Annealing curve adaptive adjustment method for titanium steel composite strip for high-strength steel processing

Abstract

The present application belongs to the technical field of metal material heat treatment, and particularly relates to a titanium steel composite strip annealing curve self-adaptive adjustment method for high-strength steel processing. The method constructs a multi-source parameter feature matrix by collecting physical properties, geometry and cold working history parameters; establishes a real-time working condition perception model to obtain a dynamic working condition data set; constructs a thermal-mechanical coupling self-adaptive control logic based on an atomic diffusion dynamics model to realize dynamic compensation of a preset annealing curve; uses an adaptive control algorithm to execute closed-loop regulation and control of the temperature zone power instruction; and finally, in combination with quality evaluation data, realizes intelligent iteration of the control strategy through a deep reinforcement learning algorithm. Through real-time and accurate intervention in the annealing process, the present application realizes micron-level control of the interface compound layer thickness, improves the consistency of the interface bonding strength, effectively reduces the residual stress, ensures the uniform refinement of the titanium layer structure, enhances the adaptability of the composite material in high-strength steel processing, and improves the production efficiency.

Description

Technical Field

[0001] This invention belongs to the field of heat treatment technology for metallic materials, and specifically relates to an adaptive adjustment method for the annealing curve of titanium-steel composite strips used in the processing of high-strength steel. Background Technology

[0002] With the continuous advancement of materials science and advanced manufacturing technology, composite materials are increasingly widely used in aerospace, chemical equipment, and high-performance machinery manufacturing. Titanium-steel composite strips, as a type of metal composite material combining the excellent corrosion resistance of titanium with the superior mechanical properties of high-strength steel, have become a key strategic material for achieving synergistic effects of structural lightweighting and functionalization. In the process of preparing titanium-steel composite materials for high-strength steel processing, the heat treatment process, especially the annealing stage, has a decisive influence on the interfacial bonding strength between composite layers, the elimination of residual stress, and the final microstructure evolution of the material.

[0003] Among these, the design of the annealing curve for the titanium-steel composite strip is crucial to ensuring that the material exhibits ideal comprehensive mechanical properties in high-strength steel processing applications. This process involves precise control of key thermal parameters such as heating temperature, holding time, and cooling rate, aiming to improve the phase transformation characteristics of the bimetallic layer by optimizing the atomic diffusion mechanism at the interface, thereby obtaining a balanced microstructure that combines high strength and high ductility.

[0004] However, current technologies for annealing titanium-steel composite strips primarily rely on manual experience or fixed process templates, making it difficult to cope with dynamic changes in operating conditions caused by fluctuations in the composition of high-strength steel and differences in work hardening levels. This results in poor consistency of interfacial bonding energy. Furthermore, due to the significant differences in thermal expansion coefficients and thermal conductivity between titanium and steel, traditional heat treatment control models lack the ability to deeply perceive and provide real-time feedback on material state evolution when facing complex nonlinear thermo-mechanical coupling states, easily leading to process defects such as interfacial delamination or uneven internal structure. In addition, a single control logic cannot achieve intelligent iteration of annealing strategies based on real-time production data, making it difficult to achieve a dynamic balance between ensuring high-performance material output and production efficiency. Summary of the Invention

[0005] The purpose of this invention is to provide an adaptive adjustment method for the annealing curve of titanium-steel composite strips used in high-strength steel processing, which can at least effectively solve one of the problems in the background art mentioned above.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0007] An adaptive adjustment method for the annealing curve of titanium-steel composite strip for high-strength steel processing includes the following specific steps:

[0008] Step 1: Construct a multi-source parameter feature matrix: By collecting the initial physical property parameters, geometric structure parameters, and cold working deformation history data of the titanium-steel composite strip and the high-strength steel substrate, an original data feature vector containing the estimated value of the interface bonding state is formed. The initial physical property parameters include the thermal conductivity, specific heat capacity, and interfacial contact thermal resistance of the titanium layer and the steel layer.

[0009] Step 2: Establish a real-time working condition perception model: Deploy a multi-channel monitoring array in the annealing furnace to acquire the surface temperature distribution field, internal stress wave signal and ambient atmospheric pressure of the titanium-steel composite strip during heating, heat preservation and cooling processes, and generate a dynamic working condition dataset through spatiotemporal alignment processing.

[0010] Step 3: Construct thermo-coupling adaptive control logic: Based on nonlinear diffusion theory and thermoelastic mechanics equations, establish an atomic diffusion dynamics model at the titanium-steel interface. Calculate the growth rate of the interface compound layer thickness under the current working conditions in real time using the model, and dynamically compensate the preset annealing temperature curve according to the growth rate.

[0011] Step 4: Adaptive adjustment of the annealing curve: The compensation results in Step 3 are converted into power control commands for each temperature zone in the annealing furnace through an adaptive proportional-integral-derivative control algorithm, thereby achieving closed-loop control of heating rate, isothermal duration and cooling gradient.

[0012] Step 5: Quality evaluation and control strategy iteration: After annealing, obtain the interfacial bonding strength and microhardness distribution data of the titanium-steel composite strip, feed the data back into the control model, and correct the adaptive adjustment parameters through a deep reinforcement learning algorithm.

[0013] Preferably, the initial physical property parameters collected in step 1 also include the function curves of the thermal expansion coefficients of the titanium layer and the steel layer as a function of temperature, wherein the thermal expansion coefficients of the titanium layer and the steel layer are respectively within their respective preset value ranges. By comparing the difference in the thermal expansion coefficients of the two, the peak value of thermal stress generated at the interface during the heating process is pre-calculated.

[0014] Preferably, the cold working deformation history data of the high-strength steel substrate in step 1 includes the total shrinkage rate, the final reduction amount, and the rolling speed during the rolling process. The total shrinkage rate is set within a preset ratio range. The dislocation density distribution inside the high-strength steel is calculated using the deformation data, which serves as the boundary condition for the recrystallization kinetics calculation during the subsequent annealing process.

[0015] Preferably, the multi-channel monitoring array in step 2 includes a predetermined number of non-contact infrared radiation temperature sensors, ultrasonic interface state detectors, and acoustic emission stress monitors. The temperature sensors have a measurement range adapted to the annealing process requirements, a preset measurement accuracy, and an emissivity compensation algorithm to eliminate the influence of the oxidation state of the titanium surface on the temperature measurement results.

[0016] Preferably, in step 2, a high-sensitivity vacuum gauge is used to monitor the argon protection pressure inside the annealing furnace. The pressure is kept constant within a preset pressure range. By maintaining a slightly positive pressure environment, external air infiltration is prevented from causing oxidation and embrittlement of the titanium layer surface.

[0017] Preferably, the atomic diffusion kinetics model at the interface in step 3 adopts a variable parameter diffusion equation based on the Arrhenius formula, wherein the activation energy parameter is dynamically corrected according to the preset mass percentage of alloying elements in the high-strength steel substrate, and the correction coefficient is within the preset correction range.

[0018] Preferably, in the process of calculating the growth rate of the interfacial compound layer thickness in step 3, the formation of the titanium-iron intermetallic compound is monitored. When the predicted compound layer thickness reaches the preset critical range, the system automatically triggers a cooling command or shortens the holding time to prevent the deterioration of the interfacial bonding strength.

[0019] Preferably, the method for dynamically compensating the preset annealing temperature curve according to the growth rate in step 3 includes: calculating the deviation between the current interface diffusion distance and the target diffusion distance; when the deviation is positive and continues to increase, reducing the set temperature of the heating zone at a preset step temperature drop rate.

[0020] Preferably, in step 4, the control cycle of the adaptive proportional-integral-derivative control algorithm is set to a preset duration. By weighted calculation of temperature deviation, deviation change rate, and deviation integral value, the duty cycle of the thyristor trigger pulse within a preset range is output to ensure that the furnace temperature following error is less than a preset error threshold.

[0021] Preferably, the control of the heating rate in step 4 is divided into multiple stages: a first preset heating rate is used in the preheating stage; a second preset heating rate lower than the first preset heating rate is used in the phase change induction stage; and pulsed temperature adjustment is performed according to the real-time diffusion rate in the diffusion control stage.

[0022] Preferably, the closed-loop control of the cooling gradient in step 4 adopts a combination of multi-stage spray cooling and air cooling. In the initial stage of cooling, the cooling rate is controlled within the range of the first preset cooling rate. When the surface temperature of the material drops below the preset temperature threshold, the second preset cooling rate is switched to perform gentle cooling to eliminate the residual thermal stress caused by the mismatch of the thermal expansion coefficients of titanium steel.

[0023] Preferably, in step 5, the interfacial bonding strength is obtained using online ultrasonic scanning technology. By analyzing the reflectivity and transmittance characteristics of ultrasonic waves at the bimetallic interface, the interfacial bonding energy density is obtained. The control index of the interfacial bonding energy density needs to meet a preset strength threshold.

[0024] Preferably, the deep reinforcement learning algorithm in step 5 adopts a proximal policy optimization logic. Its input vector includes all historical control parameters and corresponding quality evaluation indicators from steps 1 to 4. Through multiple offline simulation training and online fine-tuning, the intelligent update of the annealing curve optimization weights is achieved.

[0025] Preferably, the present invention also includes establishing a full life cycle database of the annealing process of titanium-steel composite strip, recording the steel grade batch, titanium grade specification, real-time annealing curve and final mechanical property test results of each roll of composite strip, and the data entries of the database meet the preset scale requirements.

[0026] Preferably, the method is applied to a continuous heat treatment production line and is executed by a walking beam furnace or a roller hearth furnace. The running speed of the titanium-steel composite belt is controlled within a preset speed range, and the single roll processing cycle meets the predetermined time requirements.

[0027] Preferably, the method also involves real-time simulation of the thermo-coupled stress field during the annealing process, using a multi-physics coupling simulation unit to predict the warping deformation of the composite belt, and automatically adjusting the tension value of the tension roller when the predicted warping is greater than a preset deformation threshold, with the tension adjustment amount within a preset tension range.

[0028] Preferably, the high-strength steel substrate is an alloy steel with a yield strength that meets the preset strength requirements, the titanium layer is industrial pure titanium or a titanium alloy with specific corrosion resistance, and the thickness ratio of the titanium layer to the steel layer is set within a preset range.

[0029] Preferably, the thermo-coupling adaptive control logic in step 3 further includes the prediction of the titanium layer grain size, and the calculation of the grain growth law during the annealing process through the grain growth equation to ensure that the average grain size of the titanium layer after annealing is within the preset range.

[0030] Preferably, the control of the constant temperature duration in step 4 has a preset time precision, and the end point of the heat preservation is automatically determined according to the atomic exchange saturation of the interface to prevent energy waste and interface layer embrittlement caused by excessive heat preservation.

[0031] Preferably, the quality evaluation also includes predicting the bending test of the titanium-steel composite strip at a predetermined angle, calculating the minimum radius of bending without cracking based on the microstructure characteristics after annealing, and ensuring that the product meets the deep drawing requirements of high-strength steel processing.

[0032] Preferably, when the ambient atmosphere pressure monitoring and linkage adjustment system detects that the oxygen content exceeds the first preset concentration threshold, it immediately and automatically increases the argon flow rate and issues an early warning signal until the oxygen content drops below the second preset concentration threshold.

[0033] Compared with the prior art, the present invention has the following beneficial effects:

[0034] 1. High Consistency of Interface Bonding Performance: By constructing a thermo-coupled adaptive control model, this invention achieves precise perception of the atomic diffusion process at the titanium-steel interface, completely changing the traditional mode that relies on experience to set fixed curves. Because it can compensate in real time for fluctuations in high-strength steel composition and differences in work hardening, the thickness deviation of the interface compound layer is strictly controlled within a very small preset tolerance range. This significantly improves the consistency of the interface peel strength of the titanium-steel composite strip and effectively prevents interface delamination caused by process instability.

[0035] 2. Precise Control of Microstructure Evolution: This method combines nonlinear diffusion theory with multi-source sensor monitoring to achieve rapid response to heating rate and cooling gradient. Addressing the significant differences in thermophysical properties between titanium and steel, adaptive control effectively alleviates thermo-mechanical coupling stress, significantly reducing the residual stress level within the material. This not only ensures that the mechanical properties of the high-strength steel substrate do not degrade, but also ensures that the titanium layer has a fine and uniform equiaxed structure, significantly enhancing the ductility and compatibility of the composite material during high-strength steel processing.

[0036] 3. Significant Improvement in Production Efficiency and Intelligence: Relying on deep reinforcement learning algorithms and a full lifecycle database, this invention achieves continuous self-evolution of the annealing strategy. Adaptive control logic significantly shortens the holding time of a single production cycle based on the actual reaction progress, reducing heat treatment energy consumption. Simultaneously, the fully automated control process greatly reduces the need for manual intervention, significantly improving production efficiency and providing key technological support for the large-scale, high-quality, and intelligent manufacturing of titanium-steel composite materials.

[0037] 4. Excellent adaptability to complex working conditions: The method of this invention can flexibly handle titanium-steel composite strip products with different thickness ratios and alloy compositions. Through a multi-channel monitoring array and dynamic compensation mechanism, the system can automatically identify subtle differences in raw materials and correct the process path in real time. This strong robustness enables the method to be widely used in many high-end fields, greatly expanding the application boundaries of titanium-steel composite materials. Attached Figure Description

[0038] Figure 1 This is a schematic diagram of the overall architecture of the adaptive adjustment method for the annealing curve of titanium-steel composite strip for high-strength steel processing according to an embodiment of the present invention;

[0039] Figure 2 This is a flowchart illustrating the logical flow of multi-source parameter feature matrix construction and real-time working condition perception in the adaptive adjustment method of annealing curve for titanium-steel composite strip processing of high-strength steel according to an embodiment of the present invention.

[0040] Figure 3 This is a schematic diagram of the core principle framework of atomic diffusion dynamics based on thermo-coupling in the adaptive adjustment method of annealing curve of titanium-steel composite strip for high-strength steel processing according to an embodiment of the present invention.

[0041] Figure 4 This is a flowchart illustrating the dynamic compensation and power control process of the annealing curve based on the adaptive PID algorithm in the adaptive adjustment method of the annealing curve of titanium-steel composite strip for high-strength steel processing according to an embodiment of the present invention.

[0042] Figure 5 This is a flowchart illustrating the logical process of quality evaluation and control strategy iterative correction based on deep reinforcement learning in the adaptive adjustment method for annealing curve of titanium-steel composite strip for high-strength steel processing according to an embodiment of the present invention. Detailed Implementation

[0043] Example 1

[0044] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0045] In the aforementioned adaptive adjustment method for the annealing curve of titanium-steel composite strip used for high-strength steel processing, such as Figure 1 As shown, step 1 constructs a multi-source parameter feature matrix by collecting initial physical property parameters, geometric structural parameters, and historical cold working deformation data of the titanium-steel composite strip, forming an original data feature vector containing an estimated value of the interface bonding state. In the specific implementation process, the initial physical property parameters are the basis for establishing a thermodynamic model, specifically including the thermal conductivity, specific heat capacity, and interfacial contact thermal resistance of the titanium and steel layers. Since the thermophysical properties of titanium and steel at different temperatures are highly nonlinear, the thermal conductivity and specific heat capacity are defined as continuous functions of temperature. For example, within the annealing temperature range of 20°C to 900°C, the thermal conductivity of the titanium layer exhibits a specific nonlinear growth law with increasing temperature, while the specific heat capacity of the steel layer undergoes a sudden change in the phase transition region due to the complexity of its alloy composition. The interfacial contact thermal resistance depends on the degree of physical bonding in the initial stage of composite bonding, and is initially quantitatively characterized by the attenuation coefficient of the ultrasonic pulse echo amplitude.

[0046] In the specific execution details of step 1, the initial physical property parameters collected also include the function curves of the thermal expansion coefficients of the titanium and steel layers as a function of temperature. Considering the significant difference between the thermal expansion coefficients of titanium and the high-strength steel substrate, and the fluctuation of this difference with temperature gradient, the system presets a predetermined range of values ​​for the thermal expansion coefficients of the titanium and steel layers. By comparing the difference in thermal expansion coefficients between the two through a numerical integration algorithm, the peak thermal stress generated at the interface during the heating process is pre-calculated. The calculation of the peak thermal stress not only considers linear expansion displacement but also couples the elastic modulus of the material at the corresponding temperature, thereby forming the initial predictive features of the stress distribution cloud map.

[0047] Step 1 further integrates historical cold-working deformation data of the high-strength steel substrate. This data specifically includes the total shrinkage rate, final pass reduction, and rolling speed during the rolling process. The total shrinkage rate is set within a preset range of 30% to 85%. By establishing a dislocation energy evolution equation, the deformation data is used to calculate the dislocation density distribution and its spatial heterogeneity within the high-strength steel, which serves as the boundary condition for subsequent recrystallization kinetics calculations during annealing. Since the stored energy within the high-strength steel directly drives the microstructure evolution in the early stages of annealing, the precise quantification of dislocation density provides microscopic kinetic input for the subsequent thermo-mechanical coupling model.

[0048] Specifically, the process of establishing the dislocation energy evolution equation is as follows:

[0049] First, based on the cold working deformation history data of high-strength steel substrates, a Kocks-Mecking type model based on dislocation density evolution is used to calculate the dislocation density distribution inside the material. The mathematical expression of this evolution equation is:

[0050] in:

[0051] Dislocation density (unit: ), is a key internal variable describing the degree of lattice distortion within a material.

[0052] To accumulate true strain, it is calculated from the total shrinkage rate in the cold working deformation history. ,in The total reduction rate (values ​​between 0.30 and 0.85).

[0053] The dislocation storage factor is related to the work hardening ability of a material, and its value depends on the chemical composition and deformation temperature of the high-strength steel. In this invention, The material dynamic recovery model is used for calibration, for typical high-strength steels (such as yield strength 500-1200MPa). The range of values ​​is .

[0054] The dynamic recovery coefficient reflects the rate at which dislocations are annihilated during deformation, and its expression is: ,in:

[0055] The reference recovery coefficient was determined through a high-temperature compression test, and its typical range is [value missing]. .

[0056] The actual rolling strain rate is calculated from the final pass reduction and rolling speed: ,in The rolling speed is (m / s). , These represent the thickness before and after rolling, respectively. Where is the radius of the roll. This is the absolute reduction amount.

[0057] The reference strain rate (with values ​​of...) ), The rate sensitivity index (for high-strength steel, it is usually taken as...) ).

[0058] Solving the above differential equation using numerical integration methods (such as the fourth-order Runge-Kutta method) yields the dislocation density. Adapt to the situation The evolution trajectory. Due to the different strain rates and cumulative strains experienced by the surface and core of the material during cold working, the high-strength steel substrate is discretized into several layers along the thickness direction, with each layer assigned a corresponding... and The spatial distribution and heterogeneity of dislocation density can be calculated from the value.

[0059] Furthermore, the energy stored inside high-strength steel due to cold working (stored energy) It is directly related to dislocation density, and its expression is:

[0060] in:

[0061] For body-centered cubic iron-based materials, take a constant (for body-centered cubic iron-based materials, take a constant). ).

[0062] This is the shear modulus (unit: Pa), which varies with temperature and can be obtained from material handbooks or thermodynamic calculation software (such as JMatPro).

[0063] The Burgers vector (unit: m) is taken as follows: For ferrite or martensite in high-strength steel, take... .

[0064] Calculated dislocation density distribution and storage energy This will be used as the initial boundary condition and input into the recrystallization kinetics model in step three of the subsequent annealing process to predict the nucleation and growth behavior of grains during annealing, thereby achieving coupled control of the growth of interfacial compounds and the evolution of matrix structure.

[0065] Step 2 establishes a real-time operating condition perception model. For example... Figure 2 As shown, a multi-channel monitoring array is deployed inside the annealing furnace to acquire the surface temperature distribution field, internal stress wave signal, and ambient atmospheric pressure of the titanium-steel composite strip during heating, holding, and cooling processes. A dynamic working condition dataset is generated through spatiotemporal alignment processing. The multi-channel monitoring array consists of a predetermined number of non-contact infrared radiation temperature sensors, ultrasonic interface state detectors, and acoustic emission stress monitors. Specifically, the infrared radiation temperature sensors cover a measurement range of 200°C to 1200°C, with a measurement accuracy controlled within ±1°C. Considering the characteristic that titanium surfaces are prone to oxidation at high temperatures and exhibit dynamic changes in emissivity, the system incorporates an emissivity compensation algorithm. This algorithm, combined with multi-band radiation intensity comparison results, corrects the interference of the oxide film thickness on the temperature measurement value in real time, thereby obtaining the true surface temperature distribution field.

[0066] In step 2, the ultrasonic interface condition detector uses a high-temperature coupling agent or non-contact laser ultrasound technology to capture the echo characteristic signals at the composite interface in real time. Meanwhile, the acoustic emission stress monitor is used to capture the elastic stress waves generated within the material during thermal cycling in real time. These signals are transmitted to the operating condition sensing logic center via a high-speed data acquisition card, where they undergo wavelet denoising and fast Fourier transform processing to convert them into characteristic indicators reflecting the risk of interface delamination or the degree of stress concentration.

[0067] A high-sensitivity vacuum gauge is used to monitor the argon protective pressure inside the annealing furnace in real time. This pressure is strictly maintained within a preset range of 101.5 kPa to 105.0 kPa. By maintaining this slightly positive pressure environment, external air infiltration is effectively prevented from causing oxidation and embrittlement of the titanium layer surface. When the linkage control system detects that the oxygen content exceeds the first preset concentration threshold (e.g., 50 ppm), it immediately and automatically increases the argon flow rate and issues visual and audible warning signals until the oxygen content drops below the second preset concentration threshold (e.g., 10 ppm), ensuring the absolute inertness of the annealing environment.

[0068] Step 3 involves constructing the thermo-coupled adaptive control logic. For example... Figure 3 As shown, this logic, based on nonlinear diffusion theory and thermoelastic equations, establishes an atomic diffusion kinetic model at the titanium-steel interface. The core of this model is to calculate the growth rate of the interface compound layer thickness under current operating conditions using real-time operating data. The atomic diffusion kinetic model at the interface employs a variable-parameter diffusion equation based on the Arrhenius equation.

[0069] In this step, the diffusion coefficient The expression is as follows: .in, It is the diffusion constant. As the reference activation energy, The gas constant is... To obtain the absolute temperature in real time, and The activation energy correction parameter. Dynamic correction is performed based on the preset mass percentages of alloying elements such as chromium, manganese, and molybdenum in the high-strength steel substrate. The correction coefficient is within a preset range of 0.85 to 1.15 to reflect the influence of alloying elements on the interdiffusion barrier between iron and titanium atoms.

[0070] During the calculation of the growth rate of the interfacial compound layer thickness, the system focuses on monitoring the titanium-iron intermetallic compound (such as...). , The system monitors the generation of the compound layer. When the thickness of the compound layer predicted by the diffusion model reaches a preset critical range (e.g., 1.5 micrometers to 2.5 micrometers), the system will automatically trigger a cooling command or shorten the current holding time to prevent the brittle compound from becoming too thick, which could lead to severe deterioration of the interfacial bonding strength.

[0071] The method for dynamically compensating the preset annealing temperature curve based on the growth rate in step 3 includes: calculating the deviation between the current interface diffusion distance and the target diffusion distance. When the deviation is positive and continues to increase, it means that the compound is growing too fast. The system will reduce the set temperature of the heating zone at a preset step temperature drop rate (e.g., 5 to 15 degrees Celsius per minute) to suppress diffusion through a negative feedback mechanism.

[0072] The thermo-coupled adaptive control logic also includes the prediction of titanium layer grain size. The grain growth equation is used to calculate the grain growth law during annealing, ensuring that the average grain size of the annealed titanium layer is within a preset range of 8 to 10. Furthermore, the system uses a multiphysics coupling simulation unit to predict the warpage deformation of the composite belt in real time. When the predicted warpage exceeds a preset deformation threshold (e.g., warpage exceeding 2 mm per meter width), the system automatically adjusts the tension value of the tension roller. The tension adjustment is within a preset tension range of 5% to 15%, using external tension to counteract the bending moment caused by thermal inhomogeneity.

[0073] Step 4 involves adaptive adjustment of the annealing curve. For example... Figure 4 As shown, the compensation results in step 3 are converted into power control commands for each temperature zone within the annealing furnace using an adaptive proportional-integral-derivative (PID) control algorithm, thereby achieving closed-loop control of the heating rate, isothermal duration, and cooling gradient. The control cycle of the adaptive PID control algorithm is set to 50 milliseconds to 200 milliseconds.

[0074] Specifically, the duty cycle of the output thyristor trigger pulse The calculation formula is: .in, The temperature deviation at the current moment. , , The gain parameters are optimized in real time based on the logic of step 3. This weighted calculation ensures that the furnace temperature following error is less than a preset error threshold of ±2 degrees Celsius.

[0075] The control of the heating rate is divided into three key stages. First, in the preheating stage, a first preset heating rate (e.g., 10 degrees Celsius per second) is used to allow the material to quickly pass through the low-temperature brittleness zone. Then, in the phase transformation induction stage, a second preset heating rate (e.g., 3 degrees Celsius per second) lower than the first preset heating rate is used to mitigate the volume change impact during the microstructure transformation process. Finally, in the diffusion control stage, pulsed temperature adjustment is performed based on the real-time calculated diffusion rate to precisely anchor the depth of the interfacial reaction.

[0076] The control of the isothermal duration has a time accuracy on the order of 1 second. The system automatically determines the end point of the isothermal treatment based on the predicted value of the interfacial atomic exchange saturation. Once the target binding energy density is reached, the isothermal treatment is stopped immediately and the cooling stage begins, effectively preventing energy waste and interfacial embrittlement caused by excessive isothermal treatment.

[0077] The closed-loop control of the cooling gradient employs a combination of multi-stage spray cooling and air cooling. In the initial cooling phase, to quickly lock in the high-temperature microstructure, the cooling rate is controlled within a first preset cooling rate range (e.g., 20 to 30 degrees Celsius per second). When the material surface temperature drops below a preset temperature threshold (e.g., 450 degrees Celsius), the system automatically switches to a second preset cooling rate for gentler cooling (e.g., 5 degrees Celsius per second) to eliminate residual thermal stress caused by the mismatch in the thermal expansion coefficients of titanium and steel, ensuring the straightness of the composite strip.

[0078] Step 5 involves iterating the quality evaluation and control strategy. For example... Figure 5As shown, after annealing, the interfacial bonding strength and microhardness distribution data of the titanium-steel composite strip are acquired and fed back into the control model. The interfacial bonding strength is obtained using online ultrasonic scanning technology. By analyzing the reflectivity and transmittance characteristics of ultrasound at the bimetallic interface, the energy characteristic value of the echo envelope is extracted, and the interfacial bonding energy density is obtained by inversion using an acoustic impedance mapping model. The control index of the interfacial bonding energy density must meet a preset strength threshold (e.g., greater than 150 MPa).

[0079] The quality assessment also includes predictions for bending tests at predetermined angles on titanium-steel composite strips. The system calculates the minimum radius for bending without cracking based on the microstructure characteristics, grain size distribution, and interfacial compound layer thickness after annealing. If the predicted minimum radius does not meet the deep-drawing requirements for high-strength steel processing, a strategy warning is triggered.

[0080] The core of step 5 lies in the application of a deep reinforcement learning algorithm. This algorithm employs proximal policy optimization (PPO) logic, and its input vector includes all historical control parameters from steps 1 to 4 (such as setpoints for each temperature zone, tension values, and cooling flow rates) and corresponding quality evaluation indicators (combining strength, hardness, and flatness). By constructing an action reward function and utilizing multiple offline simulation training and online fine-tuning, the intelligent updating of the annealing curve optimization weights is achieved.

[0081] This invention also establishes a full lifecycle database for the annealing process of titanium-steel composite strips. This database records in detail the high-strength steel batch, titanium grade, real-time acquired annealing curves, raw sensor waveforms, and final mechanical property test results for each roll of composite strip. The database entries meet the preset large-scale storage requirements, and through big data correlation analysis, the optimal process characteristics of different specifications of products under extreme operating conditions can be extracted.

[0082] In specific application scenarios, the method is applied to continuous heat treatment production lines, executed via walking beam furnaces or roller hearth furnaces. The running speed of the titanium-steel composite belt is controlled within a preset speed range of 5 to 25 meters per minute, and the single-roll processing cycle meets the predetermined capacity requirements. The high-strength steel base material is low-alloy high-strength steel with a yield strength meeting the preset strength requirement of 500 MPa to 1200 MPa, and the titanium layer is industrial pure titanium (e.g., ...). , Alternatively, titanium-molybdenum or titanium-palladium alloys with specific corrosion resistance can be used. The thickness ratio of the titanium layer to the steel layer is set within a preset range of 1:5 to 1:20.

[0083] Example 2

[0084] This embodiment focuses on the adaptive annealing process for ultra-thick high-strength steel-titanium composite plates (total thickness greater than 20 mm). In this application scenario, due to the large plate thickness, the temperature gradient and thermal stress distribution along the thickness direction are more complex, placing higher precision requirements on the adaptive logic of steps 3 and 4.

[0085] In step 1, the constructed multi-source parameter feature matrix includes discrete node parameters along the thickness direction. Among the initial physical property parameters, interfacial contact thermal resistance is given a higher weight because the initial bonding state during thick plate composite forming has a significant hysteresis effect on the subsequent thermodynamic response of diffusion. Historical data on cold working deformation of the high-strength steel substrate specifically records the deformation difference between the central and surface layers of the plate, and the calculated dislocation density distribution exhibits a clear gradient characteristic.

[0086] In step 2, the multi-channel monitoring array was augmented with penetrating ultrasonic monitoring points to detect real-time changes in tissue density within the thick plate. The density of acoustic emission stress monitors was increased to four per square meter to capture acoustic emission signals from internal tearing that may occur during the heating process of the thick plate from all angles. Ambient atmosphere pressure monitoring not only focused on argon pressure but also incorporated a residual oxygen content sensor with a sensitivity reaching the 1 ppm level.

[0087] In the thermo-coupling adaptive control logic of step 3, a time integral correction term is introduced into the atomic diffusion kinetics model to address the thermal inertia of the thick plate. Since the thick plate heats up slowly, the calculation of the compound growth rate must consider the cumulative effect of long-term, low-speed diffusion. The activation energy parameter correction is fine-tuned based on the precise content of microalloying elements (such as niobium, vanadium, titanium, and boron) in the substrate, with a correction coefficient accurate to 0.001.

[0088] When the model predicts that the intermetallic compound layer of titanium iron at the interface diffuses into the steel side beyond the preset depth, the temperature drop rate calculated by the system not only acts on the heating element, but also controls the speed of the circulating fan in the furnace through linkage adjustment to enhance convective heat transfer and achieve rapid response.

[0089] During the annealing curve execution in step 4, the parameters of the adaptive PID control algorithm... It is set to a non-linear proportional change with the plate thickness. The heating rate is refined into five stages, with a thermal equilibrium stage inserted between the preheating and phase change induction stages. At this stage, the heating rate is reduced to 0.5 degrees Celsius per second and held for a preset duration to ensure that the temperature difference between the inside and outside of the thick plate is less than 5 degrees Celsius, preventing the plate from warping due to excessive thermal stress.

[0090] The cooling gradient is controlled using a combined "internal cooling and external spraying" mode. In the initial control phase, the air curtain intensity of the lateral fans is adjusted first to establish a uniform cooling field across the width of the plate. When the plate surface temperature reaches 380 degrees Celsius, high-pressure fine water mist spraying is activated. The cooling rate is calculated in real time based on the thermal conductivity model of the plate thickness to ensure that the thermal stress generated during the cooling process is always lower than the yield strength of the material at that temperature.

[0091] In the evaluation stage of step 5, the online ultrasonic scanning technology is upgraded to phased array ultrasonic imaging technology to form a two-dimensional topographic map of the interface bonding state. The control index of interface bonding energy density is refined into local thresholds for different regions to address the potential microstructure inhomogeneities in large-size plates. The PPO logic of the deep reinforcement learning algorithm incorporates the edge effects of the plate into the state space, and automatically optimizes compensation strategies for edge overheating or excessively rapid cooling through process iteration.

[0092] In addition to recording basic parameters, the full lifecycle database also records the sequence of thermal images during the annealing process. By identifying features from massive amounts of thermal images, the system can automatically determine whether the heating tubes are experiencing performance aging or whether the gas distribution pipes are partially blocked, thereby achieving collaborative diagnosis of equipment health status and process parameters.

[0093] Example 3

[0094] This embodiment demonstrates the application of the described method in processing different alloy compositions (such as...). Intelligent switching and parameter correction process when titanium alloy and special weather-resistant high-strength steel are used in titanium-steel composite strips.

[0095] In step 1, when the input titanium layer parameters are changed from industrial pure titanium to... ( When this happens, the thermal conductivity function and thermal expansion curve in the multi-source parameter characteristic matrix are immediately updated globally based on the pre-stored model in the database. Because... The introduction of aluminum and vanadium into the alloy results in significant differences in specific heat capacity and thermal conductivity compared to pure titanium. The system automatically recalculates the baseline value of interfacial contact thermal resistance.

[0096] In step 2, the emissivity compensation algorithm of the infrared radiation thermometer automatically switches to alloy mode. Because... The structure of the surface oxide layer is different from that of pure titanium. The algorithm corrects the emissivity deviation at temperatures above 800 degrees Celsius by calling a preset alloy oxidation model, ensuring the accuracy of the temperature data fed back to the control logic.

[0097] In the thermo-coupling adaptive control logic of step 3, the interfacial atomic diffusion kinetics model was extended for ternary and multi-component systems. The model not only calculates the diffusion of iron and titanium but also simulates the segregation behavior of aluminum and vanadium at the interface. Activation energy parameters... The correction range is expanded to 0.75 to 1.25 to accommodate the influence of complex alloying elements on vacancy formation energy. When the system predicts the potential precipitation of complex alloyed intermetallic compounds at the interface, it issues a temperature limit command in advance.

[0098] In step 4, the adaptive PID algorithm is designed for... The lower thermal conductivity automatically reduced the proportional gain. and integral gain To prevent overshoot during the control process, the second stage of the heating rate (phase change induced stage) is significantly prolonged because... of Phase transition is a relatively slow process. The cooling gradient is adjusted to be asymmetric, meaning the cooling rate is faster above the phase transition region to preserve fine particles. The tissue undergoes a transformation, with a slower cooling rate below the phase transition zone to release the tissue stress generated by the phase transition.

[0099] In step 5, the model for obtaining the interfacial bonding strength incorporates a correction coefficient for the hardness gradient. Because... The hardness after annealing is usually higher than that of pure titanium. The acoustic impedance mismatch at the interface changes, and the system adjusts this by real-time input. The acoustic parameters are recalibrated to adjust the interfacial binding energy density. When correcting parameters, the deep reinforcement learning algorithm retrieves historical data on similar alloy proportions from a full lifecycle database as hot-start weights, improving the convergence speed of policy iteration by more than 50%.

[0100] Example 4

[0101] This embodiment describes in detail the emergency adaptive adjustment mechanism of the method when facing sudden abnormal operating conditions of the equipment (such as partial failure of the argon protection system or failure of the heating element in the temperature zone).

[0102] During the real-time operating condition sensing process in step 2, when the high-sensitivity vacuum gauge detects that the furnace pressure momentarily drops below 100.5 kPa and the residual oxygen sensor detects that the oxygen content surges above 100 ppm, the system determines that an atmospheric contamination anomaly has occurred. At this time, the logic center immediately skips the conventional adjustment logic and directly enters the anomaly handling mode.

[0103] In step 3, the thermo-coupled adaptive control logic immediately incorporates the diffusion kinetics model of oxygen atoms on the titanium surface to recalculate the oxidation thickening rate of the titanium layer. The model predicts in real time the extent of damage to the interfacial peel strength caused by the oxide layer. If the predicted oxidation damage is within a controllable range, the control command in step 4 will automatically increase the heating rate to the maximum allowable value, shorten the exposure time in the contaminated atmosphere, and significantly increase the argon purging flow rate in the cooling section to forcibly reduce the plate temperature to below 200 degrees Celsius, thus terminating the oxidation reaction.

[0104] If a sudden drop in the power of the heating element in a certain temperature zone is detected during annealing, the adaptive PID algorithm will detect the deviation between the actual temperature in that zone and the set curve. The temperature continues to increase. At this point, the adaptive control logic will utilize the power margin of adjacent temperature zones. For example, if the third temperature zone fails, the second and fourth temperature zones will automatically increase the set temperature and heating power. By adjusting the running speed (stepping speed) of the composite belt, the residence time of the material in adjacent high-temperature zones is extended. Utilizing the principle of lateral heat compensation, the cumulative thermal effect (i.e., diffusion integral value) of the titanium-steel interface is ensured to remain within the target preset range.

[0105] In step 5, the quality evaluation process involves a full-coverage online ultrasonic scan of the damaged batch of products, marking the abnormal locations. A deep reinforcement learning algorithm stores the successful parameter sequence from this anomaly handling in an "extreme operating condition knowledge base" and adjusts the weights of the original control strategy. If the final mechanical performance test shows that the bonding strength still meets the requirements, this emergency adaptive solution will be formalized as a special process path.

[0106] Example 5

[0107] This embodiment describes the digital management and closed-loop process of the entire process lifecycle in a large-scale continuous production line according to the present invention.

[0108] In step 1, the system automatically reads the original identification information of each roll of titanium-steel composite strip to be processed via RFID tags, including the furnace number of the high-strength steel, the production batch of the titanium material, and the initial pressure and temperature records during composite molding. This data is automatically filled into a multi-source parameter feature matrix to generate a unique process feature fingerprint.

[0109] During the operation of the continuous heat treatment line, the real-time sensing system in step 2 records the entire line's operating status at a frequency of 1000 data sets per second. This high-frequency data is transmitted in real time to the full lifecycle database via an industrial IoT interface. The database adopts a distributed storage architecture to ensure zero latency in data writing.

[0110] With the synergy of steps 3 and 4, the production line achieves precise "roll-to-roll" control. Even if there is a 5% thickness deviation between the front and rear ends of the same roll of composite tape, the system can adjust the temperature rise curves of each zone in real time through real-time simulation prediction and PID dynamic compensation, so that the interface bonding strength deviation over the entire length is controlled within 3%.

[0111] In step 5, in addition to physical performance evaluation, the system also introduces cost and energy consumption evaluation indicators. The deep reinforcement learning algorithm not only aims for optimal quality but also uses argon consumption and electricity consumption per unit of output as penalty terms in the reward function. Through policy iteration, the system automatically finds a balance point that ensures both interface strength and a preset high annealing energy efficiency ratio (EER).

[0112] When each roll of product leaves the factory, the full lifecycle database automatically generates a digital quality certificate containing a complete temperature cloud map, interface ultrasonic scan image, microstructure evolution trajectory, and predicted mechanical properties. Users can trace every microscopic control step in the annealing process by scanning the code on the roll.

[0113] Example 6

[0114] This embodiment describes an extended application of the method in annealing multilayer (such as steel-titanium-steel or titanium-steel-titanium) trimetallic symmetric / asymmetric composite strips.

[0115] For multi-interface systems, the feature matrix in step 1 is expanded into a multi-layer tensor structure. Contact thermal resistance parameters for multiple interfaces are added to the physical property parameters, while the geometric structure parameters must cover the thickness ratio of the interlayer and the cladding layer. For asymmetric structures (e.g., a titanium layer thickness of 1 mm on one side and 0.5 mm on the other), the system pre-calculates the prestress field generated during heating due to the imbalance of expansion torques on both sides.

[0116] In step 2, the monitoring array employs dual-sided infrared thermometry and bidirectional acoustic emission acquisition. Through spatiotemporal alignment processing, the system can identify the phase difference in thermal response between different interfaces and use it as a dynamic operating condition dataset reflecting the internal heat conduction uniformity.

[0117] The thermo-coupling control logic in step 3 evolves into a multi-interface coupling model. The system simultaneously calculates the atomic diffusion dynamics of both interfaces, selecting the side with the faster growth rate as the primary control object. The diffusion equation undergoes boundary condition reconstruction for the heterogeneous interface, considering the stress-induced diffusion effect of the interlayer metal under pressure. Activation energy parameter corrections are set according to the different degrees of alloy element segregation at the two interfaces.

[0118] In step 4, the adaptive PID control algorithm adds a "balance deviation control" term. For asymmetric multilayer boards, the system establishes a compensating temperature gradient in the thickness direction by independently adjusting the power distribution ratio of the upper and lower heaters in the furnace, thereby counteracting warping caused by uneven material expansion. This gradient control accuracy reaches 0.5 degrees Celsius per millimeter.

[0119] Step 5, quality evaluation, employs computed tomography (CT) ultrasound to extract the binding energy density of the upper and lower interfaces. The reinforcement learning algorithm uses multi-objective optimization logic during policy iteration, aiming to minimize overall processing time while ensuring all interfaces meet strength standards. The full lifecycle database meticulously records the evolution data of each layer and interface, providing data support for process optimization of multilayer composite materials.

[0120] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of this invention is defined by the appended claims and their equivalents.

Claims

1. A method for adaptive adjustment of the annealing curve of titanium-steel composite strip for high-strength steel processing, characterized in that, The method includes the following steps: Step 1: Construct a multi-source parameter feature matrix: By collecting the initial physical property parameters, geometric structure parameters of the titanium-steel composite strip and the historical data of cold working deformation of the high-strength steel substrate, an original data feature vector containing the estimated value of the interface bonding state is formed. Step 2: Establish a real-time working condition perception model: Deploy a multi-channel monitoring array composed of various types of sensors in the annealing furnace to acquire the surface temperature distribution field, internal stress wave signal and ambient atmospheric pressure of the titanium-steel composite strip during heating, heat preservation and cooling processes, and map the physical information of different dimensions to a unified time axis and spatial coordinate system through spatiotemporal alignment processing to generate a dynamic working condition dataset. Step 3: Construct thermo-coupling adaptive control logic: Based on nonlinear diffusion theory and thermoelastic mechanics equations, establish an atomic diffusion dynamics model at the titanium-steel interface. Calculate the growth rate of the interface compound layer thickness under the current working conditions in real time using the atomic diffusion dynamics model, and dynamically compensate the preset annealing temperature curve according to the growth rate of the interface compound layer thickness to achieve precise intervention in the interface reaction process. Step 4: Perform adaptive adjustment of the annealing curve: Through the adaptive proportional-integral-derivative control algorithm, the compensation results in Step 3 are converted into power control commands for each temperature zone in the annealing furnace, realizing closed-loop control of heating rate, constant temperature duration and cooling gradient, and using feedback mechanism to correct the output power of the actuator in real time. Step 5, Quality Evaluation and Control Strategy Iteration: After annealing, obtain the interfacial bonding strength and microhardness distribution data of the titanium-steel composite strip, and feed the interfacial bonding strength and microhardness distribution data back to the control model. The adaptive adjustment parameters are corrected through deep reinforcement learning algorithm to realize the intelligent updating and optimization of annealing process parameters.

2. The adaptive adjustment method for the annealing curve of titanium-steel composite strip for high-strength steel processing according to claim 1, characterized in that, The initial physical property parameters collected in step one also include the function curves of the thermal expansion coefficients of the titanium layer and the steel layer as a function of temperature, wherein the thermal expansion coefficients of the titanium layer and the steel layer are respectively within their respective preset numerical ranges; by comparing the difference in thermal expansion coefficients between the titanium layer and the steel layer through a numerical integration algorithm, the peak value of thermal stress generated at the interface during the heating process is pre-calculated; the calculation process of the peak value of thermal stress is coupled with the elastic modulus of the material at the corresponding temperature, thereby forming an initial prediction feature reflecting the evolution trajectory of stress distribution, and the initial prediction feature is stored in the multi-source parameter feature matrix.

3. The adaptive adjustment method for the annealing curve of titanium-steel composite strip for high-strength steel processing according to claim 1, characterized in that, The cold working deformation history data of the high-strength steel substrate in step one includes the total shrinkage rate, the final reduction amount, and the rolling speed during the rolling process, wherein the total shrinkage rate is set within a preset ratio range. By establishing a dislocation energy evolution equation, the dislocation density distribution and its spatial heterogeneity inside the high-strength steel are calculated using the cold working deformation history data, and this is used as the boundary condition for the recrystallization kinetics calculation in the subsequent annealing process. The stored energy inside the high-strength steel is quantified using the dislocation density distribution, providing input parameters for the micro-dynamics calculation of the thermo-coupled adaptive control logic in the subsequent steps.

4. The adaptive adjustment method for the annealing curve of titanium-steel composite strip for high-strength steel processing according to claim 1, characterized in that, In step two, the multi-channel monitoring array includes a predetermined number of non-contact infrared radiation temperature sensors, ultrasonic interface state detectors, and acoustic emission stress monitors. The measurement range of the non-contact infrared radiation temperature sensors is adapted to the requirements of the annealing process. Considering the dynamic change in emissivity of titanium surfaces at high temperatures, the real-time working condition perception model incorporates an emissivity compensation algorithm. This algorithm, combined with multi-band radiation intensity comparison results, corrects the interference of the oxide film thickness on the temperature measurement value in real time, thereby obtaining the true surface temperature distribution field.

5. The adaptive adjustment method for the annealing curve of titanium-steel composite strip for high-strength steel processing according to claim 4, characterized in that, In step two, a high-sensitivity vacuum gauge is used to monitor the argon protection pressure in the annealing furnace in real time. The argon protection pressure is kept constant within a preset pressure range. By maintaining a slightly positive pressure environment, external air infiltration is prevented from causing oxidation and embrittlement of the titanium layer surface. The real-time operating condition perception model is associated with an environmental atmosphere pressure monitoring and linkage adjustment system. When the environmental atmosphere pressure monitoring and linkage adjustment system detects that the oxygen content exceeds the first preset concentration threshold, it automatically increases the argon flow rate and issues an early warning signal until the oxygen content drops below the second preset concentration threshold.

6. The adaptive adjustment method for the annealing curve of titanium-steel composite strip for high-strength steel processing according to claim 1, characterized in that, The atomic diffusion dynamics model at the interface in step three adopts a variable parameter diffusion equation based on the Arrhenius formula. The calculation process of the diffusion coefficient includes determining the diffusion constant, the reference activation energy, the gas constant, and the absolute temperature obtained in real time. The atomic diffusion dynamics model introduces an activation energy correction parameter, which is dynamically corrected according to the preset mass percentage of alloying elements in the high-strength steel substrate. The alloying elements include chromium, manganese, and molybdenum, and the correction coefficient is within the preset correction range to reflect the influence of alloying elements on the interdiffusion barrier between iron atoms and titanium atoms.

7. The adaptive adjustment method for the annealing curve of titanium-steel composite strip for high-strength steel processing according to claim 1, characterized in that, In the process of calculating the growth rate of the interfacial compound layer thickness in step three, the system focuses on monitoring the formation of the titanium-iron intermetallic compound. When the predicted thickness of the titanium-iron intermetallic compound layer reaches the preset critical range, the system automatically triggers a cooling command or shortens the holding time to prevent the deterioration of the interfacial bonding strength. The method of dynamically compensating the preset annealing temperature curve based on the growth rate of the interfacial compound layer thickness also includes: calculating the deviation between the current interfacial diffusion distance and the target diffusion distance. When the deviation is positive and continues to increase, the set temperature of the heating zone is reduced at a preset step temperature drop rate.

8. The adaptive adjustment method for the annealing curve of titanium-steel composite strip for high-strength steel processing according to claim 1, characterized in that, In step four, the control cycle of the adaptive proportional-integral-derivative control algorithm is set to a preset duration. By weighting the temperature deviation, the rate of change of deviation, and the integral value of deviation, a power control signal within a preset range is output to ensure that the furnace temperature following error is less than a preset error threshold. The regulation of the heating rate is divided into multiple stages. In the preheating stage, a first preset heating rate is used. In the phase change induction stage, a second preset heating rate lower than the first preset heating rate is used. In the diffusion control stage, pulse temperature adjustment is performed according to the real-time diffusion rate.

9. The adaptive adjustment method for the annealing curve of titanium-steel composite strip for high-strength steel processing according to claim 1, characterized in that, In step four, the closed-loop control of the cooling gradient adopts a combination of multi-stage spray cooling and air cooling. In the initial stage of cooling, the cooling rate is controlled within the range of the first preset cooling rate. When the surface temperature of the material drops below the preset temperature threshold, the second preset cooling rate is switched to perform gentle cooling to eliminate the residual thermal stress caused by the mismatch between the thermal expansion coefficients of the titanium layer and the steel layer.

10. The adaptive adjustment method for the annealing curve of titanium-steel composite strip for high-strength steel processing according to claim 1, characterized in that, In step five, the deep reinforcement learning algorithm adopts a near-end policy optimization logic. Its input vector includes the historical control parameters and corresponding quality evaluation indicators from steps one to four. By constructing an action reward function, offline simulation training and online fine-tuning are used to achieve intelligent updating of the annealing curve optimization weights.

Images