increased by 5%. In the slow cooling stage, the fan is turned off, the roller table speed is reduced to 0.1 - 0.2 m / min, and natural cooling (rate 0.5 - 1 °C / s) is utilized to promote the growth of ferrite and the spheroidization of pearlite. In the final cooling stage, a nitrogen - carbon dioxide mixed gas (CO2 proportion 5% - 8%) is introduced, and a dense Fe3O4 oxide film is generated through gas reaction. At the same time, the roller table speed is restored to 0.5 m / min to ensure the uniformity of the oxide film.

[0205] S33. Use high - pressure water to remove phosphorus from the cooled cold - heading steel wire rod, and identify surface defects through on - line laser detection. After the detection is correct, the finished cold - heading steel wire rod is obtained.

[0206] Specifically, the surface of the wire rod after cooling is impacted by high-pressure water of 15-25 MPa, and the scale is peeled off through the water hammer effect. The water pressure is achieved by a multi-stage booster pump, and the nozzle is designed in a fan shape (covering angle 30-45°) to ensure that the water flow evenly covers the surface of the wire rod.

[0207] A laser 3D scanner is used to scan the wire rod axially at high speed to generate a three-dimensional surface topography map in real time, and defects such as scratches, folds, and pits are identified through an AI algorithm (convolutional neural network).

[0208] Detection logic: 1. Defect classification: scratches, folds, scale residue;

[0209] 2. Dynamic sorting: When a defect is detected, a pneumatic marking device is triggered to mark the position, and the defective section is imported into the repair line through PLC control of the sorting mechanism.

[0210] The present invention will be described in detail below in conjunction with specific embodiments.

[0211] Embodiment

[0212] Selection of cold heading steel: SCM435, including the following components by mass percentage (unit: %): C: 0.355; S: 0.25; Mn: 0.75; S: 0.02; P: 0.01; Cr: 1.05; Ni: 0.01; Cu: 0.15; Mo: 0.225.

[0213] In addition, the present invention is applicable to medium and low carbon cold heading steel (C = 0.10-0.45%) and microalloyed cold heading steel (containing B / Ti / Nb ≤ 0.05%), including SWRCH series (such as SWRCH35K), ML series (such as ML08Al), 10B21, etc., covering more than 90% of the cold heading steel grades in JIS, ASTM and GB standards.

[0214] The following will be described in conjunction with the operation steps of the production process. Step 1: Grinding and preheating 1.1. Grinding parameters: In the first round, a 18-mesh grinding wheel is used to remove the corner wrinkles (grinding depth 0.3 mm), and in the second round, a 25-mesh grinding wheel is switched for polishing (Ra ≤ 6.3 μm).

[0215] 1.2. Preheating parameters: Initial preheating at 600 °C (to prevent thermal stress cracks), and after heating to 1120 °C, a CO2 / N2 (volume ratio 3:7) mixed gas is introduced, and the surface hardness is increased to 250 HV (original 200 HV).

[0216] Step 2: Multi-stage temperature-controlled rolling 2.1 Rough rolling stage: temperature 1100 °C → 1000 °C (cooling rate 2 °C / s), reduction ratio 6:1 (Φ200 mm → Φ80 mm), reduction per pass 18%.

[0217] 2.2 Medium rolling stage: temperature 950 °C → 900 °C (cooling rate 1.5 °C / s), reduction ratio 4:1 (Φ80 mm → Φ38 mm), reduction rate 15%.

[0218] 2.3 Finish rolling stage: temperature 880 °C → 850 °C (cooling rate 1 °C / s), reduction ratio 2:1 (Φ38 mm → Φ18 mm), reduction rate 12%.

[0219] Step 3: Sectional controlled cooling 3.1 Rapid cooling stage: fan + water mist cooling (rate 4 °C / s), final cooling to 700 °C, austenite grain size ≤ 20 μm (traditional process ≥ 35 μm).

[0220] 3.2 Slow cooling stage: natural cooling for 300 s, ferrite ratio 76% (Avrami equation prediction error ±2%).

[0221] 3.3 Final cooling stage: introducing CO2 mixed gas (5% concentration), oxide film thickness 18 μm (traditional process ≥ 30 μm). After adopting the above production process, the average tensile strength and reduction of area of SCM435 (after the head and tail are sheared clean) are significantly decreased compared with those before optimization, decreased by 209 MPa (from 863 MPa to 654 MPa), the reduction of area is increased by 17% (from 33% to 50%), and all martensite in the structure is eliminated.

[0222] The properties of the cold heading steel wire rods of SCM435 before and after the production process optimization are shown in Table 1, where the data of the cold heading steel wire rods are measured according to the national standard GB / T 28906 - 2012.

[0223] Table 1: Properties of SCM435 before and after production process optimization

[0224] The typical microstructure characterization map before optimization is as Figure 2 shown, and the typical microstructure after optimization is as Figure 3 shown.

[0225] From the change of the mechanical properties of SCM435 under different shearing turns, it can be seen that for the SCM435 produced before optimization, the average mechanical properties per turn and the fluctuation of the reduction of area are small under different shearing turns, the fluctuation of the tensile strength is only 32 MPa, and the fluctuation of the reduction of area is only 5%. The mechanical properties of 22 mm SCM435 before optimization under different shearing turns are as Figure 4As shown, the mechanical properties of 22MM SCM435 after optimization with different numbers of shearing cycles are as follows Figure 5 as shown.

[0226] In addition, the metallographic conditions of SCM435 under different cooling conditions are as follows. As can be seen from Table 2 (measured according to the national standard GB / T 28906-2012), the average hardness and grain size level after optimization are 10 HRB lower than those before optimization, the scale is 5 microns thicker than that of the air-cooled line, and the decarburized layer thickness is the same.

[0227] Table 2: Metallography of SCM435 under Different Cooling Conditions

[0228] After passing through the overheating channel, the average tensile strength and reduction of area of SCM435 (after the head and tail are sheared cleanly) have decreased significantly compared with the existing ones, with a decrease of 209 MPa (from 863 MPa to 654 MPa), the reduction of area has increased by 17% (from 33% to 50%), all the martensite in the structure has been eliminated, and the hardness has decreased by 10 HRB (from 99 HRB to 89 HRB).

[0229] In another embodiment, B7 steel is used for cold steel wire rod processing. The performance parameters before and after optimization are shown in Table 3. Among them, the composition of B7 steel includes the following components by mass percentage: C: 0.395; S: 0.225; Mn: 0.85; S: 0.011; P: 0.012; Cr: 1.05; Mo: 0.25. The performance data is measured according to the national standard GB / T 28906-2012.

[0230] Table 3: Performance of B7 Steel before and after Optimization

[0231] After optimization, the average tensile strength of B7 (after the head and tail are sheared cleanly) has decreased significantly compared with the wire rod before optimization. The tensile strength has decreased by about 160 MPa (from 859 MPa to 699 MPa), the reduction of area has increased by 19% (from 29% to 48%), all the martensite in the structure has been eliminated, and the hardness has decreased by 7 HRB (from 100 HRB to 93 HRB). In addition, the typical structure of the cold heading wire before optimization is as Figure 6 shown, and the typical structure of the cold heading wire after optimization is as Figure 7 shown.

[0232] In summary, by means of the above technical solutions of the present invention, the present invention produces cold heading steel wire rods of different specifications by using an advanced rolling mill with a large compression ratio, and controls the cooling temperature and cooling rate of the wire rods with the help of a heat preservation tunnel furnace, realizing the production of hot-rolled cold heading steel wire rods without annealing, greatly simplifying the production process, shortening the production cycle, reducing the production cost, and improving the production efficiency; the produced cold heading steel wire rods do not need to be annealed, avoiding the changes in the organizational structure and performance that may be brought about by the annealing treatment, thereby improving the quality stability of the products; at the same time, the present invention has no strict requirements on the composition and organizational structure of the steel billet, so it can be applied to more steel types, expanding its application scope; by precisely controlling the cooling temperature and cooling rate, the strength and toughness of the products are significantly improved, and they have better use performance.

[0233] Precisely regulate the temperature changes in each stage through multi-pass temperature gradient rolling, and combine the segmented controlled cooling process to achieve the directional phase transformation of austenite to ferrite and the spheroidization of pearlite, breaking through the limitations of traditional processes; in the rolling process, through the synergistic action of differential temperature control and large reduction ratios in the rough, medium, and finish rolling stages, directly obtain ultrafine grains and uniform structures; in the cooling process, use rapid cooling to inhibit carbide coarsening, slow cooling to promote phase transformation equilibrium, supplemented by the final cooling oxide film control technology, completely eliminating the need for annealing; ultimately realizing the integrated performance improvement of high strength, high plasticity, and corrosion resistance of cold heading steel wire rods, while significantly reducing energy consumption and production costs, and promoting the upgrading of the steel hot rolling industry towards the direction of green and high efficiency.

[0234] It should be understood that although the steps in the flowchart of the accompanying drawings are shown sequentially according to the indication of the arrows, these steps do not necessarily have to be executed sequentially in the order indicated by the arrows. Unless there is a clear indication in this article, the execution of these steps has no strict order limit, and they can be executed in other orders. Moreover, at least a part of the steps in the flowchart of the accompanying drawings may include multiple sub-steps or multiple stages, and these sub-steps or stages do not necessarily have to be executed at the same time, but can be executed at different times, and their execution order does not necessarily have to be sequential, but can be executed alternately or alternately with at least a part of other steps or sub-steps or stages of other steps.

Claims

1. A production method of hot-rolled wire rods of cold-heading steel without annealing, characterized in that, Including: S1. Select a billet that meets the production specification requirements of cold heading steel wire rod, and perform grinding finishing and preheating treatment on the billet to be processed; S2. Transport the billet to a high reduction ratio rolling mill set, monitor the hot rolling process in real time to conduct hot rolling segmented control, achieve multi-pass temperature gradient rolling, and obtain cold heading steel wire rod with hot rolling forming; S3. Transport the hot-rolled cold heading steel wire rod to a heat preservation tunnel furnace, and through segmented controlled cooling and surface treatment, obtain the finished product of cold heading steel wire rod, realizing the production of hot-rolled wire rod of cold heading steel without annealing.

2. The production method of hot-rolled wire rods of cold-heading steel without annealing according to claim 1, characterized in that, The step of selecting a billet that meets the production specification requirements of cold heading steel wire rod, and performing grinding finishing and preheating treatment on the billet to be processed includes: S11. Preset the production specification of cold heading steel wire rod, select a billet with corresponding dimensions, and detect the component content of each component element in the billet by a laser spectrometer; S12. Adopt a two-round grinding wheel grinding method. In the first round, use a 16-20 mesh grinding wheel to remove the wrinkles at the corners of the billet, and in the second round, switch to a 24-26 mesh grinding wheel to perform fine polishing on the surface of the billet; S13. Place the ground billet in a preheating environment of 600-800 °C, gradually raise the preheating environment temperature to 1090-1170 °C, and then introduce a CO2 / N2 mixed gas to increase the surface hardness of the billet.

3. The production method of hot-rolled wire rods of cold-heading steel without annealing according to claim 1, characterized in that, The step of transporting the billet to a high reduction ratio rolling mill set, monitoring the hot rolling process in real time to conduct hot rolling segmented control, achieve multi-pass temperature gradient rolling, and obtain cold heading steel wire rod with hot rolling forming includes: S21. According to the production specification of cold heading steel wire rod and the billet component content, divide the rolling process into three stages: rough rolling, medium rolling and finish rolling, and use the physical constraint meta-learning algorithm to match the optimal rolling parameters; S22. Monitor the hot rolling data in the hot rolling process in real time, perform dynamic compensation on the optimal rolling parameters, and dynamically adjust the pass interval time according to the monitoring compensation result; S23. After rolling, spray a nano-Al2O3-MgO composite coating on the surface of the cold heading steel wire rod, generate a dense oxide film using the waste heat, and verify the actual production specification of the cold heading steel wire rod.

4. The production method of hot-rolled wire rods of cold-heading steel without annealing according to claim 3, characterized in that, The step of dividing the rolling process into three stages: rough rolling, medium rolling and finish rolling according to the production specification of cold heading steel wire rod and the billet component content, and using the physical constraint meta-learning algorithm to match the optimal rolling parameters includes: S211. Based on the preset production specification of cold heading steel wire rod, reverse deduce the cross-sectional dimension of the billet, and calculate the target reduction ratio of hot rolling the billet to cold heading steel wire rod; S212. Divide the rolling process of the billet into three stages: rough rolling, medium rolling and finish rolling, establish a parameter mapping table using the historical database, and output the initial reduction ratio and rolling temperature range of each stage; S213. Establish physical constraints for hot rolling of the billet based on metallurgical theory, and the physical constraints include phase transformation kinetics constraints, deformation resistance and rolling force constraints, and precipitation phase pinning constraints; S214. Set input variables and output variables, construct and train a parameter recommendation model based on meta-learning, integrate physical metallurgical constraints, and match rolling parameters corresponding to different billets; S215. Input the size and component content of the billet to be processed into the parameter recommendation model, generate rolling parameters for each stage and adjust the logic in real time to obtain the optimal rolling parameters.

5. The production method of hot-rolled wire rods of cold-heading steel without annealing according to claim 4, characterized in that, Set the input variables and output variables, construct and train a parameter recommendation model based on meta-learning, and integrate physical metallurgy constraints to match the rolling parameters corresponding to different steel billets, including: S2141. Corresponding each original task to a steel grade, set the production specifications of cold heading steel wire rods and the content of each component in the steel billet as input variables, and the rolling parameters as output variables to construct a support set. S2142. Link a physical calculation module after the output of the hidden layer to achieve physical information embedding, which is used to calculate the temperature and reduction ratio during the rolling of the steel billet at each stage. S2143. Set two physical loss term functions, namely phase transformation loss and rolling force loss, and combine the phase transformation loss and rolling force loss to set a comprehensive loss function to construct a parameter recommendation model. S2144. Pre-train the model parameters on all steel grade tasks to initially form a hot rolling knowledge base, and when facing a new steel grade, perform fine-tuning of the calculation task through single-step gradient descent.

6. The production method of hot-rolled wire rods of cold-heading steel without annealing according to claim 4, characterized in that, Input the size and component content of the steel billet to be processed into the parameter recommendation model, generate the rolling parameters at each stage and adjust the logic in real time to obtain the optimal rolling parameters, including: S2151. Obtain the carbon content of the steel billet to be processed, match the rough rolling temperature corresponding to the rough rolling stage according to the carbon content value, and calculate the rough rolling reduction ratio. The calculation formula for the rough rolling temperature is: ； where, T 粗轧 represents the rough rolling temperature in the rough rolling stage; T0 represents the preset initial temperature; [C] represents the carbon content; The calculation formula for the rough rolling reduction ratio is: ； where r 粗轧 represents the rough rolling reduction ratio in the rough rolling stage; r1 represents the initial compression ratio in the rough rolling stage output by the parameter mapping table; R total represents the target compression ratio; S2152. Obtain the titanium content, manganese content of the steel billet to be processed and the grain size after rough rolling, and calculate the medium rolling cooling rate and medium rolling reduction ratio when the steel billet enters the medium rolling stage. The calculation formula for the medium rolling cooling rate is: ； In the formula, v T represents the cooling rate during medium rolling; ΔT represents the temperature difference between rough rolling temperature and medium rolling temperature; Δt represents the time difference; [Mn] represents the manganese content; The calculation formula for the medium rolling reduction ratio is: ； where r 中轧 represents the intermediate rolling reduction rate; r2 represents the initial compression ratio in the intermediate rolling stage output by the parameter mapping table; [Ti] represents the titanium content; S2153. Calculate the Ar3 phase transformation point of the steel billet to be processed, set the finish rolling temperature of the finish rolling stage according to the Ar3 phase transformation point, and calculate the finish rolling reduction ratio of the finish rolling stage according to the carbon content. The calculation formula for the finish rolling reduction ratio is: ； where r 精轧 represents the finishing reduction ratio; r3 represents the initial compression ratio in the finishing stage output by the parameter mapping table.

7. The production method of a hot-rolled wire rod of non-annealing cold-heading steel according to claim 3, characterized in that, Real-time monitor the hot rolling data during the hot rolling process, perform dynamic compensation on the optimal rolling parameters, and dynamically adjust the pass interval time according to the monitoring compensation results, including: S221. Use the sensors inside the rolling mill to collect the hot rolling data of the steel billet rolling process in real time. S222. Calculate the difference between the actual rolling force and the target rolling force during the rolling process of each stage of the rolling mill, and calculate the reduction ratio compensation amount through the proportional-integral coefficient. S223. Calculate the temperature difference between the head and tail of the steel billet after passing through each stage. If the temperature difference between the head and tail is greater than the preset threshold, trigger the adjustment of extending the pass interval to balance the temperature drop and rolling rhythm. S224. Import the real-time hot rolling data into the historical parameter database, set the update period, and update the comprehensive loss function regularly. If there is a mutation in the hot rolling data, trigger an emergency stop signal.

8. The production method of a hot-rolled wire rod of non-annealing cold-heading steel according to claim 3, characterized in that, Transport the hot-rolled cold heading steel wire rod to the heat preservation tunnel furnace, and through segmented controlled cooling and surface treatment, obtain the finished cold heading steel wire rod to realize the production of hot-rolled cold heading steel wire rod without annealing, including: S31. Obtain and record the wire rod parameters after the cold heading steel wire rod is hot-rolled. The wire rod parameters include the finish rolling temperature, the grain size after rolling, the pre-judged ferrite ratio, and the rolling force curve. S32. Divide the cooling process of the cold heading steel wire rod into three stages: rapid cooling, slow cooling, and final cooling. Build a physical constraint meta-learning framework to generate cooling parameters for segmented control of the cooling of the cold heading steel wire rod. S33. Use high-pressure water to remove phosphorus from the cooled cold heading steel wire rod, and identify surface defects through on-line laser detection. After the detection is correct, the finished product of the cold heading steel wire rod is obtained.

9. The production method of a hot-rolled wire rod of non-annealing cold-heading steel according to claim 8, characterized in that, The step of dividing the cooling process of the cold heading steel wire rod into three stages: rapid cooling, slow cooling, and final cooling, and building a physical constraint meta-learning framework to generate cooling parameters for segmented control of the cooling of the cold heading steel wire rod includes: S321. Introduce wire rod parameters to expand the input layer, embed the phase transformation kinetics equation and the heat conduction equation to build a physical constraint layer, and build an output layer for outputting the segmented cooling rate value and the final cooling temperature. S322. Divide the cooling process of the cold heading steel wire rod into rapid cooling, slow cooling, and final cooling stages, and calculate the cooling rate and cooling time in the rapid cooling stage respectively. S323. Based on the preset segmented air-cooling strategy, use a fan-water mist combination for cooling in the rapid cooling stage, turn off the fan in the slow cooling stage, and use natural cooling by the roller table. In the final cooling stage, use natural cooling and introduce a nitrogen-carbon dioxide mixed gas to generate a dense oxide film.

10. The production method of a hot-rolled wire rod of non-annealing cold-heading steel according to claim 9, characterized in that, The step of calculating the cooling rate and cooling time in the rapid cooling stage includes: The calculation formula for the rapid cooling rate in the rapid cooling stage: ； In the formula, v 速冷 represents the rapid cooling rate in the rapid cooling stage; T 终轧 represents the finish rolling temperature; T 速冷终 represents the target temperature in the rapid cooling stage; t 速冷 represents the rapid cooling time; The calculation formula for the cooling time in the slow cooling stage: ； where t 缓冷 represents the slow cooling time; k0 represents the phase transformation rate constant; E a represents the phase transformation activation energy; R represents the gas constant; T 缓冷 represents the slow cooling temperature; X total represents the target ferrite ratio; n represents the Avrami exponent; e represents the natural constant.

Images