Aviation difficult-to-deform ring forging process path design method based on multi-field coupling

Abstract

The invention relates to the field of metal plastic forming and forging and pressing processes, and discloses an aviation difficult-to-deform ring forging process path design method based on multi-field coupling. The method comprises the following steps: generating a process initialization package through parameterized definition and boundary configuration based on a target ring forging three-dimensional model and volume constraint; performing multi-field coupling simulation configuration, including grid division, thermal and contact boundary setting, organization and field variable model binding, and generation of a transient solution control sequence; multi-field coupling calculation in the whole rolling process is executed, temperature-strain-tissue process data is collected, evaluation indexes are extracted, weights are configured, and an evaluation function input structure is formed; an optimization scheme package is output through optimization algorithm parameter setting, parameter updating and restart solving, convergence judgment and scheme freezing. According to the method, automatic inversion design of the process path is achieved, and the one-time standard reaching rate of geometric accuracy and microstructure performance of the ring forge piece is effectively improved.

Description

Technical Field

[0001] This invention relates to the field of metal plastic forming and forging processes, and in particular to a method for designing process paths for difficult-to-deform aerospace ring forgings based on multi-field coupling. Background Technology

[0002] In the field of metal plastic forming and forging processes, existing solutions for the entire rolling process of difficult-to-deform ring forgings for aerospace applications typically revolve around billet geometry settings, equipment operating parameter configurations, and finite element calculations. These solutions rely on manual experience to determine the billet cross-section and heating process, employing thermo-mechanical analysis to approximate the rolling process. The process specifications are repeatedly revised during the trial production phase. However, these methods suffer from limitations such as insufficient detail in the analysis of the target ring forging's 3D model and volume constraints, inconsistencies in the parameterized definition and boundary dimension configuration of the pre-rolled billet cross-section, and unclear descriptions of the timing and stages of process initialization. Existing methods often focus on solution paths within single or weakly coupled fields, lacking unified management of material constitutive and thermophysical property data, as well as contact boundaries, thermal boundaries, and load time axes. This leads to unstable tracking of geometric surrogate quantities and discontinuous acquisition of microstructure states throughout the rolling process, making it difficult to ensure stable transfer from the pre-rolled billet's 3D geometric parameter set and process initialization package structure to the multi-field coupled simulation configuration. Existing technologies generally suffer from common shortcomings in the joint processing of process initialization package structure, titanium alloy constitutive and thermophysical property data, and roller-to-bulk contact conditions. These shortcomings include fragmented processes and inconsistent interfaces in mesh generation and staged boundary setting, synchronization of temperature, flow, and strain field control parameters, and binding and writing back of microstructure and field variables. This makes it difficult to form a consistent process in the entire rolling process, from multi-field coupled simulation configuration to microstructure and field variable mapping table, transient solution control sequence, geometric and microstructure evaluation data, evaluation index set, evaluation function input structure, optimization iterative control package, and design variable set. As a result, there are shortcomings such as frequent repeated trials in process path design, pressure on production organization and quality consistency, and large fluctuations in processing cycle and resource input.

[0003] Furthermore, titanium alloys, due to their high specific strength, excellent corrosion resistance, and heat resistance, have become ideal structural materials in the aerospace field. Ring forgings are core load-bearing components in critical parts such as aero engines, and their performance directly affects the reliability and lifespan of the equipment. Rolling expansion is the main method for manufacturing large titanium alloy rings. The properties of titanium alloys are highly dependent on their microstructure, including the grain size of the β-transformation structure formed after β-processing, and the morphology, content, and size of the primary α phase and secondary α phase after processing the (α+β) two-phase region. However, the rolling expansion of titanium alloy rings is a complex thermo-mechanical-phase transformation coupled process carried out at high temperatures, facing severe challenges: 1. The microstructure control is extremely complex: the rolling process typically occurs in the β-phase region or the (α+β) two-phase region. In the β-phase region, deformation parameters (temperature, strain, strain rate) strictly control dynamic recrystallization and grain growth behavior; in the (α+β) two-phase region or during subsequent cooling, phase transformation kinetics determine the morphology and distribution of the α-phase. Traditional empirical design cannot quantitatively describe the process-microstructure relationship.

[0004] 2. Difficulty in ensuring uniformity of microstructure: The temperature and strain histories of different parts of the ring are significantly different during the rolling process, which can easily lead to uneven β grain size and different α phase morphology, resulting in a "gradient" in microstructure properties and affecting the overall reliability of the component.

[0005] 3. Narrow process window: Titanium alloys are sensitive to hot working parameters. Inappropriate processes can easily lead to defects such as overheating and unqualified phase composition. However, finding the optimal process window through physical trial and error is extremely costly and time-consuming. Summary of the Invention

[0006] To address the aforementioned technical problems, this invention provides a method for designing process paths for difficult-to-deform aerospace ring forgings based on multi-field coupling, comprising: Obtain the 3D model and volume constraints of the target ring forging, perform parameterized definition of the pre-rolled billet section, boundary dimension configuration and process initialization consistency check, and generate the process initialization package structure; Obtain the process initialization package structure, perform mesh generation, thermal boundary and contact boundary setting, microstructure-field variable model binding and step control setting, and generate transient solution control sequence; The transient solution control sequence is acquired and multi-field coupled calculations of the entire rolling process are performed. Temperature-strain-microstructure history is collected, evaluation indexes are extracted and weights are configured, and the input structure of the evaluation function is generated. Obtain the input structure of the evaluation function, perform optimization algorithm parameter setting, parameter update and restart solution, convergence judgment and scheme freezing processing, and generate the optimization scheme package structure.

[0007] Furthermore, the parameterization definition of the pre-rolled billet section includes: The parameterized definition of the pre-rolled billet section includes creating a pre-rolled billet section skeleton in the parameterized modeling environment, consisting of a neutral layer reference circle, inner edge control lines, and outer edge control lines. Independent parameter groups are set in the radial direction, axial direction, and circumferential direction. Volume constraints are associated through the constraint solver to form the solution objective. When the combination of parameter groups causes the volume to deviate from the volume constraint tolerance, automatic correction is triggered and the local thickening amplitude of the parameter group in the circumferential direction is adjusted first.

[0008] Furthermore, the process of configuring boundary dimensions includes: The boundary dimension configuration includes forming a list of equipment capacity constraints based on the forming equipment capacity table, the clamping space of the core roller and main roller, the effective working area of ​​the heating furnace and the maximum opening of the rolling mill, and forming a list of process safety margins based on material specifications and tooling assembly tolerances. The boundary dimensions are checked item by item for the parameter group obtained by the parameterization definition. When the minimum wall thickness is found to be lower than that specified in the safety margin list, the process reverts to the parameterization solver and locks the initial value of the inner diameter while adjusting the initial value of the outer diameter.

[0009] Furthermore, the mesh generation process includes: Mesh generation involves partitioning the annular section according to its thickness gradient, the angle range of locally thickened sections, and the chamfer range of the end face. Variable pitch sector partitioning is used in the circumferential direction, a layered refinement strategy is used in the radial direction, and local refinement and buffer bands are set in the axial transition area of ​​the end face. Adaptive local re-partitioning is triggered by the element twist degree and the minimum interior angle threshold.

[0010] Furthermore, the process of setting thermal boundaries and contact boundaries includes: The thermal boundary setting includes applying a segmented convection-radiation composite boundary to the outer surface of the billet based on the initial heating temperature and the transport-heat preservation sequence mark, and applying a time-transformed thermal boundary to the roller surface based on the roller temperature and the state of the cooling medium. The contact pressure-related local heat transfer coefficient of the clamping reference surface area is written into the thermal boundary mapping table and a boundary update event is inserted at the stage switching point. The contact boundary setting includes generating a group of candidate contact surfaces from the geometric contours of the core roller, main roller, and support roller provided by the equipment database. The contact opening and closing sequence is driven by the stage markers in the process initialization package structure. The contact form is defined as a combination of rigid-plastic forming contact and viscous-slip hybrid friction model. The friction coefficient and normal penalty parameters are retrieved from the high-temperature tribological correlation parameter table and assigned values ​​in stages. Mesh smoothing and local regeneration strategies are introduced in the contact area to handle element distortion.

[0011] Furthermore, the process of collecting temperature-strain-structure history data includes: The process acquisition includes synchronous sampling of channels defined by the field variable mapping table in the observation step. The temperature channel records the representative node temperatures near the inner edge control line, outer edge control line, and end face reference surface of the billet. The strain channel records the equivalent strain and equivalent strain increment time series described by the equivalent strain integral path in the contact zone. The strain rate channel records the equivalent strain rate trajectory in the contact zone and near the free surface. In the switching step, a forced observation is added to bind the stage marker and field variable snapshot.

[0012] Furthermore, the process of extracting evaluation indicators and configuring weights includes: The evaluation index extraction includes segmenting and merging the data according to the stage markers, extracting geometric surrogate statistics of roundness and end face flatness for each stage, strain uniformity weighted by circumferential sector and radial layered grid, and β grain size and α phase morphology stage state retrieved from the synchronous sampling entries of the organization field. The weight configuration includes reading the weight strategy table based on the key process focus, grouping the evaluation index set by stage, merging geometric, deformation and microstructure indices within the group according to a preset hierarchical relationship, mapping the weight strategy to the entries within the group, redistributing the weight of strain uniformity according to the weighted results of circumferential sector and radial layered grid, and assigning segmented weights to β grain size and α phase morphology according to the degree of overlap between the stage thermal history and microstructure history. When an entry has an anomaly label or filling mark, a weight reduction or exclusion rule is applied.

[0013] Furthermore, the process of optimizing algorithm parameter settings includes: The optimization algorithm parameter settings include adopting genetic algorithms or distributed particle swarm optimization strategies, setting the population size, crossover and mutation probability intervals or particle number, individual and group guiding factors according to the design variable dimensions and constraint list, mapping the stage priority to the stage weighting coefficient during fitness evaluation, and deriving the minimum iteration rounds based on stage coverage and item distribution density, and combining estimated costs and computing resource quotas to form an iterative batch allocation scheme.

[0014] Furthermore, the parameter update and solution restart process includes: The parameter update and restart solution process includes sampling candidate solutions from the parameter space of the optimization algorithm, decomposing the geometric parameters of the pre-rolled billet section into inner diameter, outer diameter, wall thickness distribution curve, angle range of local thickened sections, end face chamfer range, and clamping reference surface description, and writing them into the corresponding fields of the three-dimensional geometric parameter set of the pre-rolled billet. The rolling process parameters are decomposed into the core roll feed curve, main roll speed, and initial heating temperature staged description, and written into the corresponding fields of the load time axis and boundary segmentation description. Solution request entries aligned with the stage-reviewed sampling points are generated, and a constraint list comparison is performed on the candidate solutions. If a variable goes out of bounds, it is pushed back to the nearest boundary inside according to the feasible domain pushback mechanism.

[0015] Furthermore, the convergence determination and scheme freezing process includes: Convergence determination and scheme freezing involve obtaining the standardized index values ​​and weights corresponding to the variable entries through the evaluation query interface, forming an evaluation snapshot set, and simultaneously examining its cross-batch improvement magnitude, same-batch dispersion and stage coverage completeness for multi-condition judgment. When the improvement magnitude and dispersion are less than the threshold and the stage coverage completeness meets the strategy requirements, a convergence event is triggered. The scheme freezing process involves selecting representative entries from the design variable set and fixing their pre-rolled billet cross-sectional geometry parameters and rolling process parameters, while solidifying the associated stage sequence and boundary segment descriptions, and assigning a version number consisting of a date code, equipment code, and revision number to the frozen entries.

[0016] The key innovations of this invention include: (1) Construct a continuous data link driven by the process initialization package structure, multi-field coupled simulation configuration, microstructure-field variable mapping table and transient solution control sequence, and encapsulate the geometric and microstructure evaluation data into the evaluation function input structure through the evaluation index set, so as to realize the ordered transmission and traceable interface between the objects listed in claim 1.

[0017] (2) Extract the temperature field, flow field and strain field control parameters from the multi-field coupled simulation configuration, bind the dynamic recrystallization and grain growth model of the β region and the cooling phase transformation model of the (α+β) region, generate the microstructure-field variable mapping table, and form a staged mapping and write-back mechanism for microstructure variables and field variables.

[0018] (3) Based on the input structure of the evaluation function, the optimization algorithm parameter setting and iteration number configuration are carried out to obtain the optimization iteration control package. The design variable set is generated by updating the parameters and restarting the solution. Finally, the convergence judgment and scheme freezing of the design variable set are performed, and the optimization scheme package structure is output to form a closed loop link of evaluation-optimization-simulation.

[0019] The following are its main beneficial effects: (1) The integrated organization of process initialization package structure, multi-field coupled simulation configuration and transient solution control sequence ensures that the operation link from data organization to solution driving to evaluation index set and evaluation function input structure is consistent. Compared with the existing discrete process around single field or weak coupling, it can complete the continuous processing of acquisition-alignment-recording-encapsulation within the same terminology system, which is suitable for the process management scenario of the entire roll expansion process.

[0020] (2) By binding the control parameters of temperature field, flow field and strain field with the model, a microstructure-field variable mapping table is established and written back in stages. Compared with the existing independent calling path for material constitutive and microstructure evolution, geometric and microstructure evaluation data can be obtained synchronously at the observation node of the transient solution control sequence. It is suitable for the rolling process with stage switching and boundary update.

[0021] (3) Based on the optimization iteration control package to trigger parameter updates and restart the solution, the optimization scheme package structure is solidified on the basis of batch generation and convergence judgment of the design variable set. Compared with the existing trial production driven multi-round manual revision process, strategic iteration and scheme freezing can be completed under the unified metric of the evaluation function input structure. It is suitable for the application boundary of coordinated adjustment of the three-dimensional geometric parameter set of pre-rolled billet and rolling process parameters. Attached Figure Description

[0022] Figure 1 This is a flowchart illustrating a method for designing process paths for difficult-to-deform aerospace ring forgings based on multi-field coupling, as provided in an embodiment of this application. Detailed Implementation

[0023] Example 1: Refer to Figure 1 This is a flowchart illustrating a method for designing process paths for difficult-to-deform aerospace ring forgings based on multi-field coupling, provided by an embodiment of the present invention. The process may include at least steps S100-S400: S100: Obtain the three-dimensional model and volume constraints of the target ring forging, perform parameterized definition of the pre-rolled billet section, boundary dimension configuration and process initialization consistency check, and generate the process initialization package structure; S200: Obtain the process initialization package structure, perform mesh generation, thermal boundary and contact boundary setting, microstructure-field variable model binding and step control setting, and generate transient solution control sequence; S300: Acquire the transient solution control sequence and perform multi-field coupled calculation of the entire rolling process; collect temperature-strain-structure history data, extract evaluation indicators and configure weights; and generate the input structure of the evaluation function. S400: Obtain the input structure of the evaluation function, perform optimization algorithm parameter setting, parameter update and restart solution, convergence judgment and scheme freezing processing, and generate the optimization scheme package structure.

[0024] Step S100 includes at least steps S110-S130: S110. Obtain the three-dimensional model and volume constraints of the target ring forging, perform parameterized definition of the pre-rolled billet section and boundary dimension configuration, and obtain the three-dimensional geometric parameter set of the pre-rolled billet. For this step, the input sources are the target ring forging 3D model formed in the early product design phase and the volume constraints from material utilization calculation. The target ring forging 3D model includes geometric elements such as nominal inner diameter, nominal outer diameter, nominal height, chamfers and transition fillets, end face reference surfaces, and clamping reference surfaces. The volume constraints include the target net volume, machining allowance volume allocation strategy, and end trim and flash volume reservation rules. Specifically, the system first reads the target ring forging 3D model metadata and verifies topological closure, normal consistency, and geometric continuity through a model checking program. If gaps or self-intersections exist, a topology repair process is triggered. After repair, the system re-verifies and records the differences before and after repair. Subsequently, the system calculates the target blank volume based on the volume constraints and constructs a blank-finished product volume mapping table for consistency determination in the subsequent parametric solution process. Furthermore, the system creates a pre-rolled billet section skeleton in a parametric modeling environment. The skeleton consists of three parts: a neutral layer reference circle, inner edge control lines, and outer edge control lines. The neutral layer reference circle corresponds to the circumferential path of the ring's centroid. The inner and outer edge control lines describe the section thickness variation through several control nodes and segment constraints. To ensure a clear mapping relationship with the target ring forging's 3D model, the system projects the finished product's inner and outer diameters onto the pre-rolled billet section coordinate system, generating a constraint lookup table. This table records the correspondence between the finished product and the billet's inner and outer contours, the axial limit relationship of the end face reference surface, and the angular position description of the section inflection points. After completing the skeleton creation, the system initiates the pre-rolled billet section parametric definition process, setting independent parameter groups in the radial, axial, and circumferential directions. The radial direction parameter group includes the initial values ​​of the inner and outer diameters and the radial transition segment shape code; the axial direction parameter group includes the initial value of the billet height, the end face plane reference layer, and the end face chamfer range; and the circumferential direction parameter group includes the angular range and amplitude description of locally thickened or thinned sections. The aforementioned parameter set is associated with volume constraints and forms the solution objective through the constraint solver. When the combination of parameter sets causes the volume to deviate from the volume constraint tolerance, the solver triggers automatic correction, prioritizing the adjustment of the local thickening amplitude of the circumferential direction parameter set, then adjusting the initial value of the outer diameter of the radial direction parameter set, and finally outputting the parameter set state within the tolerance range. Subsequently, the system enters the boundary dimension configuration process. In this step, the boundary dimensions are defined as the upper and lower limits of the minimum wall thickness, maximum wall thickness, minimum inner diameter, maximum outer diameter, minimum end face allowance, and maximum end face allowance of the pre-rolled billet. The system generates an equipment capacity constraint list based on the forming equipment capacity table, the clamping space of the core roll and main roll, the effective working area of ​​the heating furnace, and the maximum opening of the rolling mill. Then, based on the material specifications and tooling assembly tolerances, a process safety margin list is generated. The two lists together constitute the boundary dimension configuration input.Specifically, the system checks the boundary dimensions of each parameter group obtained from the parameterized definition of the pre-rolled billet section. When the minimum wall thickness is found to be lower than the safety margin list, the system reverts to the parameterized solver, locks the initial value of the inner diameter, adjusts the initial value of the outer diameter, and simultaneously provides a revert label and a reason record. When the maximum outer diameter is found to exceed the maximum opening amount in the equipment capacity constraint list, the system triggers the section segmentation strategy, reduces the amplitude of the locally thickened section, and writes the state before and after the reduction into the change record. The processing link enters the section smoothing and transition control stage. The system performs curvature continuity checks on the inner and outer edge control lines. If a curvature abrupt change occurs, a transition segment is inserted between the control nodes, and segment constraints are used to suppress local cusps. Subsequently, the axial direction parameter group is aligned with the end face plane reference layer to ensure that there is no interference between the clamping reference surface and the end face chamfer range. After completing the above processing, the system calculates the instantaneous approximate volume of the pre-rolled billet and compares it with the billet-finished product volume mapping table. If the deviation is within the volume constraint tolerance, a three-dimensional geometric parameter set for the pre-rolled billet is formed; if the deviation still exceeds the tolerance, a secondary solution process is triggered and the parameters are adjusted according to priority. To support subsequent cross-step calls, the system stores the three-dimensional geometric parameter set of the pre-rolled billet in a data container and generates field indexes and timestamps. The fields include inner diameter, outer diameter, wall thickness distribution curve, axial height, end face chamfer range, angle range of local thickened sections, cross-sectional shape code, and clamping reference surface description. The three-dimensional geometric parameter set of the pre-rolled billet generated by the above processing is used as the output field name of this step. It is directly referenced in the multi-field coupling simulation configuration of S210 and participates in the construction of the solution domain together with the process initialization package structure. At the same time, this parameter set can be used as a reference geometric benchmark for geometric evaluation indicators in the standardization and numbering process of S320, and can be used as the source of cross-sectional geometry in the parameter update and restart solution stage of S420, thereby completing the connection with S200, S300 and S400.

[0025] S120. Extract the core roll feed curve, main roll speed and initial heating temperature from the three-dimensional geometric parameters of the pre-rolled billet, configure the initial parameters of the rolling expansion process path, and generate the process initialization package. The input source for this step is the aforementioned three-dimensional geometric parameter set of the pre-rolled billet. This set provides fields such as wall thickness distribution curves, angle ranges of locally thickened sections, axial height, and clamping reference surface descriptions. Additionally, it imports a list from the equipment database regarding the kinematic capabilities, torque capabilities, and speed ranges of the mandrel and main roll, serving as upper and lower bound references for parameter configuration. Specifically, the system first establishes a process path description framework. In this step, the process path is defined as a tripartite consisting of the mandrel feed curve, the main roll speed sequence, and the initial heating temperature. The mandrel feed curve describes the radial displacement history of the mandrel, the main roll speed sequence describes the rotational speed arrangement of the main roll at different stages, and the initial heating temperature describes the target furnace temperature before the billet enters the rolling and expanding process. The system reads the wall thickness distribution curve and the angle range of locally thickened sections from the three-dimensional geometric parameters of the pre-rolled billet. Through a path mapping module, it establishes a correspondence between sections with larger wall thicknesses and mandrel displacement control sections. The mapping rule is that sections with larger wall thicknesses correspond to gentler mandrel feed slopes, and sections with smaller wall thicknesses correspond to faster mandrel feed slopes. The mapping results are written into the draft mandrel feed curve. Subsequently, the system calls the equipment database to obtain the main roll speed range list and recommended speed levels. Combining the clamping reference surface description and axial height, it provides a draft main roll speed sequence. If the axial height is large and the locally thickened sections are widely distributed, the system inserts a segmented speed strategy into the draft main roll speed sequence. The segmented strategy is represented by stage markers and speed level numbers, and stage switching is triggered at specific displacement mileages on the mandrel feed curve. For the initial heating temperature, the system reads the heating temperature range and furnace temperature uniformity requirements of titanium alloys from the material specification library, and simultaneously reads the minimum wall thickness and maximum outer diameter from the three-dimensional geometric parameter set of the pre-rolled billet to construct a heating load assessment quantity. This assessment quantity is used to select the basic setting of the heating furnace heating program and outputs a draft initial heating temperature. Further, the system performs preliminary process path coordination on the ternary group draft. The coordination process involves coupled verification of the draft mandrel feed curve and the draft main roll speed sequence. Verification includes judging the contact area slippage trend, determining the radial deformation rate boundary, and smoothing the transition section. If a stage with a strong slippage trend is identified, the main roll speed sequence draft is lowered or the slope of the draft mandrel feed curve is reduced, and the reason for the modification is recorded. If the radial deformation rate deviates from the recommended range in the equipment database, automatic limiting is triggered and a limiting record is generated. For the initial heating temperature draft, the system links the heating furnace capacity table and clamping reference surface description to determine the heating heat transfer obstruction area under clamping conditions, and performs temperature zone fine-tuning on the initial heating temperature draft according to the proportion of obstruction area to avoid excessive temperature difference between the surface and the interior due to local thickness; at the same time, the temperature curve draft marks the time limit constraints of the heat preservation stage and the transfer from the furnace to the machine, so that the downstream multi-field coupling simulation configuration can call these timing parameters.After coordination, the system submits the draft triplet to the rule engine. The rule engine includes stage division rules, speed range selection rules, and temperature fine-tuning rules. The rule engine evaluates each input draft, and if there are any items that conflict with the equipment database, material specification library, or boundary size configuration, it provides revision suggestions and outputs the initial parameter configuration for the process path. To support tracking and backtracking, the system packages the initial parameter configuration for the process path. The package contains three main parameters: core roller feed curve, main roller speed, and initial heating temperature, as well as stage markers, limit records, revision suggestions, and timestamps. After packaging, a process initialization package is formed. The process initialization package is identified as the output field name in this step and used for consistency checks and version numbering in S130. It also serves as a core input in the multi-field coupling simulation configuration in S210, participating in mesh generation, thermal boundary, and contact boundary setting. In addition, the main parameters in this package are updated and written back during the parameter update and restart solution phase in S420, providing an entry point for subsequent closed-loop operation.

[0026] S130. Perform consistency checks and version numbering on the process initialization package, and generate the process initialization package structure. The input for this step is the aforementioned process initialization package and its associated equipment database, material specification library, and boundary dimension configuration results. The system first establishes a consistency checklist. In this step, consistency checking is defined as a multi-dimensional comparison process of three types of parameters—the core roller feed curve, the main roller speed, and the initial heating temperature—and their associated stage markers, limit records, and revision suggestions. The checklist covers four aspects: stage continuity, parameter boundary validity, cross-library constraint consistency, and timing executability. Specifically, the system performs a stage continuity check on the core roller feed curve, checking whether the start and end displacements of each stage overlap and have no empty segments. If there is overlap, the starting displacement of the subsequent stage is trimmed first according to the length of the overlap interval. If there is an empty segment, a transition segment is inserted based on the slope and boundary dimension configuration recommended value of the previous stage. Subsequently, the system performs a parameter boundary validity check on the main roller speed, comparing it with the speed range list and torque capacity in the equipment database. If the speed range of a certain stage exceeds the capacity range, the speed is downgraded without changing the stage division, and a downgrade record is generated. If the torque capacity is insufficient to support the expected deformation rate of a certain stage, the core roller feed curve is linked to reduce the slope of the corresponding stage, and the linkage relationship is recorded. For the initial heating temperature, the system compares the heating temperature range in the material specification library with the handling time limit constraint. If the temperature range exceeds the specification range, it is revised according to the upper or lower limit of the specification and a temperature range revision record is generated. If there is a conflict between the handling time limit setting and the furnace exit time of the heating program, the handling time or holding time is adjusted to form a handling-holding coordination record. After completing the above checks, the system enters the cross-library constraint consistency check. The check logic reads the clamping reference surface description and maximum outer diameter from the boundary dimension configuration results, judges the centrifugal trend risk under the main roll speed setting, and if the risk is high, it outputs a speed reduction suggestion and re-reviews the core roll feed curve slope. At the same time, it reads the angle range of the local thickened section in the three-dimensional geometric parameter set of the pre-rolled billet, judges the alignment degree between the stage mark and the angle range, and if the stage switching point falls in the middle of the thickened section, it moves the switching point to the vicinity of the section boundary to avoid frequent parameter jumps within the stage. Subsequently, the system performs a timing executability check, comparing the stage markers, main roller speed, and core roller feed curve mileage points in the process initialization package with the workshop cycle time plan. If the stage timing does not match the equipment shift response time, the stage switching point is shifted based on the equipment response time delay parameter, and a stage timing revision record is output. After completing all consistency checks, the system assigns a version number to the process initialization package that passes the check. The version number consists of a date code, an equipment code, and a revision sequence number. After the number is generated, it is written into the package's metadata and modification permissions are frozen. At the same time, the limit records, downgrade records, temperature zone revision records, handling-insulation coordination records, stage timing revision records, and linkage relationships from this check are uniformly encapsulated to form the process initialization package structure.The process initialization package structure is confirmed as the output field name in this step. This structure contains the core roller feed curve, main roller speed, and initial heating temperature, which have undergone consistency checks and version numbering freeze, along with stage markers, constraints, and change history. This structure is directly called by the multi-field coupled simulation configuration in S210 for process driving during mesh generation, thermal boundary, and contact boundary setting. Simultaneously, this structure serves as the stage switching trigger source in the step control and data logging settings of S230, and as a reference baseline for difference analysis during the parameter update and restart solution phase in S420, thus establishing a stable connection with S200, S300, and S400. In summary, the technical effects of this step are: through consistency checks and version numbering, the output process initialization package structure has clear boundary, timing, and constraint descriptions, can be directly driven by downstream solution and optimization stages, and can be stably reused in the closed-loop process.

[0027] Step S200 includes at least steps S210-S230: S210. Obtain the process initialization package structure and titanium alloy constitutive and thermal property data, perform mesh generation, thermal boundary and contact boundary setting processing, and obtain the multi-field coupled simulation configuration. The input sources for this step are the process initialization package structure formed in the previous steps and the constitutive and thermophysical property data of titanium alloy provided by the material side. The process initialization package structure includes three main parameters: core roller feed curve, main roller speed and initial heating temperature, as well as stage markers, limiting conditions and change history. In this embodiment, the constitutive and thermophysical property data of titanium alloy are defined as a process description of the rheological stress-strain-strain rate-temperature response table covering the forming temperature zone, the temperature segmented values ​​of thermal conductivity and specific heat, the working condition range values ​​of density and emissivity, and the high-temperature tribological correlation parameters. Specifically, the aforementioned process initialization package structure and titanium alloy constitutive and thermophysical property data are used as input. The pre-rolled billet three-dimensional geometric parameter set corresponding to the billet entity and the forming domain representation of the rolling mill are loaded through the geometric preprocessing module. Topological verification, patch stitching and smoothing of adjacent curved surfaces are performed on the billet entity to obtain an analytical geometry consistent with the clamping reference surface. On this basis, the mesh generation sub-process divides the grid according to the thickness gradient of the annular section, the angle range of the locally thickened section and the chamfer range of the end face. The circumferential direction adopts variable pitch sector division, the radial direction adopts a layered densification strategy, and the axial direction adopts local refinement and sets a buffer zone in the end face transition area. The mesh quality criterion in this step is defined as the interval constraint between the element twist degree and the minimum interior angle threshold. When the criterion exceeds the limit, adaptive local re-division is triggered and the revision label is recorded. Furthermore, to reflect heat transfer and contact heat exchange during the rolling process, the thermal boundary setting process reads the initial heating temperature and the handling-heating sequence mark, applies a segmented convection-radiation composite boundary to the outer surface of the billet according to the time segment from the furnace exit to the machine, and applies a time-transformed thermal boundary to the roller surface according to the roller temperature and cooling medium state provided by the equipment database; for the area where the clamping reference surface is located, the system writes the local heat transfer coefficient related to the contact pressure into the thermal boundary mapping table, and inserts a boundary update event at the stage switching point. In the contact boundary setting process, the geometric contours of the core roller, main roller, and support roller provided by the equipment database are used to generate a group of candidate contact surfaces. Then, the stage markers in the process initialization package structure drive the contact opening and closing sequence. The contact form in this step is defined as a combination of rigid-plastic forming contact and viscous-slip hybrid friction model. The friction coefficient and normal penalty parameter are retrieved from the high-temperature tribological correlation parameter table and assigned segmented values ​​according to the stage. To deal with the severe local element distortion caused by rolling contact, the system introduces a mesh smoothing and local regeneration strategy in the contact area. When the element distortion exceeds the threshold, local regeneration is triggered according to the displacement mileage of the core roller feed curve, and interpolation transfer is performed on the nodal field variables before and after regeneration.Understandably, after the aforementioned mesh generation, thermal boundary and contact boundary settings are completed, it is also necessary to construct the docking relationship between the material field parameters and load sequence required for solution control. The material field parameters are driven by the constitutive and thermal property data of titanium alloy, and the load sequence is driven by the staged description of the core roller feed curve and the main roller speed. The two are aligned on the stage time axis. When the stage switching point and the load update point do not coincide, the system automatically inserts a synchronization event in the nearest safety window and generates an alignment record. After completing the above processing, the system generates a comprehensive configuration list covering geometry, mesh, materials, thermal boundaries, contact boundaries, and load time axis, and writes stage marker index and timestamp. In this step, the comprehensive configuration list is recorded as the output field name "multi-field coupling simulation configuration," which will be used by subsequent step S220 to extract temperature field, flow field, and strain field control parameters and perform model binding. At the same time, the multi-field coupling simulation configuration will be referenced by S230 to generate step control and data recording settings. In addition, the multi-field coupling simulation configuration will serve as the starting point of the input link for S300 in the connection between the main steps, supporting the full-process acquisition of subsequent geometry and microstructure.

[0028] S220. Extract temperature field, flow field and strain field control parameters from the multi-field coupled simulation configuration, bind the β region dynamic recrystallization and grain growth model and the (α+β) region and cooling phase transformation model, and generate a microstructure-field variable mapping table. This step takes a multi-field coupled simulation configuration as input, which includes a staged description of temperature boundaries, contact-friction parameters, load time axis, and material field parameters. Specifically, the system first instantiates the sets of temperature field, flow field, and strain field control parameters in the field control manager. In this embodiment, the temperature field control parameters are defined as a time step sequence, a piecewise curve of heat transfer coefficient, an emissivity sequence, and a roller temperature trajectory; the flow field control parameters are defined as a main roller speed sequence, a core roller displacement velocity, and a support roller force boundary; and the strain field control parameters are defined as an equivalent strain increment threshold, a strain rate sampling interval, and a contact zone equivalent strain integral path description. Subsequently, the system traverses the stage markers within the multi-field coupled simulation configuration, constructs a field-parameter snapshot for each stage, establishes a one-to-one correspondence between the temperature field control parameters, flow field control parameters, and strain field control parameters and the stage time window, and records it in the control parameter time index table. When it is found that the combination of the main roller speed and the core roller feed slope in a certain stage causes the contact zone strain rate to exceed the recommended range, the system restricts the sampling interval in the strain field control parameters of that stage and writes the restriction information into the restriction record. Further, the system enters the microstructure model binding process. In this embodiment, the β-region dynamic recrystallization and grain growth model is defined as a procedural description of the dynamic recrystallization volume fraction evolution and grain size evolution based on the high-temperature plastic deformation path. The (α+β) region and cooling phase transformation model is defined as a procedural description of the phase fraction evolution and phase boundary morphology evolution covering the vicinity of the phase boundary, and is associated with the cooling path on the cooling channel. To ensure the executability of model binding, the system constructs a trigger condition table for the microstructure model based on the temperature trajectory of the temperature field control parameters and the equivalent strain increment threshold of the strain field control parameters. The trigger condition table provides a threshold for entering the β region, the (α+β) region, and the cooling zone at each stage. When the temperature trajectory and strain integral path meet the threshold conditions, the corresponding microstructure model's evolution equations are automatically activated and a staged switch flag is written. The core of model binding is to establish a two-way mapping relationship between microstructure variables and field variables. The system first defines a set of microstructure variables, including dynamic recrystallization volume fraction, average recrystallized grain size, β grain size, α phase morphology description, and cooling phase fraction; then it defines a set of field variables, including temperature, strain and strain rate, and contact state identifier. In this step, the mapping relationship is represented by a microstructure-field variable mapper. The mapper reads the control parameter time index table, samples the temperature, strain, and strain rate inputs required for updating the microstructure variables at each time step, and provides them to the microstructure model solver. At the same time, the mapper writes back the stage state of the microstructure variables to the field variable supplementary channel, which is used to form an integrated microstructure-field acquisition channel in the data recording settings of S230.Understandably, for the phase transformation process during the cooling stage, the system constructs a cooling path lookup table based on the segmented description of the thermal boundary configured in the multi-field coupled simulation. The heat transfer coefficient ranges and temperature drop rates corresponding to insulation, air cooling, spraying, or roller cooling are written into the lookup table, and an index is established with the cooling phase transformation model. When a stage switch triggers a change in the cooling path, the mapper reads the new heat transfer range from the lookup table and updates the input to the microstructure model. To handle anomalies in actual calculations, the system sets a boundary protector in the microstructure-field variable mapper. When the temperature sampling value exceeds the coverage range of the material data, the boundary protector triggers an extrapolation limit and records the extrapolation label. When the contact state causes the strain integration path to be interrupted, the boundary protector performs a smooth transition using the equivalent strain increment of the most recent available time step, avoiding discontinuities in microstructure variable updates. After completing the above binding, the system generates a tissue-field variable mapping table covering the entire time domain of the phase. In this embodiment, the mapping table contains structured entries including tissue variable names, corresponding field variable names, sampling-write-back strategies, trigger thresholds, and phased switch markers. In this step, the set of structured entries is recorded as the output field name tissue-field variable mapping table, which is directly called by S230 when performing step control and data recording settings for multi-field coupling simulation configuration and tissue-field variable mapping table. At the same time, the tissue-field variable mapping table serves as the field mapping basis for S300 acquisition and processing in cross-main steps, supporting the consistency of subsequent extraction and numbering from geometric and micro-organism evaluation data.

[0029] S230. Perform step control and data recording settings on the multi-field coupled simulation configuration and tissue-field variable mapping table to generate transient solution control sequence; This step uses the multi-field coupled simulation configuration and the tissue-field variable mapping table as parallel inputs. Specifically, the system first creates a time stepping manager. In this embodiment, the time step is defined as a non-uniform time increment sequence around the stage marker tissue, containing three types of nodes: normal steps, switching steps, and observation steps. The normal steps are used for regular solution progression, the switching steps are used to load boundary and load updates caused by stage switching, and the observation steps are used to enhance the synchronous sampling of tissue variables and key field variables. The time stepping manager reads the load time axis and thermal boundary segmentation description of the multi-field coupled simulation configuration, registers the time points at the stage markers as switching steps, and inserts observation steps in the corresponding intervals according to the trigger thresholds in the tissue-field variable mapping table. When the load time axis is out of sync with the thermal boundary update, the time stepping manager inserts a bridging step between them. The length of the bridging step is given by the alignment record, and a bridging label is written into the control sequence. Subsequently, the system constructs a set of control instructions for the solution process. In this step, the control instructions are defined as an ordered combination of solver state, boundary refresh, contact update, mesh regeneration, and data acquisition actions for each time step. The control instruction set first sets the solver state to thermo-mechanical coupling and contact activation for the normal step, and refreshes the thermal boundary and friction parameters in sequence. For the switching step, the boundary refresh priority is set, and contact opening and closing and local mesh regeneration are triggered after the refresh is completed. For the observation step, the data acquisition priority is set to ensure that the snapshot of the organization variable and the peak value of the contact area field variable are recorded when the evolution event occurs. Understandably, the data recording setup process reads the organization variable names and sampling-write-back strategies from the organization-field variable mapping table. Organization variable acquisition channels are created synchronously at each observation and switching step, and the sampling point selection rules and interpolation strategies for field variables are written into the channel attributes. Simultaneously, the process deploys a geometry tracker on the set of mesh nodes near the inner edge control line, outer edge control line, and end face reference surface. The geometry tracker records the geometric surrogate quantities of roundness and end face flatness in a time series. This record will be incorporated into the geometric and microstructure evaluation data when the S300 acquires the transient solution control sequence and performs multi-field coupled calculations throughout the rolling process. To enhance the robustness of the calculation process, the system inserts anomaly capture and rollback strategies into the control sequence. When element distortion exceeds a threshold and local mesh regeneration fails to recover, the system rolls back to the previous normal step and shortens the current time step length, while simultaneously reducing the main roll speed update amplitude. When the contact solution fails to converge, the normal penalty parameter is temporarily increased and the update rate of the core roll feed slope is reduced; these adjustments are written into the solution adjustment record. Furthermore, the system performs timing fine-tuning on the switching steps of the control sequence based on the stage timing revision records in the process initialization package structure to ensure consistency between the stage switching trigger source and the equipment response time delay parameters. At the same time, the system adds data compression and archiving parameters to the control sequence, stipulating that high-frequency sampling of the observation step is retained at the original precision, and ordinary steps are downsampled using a segmented averaging strategy. In this embodiment, the archiving format is defined as a time series dataset grouped by stage number and layered by variable type.After all configurations are completed, the time stepping manager combines control commands and data recording settings into an executable timing script. The script specifies the solution actions, boundary and load refreshes, contact and mesh processing, organization and field variable acquisition, and anomaly rollback rules for each time step. In this step, the timing script is recorded as the output field name "Transient Solution Control Sequence," which is directly called by S310 when it acquires the transient solution control sequence and performs multi-field coupling calculations throughout the entire roll-expansion process, serving as a unified driver for the four types of processes: thermo-mechanical-contact-organic. Simultaneously, the transient solution control sequence corresponds to the parameter updates and solution restarts of S400 in cross-main steps. After S420 updates the parameters and writes them back to the load time axis and boundary segmentation description, a new round of timing scripts can be quickly reconstructed based on the structured definition of this control sequence. In summary, the technical effects of this step are as follows: By configuring the multi-field coupled simulation and setting up the step control and data recording of the tissue-field variable mapping table, an executable timing script covering staged solution, boundary and load updates, contact and mesh processing, and tissue-field synchronous acquisition is formed, providing a stable and integrated driving foundation for subsequent calculations and evaluations.

[0030] Step S300 includes at least steps S310-S330: S310. Obtain the transient solution control sequence and perform multi-field coupled calculation of the entire rolling process. Perform temperature-strain-strain rate history acquisition and processing to obtain geometric and microstructure evaluation data. The input for this step is the transient solution control sequence formed in the previous steps. In this embodiment, the transient solution control sequence is a set of time-series scripts for each time step, including solver state, boundary and load refresh, contact and mesh processing, tissue and field variable acquisition, and anomaly rollback rules. It includes three types of nodes: normal steps, switching steps, and observation steps, and has stage marker indexes and alignment records. Specifically, the aforementioned transient solution control sequence is loaded into the computation manager as the driving input. The computation manager triggers the solution action one by one according to the type of time step: when a normal step is read, the thermo-mechanical-contact simultaneous solution is executed, and the thermal boundary and friction parameters in the multi-field coupled simulation configuration are refreshed in the current step; when a switching step is read, the boundary update is applied first, then the contact opening and closing and local mesh regeneration are executed, and the solution for the current step is advanced after the update is completed; when an observation step is read, under the condition of maintaining the same boundary and contact state as the previous normal step, the tissue-field synchronous acquisition is triggered first, and then the solution for the current step is advanced. To ensure timeline consistency, the calculation manager first reads the alignment record and concatenates asynchronous events between the load timeline and the boundary segment description through bridging steps inserted by bridging tags. The length of the bridging step is set according to the time quantization data given in the alignment record, and the bridging event is written to the runtime log after execution. Furthermore, before each time step, the calculation manager retrieves contact candidate surface groups and local heat transfer intervals based on stage markers. Within the contact candidate surface group, a combined expression of rigid-plastic forming contact and viscous-slip hybrid friction model is enabled, and the solver's internal constraints are configured with normal penalty parameters and segmented friction coefficients. Within the local heat transfer interval, convective and radiative heat transfer is expressed as constant or segmented constant loads within a time period based on the thermal boundary segment description. The roller surface temperature trajectory is refreshed in the switching step and remains frozen in the observation step for stable sampling. For mesh processing, the computation manager defines a distortion threshold in the contact zone. When the element distortion exceeds the threshold, local mesh regeneration triggered by the transient solution control sequence is automatically invoked. The nodal temperature, strain, and strain rate before regeneration are interpolated and mapped to the new mesh in the solver, and the interpolation error is evaluated as a residual. The evaluation results are written into the solver adjustment record for subsequent auditing and recalculation reference. Understandably, the temperature-strain-strain rate history acquisition and processing follows the priority arrangement of the transient solution control sequence for observation steps and switching steps: In the observation step, the system performs synchronous sampling on the channels defined in the microstructure-field variable mapping table. The temperature channel records the representative nodal temperatures near the inner edge control line, outer edge control line, and end face reference surface of the billet, using fixed time intervals for sampling and adding labels in each observation step; the strain channel records the time series of equivalent strain and equivalent strain increment, and the sampling path follows the equivalent strain integral path description in the contact zone; the strain rate channel records the equivalent strain rate trajectory of representative element points in the contact zone and near the free surface, and inserts a micro-step sampling before and after the switching step to capture abrupt changes.During the switching step, after completing boundary refresh and contact opening / closing, the system adds a forced observation, binding the temperature, strain, and strain rate snapshots of the current step with the stage marker. To maintain the continuity of geometric tracking, the geometric tracker updates the geometric surrogate quantities of roundness and end face flatness at each time step: roundness is recorded by accumulating the radius or thickness difference between the inner and outer control lines at several circumferential division points via time series, and end face flatness is recorded by the axial coordinate deviation sequence of mesh nodes near the end face reference surface; when local mesh regeneration occurs, the geometric tracker performs node mapping before and after regeneration, uses a shape function interpolation strategy to transfer historical geometric surrogate quantities to the new mesh node set, and writes the transfer error as an additional field into the time series metadata. In terms of anomaly handling, when the contact solution fails to converge within the limited iteration step, the computation manager reverts to the previous normal step according to the anomaly rollback strategy in the transient solution control sequence, shortens the current time step length, reduces the main roller speed update amplitude, and triggers an additional observation step to record the microstructure-field state before rollback. When the temperature field value exceeds the limit or is accessed outside the material data coverage range, the boundary protector enables extrapolation constraints, halves the temperature update in this time step, and records the extrapolation label simultaneously before proceeding to the next solution step. After completing the advancement and acquisition of the entire time axis, the data archiver retains the observation step at its original accuracy according to the archiving parameters of the transient solution control sequence, performs segmented averaging downsampling on the normal steps, and establishes independent data groups for each stage. Within each group, the temperature-strain-strain rate history and geometric surrogate sequence are stored hierarchically according to variable type, and all sampling entries are associated with stage markers, bridging labels, extrapolation labels, and solution adjustment records. Finally, the computation manager encapsulates the aforementioned time-series data, geometric tracing results, and microstructure-field synchronous sampling entries into a unified package, named the output field "Geometric and Microstructure Evaluation Data". This output field is provided to subsequent steps via the data bus at the end of this step. In this embodiment, the geometric and microstructure evaluation data is directly used by S320 to extract roundness, end-face flatness, strain uniformity, β grain size, and α phase morphology from the geometric and microstructure evaluation data, and to perform standardization and numbering. At the same time, it is passed internally by S300 as input across main steps, and is used in the parameter update and restart solution phase of S400 to compare the process differences between the old and new schemes, forming the computational data basis for the closed-loop link.

[0031] S320. Extract roundness, end face flatness, strain uniformity, β grain size and α phase morphology from geometric and microstructure evaluation data, perform standardization and numbering processing, and generate a set of evaluation indicators. The input for this step is the aforementioned output field of geometric and microstructure evaluation data. In this embodiment, the data includes temperature-strain-strain rate history, geometric surrogate quantity sequence, and microstructure-field synchronous sampling entries for each stage and time step, and carries stage markers, bridging labels, extrapolation labels, and solution adjustment records. Specifically, the index extraction manager first segments and merges the data according to the stage markers, and establishes a candidate sampling set within each stage's time window. In this candidate sampling set, roundness and end-face flatness are geometric indices, derived from the geometric surrogate quantity sequence of nodes near the inner edge control line, outer edge control line, and end-face reference surface recorded by the geometric tracker. The system smooths the difference sequence of radius or thickness in the circumferential direction according to predetermined division points, and extracts representative statistics based on the sampling position of the observation step within the stage. Strain uniformity is a deformation index, derived from the equivalent strain time series near the contact area and free surface. The system performs grid-weighted summarization in the stage according to circumferential sectors and radial layers to obtain a uniformity expression in the spatial-temporal dimension. For the two types of microstructure indicators, β grain size and α phase morphology, the data are derived from the microstructure-field synchronous sampling entries. The system retrieves the corresponding stage states based on the microstructure variable names in the microstructure-field variable mapping table and the sampling-write-back strategy, and then aligns them with the observation steps on the time axis to construct microstructure statistical entries within and between stages. Furthermore, to ensure consistent processing of indicators with different scales and dimensions in subsequent steps, the system enters a standardization process: for geometric and deformation indicators, the system first performs detrending processing within a stage according to a time window, and then performs interval normalization between stages; for microstructure indicators, the system first performs threshold trimming according to the material data coverage range, and then performs segmented stretching between stages, so that different stages have a consistent scaling expression within the same comparison interval; all indicators retain the original numerical links during this process, and the standardization method identifier and parameter source are recorded in the indicator entries. In terms of numbering, the system assigns a number to each indicator, which consists of a step code, a stage code, a variable code, and a time slice number. The system also records the association with stage markers, bridging labels, extrapolation labels, and solution adjustment records in the entries. When a rollback event is detected in a certain stage, the numbering process prioritizes retaining the observation step number before the rollback as a reference number and adds the observation step number after the rollback to the replacement mapping table. The replacement mapping relationship is written into the entries to ensure the traceability of subsequent comparisons.Understandably, the reliability and availability of the indicators are maintained in this embodiment through an anomaly labeling and removal strategy: when the geometric surrogate quantities of roundness and end face flatness show an abnormal peak near the switching step and the peak completely coincides with the mesh regeneration event, the system performs a smooth replacement of the statistics at that time point without deleting the original entry, and adds an anomaly label to the entry metadata; when strain uniformity shows continuous missing measurements in a certain sector and the missing measurements are associated with the contact non-convergence event, the system uses the spatial weighted average of the most recent valid time slice to fill in the missing measurements and writes a filling mark in the entry for subsequent auditing and recalculation reference. After extraction, standardization, and numbering, the system encapsulates all indicator items in a structured manner, forming a summary structure containing indicator name, number, standardization method, stage index, and metadata link, and names it as the output field name evaluation indicator set. At the end of this step, the evaluation indicator set is passed to the subsequent step S330 to configure the weights and encapsulate the evaluation indicator set in a structured manner, generating an evaluation function input structure call. At the same time, it serves as the input across the main steps for S400 to perform effect judgment and difference recording in the comparison stage after the optimization algorithm parameter setting and iteration number configuration, thus forming a stable connection with both S300 internally and S400 externally.

[0032] S330. Perform weight configuration and structured encapsulation on the evaluation index set to generate the evaluation function input structure; The input source for this step is the aforementioned output field evaluation index set. In this embodiment, this set includes multiple index items consisting of roundness, end face flatness, strain uniformity, β grain size, and α phase morphology, along with their numbers, standardization methods, stage indexes, and metadata links. Specifically, the weight configuration manager first reads the weight strategy table based on the process focus. The weight strategy table originates from the design phase's emphasis on geometric consistency and microstructure consistency, and includes allocation rules for stage and index dimensions. After reading the strategy, the system groups the evaluation index set by stage and merges geometric, deformation, and microstructure indices within each group according to a preset hierarchical relationship. During the merging process, the numbers and metadata links are not changed; only hierarchical indexes are added to the group-level entries to ensure a clear path for subsequent weight superposition. Furthermore, the system maps the weighting strategy to entries within a group: for geometric indicators, the weights of roundness and end-face flatness are assigned to their respective numbered entries; for deformation indicators, the weight of strain uniformity is redistributed according to the weighted results of the circumferential sector and radial layering grids; for microstructure indicators, the weights of β grain size and α phase morphology are segmented and weighted according to the degree of overlap between the stage thermal history and microstructure history. When there is a stage priority in the strategy, the system first assigns the total stage weight, and then proceeds to the entry-level allocation within the group. The allocation results are recorded in the weight mapping table, and a weight field is added to the metadata of each entry. To maintain the traceability of weight configuration, the system establishes a bidirectional link between the weight mapping table and the original indicator entries. When an entry is found to have an abnormal annotation or a filling mark, a weight reduction rule or exclusion rule is applied to that entry. The result of the weight reduction or exclusion is added to the entry metadata as a disposal mark, ensuring the transparency and verifiability of the weight configuration. After weight configuration is completed, the system enters a structured encapsulation process. Based on the input habits of the optimization solver, the process organizes the entries into two views: a stage sequence view and an indicator sequence view. The stage sequence view is arranged along the stage timeline, with each stage containing entry numbers, standardized values, weights, and metadata links. The indicator sequence view is arranged by indicator type, with each indicator type aggregating cross-stage numbers, standardized values, and weights. The two types of views are mutually accessible through numbers and stage indexes, and include a weight mapping table and a strategy source pointer for easy downstream calls. Understandably, the output of the structured encapsulation needs to carry the interface identifier of the optimization loop. After encapsulation, the system adds an interface field to the structure and includes the timestamp and version number of this encapsulation in the metadata. Subsequently, the system names the structure as the output field name evaluation function input structure, writes it to the data bus as the output product of this step, and directly provides it to S410 for use when setting optimization algorithm parameters and configuring the number of iterations. At the same time, the field layout of the evaluation function input structure is consistent with the parameter inversion process of S400 in the cross-main step, which facilitates S420 to write back and compare after parameter update and restart solution, forming a self-consistent evaluation-optimization closed loop.

[0033] Furthermore, to achieve a unified quantitative evaluation of the geometric accuracy and microstructure properties of ring forgings, this invention introduces a comprehensive evaluation function when generating the input structure of the evaluation function. The comprehensive evaluation function, expressed in a weighted summation form, is as follows: in: Let be the comprehensive evaluation function, and be the objective function of the optimization algorithm; These are the weighting coefficients; This refers to the roundness error (quantized value). This refers to the end face flatness error (quantized value). The coefficient of uniformity of strain; For β grain size compliance rate; The α phase morphology and content compliance rate.

[0034] Weighting coefficient to This reflects the relative importance of geometric accuracy and microstructure performance under different service scenarios, among which the weight of microstructure index is... , This evaluation function is typically assigned a high value to ensure that the optimization algorithm prioritizes improving microstructure uniformity and grain size matching during the search process. It serves as the quantification basis for the optimization algorithm's objectives, mapping multi-dimensional performance indicators to a unified evaluation space and ensuring the consistency of convergence direction between subsequent genetic algorithms and particle swarm optimization.

[0035] In summary, the technical effects of this step are as follows: By configuring and structurally encapsulating the weights of the evaluation index set, the output evaluation function input structure has a dual organizational form of stage view and index view, and carries weight mapping and metadata links, which can be directly called by subsequent optimization steps and stably reused in the closed-loop process.

[0036] Step S400 includes at least steps S410-S430: S410. Obtain the input structure of the evaluation function, perform optimization algorithm parameter setting and iteration number configuration processing, and obtain the optimization iteration control package; The input source for this step is the evaluation function input structure formed in the previous step. This structure carries two organizational forms: a stage sequence view and an index sequence view, along with a weight mapping table and a strategy source pointer. It is used to characterize the standardized values, numbers, and stage indices of various indices, such as roundness, end-face flatness, strain uniformity, β grain size, and α phase morphology. Specifically, the aforementioned evaluation function input structure is loaded into the optimization configuration manager. The optimization configuration manager first reads the weight mapping table and constructs a cumulative expression of the target evaluation metric. It establishes a one-to-one correspondence between the stage time axis in the stage sequence view and the entry numbers in the index sequence view, forming an evaluation query interface that can be called by the search engine. Subsequently, the optimization configuration manager loads stage priorities and constraint boundaries from the strategy source pointer according to process focus and equipment operation constraints, generating a constraint list. This constraint list covers the upper and lower bounds of the pre-rolled billet cross-sectional geometric parameters, the adjustable range of the rolling process parameters, and the stage sequence switching window. Furthermore, the configuration manager enters the optimization algorithm parameter setting process. In this embodiment, alternative strategies such as Genetic Algorithm (GA) or Particle Swarm Optimization (PSO) are preferred. When Genetic Algorithm (GA) is selected, the configuration manager sets the population size, crossover and change probability ranges and termination criteria according to the design variable dimensions and constraint list, and maps the stage priority to the stage weighting coefficient during fitness evaluation. When Particle Swarm Optimization (PSO) is selected, the configuration manager sets the number of particles, individual and population guiding factors and step size decay curve according to the variable dimensions and stage window, and registers the constraint list as an out-of-bounds penalty and feasible region return mechanism. All selections have clear strategy flags and version timestamp records in the running parameters. Understandably, the iteration count configuration process unfolds immediately after the algorithm parameter setting. The configuration manager derives the suggested minimum iteration count from the stage coverage and item distribution density of the evaluation function input structure, and combines this with the estimated cost and computing resource quota in the equipment database to form an allocation scheme for iteration batches and the number of evaluations per batch. When there is a narrow stage switching window, the configuration manager compresses the time arrangement of iteration batches and sets mandatory evaluation sampling points aligned with stages in the scheme to ensure complete coverage of stage-based evaluation during the iteration process. To enhance the executability of the optimization-simulation closed loop, the configuration manager also generates a solution triggering strategy that interfaces with the simulation end. In this embodiment, the triggering strategy is defined as generating a batch of solution requests and pushing them to the load time axis and boundary segmentation description interface of the multi-field coupled simulation end when the design variables are updated to a specified batch or reach the mandatory evaluation sampling point of the stage. The triggering strategy is stored as a trigger table in the configuration file and indexed with the sampling plan of the search engine.Regarding anomalies and auditing, after configuring the optimization algorithm parameters and iteration count, the configuration manager generates anomaly handling strategies and an audit flag set. The anomaly handling strategies include search radius scaling rules for convergence stagnation and penalty coefficient gain rules for frequent boundary violations. The audit flag set records the evaluation query calls, constraint hit counts, and return event frequency for each iteration. After completing the above processing, the optimization configuration manager packages the algorithm parameters, iteration batch plan, solution triggering strategy, anomaly handling strategy, and audit flag set together, and writes them, along with the evaluation query interface and weight mapping reference, into an executable control file, recording it as the output field name "Optimization Iteration Control Package." The optimization iteration control package is directly provided to subsequent steps after this step. S420 extracts the pre-rolled billet cross-section geometric parameters and rolling process parameters from the optimization iteration control package and updates and restarts the solution. Simultaneously, the optimization iteration control package maintains a consistent data interface with the evaluation chain of S300 during cross-main step connections, supporting closed-loop linkage.

[0037] S420. Extract the geometric parameters of the pre-rolled billet section and the rolling process parameters from the optimization iteration control package, update the parameters and restart the solution to generate a set of design variables. The input for this step is the aforementioned output field optimization iteration control package. This control package contains algorithm parameters, iteration batch plans, solution triggering strategies, and exception handling strategies, and the weight mapping references can be accessed through the evaluation query interface. Specifically, the optimization iteration control package is loaded into the variable orchestration manager. The variable orchestration manager first determines the range of variable subsets to be called in this round according to the iteration batch plan, and reads the stage-specific evaluation sampling points to be covered in this batch from the sampling plan of the control package. Subsequently, the variable orchestration manager parses the acceptable variable structure and time alignment requirements of the simulation end from the solution triggering strategy, and constructs a variable mapping template. In this embodiment, the variable mapping template is defined as a structured mapping relationship that maps the geometric parameters of the pre-rolled billet section and the rolling process parameters to the three-dimensional geometric parameter set, load time axis, and boundary segment description of the pre-rolled billet in the simulation end. Furthermore, the variable orchestration manager enters the parameter extraction and update process, sampling a set of candidate solutions from the algorithm parameter space of the optimization iteration control package. The geometric parameters of the pre-rolled billet section in the candidate solutions are decomposed into inner diameter, outer diameter, wall thickness distribution curve, angle range of local thickening section, end face chamfer range, and clamping reference surface description. The rolling process parameters are also decomposed into staged descriptions of mandrel feed curve, main roll speed, and initial heating temperature. For each candidate solution, the variable orchestration manager writes the geometric parameters of the pre-rolled billet section into the corresponding fields of the three-dimensional geometric parameter set of the pre-rolled billet according to the variable mapping template, writes the rolling process parameters into the corresponding fields of the load time axis and boundary segment description, and generates a solution request entry aligned with the stage-required evaluation sampling point. Understandably, the parameter update and solution restart process is divided into three stages: pre-verification, synchronous distribution, and result registration. In the pre-verification stage, the variable orchestration manager performs constraint list comparison on the candidate solutions. If a variable is found to be out of bounds, it is returned to the nearest boundary according to the feasible domain return mechanism in the control package, and the out-of-bounds and return labels are recorded. In the synchronous distribution stage, the variable orchestration manager combines several candidate solutions into a distribution batch according to the batch threshold of the solution triggering strategy. The updated values ​​of the three-dimensional geometric parameter set of the pre-rolled billet, the load time axis, and the boundary segment description are written to the remote configuration through the interface of the multi-field coupled simulation end. Then, a solution restart request is triggered and the restart timestamp is registered. In the result registration stage, the variable orchestration manager waits for the simulation end to return the reference pointer of the geometric and microstructure evaluation data of this batch. The pointer is then mapped to the variables of this batch to facilitate the subsequent evaluation query interface connection. Regarding anomalies and adaptation, when consecutive out-of-bounds resubmission events occur in a batch, the variable orchestration manager calls the anomaly handling strategy to increase the penalty coefficient gain and reduces the search radius in the next batch. When a candidate solution is backed up in the simulation due to non-convergence of contact, the variable orchestration manager automatically reduces the update amplitude of the main roller speed associated with the candidate solution and records the solution adjustment mark in the variable entry to ensure that the data source can be distinguished during subsequent comparisons.After completing a batch distribution and result registration, the variable arrangement manager summarizes the variable entries of this batch, forming a structured variable table with stage index, out-of-bounds and return tags, solution adjustment markers and timestamps, and names the variable table as the output field name design variable set. At the end of this step, the design variable set is directly passed to subsequent steps for invocation. S430 performs convergence judgment and scheme freezing on the design variable set. At the same time, the design variable set is interconnected with the multi-field coupling calculation results of the entire process of S310 in cross-main steps through pointer mapping, which facilitates the cyclical advancement of optimization-simulation-evaluation.

[0038] S430. Perform convergence determination and scheme freezing on the design variable set, and generate the optimized scheme package structure; The input source for this step is the aforementioned output field design variable set. The design variable set contains batch-organized variable entries and their mapping pointers to geometric and micro-organization evaluation data, along with stage indexes, out-of-bounds and resubmission labels, solution adjustment markers, and timestamps. Specifically, the design variable set is loaded into the convergence decision manager. The convergence decision manager first obtains the standardized index value and weight corresponding to each variable entry through the evaluation query interface, aligns the stage sequence view and the index sequence view on the same time axis, forming an evaluation snapshot set suitable for this round of decision-making. Subsequently, the convergence decision manager constructs a convergence detection process based on the termination criteria set in the optimization iteration control package. In this embodiment, convergence detection is defined as a multi-condition judgment process that simultaneously examines the cross-batch improvement magnitude, intra-batch dispersion, and stage coverage completeness of the evaluation snapshot set. When the improvement magnitude is less than a threshold and the dispersion is less than a threshold, and the stage coverage completeness meets the strategy requirements, a convergence event is triggered. When incomplete stage coverage or excessive dispersion occurs, a continue iteration event is triggered, and a correction suggestion is given for the sampling plan of the next batch. Furthermore, the convergence decision manager differentiates the processing of entries with out-of-bounds and return tags and solution adjustment marks: for variable entries that frequently involve out-of-bounds and return, a weight reduction or elimination strategy is adopted to prevent outliers from causing deviations in the decision results; for entries with solution adjustment marks due to non-convergence, a neighborhood substitution strategy is adopted, using the average index of adjacent entries in the same batch to replace the missing or outlier values ​​of the entry, and retaining the substitution source in the audit record. After the convergence event is triggered, the convergence determination manager enters the scheme freezing process. In this embodiment, scheme freezing is defined as the process of selecting representative items from the design variable set and fixing the corresponding pre-rolled billet cross-sectional geometric parameters and rolling process parameters, while simultaneously solidifying the associated stage timing and boundary segmentation descriptions. When the scheme is frozen, the manager first reads the audit mark set and weight mapping table to confirm that the evaluation contribution and weight distribution of the selected items in all stages meet the strategy requirements. Then, the pre-rolled billet three-dimensional geometric parameter set, load time axis, and boundary segmentation description in the item are promoted from the candidate state to the frozen state, and a version number is assigned to the frozen item. The version number consists of a date code, an equipment code, and a revision number, which facilitates the tracing of history. Understandably, the archived output after the scheme is frozen is executed by the scheme solidifier. The scheme solidifier packages the three-dimensional geometric parameter set of the pre-rolled billet, the core roll feed curve, the main roll speed and the initial heating temperature in the frozen state, together with the stage timing, boundary segment description, weight mapping reference and audit record, into a structured resource containing four types of sub-packages: geometry, process, timing and audit. The version number and timestamp are written into the metadata. In this embodiment, this structured resource is named the output field name optimization scheme package structure.

[0039] Furthermore, in the entire multi-field coupled modeling and optimization design process, this invention constructs an integrated framework for the entire process from pre-rolled billet design to microstructure control, based on the material characteristics and forming rules of titanium alloy ring forgings.

[0040] Specifically, it includes the following steps: (1) Parametric modeling and process initialization of pre-rolled billet: Based on the three-dimensional model of the target ring, a parametric pre-rolled billet model is designed. For titanium alloy materials, the heating specifications of the billet (heating to the β phase region or (α+β) two-phase region), holding time, rolling temperature range and cooling method (such as air cooling, wind cooling, etc.) are specified to provide a thermal path reference for process initialization.

[0041] (2) High-fidelity simulation of titanium alloy-specific multi-field coupling: A high-precision multi-field coupling finite element model including phase transformation effect is established to realize the synchronous coupling calculation of the three macroscopic fields (temperature field, flow field, strain field), and the flow stress is dynamically updated as a function of temperature, strain and strain rate. A titanium alloy microstructure evolution model is introduced: When the unit temperature is higher than the β transformation temperature, dynamic recrystallization and grain growth process are triggered; when the temperature is lower than the β transformation temperature, the spheroidization and precipitation behavior of the primary α phase is predicted; during the cooling stage, the precipitation of the secondary α phase is simulated based on the TTT / CCT curve or the phase transformation kinetic model, realizing the strong bidirectional coupling between microstructure evolution and flow properties.

[0042] (3) Comprehensive evaluation of geometric and microstructure properties: After the simulation is completed, the system automatically extracts geometric and microstructure properties to comprehensively evaluate the roundness, end face flatness, β grain size, primary α phase morphology and microstructure uniformity of the ring. The weight of the comprehensive evaluation function F is tilted towards the microstructure properties, so that the optimization objective is performance-oriented.

[0043] (4) Intelligent optimization iteration for microstructure control: Taking the comprehensive evaluation function F as the optimization target, the design variables are searched in multiple dimensions using global optimization strategies such as genetic algorithms. The optimization variables include the geometry of the pre-rolled billet and temperature-strain path parameters (such as the deformation of the β phase region, the final rolling temperature and the cooling rate) to achieve process inversion with controllable microstructure.

[0044] (5) Optimal solution output: The final output is the optimal pre-rolled billet model and complete rolling process specification that match the service requirements of the ring, realizing "tailor-made" forming control, so that the geometric accuracy and microstructure of the product meet the standards in one go.

[0045] In terms of closed-loop integration, the optimized solution package structure is fed back to S110 and S120 as a reference baseline for the parameterized definition of the pre-rolled billet section and the initial parameter configuration of the rolling process path, for the rapid start-up of the next round of projects or variants. On the other hand, this structure can directly replace the process initialization package structure in the construction of multi-field coupled simulation configuration during mesh generation, thermal boundary and contact boundary setting in S210, supporting rapid recalculation or acceptance verification. At the same time, the optimized solution package structure forms a stable interface between S400 and S300. When subsequent comparison verification or small-scale re-optimization is required, an audit branch can be derived without changing the version number, ensuring the consistency of traceability and reuse.

[0046] In summary, the technical effects of this step are as follows: through convergence determination and scheme freezing, the output optimized scheme package structure solidifies the geometric, process and timing information in the same version, and establishes a clear mapping with the evaluation and simulation link, which facilitates reuse and iterative expansion in the closed-loop system.

Claims

1. A method for designing process paths for difficult-to-deform aerospace ring forgings based on multi-field coupling, characterized in that, include: Obtain the 3D model and volume constraints of the target ring forging, perform parameterized definition of the pre-rolled billet section, boundary dimension configuration and process initialization consistency check, and generate the process initialization package structure; Obtain the process initialization package structure, perform mesh generation, thermal boundary and contact boundary setting, microstructure-field variable model binding and step control setting, and generate transient solution control sequence; The transient solution control sequence is acquired and multi-field coupled calculations of the entire rolling process are performed. Temperature-strain-microstructure history is collected, evaluation indexes are extracted and weights are configured, and the input structure of the evaluation function is generated. Obtain the input structure of the evaluation function, perform optimization algorithm parameter setting, parameter update and restart solution, convergence judgment and scheme freezing processing, and generate the optimization scheme package structure.

2. The method according to claim 1, characterized in that, The parameterization definition of the pre-rolled billet section includes: The parameterized definition of the pre-rolled billet section includes creating a pre-rolled billet section skeleton in the parameterized modeling environment, consisting of a neutral layer reference circle, inner edge control lines, and outer edge control lines. Independent parameter groups are set in the radial direction, axial direction, and circumferential direction. Volume constraints are associated through the constraint solver to form the solution objective. When the combination of parameter groups causes the volume to deviate from the volume constraint tolerance, automatic correction is triggered and the local thickening amplitude of the parameter group in the circumferential direction is adjusted first.

3. The method according to claim 1, characterized in that, The process of configuring boundary dimensions includes: The boundary dimension configuration includes forming a list of equipment capacity constraints based on the forming equipment capacity table, the clamping space of the core roller and main roller, the effective working area of ​​the heating furnace and the maximum opening of the rolling mill, and forming a list of process safety margins based on material specifications and tooling assembly tolerances. The boundary dimensions are checked item by item for the parameter group obtained by the parameterization definition. When the minimum wall thickness is found to be lower than that specified in the safety margin list, the process reverts to the parameterization solver and locks the initial value of the inner diameter while adjusting the initial value of the outer diameter.

4. The method according to claim 1, characterized in that, The mesh generation process includes: Mesh generation involves partitioning the annular section according to its thickness gradient, the angle range of locally thickened sections, and the chamfer range of the end face. Variable pitch sector partitioning is used in the circumferential direction, a layered refinement strategy is used in the radial direction, and local refinement and buffer bands are set in the axial transition area of ​​the end face. Adaptive local re-partitioning is triggered by the element twist degree and the minimum interior angle threshold.

5. The method according to claim 1, characterized in that, The process of setting thermal boundaries and contact boundaries includes: The thermal boundary setting includes applying a segmented convection-radiation composite boundary to the outer surface of the billet based on the initial heating temperature and the transport-heat preservation sequence mark, and applying a time-transformed thermal boundary to the roller surface based on the roller temperature and the state of the cooling medium. The contact pressure-related local heat transfer coefficient of the clamping reference surface area is written into the thermal boundary mapping table and a boundary update event is inserted at the stage switching point. The contact boundary setting includes generating a group of candidate contact surfaces from the geometric contours of the core roller, main roller, and support roller provided by the equipment database. The contact opening and closing sequence is driven by the stage markers in the process initialization package structure. The contact form is defined as a combination of rigid-plastic forming contact and viscous-slip hybrid friction model. The friction coefficient and normal penalty parameters are retrieved from the high-temperature tribological correlation parameter table and assigned values ​​in stages. Mesh smoothing and local regeneration strategies are introduced in the contact area to handle element distortion.

6. The method according to claim 1, characterized in that, The process of collecting temperature-strain-structure history data includes: The process acquisition includes synchronous sampling of channels defined by the field variable mapping table in the observation step. The temperature channel records the representative node temperatures near the inner edge control line, outer edge control line, and end face reference surface of the billet. The strain channel records the equivalent strain and equivalent strain increment time series described by the equivalent strain integral path in the contact zone. The strain rate channel records the equivalent strain rate trajectory in the contact zone and near the free surface. In the switching step, a forced observation is added to bind the stage marker and field variable snapshot.

7. The method according to claim 1, characterized in that, The process of extracting evaluation indicators and configuring weights includes: The evaluation index extraction includes segmenting and merging the data according to the stage markers, extracting geometric surrogate statistics of roundness and end face flatness for each stage, strain uniformity weighted by circumferential sector and radial layered grid, and β grain size and α phase morphology stage state retrieved from the synchronous sampling entries of the organization field. The weight configuration includes reading the weight strategy table based on the key process focus, grouping the evaluation index set by stage, merging geometric, deformation and microstructure indices within the group according to a preset hierarchical relationship, mapping the weight strategy to the entries within the group, redistributing the weight of strain uniformity according to the weighted results of circumferential sector and radial layered grid, and assigning segmented weights to β grain size and α phase morphology according to the degree of overlap between the stage thermal history and microstructure history. When an entry has an anomaly label or filling mark, a weight reduction or exclusion rule is applied.

8. The method according to claim 1, characterized in that, The process of optimizing algorithm parameter settings includes: The optimization algorithm parameter settings include adopting genetic algorithms or distributed particle swarm optimization strategies, setting the population size, crossover and mutation probability intervals or particle number, individual and group guiding factors according to the design variable dimensions and constraint list, mapping the stage priority to the stage weighting coefficient during fitness evaluation, and deriving the minimum iteration rounds based on stage coverage and item distribution density, and combining estimated costs and computing resource quotas to form an iterative batch allocation scheme.

9. The method according to claim 1, characterized in that, The process of parameter update and restarting the solution includes: The parameter update and restart solution process includes sampling candidate solutions from the parameter space of the optimization algorithm, decomposing the geometric parameters of the pre-rolled billet section into inner diameter, outer diameter, wall thickness distribution curve, angle range of local thickened sections, end face chamfer range, and clamping reference surface description, and writing them into the corresponding fields of the three-dimensional geometric parameter set of the pre-rolled billet. The rolling process parameters are decomposed into the core roll feed curve, main roll speed, and initial heating temperature staged description, and written into the corresponding fields of the load time axis and boundary segmentation description. Solution request entries aligned with the stage-reviewed sampling points are generated, and a constraint list comparison is performed on the candidate solutions. If a variable goes out of bounds, it is pushed back to the nearest boundary inside according to the feasible domain pushback mechanism.

10. The method according to claim 1, characterized in that, The convergence determination and scheme freezing process includes: Convergence determination and scheme freezing involve obtaining the standardized index values ​​and weights corresponding to the variable entries through the evaluation query interface, forming an evaluation snapshot set, and simultaneously examining its cross-batch improvement magnitude, same-batch dispersion and stage coverage completeness for multi-condition judgment. When the improvement magnitude and dispersion are less than the threshold and the stage coverage completeness meets the strategy requirements, a convergence event is triggered. The scheme freezing process involves selecting representative entries from the design variable set and fixing their pre-rolled billet cross-sectional geometry parameters and rolling process parameters, while solidifying the associated stage sequence and boundary segment descriptions, and assigning a version number consisting of a date code, equipment code, and revision number to the frozen entries.

Images