Additive-welding integrated manufacturing cooperative control system for TA15 titanium alloy material aviation complex component

Abstract

The invention relates to an additive-welding integrated manufacturing cooperative control system for a TA15 titanium alloy material aviation complex component. The additive-welding integrated manufacturing cooperative control system comprises a state sensing module, a control module and a control module, wherein the state sensing module is used for collecting multi-source physical process state data of the aviation complex component in the welding increasing process; the state modeling module is used for performing fusion modeling to obtain a high-dimensional state data set; the strategy generation module is used for constructing a multi-source cooperative control strategy based on the high-dimensional state data set; the anomaly identification module is used for identifying an abnormal action under a multi-source cooperative control strategy and inserting physical action compensation; the action coordination module is used for generating a physical action instruction based on a multi-source coordination control strategy and physical action compensation and performing dynamic coordination; the auxiliary control module is used for matching and correcting physical action instructions of historical tasks according to the working condition similarity when a new task is switched; and the optimization updating module is used for collecting physical action instructions to carry out performance evaluation and carrying out feedback optimization on a multi-source cooperative control strategy after each task is finished.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of advanced aviation manufacturing equipment and its online intelligent control technology, in particular to a kind of additive-welding integrated manufacturing collaborative control system for TA15 titanium alloy material aviation complex component. BACKGROUND

[0002] With the continuous evolution of aviation component manufacturing towards complex curved surface, integrated forming and multi-functional integration, traditional segmented processing and multi-process assembly mode gradually exposes many bottlenecks in geometric precision, metallurgical quality and manufacturing efficiency. Especially in the manufacturing process of large-size complex components, thin-walled dissimilar material joint structures and multi-curvature channel parts, there are problems such as uneven heat input, path deformation accumulation, and difficulty in molten pool control, which seriously restrict the component forming quality and manufacturing stability. Large-size complex component manufacturing represented by TA15 titanium alloy is particularly prominent. The traditional "additive-transformation welding" mode is difficult to realize unified regulation of the whole process of additive and welding due to the fragmentation of heat field control, clamping precision error and process continuity interruption, which leads to the accumulation of residual stress under the action of multiple thermal cycles, and further causes quality risks such as structural deformation, crack initiation or interface strength fluctuation.

[0003] Most of the current equipment still adopts the mode of separate control of functional units on the process chain, lacking the mechanism of energy fusion, path coordination and state compensation at the system level. Although some additive platforms have integrated welding units, they usually only exist as an attached function, lacking the ability to fine control the dissimilar material connection area, continuous corner area or boundary heat accumulation area. In addition, the molten pool state monitoring means is limited, mainly focusing on single-point temperature feedback or light intensity estimation, which cannot realize process dynamic feedback and closed-loop regulation under multi-source perception. This control mode is difficult to cope with problems such as unstable molten pool, insufficient interface wetting or uncontrollable metallurgical defects in the manufacturing of high complexity components, affecting the consistency and repeatability of the forming process.

[0004] Especially in the manufacturing of typical aviation structures such as wing rib frames, cabin frame or engine shell type components, including the use of TA15 titanium alloy material, not only the problem of complex path and space reconstruction exists, but also the coupling of weld forming quality, component residual stress field and overall structure size control is required. Under the traditional mode, the factors such as path execution lag, heat source switching mutation and single cooling control are superimposed, which can easily lead to component warping, interlayer debonding or mechanical property degradation. Due to the lack of multi-mode physical state perception and real-time dynamic compensation mechanism, the existing manufacturing system often relies on empirical parameter selection and manual path intervention when processing such components, and the manufacturing process lacks intelligent and adaptive ability, which seriously restricts the technical breakthrough of integrated manufacturing of aviation components with high quality and high efficiency. In summary, an integrated additive-welding manufacturing system is needed, which can realize the coordinated regulation of additive and welding processes in a unified platform, has the ability of multi-source state perception, dynamic path correction and thermal field stability control, especially for multi-curvature, dissimilar material connection and high residual stress sensitive structure, to realize the system level collaborative optimization of manufacturing precision, metallurgical quality and forming efficiency, to meet the complexity and flexible adaptation requirements of new generation aviation component manufacturing. SUMMARY

[0005] The purpose of the present application is to solve the problems of TA15 titanium alloy material aviation complex component in the process of integrated additive and welding manufacturing, such as uneven heat input, poor interlayer structure continuity, redundant energy accumulation in path splicing area, and difficulty in multi-source process coupling regulation, and to provide an integrated additive-welding manufacturing collaborative control system for TA15 titanium alloy material aviation complex component.

[0006] To achieve the above purpose, the present application provides the following scheme:

[0007] An integrated additive-welding manufacturing collaborative control system for TA15 titanium alloy material aviation complex component, comprising:

[0008] A state perception module for collecting multi-source physical process state data of aviation complex structure in the process of additive welding process;

[0009] A state modeling module for modeling the multi-source physical process state data to obtain a high-dimensional state data set;

[0010] A strategy generation module for constructing a multi-source collaborative control strategy based on the high-dimensional state data set;

[0011] An abnormality identification module for identifying abnormal actions in the additive welding process under the multi-source collaborative control strategy and inserting physical action compensation;

[0012] An action coordination module is configured to generate physical action instructions based on the multi-source collaborative control strategy and physical action compensation, and to perform dynamic coordination on the physical action instructions by using dynamic conflict detection.

[0013] An auxiliary control module is configured to switch a new task, match physical action instructions of historical tasks according to working condition similarity, and correct by using real-time physical feedback.

[0014] An optimization updating module is configured to collect physical action instructions to perform performance evaluation when each task ends, and to perform feedback optimization on the multi-source collaborative control strategy according to the evaluation result.

[0015] Optionally, the fusion modeling of the multi-source physical process state data comprises:

[0016]

[0017] wherein, X represents a process state vector, W1-W5 are adjustable weighting coefficients, respectively. represents a nonlinear activation function, T m is a molten pool temperature, d l is an interlayer penetration depth, w s is a weld width, v s is a path speed, and Δz is a component thermal deformation amount.

[0018] Optionally, the multi-source collaborative control strategy comprises:

[0019] A path control unit is configured to perform real-time vector compensation of an additive path or a welding track based on component structure offset.

[0020] A laser heat source control unit is configured to perform adaptive adjustment of laser energy density based on real-time temperature feedback.

[0021] A cooling control unit is configured to perform dynamic duty cycle adjustment based on thermal load distribution.

[0022] Optionally, the path control unit performing real-time vector compensation of an additive path or a welding track based on component structure offset comprises:

[0023]

[0024] s corr = Δs + α z · Δz.

[0025] wherein, Δs represents a path fine adjustment correction vector, s ref represents a target path position vector, s act represents a current actual path position vector; K p is a proportional adjustment coefficient of position offset; K iFor integral adjustment coefficient, s corr For final vector compensation path instruction, Δz is component space displacement error, α z For path influence coefficient of Δz.

[0026] Optionally, the laser heat source control unit includes the following steps for laser energy density adaptive adjustment based on real-time temperature feedback:

[0027] Q in = P laser · τ scan + K T · ΔT;

[0028] Wherein, Q in represents the actual heat input energy per unit scanning path segment, P laser is the current laser output power, τ scan is the scanning time length of the current path; ΔT is the difference between the current molten pool center temperature T meas and the target set temperature T set , and K T is the temperature adjustment gain coefficient.

[0029] Optionally, the cooling control unit includes the following steps for dynamic duty cycle adjustment based on heat load distribution:

[0030] Calculate the heat removal target value per unit time, control the heat removal target value to reach the preset temperature control requirement by dynamically adjusting the cooling control duty cycle within the preset optimal duty cycle range;

[0031] Wherein, the cooling control duty cycle is:

[0032]

[0033] The heat removal target value is:

[0034]

[0035] Wherein, λ c is the cooling control duty cycle, t cool is the actual cooling duration within the manufacturing cycle, t cycle is the manufacturing cycle, ΔQ represents the heat removal target value per unit time, γ1 and γ2 are empirical adjustment coefficients, is the temperature rise rate at the current component position, q h is the heat flux density at the current component position.

[0036] Optionally, identifying abnormal actions in the increased welding process under the multi-source collaborative control strategy includes:

[0037] The previous stage action execution result is compared with the current state vector by calculating a temperature deviation residual vector, a path state disturbance function, a layer height mutation index and a redundant heat accumulation trend quantity, and an abnormal action and corresponding abnormal type and abnormal level are determined according to a comparison result.

[0038] Optionally, dynamically coordinating the physical action instruction by using the dynamic conflict detection comprises:

[0039] Based on the component topology structure, path scheduling relationship and inter-module coupling logic, an action conflict graph is constructed in real time, all edge sets in the conflict graph are traversed, and laser power-path trajectory coupling conflict, support load-trajectory displacement non-synergistic conflict and cooling airflow-power coupling disturbance are detected and corrected.

[0040] Optionally, the physical action instruction is collected for performance evaluation, and a multi-source synergistic control strategy is feedback optimized according to an evaluation result.

[0041] A residual evaluation vector is constructed by integrating encoding of a molten pool thermal state, interlayer geometric accuracy, an actual path execution trajectory and component deformation behavior.

[0042] When any component in the residual evaluation vector exceeds a preset upper limit of tolerance, a multi-target coordination compensation strategy is executed to complete optimization of the multi-source synergistic control strategy.

[0043] Optionally, the multi-target coordination compensation strategy comprises:

[0044] If a molten pool temperature relative deviation is greater than a preset upper limit of molten pool temperature relative deviation, a current laser output power is reduced; if a spatial geometric offset is greater than a preset upper limit of spatial geometric offset, a path displacement compensation amount is adjusted; if an interlayer melt depth target error is greater than a preset upper limit of interlayer melt depth target error, a laser power or scanning speed is increased; and if a path speed execution error is greater than a preset upper limit of path speed execution error, a path trajectory beat is re-planned.

[0045] The present application has the following beneficial effects:

[0046] (1) The application realizes the integration of additive and welding process segments, and closed-loop control, significantly improving the consistency of the overall forming of the component and the continuity of the manufacturing process. Traditional complex component manufacturing often models and executes additive and welding processes separately, which is obviously fragmented in terms of path continuity, heat input balance and microstructure transition control, making it difficult to ensure quality consistency in high complexity structures. The application establishes a unified multi-source state perception system and a cross-segment control mechanism to integrate energy input, topography formation in the additive process and fusion behavior, thermal deformation management in the welding process into a coordinated execution unit. Through path segment identification and energy density dynamic control strategy, the smooth matching of laser power, scanning speed and cooling rhythm in the additive-welding transition zone is realized, avoiding the problems of microstructure discontinuity, forming deviation or structure instability caused by process switching, and enhancing the physical coherence and structural homogeneity of the manufacturing process.

[0047] (2) The application constructs a parameter adaptive adjustment mechanism based on the fusion of process physical state vector, which significantly enhances the system's adaptability and manufacturing stability under different component tasks and various process conditions. In traditional manufacturing processes, different workpieces often require different path strategies and energy configurations due to size characteristics, thermal sensitivity or changes in assembly interface position. Existing systems rely on manual parameter adjustment or offline test debugging, which is low in response efficiency and high in error rate. The physical state vector fusion model proposed in the application can fuse key process parameters such as molten pool temperature, penetration, weld width and scanning speed into a unified process state code, and realize accurate representation of the process state through nonlinear mapping. On this basis, the system can quickly complete the automatic matching of path parameters and energy configuration according to the similarity between the current process state and historical manufacturing data, supporting rapid deployment and strategy reuse during task switching. This mechanism significantly reduces the debugging time and trial-and-error cost, improves the dynamic response efficiency and manufacturing accuracy of the system in multi-component, multi-process and variable task scenarios.

[0048] (3) The application introduces cross-module temperature-path coupling compensation and partitioned thermal control mechanism, effectively alleviates the problem of welding area heat accumulation and deformation instability, and improves the control level of structural residual stress. In the integrated manufacturing of complex components, the problems of long path, concentrated heat input and complex cooling path are often encountered, which can easily cause overheating, shrinkage stress concentration and morphology distortion in the path overlap area or splicing area. The application proposes a partitioned thermal control model based on path segment identification, which divides the manufacturing path into stable area, transition area and splicing sensitive area, and assigns different energy density and scanning strategy according to the area; and cooperates with adjustable cooling strategy to provide flexible cooling control in the corner of the component, the splicing segment or the multi-layer stacking area, realizes the dynamic balance of heat source input and heat diffusion. In addition, the system can sense the molten pool temperature and morphology trend in real time, trigger the micro compensation mechanism based on path-energy coupling, effectively eliminate the structural deviation and deformation risk caused by heat stress accumulation, and ensure the structural stability and morphology consistency of the manufactured components from the source. BRIEF DESCRIPTION OF DRAWINGS

[0049] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed in the embodiments. Obviously, the drawings described below are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0050] Figure 1 A work flow chart of the additive-welding integrated manufacturing collaborative control system for TA15 titanium alloy material aviation complex components. DETAILED DESCRIPTION

[0051] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0052] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the present application will be further described in detail below with reference to the drawings and specific embodiments.

[0053] Embodiment one:

[0054] The embodiment provides an additive-welding integrated manufacturing collaborative control system for TA15 titanium alloy material aviation complex components, which comprises:

[0055] A state sensing module is used to collect multi-source physical process state data of the aviation complex structure during the additive-welding process.

[0056] a state modeling module configured to fuse and model the multi-source physical process state data to obtain a high-dimensional state data set;

[0057] a strategy generation module configured to construct a multi-source collaborative control strategy based on the high-dimensional state data set;

[0058] an abnormality identification module configured to identify abnormal actions in the welding process under the multi-source collaborative control strategy and insert physical action compensation;

[0059] an action coordination module configured to generate physical action instructions based on the multi-source collaborative control strategy and the physical action compensation and perform dynamic coordination on the physical action instructions by using dynamic conflict detection;

[0060] an auxiliary control module configured to match physical action instructions of historical tasks according to work condition similarity when switching to a new task and correct the physical action instructions by using real-time physical feedback;

[0061] an optimization update module configured to collect physical action instructions to perform performance evaluation when each task is completed and perform feedback optimization on the multi-source collaborative control strategy according to the evaluation results.

[0062] Specifically, the multi-source process state data set includes a molten pool temperature T m , an interlayer penetration depth d l , a weld width w s , a path speed v s , and a component thermal deformation Δz. The system constructs a process state vector X based on the data set to describe the instantaneous physical working condition and topography evolution state in the additive-welding integrated manufacturing process.

[0063] The fusion process is based on feature reconstruction and dynamic weight adjustment mechanism of multi-sensor data, adopts weighted linear superposition and nonlinear activation function to constitute a mapping relationship, and fuses and models the multi-source physical process state data, including:

[0064]

[0065] wherein X represents the process state vector, W1-W5 are adjustable weighting coefficients; represents a nonlinear activation function, T m is the molten pool temperature, d l is the interlayer penetration depth, w s is the weld width, v s is the path speed, and Δz is the component thermal deformation.

[0066] The fusion state vector X supports the early identification and determination of abnormal states such as insufficient penetration, unstable path, and unbalanced heat transport in the manufacturing process, and serves as a key input variable for the process intervention action generation module, providing a basis for subsequent energy regulation, path optimization, and feedback correction.

[0067] Further, the multi-source collaborative control strategy includes:

[0068] A path control unit for real-time vector compensation of additive paths or welding trajectories based on component structure offset;

[0069] A laser heat source control unit for adaptive adjustment of laser energy density based on real-time temperature feedback;

[0070] A cooling control unit for dynamic duty cycle adjustment based on heat load distribution.

[0071] The path control unit has the ability to compensate for real-time vectors based on component structure offset, and can dynamically fine-tune additive paths or welding trajectories to suppress local path deviations caused by uneven heat input, equipment hysteresis, or platform inertia.

[0072] The path compensation control model uses a proportional-integral type offset control method, and the offset correction amount Δs is calculated from the instantaneous vector difference between the current detected path and the reference planned path, and the control expression is:

[0073]

[0074] Where Δs represents the path fine-tuning correction vector, s ref represents the target path position vector set in the system, s act represents the current actual path position detected by the sensor; K p is the proportional adjustment coefficient of position offset, used to quickly respond to deviation changes; K i is the integral adjustment coefficient, used to offset slow time-varying errors.

[0075] To ensure accurate transition of the multi-axis platform under complex path conditions, the system further introduces a compensation feedback mechanism for component space displacement error Δz, which is specifically expressed as:

[0076] s corr = Δs + α z · Δz;

[0077] Where s corr is the final generated vector compensation path instruction, α z is the path influence coefficient of space deformation Δz, used to characterize the disturbance degree of deformation on trajectory control, and ensure that the system can still maintain path execution accuracy and weld seam continuity in the presence of thermal deformation.

[0078] The path fine-tuning and displacement feedback coupling mechanism can trigger trajectory correction actions in advance when the system detects component deformation trends, avoid the accumulation and diffusion of micro-scale errors in continuous manufacturing processes, and improve the spatial consistency and connection robustness of component shaping.

[0079] The laser heat source control unit has a laser energy density adaptive adjustment mechanism based on real-time temperature feedback, which can dynamically respond to heat input under different working conditions and inhibit overburning or insufficient fusion phenomena.

[0080] The laser energy density set value Q in Based on the thermal feedback error ΔT and the interlayer scanning time τ scan Dynamic calculation, the control model expression is:

[0081] Q in = P laser ·τ scan + K T ·ΔT;

[0082] Where Q in represents the actual heat input energy per unit scanning path segment, P laser is the current laser output power, τ scan is the scanning time of the path segment; ΔT is the difference between the current molten pool center temperature T meas and the target set temperature T set , that is:

[0083] ΔT = T set -T meas ;

[0084] K T is the temperature regulation gain coefficient, which is used to adjust the system laser power distribution according to the molten pool thermal field feedback.

[0085] Through the energy density self-regulation mechanism, the system can stably control the molten pool size and energy coverage area under complex thermal conditions, and ensure the consistency of weld structure and interlayer metallurgical quality.

[0086] The cooling control unit has a dynamic duty cycle adjustment mechanism for thermal load distribution, which can adaptively adjust the cooling gas on-off rhythm according to the local thermal accumulation state of the component, and realize balanced regulation and control of the component thermal field.

[0087] The system takes the manufacturing period t cycle as the time reference, defines the cooling control duty cycle λ c , and the following:

[0088]

[0089] Where λc denotes the time proportion of the cooling unit being on in a single cycle, t cool denotes the duration of actual cooling in the cycle.

[0090] The system combines the temperature rise rate at the current component position and the heat flux density q h to comprehensively build a temperature control demand function:

[0091]

[0092] wherein γ1, γ2 are empirical adjustment coefficients, and ΔQ represents the target value of heat removal per unit time. The system maps the optimal duty cycle interval according to ΔQ and the preset cooling capacity library and dynamically adjusts the cooling control strategy.

[0093] This mechanism is particularly suitable for high-density heat input sections (such as narrow gap welding areas, path intersection areas), and can effectively suppress thermal instability, material overheating, and weld ablation without interfering with the main energy input.

[0094] Further, identifying abnormal actions in the welding process under the multi-source collaborative control strategy includes:

[0095] By calculating the temperature deviation residual vector, path state disturbance function, layer height mutation index, and redundant heat accumulation trend, the execution results of the previous stage actions are compared with the current state vector, and the abnormal actions and corresponding abnormal types and abnormal levels are determined according to the comparison results.

[0096] Further, dynamically coordinating the physical action instructions using the dynamic conflict detection includes:

[0097] Based on the component topology structure, path scheduling relationship, and inter-module coupling logic, an action conflict graph is constructed in real time, all edge sets in the conflict graph are traversed, and laser power-path trajectory coupling conflicts, support load-trajectory displacement non-collaborative conflicts, and cooling airflow-power coupling disturbances are detected and corrected.

[0098] Further, when switching to a new task, the manufacturing initial parameters are generated through a similarity matching mechanism, and the current process state vector is Y cur , the historical working condition vector set is Y hist , and the matching process is defined as:

[0099]

[0100] wherein D(·) is a process state distance function, which can use Euclidean distance or cosine similarity, and Y * is used to quickly generate initial parameter configurations for integrated welding manufacturing.

[0101] Further, the physical action instruction is collected for performance evaluation, and a feedback optimization is performed on the multi-source collaborative control strategy according to an evaluation result, and the feedback optimization comprises:

[0102] The molten pool thermal state, the interlayer geometric accuracy, the actual execution trajectory of the path and the component deformation behavior are integrated and coded to construct a residual error evaluation vector.

[0103] When any component in the residual error evaluation vector exceeds a preset upper limit of tolerance, a multi-objective coordination compensation strategy is executed to complete the optimization of the multi-source collaborative control strategy.

[0104] The multi-objective coordination compensation strategy comprises:

[0105] If the relative deviation of the molten pool temperature is greater than a preset upper limit of the relative deviation of the molten pool temperature, the current laser output power is reduced; if the spatial geometric offset is greater than a preset upper limit of the spatial geometric offset, the path displacement compensation amount is adjusted; if the interlayer melt depth target error is greater than a preset upper limit of the interlayer melt depth target error, the laser power or the scanning speed is increased; and if the path speed execution error is greater than a preset upper limit of the path speed execution error, the path trajectory beat is re-planned.

[0106] Specifically, after each round of manufacturing intervention, a multi-source state feedback vector is automatically constructed for manufacturing deviation quantification and correction effect evaluation, and supports subsequent control strategy closed-loop optimization.

[0107] The system integrates and codes the molten pool thermal state, the interlayer geometric accuracy, the actual execution trajectory of the path and the component deformation behavior to construct a residual error evaluation vector ΔS:

[0108] ΔS=[ΔT m ,Δd l ,Δz,Δv s ];

[0109] Wherein ΔT m represents the relative deviation of the molten pool temperature, T meas -T target ; Δd l represents the interlayer melt depth target error, d target -d meas ; Δz represents the spatial geometric offset; and Δv s represents the path speed execution error.

[0110] The error vector is used as the core process feedback basis for evaluating the effect of the manufacturing strategy, and is used for the calling logic judgment condition of the post-processing module, the path re-planning module and the heat input redistribution module in the system.

[0111] When any component Δs i exceeds its tolerance upper limit δ ii.e. meet:

[0112] |Δs i |>δ i ;

[0113] The system immediately executes rollback logic or multi-target process parameter linkage compensation.

[0114] Upon detecting that any component Δs i of the error vector exceeds its corresponding tolerance δ i , the following multi-dimensional adaptive compensation strategy is executed: if ΔT m >δ T , then P laser is reduced; if Δz>δ z , then the path displacement compensation amount δ s is adjusted; if Δd l >δ d , then the laser power or scanning speed is increased; if Δv s >δ v , then the path trajectory beat is re-planned.

[0115] Each δ i above represents a safe tolerance range for the corresponding state parameter. Through this multi-target coordinated compensation mechanism, the system can achieve manufacturing stability and robustness guarantee under complex component conditions.

[0116] Further, the hardware execution device of the system adopts a modular, quick-mount structure design, which can realize quick replacement and process switching between additive and welding tasks, significantly improving the system's adaptability and workshop scheduling efficiency.

[0117] The hardware execution device at least includes: a multi-type wire feeder head (side feeding / coaxial) quick-release module; a switching interface of a laser and electric arc head composite welding device; an independent gas cooling passage quick-plug assembly; a redundant drive card slot and a redundant bus port in the control cabinet.

[0118] All connection methods between modules adopt standardized mechanical interfaces and CANopen / EtherCAT industrial communication protocols to realize low-delay interconnection and quick cold start. Through a pre-defined interface definition file, the system can automatically identify the current configuration and load the corresponding process parameter template after replacing the module.

[0119] The manufacturing parameter configuration library supports modular expansion, and the configuration relationship is constructed as a multi-node working condition graph G=(V, E), where V represents historical working condition nodes, and E represents parameter adjustment paths in the manufacturing process; the system updates the node characteristics adaptively through real-time physical feedback, supporting quick parameter reuse and strategy evolution under multi-material, multi-component, and multi-task conditions.

[0120] The hardware execution device is particularly suitable for the multi-component, multi-material, and multi-process collinear manufacturing requirements in aviation manufacturing tasks, and ensures that the system does not cause large-scale configuration delay due to equipment replacement.

[0121] Further, the system also has a self-learning capability for historical process data, can construct a state-response mapping library based on multiple rounds of manufacturing results, and form a callable process optimization experience model. After each round of manufacturing task of the system, the following is automatically archived:

[0122] The current control parameter input vector is:

[0123] P=[P laser ,τ scan ,v s ,λ c ];

[0124] The corresponding error vector feedback result is:

[0125] ΔS=[ΔT m ,Δd l ,Δz,Δv s ];

[0126] The system adopts a sliding window evaluation strategy, extracts the parameter combination with the optimal correction effect in the last N rounds of manufacturing as a reference template, and iteratively updates the process parameter recommendation model. When similar structures, materials or component types are encountered again, the system preferentially calls the template to reduce the parameter exploration time of the new round.

[0127] This learning mechanism supports the system to change from "passive response" to "predictive correction", forms a knowledge graph and parameter optimization path for typical process scenarios, and improves the overall process adaptation intelligence level of the system.

[0128] Further, the system is also equipped with a multi-level human-computer interaction interface and state visualization function, supports system state observation, intervention, traceability, and has an operation safety protection mechanism and parameter modification permission level control.

[0129] The system supports workstation-level human-computer interaction and safety feedback monitoring, and displays process state parameters such as molten pool temperature, interlayer penetration, path speed, and cooling duty ratio in real time, provides an over-threshold alarm and manual intervention interface, and ensures the operation safety and strategy flexibility of the welding collaborative manufacturing process.

[0130] The interface functions include: multi-curve comparison: real-time display of core physical parameters such as T m , d l , v s , Q in ; and atlas type path error display: comparison of s act and s ref; Abnormal prediction heat map: Real-time prediction of overheating risk area according to path heat input model; Intervention suggestion prompt: Prompt operator whether to manually trigger intervention based on experience rule base;

[0131] The system supports three operation mode switching: automatic execution mode, suggestion assistance mode and manual intervention mode, and has a safety locking area to avoid accidental manual instruction interference in high temperature section. All parameter modifications need to pass through operator permission verification, and the system automatically records the log of each manual operation for quality traceability and responsibility archiving.

[0132] The interface system enhances the controllability, safety and human-machine collaboration adaptation ability of the manufacturing process, meeting the comprehensive expectations of "transparent production", "real-time tracking" and "collaborative intervention" in aviation manufacturing front line.

[0133] Example two:

[0134] Combined Figure 1 A practical application of a kind of additive-welding integrated manufacturing collaborative control system for TA15 titanium alloy material aviation complex components is introduced, including the following contents:

[0135] S1, in the additive and welding integrated manufacturing process, in order to realize the fine perception and multi-dimensional data-driven modeling of key process behavior, first deploy high-density, multi-modal physical state acquisition network, covering heat source behavior, wire feeding, trajectory execution, weld formation, residual stress evolution and cooling disturbance and other core process dimensions.

[0136] In actual assembly construction, additive nozzle and composite light source as heat input center, the system arranges high frame rate (≥8000fps) infrared thermal imager array around it, to capture the molten pool center temperature T m , front and back temperature difference ΔT and thermal gradient distribution Form the primary thermal field state sequence; this sequence is synchronized with the real-time profile (width w s , height h l ) of the weld obtained by the laser profiler, to calibrate the time lag relationship between heat input and forming response.

[0137] In terms of path and motion state, the system is carried by five-axis or six-axis parallel platform with wire feeder or multifunctional composite nozzle, through multi-channel encoder array to return actual trajectory position q(t) at each moment, and derive trajectory speed v s and differential curvature κ, and then combine the tension feedback and linear speed encoding value f w of the wire feeding system to construct dynamic path execution feature flow P g = v s , f w, κ. Can accurately capture the spatio-temporal misalignment behavior within the "path drift-pool bias-shape deviation" chain, providing key inputs for trajectory correction in collaborative control.

[0138] To comprehensively restore the spatial response characteristics of the forming behavior, the system arranges high-resolution laser scanning modules on the construction side wall, corner, and remelting area to extract the weld cross-sectional width w s , the pool back edge layer height residual h l , and the component profile roughness R a , dynamically compensating for the weld distortion caused by heat input instability and path disturbance. At the same time, to track the coupling path between heat, force, and appearance, the system installs an array of fiber Bragg strain sensors at key structural parts (such as high-stress splicing ribs and arc-shaped support pieces) to measure the local thermal strain ε(t) in the workpiece, and through the constitutive model, to restore the transient residual stress distribution σ(t) and the component dynamic deformation Δz(t), which are synchronously sent to the state modeling module for closed-loop correction of the heat input strategy.

[0139] In terms of the cooling system, the system integrates a micro-thermocouple and a differential pressure flowmeter at the front end of the nozzle to accurately measure the cooling gas flow rate q g and outlet temperature T g , while collecting the local environmental temperature T e , humidity H e , and flow field disturbance signals at the work station, to construct the cooling environment vector E c ={q g , T g , T e}. This information is of great significance in judging the execution effect of the temperature control strategy and the redundant heat energy dissipation behavior, especially in thin-walled components and high-power multi-layer welding sections, and has irreplaceable monitoring value.

[0140] After all the above signals are synchronized by a high-frequency hardware clock, a sliding window drift compensation algorithm is used to time-align the cross-modal data stream, and through Z-score standardization, the numerical scale is unified, and low-confidence points (outliers detected by the Hampel algorithm) and repeated samples are removed. Finally, four types of basic state vectors are formed:

[0141] Thermal field state vector:

[0142] Path execution state vector: P g ={v s , f w , κ};

[0143] Forming appearance state vector: M s ={w s , h l , R a};

[0144] Environment cooling state vector: E c = {q g , T g , T e} ;

[0145] To unify the above multi-source heterogeneous physical states and improve the input adaptability of the subsequent strategy generation module, the system introduces a weighted fusion mapping to construct a high-dimensional embedding vector Y, expressed as follows:

[0146]

[0147] where W1-W4 are fusion weight matrices, which are updated in the reverse direction by control strategy feedback in the online training phase to enhance the recognition ability of the coupling mode for abnormal states; the activation function The Swish function is used to ensure the expression power of nonlinear mapping while preserving the continuity of gradient propagation to support the stable convergence of the subsequent policy network. The typical dimension of the fusion vector Y is expanded from 30-40 dimensions in the original channel to 256 dimensions, which is used to fully express the coupling mode between each physical domain.

[0148] To further compress the feature space, improve the operation efficiency, and suppress the influence of redundant dimensions on strategy learning, the system introduces kernel principal component analysis (KPCA) based on kernel functions, which projects the fusion vector Y into a low-dimensional embedding space using a Gaussian kernel, and outputs the final state embedding vector as the main input of the strategy generation module.

[0149] This high-dimensional state vector not only covers the thermal input dynamics, path stability, forming quality, and environmental cooling feedback in the manufacturing process, but also has good time sequence traceability and channel explainability. Through this perception and fusion mechanism, the system provides a solid and complete physical state basis for the subsequent energy density adjustment, wire feeding-path-cooling collaborative strategy optimization, effectively supporting the stable construction and quality guarantee of complex weld built-up components under high beat and multiple disturbances.

[0150] S2, after completing the unified modeling of the thermal field, path, forming, and cooling states in the previous stage, the system takes the state vector X as input to further construct a process action generation strategy to coordinate the three key physical variables of laser power adjustment, path scanning speed fine-tuning, and cooling intensity compensation, achieving adaptive control of molten pool temperature control, morphology stability, and energy efficiency in the manufacturing process. During operation, the system automatically extracts the current state vector X t for each additive path or weld segment, and generates a set of physical action outputs A t = {ΔP l , Δv s , Δλ c} respectively represent the micro-adjustment amount of the current laser power, scanning speed and cooling duty cycle.

[0151] In each action generation cycle, the system will determine whether the additive area or the welding interface is overheated, the trajectory is deviated or the height of the formed layer is abnormal, etc. according to the current component position, interlayer thermal accumulation and path topology. If the behavior deviating from the process target is observed, the physical action instruction is used to adjust the core process channel, so that the system state gradually converges to the target stable interval. The strategy behavior does not depend on human intervention, and the control logic is automatically completed in the system self-sensing and feedback loop, and through the instruction path linkage, laser power subdivision adjustment and gas cooling ramp cooperation, etc., a stable, continuous and non-mutated control output is formed.

[0152] The system adopts a multi-target action selection strategy, and the pool temperature deviation e T = |T m -T set |, the scanning speed error e v = |v s -v set |, the forming deviation index Δh l , the residual stress increment Δσ and the invalid heat consumption proportion ρ loss are used as the main reference quantities to construct a comprehensive performance evaluation expression:

[0153] r t = β1·Q form - β2·e T - β3·Δh l - β4·Δσ - β5·ρ loss ;

[0154] Wherein, Q form represents the appearance forming quality judgment value of the current layer or welding area, which is determined by the weight of factors such as weld uniformity, layer height stability and transition boundary continuity; the meanings of the remaining variables correspond to the actual monitored physical quantities one by one, and the weight coefficients β1-β5 are set according to the priority control target of different component types and process stages in the manufacturing process, and can be automatically adjusted according to the state drift during operation, to ensure that the system forms a controllable balance between precision control, stable operation and energy efficiency constraints.

[0155] After the physical action output, the system further performs boundary clipping and dynamic buffer correction on the action quantity to avoid physical risks such as pool collapse or appearance instability caused by power or path mutation. For example, when the power adjustment amount ΔP l exceeds the set interval [ΔP l min , ΔP l max ], the system automatically performs clipping operation, and introduces a first-order compensation ramp in the cooling channel:

[0156]

[0157] Ensure the cooling dynamic phase matching with power adjustment, avoid over-cooling impact or micro-crystallization discontinuity caused by thermal hysteresis. In terms of path adjustment, the system simultaneously introduces path smoothing module to fine-tune the speed s Implement inertia filtering operation to maintain the consistency of path trajectory and the smoothness of platform motion.

[0158] The entire strategy generation and action execution process keeps the process physical feedback as a closed loop basis. After completing a path or a layer of additive operation, the system records the action execution results and physical state deviation, forms a "state-action-feedback" ternary dynamic sample, and continuously updates the next stage physical strategy reference. Through the strategy generation and dynamic compensation mechanism, the system can realize the physical level flexible coordination of laser power, scanning trajectory and cooling condition in the welding and additive integrated manufacturing process without external auxiliary modeling, support the stable construction and continuous quality assurance of components under the condition of variable tempo, multi-contour and multi-source thermal interference.

[0159] S3, on the basis of the system having state perception and adaptive physical strategy generation capability, further establish a rapid identification and hierarchical intervention mechanism for key process abnormalities in additive and welding components, to realize the closed-loop control and multi-layer correction of typical problems such as path error, molten pool instability, overheated layer accumulation, metallurgical discontinuity, etc. In each construction of the welding and additive process, the system periodically compares the current state vector X t With the last stage action execution results, if there are signs such as state variable fluctuation out of limits, decoupling trend between path-temperature-stress expanding or residual variable convergence instability, the system will quickly locate the abnormal level according to the abnormal type and dynamically insert the physical action compensation strategy.

[0160] This abnormal identification mechanism mainly relies on the real-time calculation of the following four types of key physical feature time series residual distribution:

[0161] Temperature deviation residual vector: ΔT r = T m (t) - T pred (t) ;

[0162] Path state disturbance function: Δq = |q(t) - q cmd (t) |2;

[0163] Layer height mutation index: δh = |h l (t) - h l-1 (t) |;

[0164] Redundant heat accumulation trend:

[0165] wherein, T pred (t) is the theoretical molten pool temperature calculated by the previous stage power and path prediction model, q cmd (t) is the trajectory planning target point, T base is the baseline temperature of the base building platform, and all residual amounts are dynamically updated in time sliding windows.

[0166] After identifying local anomalies, the system automatically classifies the anomaly level L i ∈{ slight, moderate, severe} according to the deviation level and evolution trend, and calls the corresponding intervention module. Taking the local high temperature retention of the weld as an example, if and accompanied by δh>∈ h , the system immediately inserts a laser power slow-down instruction and adds a path residence time adjustment Δt pause , and simultaneously increases the cooling gas flow rate interval Δq g , forming a concurrent compensation of the "power-path-cooling" three channels.

[0167] If the path speed drift is identified and the position deviation Δq>∈ q , the system will simultaneously adjust the acceleration limiter in the path control channel, reduce the current layer speed interval, and fine-tune the scanning start and end coordinates to prevent the laser from applying energy under unstable trajectories, causing the molten pool to be offset and the morphology to be irregularly expanded. In stress-dominated anomalies, if and the trend continues to grow, the system will limit the input power upper limit P l max of the current layer and prolong the cooling interval time t cool between adjacent components to weaken the heat input density and disperse the residual thermal gradient.

[0168] The system maps all inserted actions to the industrial instruction set through a dedicated logic layer and records the corresponding response effect. If the compensation action takes effect within the period, the state residual significantly converges, and the current strategy remains unchanged; if the residual does not decrease but increases or new abnormal points appear, the system raises the intervention level and links the next layer path strategy and cooling intensity reinitialization, establishing a multi-round dynamic closed loop.

[0169] In multi-process continuous construction tasks, the system also builds an abnormal behavior working condition library, records the abnormal trigger mode and response success rate under different component topologies, path morphologies, and environmental disturbances, and automatically matches the optimal compensation combination under similar working conditions when the task is switched or the path is updated. The working condition library uses a state distance-based retrieval expression:

[0170]

[0171] where D(·) is a state similarity function, Euclidean distance or Dynamic Time Warping (DTW) metric, used to quickly filter the most similar successful intervention cases from the large-scale historical records.

[0172] Through the abnormality identification and compensation mechanism, the system not only realizes the fast response to local deviation and process tolerance rollback in the welding process, but also builds a sustainable intervention logic based on state evolution, which supports the system to maintain process steady-state operation in the environment of complex component multi-round interlayer stacking and high coupling of heat-force-morphology, significantly improving the construction consistency and process stability.

[0173] S4, after completing the hierarchical identification of abnormal states and action compensation, the system enters the automatic generation of physical action instructions, conflict detection and multi-process execution coordination phase to ensure the dynamic consistency between different execution modules, the collaborative rationality of resource allocation and the physical action continuity under the construction task of complex components in the integrated additive-welding process. After each round of action strategy update, the system updates the state vector X t and the response action A t ={ΔP l ,Δv s ,Δλ c} converts the action parameters into a standardized industrial instruction structure and encapsulates it as a five-tuple as follows:

[0174] Instruction={Module ID ,Action Type ,Parameter,Target Value ,Tolerance Band};

[0175] where Module ID represents the current instruction target module number (such as laser control unit, cooling valve or trajectory execution platform); Action Type specifies the physical control category, such as "laser power setting", "cooling valve opening adjustment" or "path node speed correction"; Parameter is the specific action variable key value, such as ΔP l ,Δv s ,etc.; Target Value and Tolerance Band define the instruction target value and action tolerance range respectively, ensuring execution safety redundancy.

[0176] After all the instructions are generated, the system performs dynamic conflict detection on the instruction set to be issued through the action coordination scheduling module. Based on the component topology, path scheduling relationship and inter-module coupling logic, the system constructs an action conflict graph G=(V, E) in real time, where the nodes V represent the action module set and the edges E represent the coupling or mutual exclusion relationship. The system traverses all edge sets in the conflict graph to quickly detect and correct the following three types of conflicts:

[0177] Laser power-path trajectory coupling conflict: If there is an acceleration peak in the current segment of the path control module, and the laser module instruction specifies a power rising trend, local overheating may occur. The system automatically inserts a "dynamic power ramp" strategy, setting the power slope to:

[0178]

[0179] where a is the linkage buffer factor, which constrains the time domain interaction of the two channels.

[0180] Support load-trajectory displacement non-synchronization conflict: When there is a cantilever structure or a closed section of the side wall in the component, if the trajectory turning is sharp and the path beat adjustment lags behind, the support force arm module may not be in place in time. The system detects the path corner change rate Δκ and the support compensation bandwidth Δθ s whether they coincide, if the difference exceeds the buffer threshold δ θ , the main path execution trigger signal is delayed to ensure that the support structure completes the posture closed-loop adjustment.

[0181] Cooling air flow-power coupling disturbance: In thin-walled structures or continuous high-power scanning segments, if the cooling air pressure is not adjusted in time, local crystalline stress concentration may occur. The system detects the power instruction ΔP l and the cooling target Δq g whether they match the expected heat flux ratio Φ expected , if the error is greater than δ Φ , the system inserts a buffer correction action:

[0182]

[0183] Compensate for dynamic thermal field imbalance by slightly increasing the pressure.

[0184] After all the conflict detection and correction are completed, the system writes the action instructions to the lower controller through the OPC UA or EtherCAT industrial protocol, and at the same time, it attaches a state callback channel that returns the execution feedback after each action execution, which is used for subsequent action optimization and instruction update. The system dynamically distinguishes between high-priority (such as path offset correction) and low-priority (such as cooling redundancy adjustment) actions during instruction scheduling, and implements preemptive protection for critical physical interventions through a priority scheduling table.

[0185] In actual operation, the system synchronously enables the action execution monitoring channel, and records the following three types of physical execution curves in real time:

[0186] Laser power dynamic trajectory P l (t);

[0187] Cooling gas response speed q g (t);

[0188] Scan path speed distribution v s (t);

[0189] All curves are evaluated by sliding window regression and standard error. If it is found that the execution result deviates from the set target by more than the tolerance δ exec , the system immediately triggers the fine action compensation, such as dynamically reducing the power loading slope, adjusting the path acceleration factor, or increasing the cooling valve response frequency, to form a flexible closed-loop correction to the original instruction.

[0190] Through this stage, the system completes the whole-link closed-loop control process from physical action generation, conflict identification, coupling coordination to execution monitoring and fine-tuning compensation, ensuring that under the conditions of high complex path, high energy density heat input and multi-source process dynamics, laser-path-cooling modules are consistent and stable, providing stable, flexible and safe physical action support for high-strength, multi-morphology welding additive manufacturing processes.

[0191] S5, with the online of new workpiece construction tasks or real-time adjustment of production line configuration, the system enters the rapid adaptation and dynamic initialization stage of auxiliary action strategy, ensuring that when the path topology, forming beat, heat input scheduling and other parameters change significantly, the original strategy can be migrated and optimized in time to maintain the efficiency and stability of the additive-welding integrated process. To this end, the system designs a task initialization mechanism based on similar working condition mapping and strategy fusion, and a dynamic evolution model to continuously optimize the adaptive ability and migration robustness of the strategy.

[0192] In the initial stage of task switching, the system first takes the initial state X new of the new task process (including path type, heat input initial configuration, cooling arrangement mode, etc.) as input, and performs similarity matching of historical working condition nodes v i ∈V in the component-level working condition graph G=(V,E). The similarity function D(X new ,X i ) considers the following dimensions:

[0193] Path feature similarity: based on differential curvature κ, trajectory density and interlayer stacking direction to construct spatial matching indicators;

[0194] Heat input mode matching degree: according to the historical power density scheduling curve P l(t) segment-wise alignment with current power setting target;

[0195] Cooling environment difference: contrast cooling airflow rate q g , ambient temperature T e and airflow disturbance frequency spectrum ω c statistical distribution;

[0196] Forming response coupling index: analyze the deviation of weld initial cross-section morphology {w s , h l} from historical working condition average.

[0197] The above multi-scale similarity is weighted combined to form the global matching distance:

[0198]

[0199] Where β i is the weighting factor of each matching dimension, which is automatically adjusted according to the sensitivity of the component manufacturing stage.

[0200] The system selects the top n nodes {X (1) ,…,X (n)}, extracts their historical control action sequences {A (1) ,…,A (n)} from the corresponding strategy library, and generates the initial action configuration of the new task using linear weighted fusion:

[0201] Where,

[0202] Where ∈ is a small positive number to avoid division by zero exception. This fusion strategy can effectively eliminate the bias risk of a single working condition on the initialization of new task strategy, while introducing the diversity of physical behavior at the beginning of strategy generation, improving the subsequent adjustment stability.

[0203] However, for new tasks with obvious new features (such as complex irregular curves, long cantilever molten pool unstable area, etc.), historical strategy fusion may not be enough to cover the actual behavior distribution. Therefore, the system introduces a strategy fine-tuning mechanism. Based on A init , the system calculates the residual error between the actual physical behavior and the expected strategy output using the state-result pair (X t , O t ) returned by the new task at the initial stage:

[0204]

[0205] Where represents the loss function constructed based on physical feedback (such as maximum residual stress, weld morphology deviation, etc.), and fπ is the strategy execution mapping function, O tOutput the current physical execution.

[0206] When the residual exceeds the threshold value δ A , the system updates by small step size parameters:

[0207]

[0208] To achieve dynamic adaptation of the strategy to the behavior of new features. This process does not depend on global retraining of the strategy, has high responsiveness and low resource overhead, and is suitable for short-cycle task frequently changing production line scenarios.

[0209] Finally, the new task strategy A new contains both the fusion result of historical experience and the real-time physical feedback correction term, and has completeness and dynamic adaptability. The action instruction set generated under this strategy will be directly transmitted to the scheduling module and enter the issuing and execution phase.

[0210] Through the completion of this phase, the system ensures that after each task update or component process iteration, a control strategy with high matching degree, fast convergence speed, and strong physical adaptability can be built in a very short time. This mechanism greatly improves the system's ability to respond to complex components, high-frequency task switching, and non-standard heat path structures, ensuring flexible construction and high-reliability execution of the additive-welding integrated process.

[0211] S6, After completing the execution cycle of the additive-welding integrated process, the system enters the history collection and performance evaluation phase of auxiliary control actions, aiming to build a closed-loop learning path through the collection, analysis, and structured modeling of multi-dimensional physical feedback data, continuously optimizing the robustness, adaptability, and component quality assurance capability of subsequent control strategies. This phase is a key component of strategy evolution and also forms the basis for dynamic updating of working condition maps and self-repairing of the strategy library.

[0212] The system automatically starts the data backtracking mechanism at the task end node, synchronously collecting the whole-process physical response around the following four types of core channels:

[0213] Heat input channel: Collecting the heat source power density P l (t) at each moment during the additive-welding process, the center temperature T m (t) of the molten pool, and the thermal gradient distribution

[0214] Path execution channel: Recording the path trajectory position q(t), trajectory speed v s (t), and curvature change κ(t), reflecting the path control and actual deviation;

[0215] Forming response channel: Including the weld appearance cross-sectional features {w s (t), h l (t), and Ra (t)} and its wave spectrum for evaluating the forming quality;

[0216] Residual stress and structural response channel: Transient strain ε(t) and inverted residual stress σ(t), component deformation Δz(t) are returned by fiber optic strain sensors.

[0217] All data are uniformly structured into task feedback datasets:

[0218]

[0219] where X t is the state vector at each time, A t is the action output O t generated by the strategy, and is the physical response indicator (such as molten pool temperature stability, weld bead boundary consistency, component deformation peak, etc.).

[0220] The system defines the following performance evaluation indicator function in the dataset

[0221]

[0222] where: is the forming stability measure, based on the forming profile mean square deviation and the weld bead continuity indicator; is the quality achievement measure, evaluating the forming area defect rate and the residual stress threshold overrun ratio; is the execution efficiency measure, based on the total execution time, the average power regulation efficiency, and the cooling strategy energy consumption ratio; λ i is the weighting coefficient, set according to the attention degree of different dimensions by the production line.

[0223] The performance evaluation function reflects the strategy iteration effect, and the performance gain rate is defined as:

[0224]

[0225] When the performance gain rate is positive and steadily increasing in consecutive task cycles, it indicates that the strategy evolution process is effective, otherwise the system will automatically trigger the strategy version rollback mechanism to restore to the previous stable performance strategy version.

[0226] In order to identify possible physical instability factors in the control strategy, the system further calculates the action abnormality rate ρ f :

[0227]

[0228] where: N faultFor the identified abnormal behavior (such as molten pool instability, path deviation out of limit, weld seam fluctuation frequency anomaly); N total For the total number of sampling moments in the task.

[0229] When p f Exceeding the preset threshold value δ f The system marks the strategy as "low stability" and rejects the call or only uses it in low-priority conditions in the next task cycle.

[0230] All feedback data will be used for graph node embedding update. For example, in the graph neural network update mode, the historical working condition node v i Corresponding embedding representation The update is completed by the following recursion:

[0231]

[0232] Wherein, is the set of adjacent working condition nodes, W (l) is a learnable weight, and σ(·) is an activation function. This mechanism can dynamically press the strategy features of the physical feedback of the new task into the graph node representation, enhancing the strategy transfer ability and working condition matching accuracy.

[0233] On the human-computer interaction side, the system supports "strategy review" and "abnormal analysis" functions in the later stage of the task, providing visual modules such as clamp power curve, path tracking deviation graph, and weld forming animation playback to facilitate operators to identify process abnormal evolution path. The system can also highlight the maximum residual stress area, path shock frequency band, and weld boundary offset point on the interface to assist manual evaluation of strategy quality and execution safety.

[0234] Through the completion of this stage, the system realizes the whole process closed loop from physical action closed loop execution, to multi-dimensional feedback structure, to strategy safety evaluation and node knowledge update. Strategy update no longer depends on pure manual adjustment or offline testing, but relies on high-frequency, stable, and adaptive action optimization channel formed by system-level self-feedback, supporting the whole process control of welding-additive intelligent construction under long-period, multi-configuration tasks.

[0235] The above-described embodiments are only descriptions of the preferred modes of the present application and do not limit the scope of the present application. Without departing from the design spirit of the present application, various modifications and improvements to the technical solutions of the present application made by those skilled in the art shall fall within the protection scope determined by the claims of the present application.

Claims

1. A TA15 titanium alloy material aviation complex component oriented additive-welding integrated manufacturing collaborative control system, characterized in that, The method comprises the following steps: a state sensing module for collecting multi-source physical process state data of an aero complex structure during the welding process; a state modeling module for modeling the multi-source physical process state data to obtain a high-dimensional state data set; a strategy generation module for constructing a multi-source collaborative control strategy based on the high-dimensional state data set; an abnormality identification module for identifying abnormal actions in the welding process under the multi-source collaborative control strategy and inserting physical action compensation; an action coordination module for generating physical action instructions based on the multi-source collaborative control strategy and physical action compensation, and dynamically coordinating the physical action instructions using dynamic conflict detection; an auxiliary control module for switching to a new task, matching the physical action instructions of historical tasks according to the working condition similarity, and correcting them using real-time physical feedback; an optimization update module for collecting physical action instructions for performance evaluation after each task is completed, and feeding back the multi-source collaborative control strategy according to the evaluation results.

2. The TA15 titanium alloy material aerospace complex component oriented integrated additive-welding manufacturing collaborative control system according to claim 1, characterized in that, The modeling of the multi-source physical process state data includes: Wherein, X represents the process state vector, W1~W5 are adjustable weighting coefficients respectively; represents a nonlinear activation function, T m is the molten pool temperature, d l is the interlayer penetration, w s is the weld width, v s is the path speed, and Δz is the component thermal deformation.

3. The TA15 titanium alloy material aerospace complex component oriented integrated additive-welding manufacturing collaborative control system according to claim 1, characterized in that, The multi-source collaborative control strategy includes: a path control unit for real-time vector compensation of additive paths or welding trajectories based on component structure offsets; a laser heat source control unit for adaptive adjustment of laser energy density based on real-time temperature feedback; a cooling control unit for dynamic duty cycle adjustment based on thermal load distribution.

4. The TA15 titanium alloy material aerospace complex component oriented integrated additive-welding manufacturing collaborative control system according to claim 3, characterized in that, The path control unit performs real-time vector compensation of additive paths or welding trajectories based on component structure offsets, which includes: s corr = Δs + α z · Δz; where Δs represents the path fine-tuning correction vector, s ref represents the target path position vector, s act represents the current actual path position vector; K p is the proportional adjustment coefficient of position offset; K i is the integral adjustment coefficient, s corr is the final vector compensation path instruction, Δz is the component space displacement error, α z is the path influence coefficient of Δz.

5. The TA15 titanium alloy material aerospace complex component oriented integrated additive-welding manufacturing collaborative control system according to claim 3, characterized in that, The laser heat source control unit performs adaptive adjustment of laser energy density based on real-time temperature feedback, which includes: Q in = P laser · τ scan + K T · ΔT; wherein Q in represents the actual heat input energy of a unit scanning path segment, P laser is the current laser output power, τ scan is the scanning time length of the current path; ΔT is the difference between the current molten pool center temperature T meas and the target set temperature T set , and K T is the temperature adjustment gain coefficient.

6. The TA15 titanium alloy material aero complex component oriented integrated additive-welding manufacturing synergic control system according to claim 3, characterized in that, The cooling control unit performs dynamic duty cycle adjustment based on thermal load distribution, which includes: calculating the heat removal target value in the current unit time, dynamically adjusting the cooling control duty cycle within the preset optimal duty cycle range, and controlling the heat removal target value to meet the preset temperature control requirements; wherein the cooling control duty cycle is: the heat removal target value is: where λ c is the cooling control duty cycle, t cool is the duration of actual cooling performed within the manufacturing cycle, t cycle is the manufacturing cycle, AQ represents the target value of heat removal per unit time, γ1, γ2 are empirical adjustment coefficients, is the temperature rise rate at the current component position, q h is the heat flux density at the current component position.

7. The TA15 titanium alloy material aero complex component oriented integrated additive-welding manufacturing synergic control system according to claim 1, characterized in that, Identifying abnormal actions in the welding process under the multi-source collaborative control strategy includes: By calculating the temperature deviation residual vector, the path state disturbance function, the layer height mutation index, and the redundant heat accumulation trend, the action execution result of the previous stage is compared with the current state vector, and the abnormal action and the corresponding abnormal type and abnormal level are determined according to the comparison result.

8. The TA15 titanium alloy material aero complex component oriented integrated additive- welding manufacturing synergic control system according to claim 1, characterized in that, Using the dynamic conflict detection to dynamically coordinate the physical action instructions includes: Based on the component topology structure, path scheduling relationship and module coupling logic, an action conflict graph is constructed in real time, all edge sets in the conflict graph are traversed, and laser power-path trajectory coupling conflicts, support load-trajectory displacement non-collaborative conflicts, and cooling air flow-power coupling disturbances are detected and corrected.

9. The TA15 titanium alloy material aero complex component oriented integrated additive-welding manufacturing synergic control system according to claim 1, characterized in that, Collecting the physical action instructions for performance evaluation and feeding back the multi-source collaborative control strategy according to the evaluation results includes: Integrating and encoding the molten pool thermal state, layer geometric accuracy, actual path execution trajectory and component deformation behavior to construct a residual evaluation vector; When any component of the residual evaluation vector exceeds a preset upper limit of tolerance, a multi-target coordinated compensation strategy is executed to complete optimization of the multi-source collaborative control strategy.

10. The TA15 titanium alloy material aerospace complex component oriented integrated additive-welding manufacturing collaborative control system according to claim 9, characterized in that, The multi-target coordinated compensation strategy includes: If the relative deviation of the molten pool temperature is greater than a preset upper limit of the relative deviation of the molten pool temperature, the current laser output power is reduced; if the spatial geometric offset is greater than a preset upper limit of the spatial geometric offset, the path displacement compensation amount is adjusted; if the interlayer melt depth target error is greater than a preset upper limit of the interlayer melt depth target error, the laser power or the scanning speed is increased; and if the path speed execution error is greater than a preset upper limit of the path speed execution error, the path trajectory beat is re-planned.

Images