Creep-fatigue interaction damage identification method and system for HR3C dissimilar steel welding seam

Abstract

The invention discloses an HR3C dissimilar steel weld creep-fatigue interaction damage identification method and system, and relates to the technical field of material detection, and the method comprises the steps: calibrating a first dominant region and a second non-dominant region of a weld structure by a three-dimensional window for an HR3C dissimilar steel element; connecting the composite sensor to execute staggered sensing sampling of the first dominant region, and determining first interactive damage by performing mode-damage bidirectional analysis and mutual verification; a diffusion linear relation is introduced, diffusion direction intelligent generation based on a second non-dominant region is executed, and second interaction damage is determined; and according to the welding seam structure, digital space splicing is conducted on the first interaction damage and the second interaction damage to serve as a damage recognition map, and display early warning is conducted on a terminal interface. According to the method, the technical problem that the creep and fatigue interaction damage of the weld joint area cannot be accurately identified at the same time in the material detection process in the prior art is solved, and the technical effect of improving the integrity and precision of weld joint creep-fatigue interaction damage identification in material detection is achieved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of material detection, in particular to a HR3C dissimilar steel weld joint creep-fatigue interactive damage identification method and system. BACKGROUND

[0002] The HR3C dissimilar steel weld joint is often in the working condition of high temperature creep and alternating or superimposed cyclic load during service, and the damage form of the weld joint area has significant spatiotemporal evolution characteristics. For the material detection of such weld joint, it is difficult to comprehensively obtain the damage state in different temperature-stress intervals, usually relying on fixed point sensing means or damage evaluation method based on single load response. At the same time, the creep damage and fatigue damage have significant differences in the mechanism and evolution path, and when they exist interactively in the same area, their characteristics often appear to be mutually masked or nonlinearly interlaced. The traditional method is difficult to realize the cooperative identification and accurate distinction of the creep-fatigue interactive damage, resulting in insufficient coverage and deviation of the weld joint health state evaluation. SUMMARY

[0003] The present application provides a HR3C dissimilar steel weld joint creep-fatigue interactive damage identification method and system, which is used to solve the technical problem that the creep and fatigue interactive damage in the weld joint area cannot be accurately identified simultaneously in the material detection process in the prior art.

[0004] In view of the above problems, the present application provides a HR3C dissimilar steel weld joint creep-fatigue interactive damage identification method and system.

[0005] In a first aspect of the present application, a HR3C dissimilar steel weld joint creep-fatigue interactive damage identification method is provided, which comprises:

[0006] For the HR3C dissimilar steel element, the first dominant area and the second non-dominant area of the weld joint structure are calibrated by a three-dimensional window, wherein the first dominant area covers at least one dominant element of high temperature creep and cyclic load, and each area has a unified temperature-stress mode; a composite sensor is connected to perform staggered sensing sampling of the first dominant area, the first interactive damage is determined by performing bidirectional analysis and mutual inspection of mode-damage, the diffusion linear relationship is introduced, the diffusion to intelligence generation based on the second non-dominant area is performed, and the second interactive damage is determined, wherein the diffusion linear relationship is the superposition of stress diffusion dimension and temperature diffusion dimension; according to the weld joint structure, the first interactive damage and the second interactive damage are digitally spliced in space as a damage identification map, which is displayed and warned on a terminal interface.

[0007] In a second aspect of the present application, a HR3C dissimilar steel weld joint creep-fatigue interactive damage identification system is provided, which comprises:

[0008] The calibration module is used for calibrating a first dominant area and a second non-dominant area of a weld structure in a three-dimensional window for HR3C dissimilar steel elements, wherein the first dominant area covers at least one dominant element of high-temperature creep and cyclic load, and each area has a unified temperature-stress mode; the first interactive damage determination module is used for connecting a composite sensor, performing time-lapse sensing sampling of the first dominant area, determining the first interactive damage by performing bidirectional analysis and mutual test of mode-damage, and the second interactive damage determination module is used for introducing a diffusion linear relationship, performing diffusion-based intelligent generation based on the second non-dominant area, and determining the second interactive damage, wherein the diffusion linear relationship is the superposition of stress diffusion dimensions and temperature diffusion dimensions; the splicing module is used for digitally splicing the first interactive damage and the second interactive damage in a space according to the weld structure, taking the interactive damage as a damage identification atlas, and displaying a warning on a terminal interface.

[0009] One or more technical solutions provided in the present application have at least the following technical effects or advantages:

[0010] The present application calibrates a first dominant area and a second non-dominant area of a weld structure in a three-dimensional window for HR3C dissimilar steel elements, wherein the first dominant area covers at least one dominant element of high-temperature creep and cyclic load, and each area has a unified temperature-stress mode; a composite sensor is connected to perform time-lapse sensing sampling of the first dominant area, and the first interactive damage is determined by performing bidirectional analysis and mutual test of mode-damage; a diffusion linear relationship is introduced to perform diffusion-based intelligent generation based on the second non-dominant area to determine the second interactive damage, wherein the diffusion linear relationship is the superposition of stress diffusion dimensions and temperature diffusion dimensions; the first interactive damage and the second interactive damage are digitally spliced in a space according to the weld structure to take the interactive damage as a damage identification atlas, and a warning is displayed on a terminal interface. The present application solves the technical problem that the existing material detection process cannot accurately identify the creep and fatigue interactive damage of the weld area at the same time, and achieves the technical effect of improving the integrity and precision of weld creep-fatigue interactive damage identification in material detection by constructing a three-dimensional partition identification model covering the dominant area and the non-dominant area, and combining time-lapse sensing sampling and diffusion-based intelligent generation strategy to realize double-zone damage identification. BRIEF DESCRIPTION OF DRAWINGS

[0011] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiment description. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without any creative effort on the basis of these drawings.

[0012] Figure 1 The HR3C dissimilar steel weld creep-fatigue interactive damage identification method flowchart provided in the embodiments of the present application;

[0013] Figure 2 This is a schematic diagram of the HR3C dissimilar steel weld creep-fatigue interactive damage identification system provided in the embodiments of this application.

[0014] Explanation of reference numerals in the attached drawings: Calibration module 11, First interactive damage determination module 12, Second interactive damage determination module 13, and splicing module 14. Detailed Implementation

[0015] This application provides a method and system for identifying creep-fatigue interactive damage in HR3C dissimilar steel welds. It addresses the technical problem in existing materials testing processes where creep and fatigue interactive damage in weld areas cannot be accurately identified simultaneously. By constructing a three-dimensional partitioned identification model covering both dominant and non-dominant areas, and combining staggered-time sensing sampling with a diffusion-oriented intelligent generation strategy, dual-zone damage identification is achieved, thereby improving the completeness and accuracy of creep-fatigue interactive damage identification in materials testing.

[0016] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0017] It should be noted that any variation of the terms "comprising" and "having" is intended to cover non-exclusive inclusion, for example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units that are explicitly listed, but may include other steps or modules that are not explicitly listed or that are inherent to such processes, methods, products, or devices.

[0018] Example 1, as Figure 1 As shown, this application provides a method for identifying creep-fatigue interactive damage in HR3C dissimilar steel welds, the method comprising:

[0019] Step S100: For HR3C dissimilar steel components, the first dominant region and the second non-dominant region of the weld structure are calibrated using a three-dimensional window. The first dominant region covers at least one dominant factor, namely high-temperature creep and cyclic loading, and each region has a uniform temperature-stress mode.

[0020] In the embodiment of the present application, when the region calibration is performed on the HR3C dissimilar steel weld, first, a three-dimensional modeling software is used to establish a complete geometric model containing the weld structure and the surrounding base material, and the model is divided into a spatial grid fine enough. In the model, the boundary information under the actual service condition is input, including the action range of the heat source, the environmental temperature, the direction and size of the mechanical load, and the welding residual stress and the heat treatment history are superimposed, so as to constitute a true restoration of the service environment of the weld.

[0021] Then, by loading the above-mentioned thermal load and force load working conditions, the model is subjected to thermal-structural joint calculation, and the temperature change and stress change of the weld region at different time periods are output. In the output results, the temperature interval and stress amplitude of different positions of the weld are marked, and these information is arranged into a spatial distribution diagram. The distribution diagram clearly reflects which regions are continuously under stress at high temperature, and which regions have stable temperature or light load.

[0022] Then, according to the distribution diagram, the weld region is spatially divided. For those regions with high temperature or subjected to alternating load, it is judged that they may be in a high temperature creep or cyclic fatigue working state, and they are classified into the first dominant region. The region is usually close to the weld center or the fusion zone, and is the key region with the most concentrated damage. The remaining regions without high temperature or alternating load conditions are classified into the second non-dominant region, which is used to represent the process of damage diffusion or boundary transition behavior.

[0023] After completing the region division, in order to ensure the consistency of subsequent processing, the temperature data and stress data in the two regions are standardized. By unifying the temperature unit, stress unit and data arrangement format, the temperature-stress data in the two regions can be analyzed and compared in the same system. This processing step forms a unified temperature-stress mode.

[0024] Further, the method provided by the embodiment of the present application further comprises, before performing the staggered sensing sampling of the first dominant region:

[0025] According to the temperature-stress mode rule of the HR3C dissimilar steel element working environment, the staggered sampling rules of each sub-region in the first dominant region are set; according to the staggered sampling rules, the composite sensor is deployed, and each sub-region in the first dominant region is independently sensed.

[0026] In the embodiment of the present application, when the weld damage monitoring configuration of the HR3C dissimilar steel element is performed, first, combined with the high temperature and high load operating environment in which it is located, a thermal-mechanical response model of the weld area is constructed based on the temperature records, load variation records and weld service life cycle data collected during long-term operation. Through thermal-mechanical simulation and on-site monitoring data verification, the temperature rising rate, temperature fluctuation period, equivalent stress peak frequency and stress falling time of each partition in the first dominant zone of the weld during continuous operation are extracted, and then the temperature-stress mode law of each partition is formed. The temperature-stress mode law is used to represent the change trend and response characteristics of each partition under the combined action of thermal load and force load in the real working condition.

[0027] Based on the extracted temperature-stress mode law, a differentiated data acquisition plan is formulated for the multiple spatial partitions in the first dominant zone, and a staggered sampling rule with time complementarity and response priority is formed. The staggered sampling rule specifies the sampling time interval and sampling starting point of each partition, specifically including setting the sampling period to once every 30 seconds for the area with frequent temperature changes and severe stress response; for the area with stable temperature and stress changes, the sampling period can be relaxed to once every 5 minutes; in addition, to avoid the concentration of system load caused by simultaneous sampling of multiple areas at the same time, all sampling tasks are staggered in time axis, thereby realizing the sampling control logic of spatial asynchronous and time peak-shifting.

[0028] After completing the staggered sampling rule formulation, the composite sensor is deployed according to the sampling requirements of each partition. The composite sensor has multi-physical quantity measurement function and can simultaneously acquire key monitoring data such as temperature value, stress and strain response and their change rate. During deployment, according to the structural characteristics and stress concentration position of each partition in the first dominant zone, reasonable installation points are selected, and the composite sensor is fixed in place through weld mounting or embedded structure installation method, ensuring its stability and anti-interference ability in long-term service environment. Each composite sensor is bound to the corresponding partition in the system initialization stage, and the staggered sampling rule of the partition is loaded as the local control parameter.

[0029] During the acquisition process, the composite sensors in each partition of the first dominant zone have sensing autonomy function, that is, each sensor independently controls the sampling start and data upload process according to the loaded staggered sampling rule. When the sensor detects a sudden increase in temperature change or stress amplitude exceeding the preset threshold, it can automatically temporarily increase the sampling frequency or report local state abnormality, thereby realizing rapid response to high-risk state; in the stable stage of working state, the sensor automatically switches to low-frequency sampling or data packaging upload mode to improve data utilization and reduce system power consumption.

[0030] Step S200: connecting the composite sensor, performing the time-lapse sensing sampling of the first dominant zone, determining the first interaction damage by performing the bidirectional analysis and mutual verification of the mode-damage.

[0031] In the embodiment of the present application, after connecting the composite sensor, first set up the peripheral interface on the sensing side of the communication bus, and deploy the data processing element with real-time processing capability to form a stable data channel. After triggering the time-lapse sampling rule of the partition to which the composite sensor belongs, the composite sensor starts to perform the strain-temperature data collection in the form of three-dimensional stacking, and at the same time, the damage data under the running state of the weld is obtained. The collected data flows into the data processing element through the peripheral interface, and the mode data and the damage data are structurally processed, the feature vector representing the running state of the weld is extracted, and the time stamp and the location code are combined for identification, so as to construct the time-lapse sensing data with clear time sequence and spatial positioning.

[0032] Then, the preset bidirectional identifier model is called to perform the bidirectional analysis of the mode-damage of the above time-lapse sensing data. Specifically, the positive identification branch is responsible for damage judgment on the extracted mode feature vector, and respectively identifies the creep damage data and the fatigue damage data; then the damage feature vector and the two types of damage data identified are input into the back derivation branch, the first verification node is used to judge whether the damage feature exceeds the directly superimposed value, if the condition is met, the second calculation node is used to calculate the difference value as the interaction incremental damage. Finally, the damage feature vector, the creep damage data, the fatigue damage data and the interaction incremental damage are integrated to determine the first interaction damage.

[0033] Further, in the method provided by the embodiment of the present application, performing the time-lapse sensing sampling of the first dominant zone further includes:

[0034] The data processing element is deployed in the peripheral interface, wherein the peripheral interface is located at the sensing side of the communication bus; the composite sensor triggers the time-lapse sampling rule, performs the mode data collection under the 3D stacked strain-temperature, and performs the damage data collection under the weld state detection; for the mode data and the damage data, the data processing element flowing through the peripheral interface performs the feature vector extraction and the identification based on the time stamp and the location code to determine the time-lapse sensing data.

[0035] In the embodiment of the present application, when deploying the data processing element in the peripheral interface, first install the data processing element in the communication bus close to the sensor side to ensure that it can directly receive the data stream uploaded by the composite sensor. The peripheral interface forms a wired connection with the composite sensor through the bus node, has the local data preprocessing and high-speed forwarding capability, and provides a low-delay entrance for subsequent data processing.

[0036] The composite sensor is activated according to a preset staggered sampling rule, and when a set sampling time point is reached, a data collection program is automatically started to perform mode data collection under 3D stacked strain-temperature. In the collection, the composite sensor synchronously collects strain changes of the weld area in X, Y and Z directions, and simultaneously obtains the current temperature value and the temperature change rate, matches the strain data in each direction with the temperature data at a unified time point, and constructs a data segment of a three-dimensional stacked structure as mode data.

[0037] Meanwhile, the composite sensor starts the weld state detection in parallel to collect damage data. During the damage data collection process, the system monitors the nonlinear fluctuation of the strain curve, the mutation node in the stress change and the deviation trend between the temperature and the strain in real time, extracts the preliminary signals representing the internal defects or material degradation of the weld, and forms the damage data.

[0038] After the collection is completed, the mode data and the damage data flow into the peripheral interface through the communication bus and are processed by the data processing element deployed on the interface. The data processing element extracts the feature vectors of the received mode data and damage data, including strain change rate, temperature fluctuation amplitude, load response lag value, abnormal point duration, etc. Subsequently, the processed data is assigned a time stamp and a location code, the time stamp is used to indicate the accurate time point of data collection, and the location code is used to record the spatial position and sensor number of the data source.

[0039] Finally, all processing results are integrated into staggered sensing data with explicit time identification and spatial positioning.

[0040] Further, the method provided by the application embodiment further comprises:

[0041] A damage identification module is developed in the weld detection system, wherein the identification module includes a bidirectional identifier and a black box generator; wherein the construction of the bidirectional identifier includes: taking sample mode data as input and creep damage and fatigue damage as output to train a positive identification branch; based on the damage data and the output of the positive identification branch, a first verification node and a second calculation node are deployed to train an inverse deduction branch, wherein the interactive incremental damage is taken as a calculation target; the positive identification branch and the inverse deduction branch are cascaded as the bidirectional identifier.

[0042] In the application embodiment, when developing the damage identification module in the weld detection system, first, an identification structure including a bidirectional identifier and a black box generator is constructed to model and identify the damage state of the HR3C dissimilar steel weld under the coupling action of creep and fatigue. The bidirectional identifier in the identification module is used to realize the positive judgment of the mode data to the damage result and the inverse deduction of the damage data to the interactive increment, and the black box generator is used to realize the intelligent generation of damage compensation in the non-dominant area.

[0043] In the process of constructing the bidirectional recognizer, first, a supervised learning method is used to establish a positive recognition branch. The specific steps are as follows: the sample pattern data obtained by the time-lapse sensing sampling is taken as the input, the input data contains three-dimensional stacked strain-temperature joint features, and the standard input vector is formed through normalization and feature extraction processing; the creep damage and fatigue damage results labeled by historical experiments are taken as the output label, a training set is constructed, and a neural network structure is used for training, and a cross-entropy loss function is used for iterative optimization during training. Through the training process, a positive recognition branch with classification ability is obtained, which can output creep damage data and fatigue damage data for any input pattern feature vector.

[0044] Then, a residual calculation method is used to construct a first verification node. The node superimposes the creep damage data and fatigue damage data output by the positive recognition branch to form a basic damage superposition value; at the same time, the damage feature vector in the current time-lapse sensing data is extracted and compared with the above superposition value to calculate the residual value. If the residual value is below the set threshold, it is determined that the damage result is reliable; if the residual value exceeds the threshold, it indicates that there is a damage part that exceeds the basic superposition, triggering the reverse modeling process. After triggering the reverse modeling, a least squares regression method is used to construct a second calculation node. The difference between the damage feature vector and the sum of the creep damage data and the fatigue damage data is taken as the target quantity, and the regression model is trained to fit and predict the interactive incremental damage. The input is the damage feature vector, and the output is the residual term, i.e. the interactive incremental damage, which is used to compensate for the damage influence that the positive recognition branch cannot cover.

[0045] After completing the above two branches, a model cascading method is used to combine the positive recognition branch and the reverse derivation branch to form a complete bidirectional recognizer. The structure completes the preliminary judgment of creep and fatigue in the forward path, and verifies and corrects the damage result in the reverse path, realizing the closed-loop processing capability from data input, result recognition to interactive damage calculation.

[0046] Further, the method provided by the application embodiment further comprises:

[0047] The first dominant region is taken as a known quantity, the second non-dominant region is taken as a generated quantity, and a diffusion linear relationship is taken as a constraint for adversarial training to determine a generation-discrimination component, wherein the diffusion direction is the diffusion from the first dominant region to the second non-dominant region; and the generation-discrimination component is disassembled, and the generation component is taken as the black box generator.

[0048] In the embodiment of the present application, in the process of constructing the black box generator, first, the first dominant area is taken as a known quantity, the creep damage data, fatigue damage data and interactive incremental damage data identified by the bidirectional identifier in the region are extracted, the temperature distribution and stress distribution of the corresponding space position are combined to form a multi-dimensional structure input, and the first dominant area damage input tensor with spatial continuity and physical consistency is unified. The tensor is taken as the input data of the generation network, which completely expresses the damage evolution state of the dominant area under the action of thermal-mechanical coupling.

[0049] Then, the second non-dominant area is taken as a generated quantity, a spatial coordinate mapping aligned with the first dominant area is established according to the preset structure grid, the boundary condition information of the region in the same period is extracted, including temperature conduction boundary, structure load response and material properties, etc., and the second non-dominant area target output tensor is constructed as the reference target output in the generator training stage. The output tensor represents the expected damage state of the current region without monitoring points, and is a supervision signal when training the generator.

[0050] On the basis of establishing the input and output tensors, a generation-discrimination component is constructed by using a generative adversarial training method. The generator takes the first dominant area damage input tensor as the input to predict the damage distribution atlas of the second non-dominant area; the discriminator receives the output atlas of the generator and the target output tensor, judges whether the input is from the real distribution, and optimizes the generator parameters through back propagation to improve the fitting ability of the predicted atlas. In the training process, a diffusion linear relationship is introduced as a physical constraint of the generator structure. The diffusion linear relationship is formed by the gradient characteristics of the stress diffusion dimension and the temperature diffusion dimension, and is embedded in the middle layer of the generator in the form of a weight matrix in the network, which limits the spatial evolution direction and rate of the output. The diffusion direction is explicitly set as the diffusion path from the first dominant area to the second non-dominant area, ensuring that the generated damage atlas conforms to the physical conduction process caused by thermal-mechanical coupling in the real weld structure, and has structural rationality.

[0051] After the generator and the discriminator are alternately trained and reach the set training convergence condition, the generation-discrimination component is separated in structure by using a model disassembly method. The part of the generator that has been trained and has prediction ability is extracted separately, the structure parameters, input interface and output logic are retained, and it is defined as the final black box generator.

[0052] Further, in the method provided by the embodiment of the present application, the first interactive damage is determined by performing bidirectional analysis and mutual inspection of the mode-damage, and the method further comprises:

[0053] According to the positive identification branch in the bidirectional identifier, the mode feature vector in the time-lapse sensing data is subjected to damage decision to determine the creep damage data and the fatigue damage data; the damage feature vector in the time-lapse sensing data, the creep damage data and the fatigue damage data are introduced into the inverse derivation branch, based on a first verification node, it is judged whether the damage feature vector is greater than the direct superposition of the creep damage data and the fatigue damage data, if the verification is successful, based on a second calculation node, the difference between the damage feature vector and the superposition data is calculated as an interactive incremental damage; the damage feature vector, the creep damage data, the fatigue damage data and the interactive incremental damage are integrated and added to the first interactive damage.

[0054] In the embodiments of the present application, for the time-lapse sensing data obtained by time-lapse sensing sampling, first, the mode feature vector is extracted and input into the positive identification branch in the bidirectional identifier to perform a damage discrimination process based on a neural network. The positive identification branch uses a convolutional neural network structure to perform hierarchical feature extraction and classification mapping on the input mode feature vector, and outputs the creep damage data and the fatigue damage data. Among them, the creep damage data is used to represent the deformation accumulation trend of the weld structure under the action of high temperature sustained load; the fatigue damage data is used to reflect the crack initiation and propagation risk under the action of periodic alternating load. This process is damage decision, which realizes the positive identification from state feature to damage type.

[0055] Subsequently, the creep damage data and the fatigue damage data obtained by the above identification are input into the inverse derivation branch in the bidirectional identifier together with the damage feature vector corresponding to the current sensing period, and enter the reverse checking process based on verification and residual calculation. In this process, first, the first verification node is called, and the creep damage data and the fatigue damage data are subjected to item-by-item addition operation by a numerical superposition judgment method to obtain a direct superposition result, i.e. the damage total that can be completely explained by the positive identification branch in theory. Subsequently, the first verification node compares the superposition result with the actual input damage feature vector dimension by dimension, if it is determined that the damage feature vector is greater than the direct superposition result, it is confirmed that there is an additional damage component that is not covered by the positive identification branch, and the verification is passed, and the subsequent step is entered.

[0056] After the verification is passed, the second calculation node is activated, which uses the least squares regression method to model the residual between the damage feature vector and the superposition result, calculates the numerical difference, and defines the difference as an interactive incremental damage. The interactive incremental damage represents the nonlinear superposition effect between creep and fatigue under actual service conditions, such as material microstructure mutation caused by thermal-mechanical coupling, alternating response amplification and other phenomena, which belongs to a composite response independent of conventional single damage mechanism.

[0057] Finally, the damage feature vectors obtained in the current cycle, the identified creep damage data and fatigue damage data, and the calculated interaction incremental damage are merged according to the spatial position and time index to generate a unified first interaction damage data package.

[0058] Step S300: introducing a diffusion linear relationship, performing diffusion-oriented intelligent generation based on the second non-dominant zone, and determining the second interaction damage, wherein the diffusion linear relationship is the superposition of the stress diffusion dimension and the temperature diffusion dimension.

[0059] In the embodiments of the present application, the diffusion linear relationship is introduced to carry out the diffusion-oriented intelligent generation process of the second non-dominant zone for determining the second interaction damage. Specifically, first, the spatial phase between the first dominant zone and the second non-dominant zone is determined according to the positional relationship of the two zones in the spatial model, and a unified digital space is constructed. In the digital space, the damage state of the first dominant zone is identified based on the first interaction damage result, and a generation label is added to each node in the second non-dominant zone as a target area for damage prediction.

[0060] Subsequently, the initialized digital space is imported as input into the trained black box generator, and the diffusion linear relationship is applied as a constraint condition in the generation process. The diffusion linear relationship is composed of the numerical gradients of the stress diffusion dimension and the temperature diffusion dimension, and is used to control the extension path and conduction intensity of the damage information from the first dominant zone to the second non-dominant zone. The generator completes the damage prediction output of each node in the second non-dominant zone under the constraint, and finally generates complete second interaction damage data.

[0061] Further, the method provided by the embodiments of the present application, which performs diffusion-oriented intelligent generation based on the second non-dominant zone to determine the second interaction damage, further comprises:

[0062] determining a digital space based on the spatial phase of the first dominant zone and the second non-dominant zone, identifying the damage of the first dominant zone based on the first interaction damage, adding a generation label to the second non-dominant zone, initializing the digital space, importing the initialized digital space into the black box generator, and performing intelligent generation of the damage of the second non-dominant zone based on the diffusion linear relationship as a constraint to determine the second interaction damage.

[0063] In the embodiments of the present application, first, the spatial phase between the first dominant zone and the second non-dominant zone is determined according to the spatial arrangement relationship of the two zones in the three-dimensional structure of the weld, i.e., the relative position and grid mapping relationship of the two zones in the unified structure coordinate system. Through the spatial phase matching, a continuous grid model covering the dominant zone and the non-dominant zone is established to form an overall operable digital space.

[0064] Subsequently, based on the identified first interaction damage, an damage identification operation is performed on each grid node of the first dominant zone in the digital space. Specifically, the creep damage data, fatigue damage data and interaction incremental damage data of each node are labeled as known damage states, forming a damage input layer with spatial position and damage type dual attributes. At the same time, the corresponding second non-dominant zone area is marked in the digital space, and a generated label is assigned to it, which is used to indicate that the area is a damage prediction area to be generated. Through this process, the initialization digital space is completed.

[0065] After the above initialization is completed, the initialized digital space is input into the trained black box generator, which internally embeds a diffusion linear relationship constraint mechanism. The diffusion linear relationship is composed of stress diffusion dimension and temperature diffusion dimension, the former reflects the stress gradient conduction trend from the dominant zone to the non-dominant zone, and the latter describes the spatial distribution characteristics of thermal influence with distance. Both are encoded as weight matrices in a linear superposition form and embedded in the intermediate layer neural structure of the black box generator as constraint information to adjust the output distribution of the generator.

[0066] Under the guidance of this physical constraint, the black box generator performs damage generation under the spatial diffusion logic for all nodes with generated labels in the second non-dominant zone, automatically predicts the creep damage, fatigue damage and interaction coupling effect of each node from the dominant zone to the target area, and generates a damage output tensor covering the entire target area.

[0067] Finally, according to the output result of the black box generator, the intelligent judgment of the damage state of the second non-dominant zone is completed, and the second interaction damage is determined, which is the reasoning result of the area without sensor arrangement, and together with the first interaction damage constitutes a complete weld interaction damage identification atlas, supporting the subsequent digital space splicing, transition point identification and global damage warning process.

[0068] Step S400: According to the weld structure, the first interaction damage and the second interaction damage are spliced in the digital space as a damage identification atlas, which is displayed and warned on the terminal interface.

[0069] In the embodiment of the present application, first, according to the digital space model corresponding to the weld structure, the first interactive damage and the second interactive damage are spatially positioned by using the time stamp and the position code, and the digital space splicing operation is performed under a unified coordinate system to generate a spliced damage atlas. Subsequently, the window boundary region in the spliced atlas is identified, and the damage data transition condition therein is detected. If the transition data is detected and the amplitude thereof does not exceed a preset transition value, point-level smoothing processing is performed at the corresponding boundary; if the transition amplitude exceeds the preset value, region-level smoothing processing is performed in the region where the boundary is located. Through the identification and hierarchical optimization of the transition data, a damage identification atlas with higher continuity and reflecting the actual state of the weld is finally formed, and is displayed in a graphical manner in real time on a terminal interface, so that the visual warning of the damage in the key weld region is realized.

[0070] Further, the method provided by the embodiment of the present application, in which the digital space splicing is performed as the damage identification atlas, further comprises:

[0071] According to the identification of the time stamp and the position code, the first interactive damage and the second interactive damage are spliced in the digital space to determine a spliced damage atlas; the window boundary of the spliced damage atlas is identified, the damage data transition is identified, and the transition data is located; the transition data in the spliced damage atlas is optimized to determine the damage identification atlas.

[0072] Further, the method provided by the embodiment of the present application further comprises:

[0073] If there is transition data and it is less than or equal to a preset transition value, point smoothing processing of the transition boundary is performed; if there is transition data and it is greater than the preset transition value, region smoothing processing of the transition boundary is performed.

[0074] In the embodiment of the present application, first, based on the characteristic data of the first interactive damage and the second interactive damage, the time stamp and the position code (i.e. the identification information recording the data acquisition time and the spatial coordinates) of each damage point are extracted in the digital space model constructed uniformly, so that the synchronous positioning and spatial reconstruction of the two types of damage data are realized. In this way, the two types of damage data are spliced according to the physical position and the time sequence to generate a spliced damage atlas.

[0075] Subsequently, a sliding window scanning method is used to identify the boundary of the spliced damage atlas, and according to the fluctuation of the local data value in the window, transition identification is performed. Specifically, by comparing the value difference between the center point of the window and its neighborhood data, the jump point with significant change is extracted as the transition data. For example, if the damage value of a certain window center point is 0.12, and the values of the two adjacent points before and after it are 0.05 and 0.06 respectively, the value difference exceeds the preset transition value 0.04, and it can be determined that the center point is a transition point, which is used as a representative value of abnormal change to participate in subsequent processing.

[0076] The identified transition data is classified, if the transition amplitude is lower than or equal to a preset transition value, a moving average point smoothing method is adopted, a symmetric window is constructed with the transition point as the center, the average value of the adjacent sampling points is calculated, and the average value is assigned back to the center point, the local noise or small mutation is smoothed and corrected, so as to improve the data stability and the continuity of the atlas.

[0077] If the transition amplitude exceeds the preset transition value, the transition region is modeled and repaired by a local linear fitting region smoothing method, the local region where the transition point is located is selected as the fitting input, a linear function is used to construct the trend boundary, a continuous curve is generated by fitting to replace the original mutation curve, the smooth transition of the surrounding structure is maintained, and the identification accuracy of the whole graph and the interpretation ability of the mutation region are improved.

[0078] Finally, the spliced damage atlas after the above smoothing optimization is output as a damage identification atlas, transmitted to a terminal interface module, and a visual graph is generated by a graph rendering method, which is used to realize dynamic early warning display of the weld structure damage, and assist the operator to intuitively master the damage region distribution and severity.

[0079] In the embodiments of the present application, the above-mentioned embodiments have at least the following technical effects:

[0080] The present application calibrates the first dominant region and the second non-dominant region of the weld structure by a three-dimensional window for HR3C dissimilar steel elements, wherein the first dominant region covers at least one dominant element of high temperature creep and cyclic load, and each region has a unified temperature-stress mode; a composite sensor is connected to perform staggered sensing sampling of the first dominant region, the first interactive damage is determined by performing bidirectional analysis and mutual inspection of mode-damage; the diffusion linear relationship is introduced to perform intelligent generation based on the second non-dominant region, and the second interactive damage is determined, wherein the diffusion linear relationship is the superposition of stress diffusion dimension and temperature diffusion dimension; according to the weld structure, the first interactive damage and the second interactive damage are digitally spliced in space as a damage identification atlas, which is displayed and warned on the terminal interface. The present application solves the technical problem that the creep and fatigue interactive damage of the weld region cannot be accurately identified at the same time in the material detection process in the prior art, by constructing a three-dimensional partition identification model covering the dominant region and the non-dominant region, and combining staggered sensing sampling and intelligent generation strategy to realize double-region damage identification, the technical effects of improving the integrity and precision of the weld creep-fatigue interactive damage identification in material detection are achieved.

[0081] Embodiment two, based on the same inventive concept as the HR3C dissimilar steel weld creep-fatigue interactive damage identification method in the foregoing embodiments, such as Figure 2As shown, the present application provides a HR3C dissimilar steel weld joint creep-fatigue interaction damage identification system, and the system and method embodiments in the present application are based on the same inventive concept. Wherein, the system comprises:

[0082] a calibration module 11 for calibrating a first dominant zone and a second non-dominant zone of a weld structure in a three-dimensional window for a HR3C dissimilar steel element, wherein the first dominant zone covers at least one dominant factor of high temperature creep and cyclic load, and each zone has a unified temperature-stress pattern; a first interaction damage determination module 12 for connecting a composite sensor, performing staggered sensing sampling of the first dominant zone, and determining a first interaction damage by performing bidirectional analysis and mutual verification of pattern-damage; a second interaction damage determination module 13 for introducing a diffusion linear relationship, performing diffusion-based intelligent generation based on the second non-dominant zone, and determining a second interaction damage, wherein the diffusion linear relationship is the superposition of stress diffusion dimension and temperature diffusion dimension; a splicing module 14 for digitally splicing the first interaction damage and the second interaction damage according to the weld structure, as a damage identification atlas, and displaying a warning on a terminal interface.

[0083] Further, the system is also used to implement the following functions:

[0084] According to the operation environment of the HR3C dissimilar steel element, the mode law of temperature-stress is used to set the staggered sampling rules of each subzone in the first dominant zone; according to the staggered sampling rules, the composite sensor is deployed, and each subzone in the first dominant zone is autonomously sensed.

[0085] Further, the system is also used to implement the following functions:

[0086] A data processing element is disposed at a peripheral interface, wherein the peripheral interface is located at a sensing side of a communication bus; the composite sensor triggers the staggered sampling rules, performs mode data collection under 3D stacked strain-temperature, and performs damage data collection under weld joint state detection; for the mode data and the damage data, the data processing element of the peripheral interface is communicated, feature vector extraction and identification based on time stamp and location code are performed, and staggered sensing data is determined.

[0087] Further, the system is also used to implement the following functions:

[0088] Develop a damage identification module in the weld detection system, wherein the identification module includes a bidirectional identifier and a black box generator; wherein the construction of the bidirectional identifier includes: taking sample pattern data as input, taking creep damage and fatigue damage as output, training a positive identification branch; based on damage data and positive identification branch output, deploying a first verification node and a second calculation node, training an inverse derivation branch, wherein the interactive incremental damage is taken as the calculation target; cascade the positive identification branch and the inverse derivation branch as the bidirectional identifier.

[0089] Further, the system is also used to realize the following functions:

[0090] Taking the first dominant area as a known quantity and the second non-dominant area as a generated quantity, and taking the diffusion linear relationship as a constraint, the system is used to perform adversarial training to determine a generation-discrimination component, wherein the diffusion direction is the diffusion from the first dominant area to the second non-dominant area; the generation-discrimination component is disassembled, and the generation component is taken as the black box generator.

[0091] Further, the system is also used to realize the following functions:

[0092] According to the positive identification branch in the bidirectional identifier, the pattern feature vector in the time-lapse sensing data is subjected to damage decision to determine creep damage data and fatigue damage data; the damage feature vector in the time-lapse sensing data, the creep damage data and the fatigue damage data are introduced into the inverse derivation branch, and based on the first verification node, it is determined whether the damage feature vector is greater than the direct superposition of the creep damage data and the fatigue damage data, if the verification is successful, based on the second calculation node, the difference between the damage feature vector and the superposition data is taken as the interactive incremental damage; the damage feature vector, the creep damage data, the fatigue damage data and the interactive incremental damage are integrated and added to the first interactive damage.

[0093] Further, the system is also used to realize the following functions:

[0094] Taking the spatial phase of the first dominant area and the second non-dominant area, the system is used to determine a digital space; based on the first interactive damage, the system is used to perform damage identification on the first dominant area and add a generation label to the second non-dominant area to initialize the digital space; the initialized digital space is introduced into the black box generator, and the system is used to perform damage intelligent generation on the second non-dominant area based on the diffusion linear relationship as a constraint to determine the second interactive damage.

[0095] Further, the system is also used to realize the following functions:

[0096] According to the identification of the timestamp and the location code, the first interaction lesion and the second interaction lesion are spliced in the digital space to determine a spliced lesion atlas; a window boundary of the spliced lesion atlas is identified, lesion data transition identification is performed, and transition data is located; the transition data in the spliced lesion atlas is optimized to determine the lesion identification atlas.

[0097] Further, the system is also used to realize the following functions:

[0098] If there is transition data and it is less than or equal to a preset transition value, point smoothing processing of the transition boundary is performed; if there is transition data and it is greater than the preset transition value, region smoothing processing of the transition boundary is performed.

[0099] It should be noted that the above-mentioned sequence of the embodiments of the present application is only for description, and does not represent the advantages and disadvantages of the embodiments. The above-mentioned specific embodiments of the present application are described. The processes depicted in the drawings do not necessarily require the specific order and continuous order shown to achieve the desired results. In some embodiments, multi-task processing and parallel processing are also possible or can be advantageous.

[0100] The above only describes the preferred embodiments of the present application and does not limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

[0101] The present application is only an exemplary description of the present application, and should be considered to cover any and all modifications, changes, combinations or equivalents within the scope of the present application. Obviously, those skilled in the art can make various modifications and changes to the present application without departing from the scope of the present application. Thus, if these modifications and changes of the present application belong to the scope of the present application and its equivalents, the present application intends to include these modifications and changes.

Claims

1. A method for identifying creep-fatigue interactive damage in HR3C dissimilar steel welds, characterized in that, The method includes: For HR3C dissimilar steel components, a three-dimensional window is used to define the first dominant zone and the second non-dominant zone of the weld structure. The first dominant zone covers at least one dominant factor, high-temperature creep and cyclic loading, and each zone has a uniform temperature-stress mode. Connect the composite sensor, perform staggered sensing sampling of the first dominant region, and determine the first interactive damage by performing bidirectional analysis and cross-verification of mode and damage; Introducing a diffusion linear relationship, performing diffusion-oriented intelligent generation based on the second non-dominant region, and determining the second interactive damage, wherein the diffusion linear relationship is the superposition of the stress diffusion dimension and the temperature diffusion dimension; Based on the weld structure, the first interactive damage and the second interactive damage are digitally spatially stitched together to form a damage identification map, which is then displayed and alerted on the terminal interface.

2. The method for identifying creep-fatigue interactive damage in HR3C dissimilar steel welds as described in claim 1, characterized in that, Before performing staggered sensing sampling of the first dominant region, the following steps are included: Based on the operating environment of HR3C dissimilar steel components, and taking into account the temperature-stress pattern, staggered sampling rules are set for each zone within the first dominant area. The composite sensor is deployed according to the staggered sampling rule, wherein each zone within the first dominant region is autonomously sensed.

3. The method for identifying creep-fatigue interactive damage in HR3C dissimilar steel welds as described in claim 2, characterized in that, Performing staggered sensing sampling in the first dominant region includes: A data processing element is deployed at a peripheral interface, wherein the peripheral interface is located on the sensing side of the communication bus; The composite sensor triggers the staggered sampling rule to perform 3D stacked strain-temperature pattern data acquisition and damage data acquisition under weld condition detection. For the pattern data and damage data, the data processing element that communicates through the peripheral interface performs feature vector extraction and identification based on timestamps and location codes to determine the time-lapse sensing data.

4. The method for identifying creep-fatigue interactive damage in HR3C dissimilar steel welds as described in claim 3, characterized in that, A damage identification module is developed within a weld inspection system, wherein the identification module includes a bidirectional identifier and a black box generator; The construction of the bidirectional identifier includes: Using sample pattern data as input and creep damage and fatigue damage as output, train the positive recognition branch; Based on damage data and the output of the positive recognition branch, a first verification node and a second calculation node are deployed to train the inverse derivation branch, with interactive incremental damage as the calculation target; The positive recognition branch and the inverse derivation branch are cascaded together to form the bidirectional recognizer.

5. The method for identifying creep-fatigue interactive damage in HR3C dissimilar steel welds as described in claim 4, characterized in that, The construction of the black-box generator includes: Using the first dominant region as a known quantity and the second non-dominant region as a generated quantity, adversarial training is conducted with the linear diffusion relationship as a constraint to determine the generator-discriminator component. The diffusion direction is from the first dominant region to the second non-dominant region. Disassemble the generation-discrimination component and use the generation component as the black-box generator.

6. The method for identifying creep-fatigue interactive damage in HR3C dissimilar steel welds as described in claim 5, characterized in that, By performing bidirectional pattern-damage analysis and cross-verification, the first interactive damage was identified, including: Based on the positive recognition branch in the bidirectional recognizer, damage decision is made on the pattern feature vector in the time-staggered sensing data to determine creep damage data and fatigue damage data. The damage feature vector in the time-staggered sensing data, the creep damage data, and the fatigue damage data are imported into the inverse derivation branch. Based on the first verification node, it is determined whether the damage feature vector is greater than the direct superposition of the creep damage data and the fatigue damage data. If the verification is successful, based on the second calculation node, the difference between the damage feature vector and the superposition data is calculated as the interactive incremental damage. The damage feature vector, creep damage data, fatigue damage data, and interactive incremental damage are integrated and added to the first interactive damage.

7. The method for identifying creep-fatigue interactive damage in HR3C dissimilar steel welds as described in claim 6, characterized in that, Perform diffusion-based intelligent generation based on the second non-dominant region to determine the second interaction impairment, including: The digital space is determined by the spatial phase between the first dominant region and the second non-dominant region; Based on the first interactive damage, damage is identified in the first dominant region, generation tags are added to the second non-dominant region, and the digitization space is initialized. The initialized digital space is imported into the black box generator, and damage is intelligently generated for the second non-dominant region under the constraint of the diffusion linear relationship to determine the second interactive damage.

8. The method for identifying creep-fatigue interactive damage in HR3C dissimilar steel welds as described in claim 1, characterized in that, Digital spatial stitching is performed to create a damage identification atlas, including: Based on the identification of timestamps and location codes, the first interactive damage and the second interactive damage are spliced ​​together in the digital space to determine the spliced ​​damage map; Identify the window boundaries of the spliced ​​damage map, identify damage data transitions, and locate the transition data; Optimize the transition data within the spliced ​​damage map to determine the damage identification map.

9. The method for identifying creep-fatigue interactive damage in HR3C dissimilar steel welds as described in claim 8, characterized in that, If transition data exists and is less than or equal to the preset transition value, perform point smoothing processing on the transition boundary; If transition data exists and is greater than the preset transition value, perform regional smoothing processing on the transition boundary.

10. A creep-fatigue interactive damage identification system for dissimilar steel welds in HR3C, characterized in that, The system is used to execute the creep-fatigue interactive damage identification method for HR3C dissimilar steel welds as described in any one of claims 1-9, and the system includes: The calibration module is used to calibrate the first dominant zone and the second non-dominant zone of the weld structure for HR3C dissimilar steel components using a three-dimensional window. The first dominant zone covers at least one dominant factor, high-temperature creep and cyclic loading, and each zone has a uniform temperature-stress mode. The first interactive damage determination module is used to connect to the composite sensor, perform staggered sensing sampling of the first dominant region, and determine the first interactive damage by performing bidirectional analysis and cross-verification of mode and damage. The second interactive damage determination module is used to introduce a diffusion linear relationship, perform intelligent generation of diffusion direction based on the second non-dominant region, and determine the second interactive damage, wherein the diffusion linear relationship is the superposition of stress diffusion dimension and temperature diffusion dimension. The splicing module is used to digitally splice the first interactive damage and the second interactive damage according to the weld structure, and use it as a damage identification map to display and warn on the terminal interface.

Images