Segmented weld joint detection method and system based on weld joint life prediction

Abstract

The invention discloses a welding seam segment detection method and system based on welding seam life prediction, and relates to the field related to data detection.The method comprises the steps that a welding seam segment rule is defined, heterogeneous data of different detection devices are collected and obtained for registration, a registration data set is obtained, feature extraction is conducted on the registration data set according to the welding seam segment rule, and a welding seam segment detection result is obtained; according to the method, a feature framework is constructed, a weld joint life prediction model is constructed according to the feature framework, a plurality of residual life prediction results are obtained, segmentation priority ranking is performed on weld joints, and a detection decision report is generated, so that the technical problems of non-uniform segmentation standards, low life prediction precision and difficulty in multi-source data integration caused by isomerism of weld joint detection data are solved; the technical effects that the prediction accuracy is improved by unifying segmentation rules and fusing physical and data models, the priority detection report is generated, and efficient maintenance and accident risk prevention and control are achieved are achieved.

Description

Technical Field

[0001] The present application relates to data detection related fields, and in particular to a weld segmentation detection method and system based on weld life prediction. Background Art

[0002] Traditional weld inspection relies on single inspection technologies such as radiography (RT) and ultrasonic testing (UT). These segmentation rules are limited by the physical characteristics of the equipment and differences in industry standards, resulting in inconsistent data coordinate systems and chaotic segmentation granularity. Furthermore, existing life prediction methods often use isolated models that rely on idealized assumptions, making it difficult to cope with complex working conditions. Due to the heterogeneity of inspection data, they face difficulties in feature alignment and insufficient generalization capabilities. This results in low reliability in prediction results, making it difficult to support predictive maintenance decisions. Furthermore, multi-source inspection reports lack a unified integration framework, forcing operations and maintenance personnel to manually calibrate defect locations and risk levels, which is inefficient and prone to human error. Summary of the Invention

[0003] This application solves the technical problems of inconsistent segmentation standards, low life prediction accuracy and difficulty in integrating multi-source data caused by the heterogeneity of weld inspection data by providing a weld segmentation inspection method and system based on weld life prediction. It achieves the technical effect of improving prediction accuracy by unifying segmentation rules, integrating physical and data models, and generating priority inspection reports to achieve efficient maintenance and accident risk prevention and control.

[0004] The present application provides a weld segmentation detection method based on weld life prediction. The method is applied to a weld segmentation detection system based on weld life prediction, and includes: defining weld segmentation rules; collecting and acquiring heterogeneous data from different detection equipment for registration to obtain a registration data set; extracting features from the registration data set according to the weld segmentation rules, constructing a feature framework, and constructing a weld life prediction model based on the feature framework to obtain multiple remaining life prediction results; and prioritizing the welds by segmentation according to the multiple remaining life prediction results to generate a detection decision report.

[0005] In a possible implementation, the process of defining the weld segmentation rule performs the following processing: traversing the target weld to collect data, determining the weld starting point based on the weld data set, and establishing a one-dimensional X-axis coordinate system along the weld direction with the weld starting point as the origin; performing adjacent overlap analysis according to the one-dimensional X-axis coordinate system and setting the adjacent overlap ratio; performing segmentation preset based on the adjacent overlap ratio, automatically encrypting the segment length according to the preset result, and defining the weld segmentation rule.

[0006] In a possible implementation, heterogeneous data from different detection devices are collected and registered to obtain a registered data set, and the following processing is performed: heterogeneous data from different detection devices are collected, unstructured detection is performed based on the weld data set, and multiple weld key points are extracted; the multiple weld key points are mapped to the one-dimensional X-axis coordinate system to obtain multiple coordinate groups; the multiple coordinate groups are normalized to obtain normalized coordinate values; and the normalized coordinate values ​​are added to the registered data set.

[0007] In a possible implementation, the registration data set is feature extracted according to the weld segmentation rule, a feature framework is constructed, and the following processing is performed: the normalized coordinate values ​​are matched according to the weld segmentation rule to generate segmented coordinate matching results; segmentation basic attributes are extracted, the segmented coordinate matching results are retrieved according to the segmentation basic attributes, and defect attribute data are determined; the defect attribute data is used as an index to search for welding process parameters and determine associated process parameters; the multiple weld key points are feature analyzed and reorganized according to the associated process parameters to construct the feature framework.

[0008] In a possible implementation, a weld life prediction model is constructed based on the feature framework to obtain multiple remaining life prediction results, and the following processing is performed: fracture mechanics calculations are performed based on the registration data set to obtain multiple initial life prediction values ​​for multiple segments; the feature framework is trained using a data-driven algorithm to construct a weld life prediction model; the registration data set is synchronized to the weld life prediction model for analysis to obtain multiple life prediction values ​​for multiple segments; the multiple initial life prediction values ​​are weightedly fused with the multiple life prediction values ​​to generate the multiple remaining life prediction results.

[0009] In a possible implementation, the feature framework is trained using a data-driven algorithm to construct a weld life prediction model, and the following processing is performed: structured analysis is performed based on the feature framework, a structured data set is extracted, and the structured data set is used as a training data set; the training data set is time-series processed using an LSTM model to construct a remaining life decay curve; the training data set is reverse cross-validated based on the remaining life decay curve, and the weld life prediction model is constructed based on the validation results.

[0010] In a possible implementation, the welds are prioritized by segmentation according to the multiple remaining life prediction results, an inspection decision report is generated, and the following processing is performed: a weld risk analysis is performed based on the multiple remaining life prediction results, and multiple weld segment risk levels are defined; an environmental risk factor is retrieved, and the multiple weld segment risk levels are superimposed on the environmental risk factor to obtain a comprehensive risk score; the weld segments are sorted according to the comprehensive risk score, and a priority sorting result is formulated; a trend analysis is performed on the priority sorting result and historical inspection data, and a long-term maintenance strategy is output, and the long-term maintenance strategy is added to the inspection decision report.

[0011] The present application also provides a weld segmentation detection system based on weld life prediction, including: a rule definition module, used to define weld segmentation rules; a data alignment module, used to collect and obtain heterogeneous data from different detection equipment for alignment, and obtain a alignment data set; a feature extraction module, used to extract features of the alignment data set according to the weld segmentation rules, construct a feature framework, and construct a weld life prediction model based on the feature framework to obtain multiple remaining life prediction results; a segmentation sorting module, used to sort the welds by segmentation priority according to the multiple remaining life prediction results, and generate a detection decision report.

[0012] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0013] The weld segmentation detection method and system based on weld life prediction provided in this application relate to the field of data detection technology. They solve the technical problems of inconsistent segmentation standards, low life prediction accuracy and difficulty in integrating multi-source data caused by the heterogeneity of weld detection data. They achieve the technical effect of improving prediction accuracy by unifying segmentation rules, integrating physical and data models, and generating priority detection reports to achieve efficient maintenance and accident risk prevention and control. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] To more clearly illustrate the technical solutions of the embodiments of the present application, the accompanying drawings of the embodiments of the present application will be briefly introduced below. Flowcharts are used in this application to illustrate the operations performed by the system according to the embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed in precise order. On the contrary, various steps may be processed in reverse order or simultaneously as needed. Furthermore, other operations may be added to these processes, or one or more operations may be removed from these processes.

[0015] Figure 1 A schematic flow chart of a weld segment detection method based on weld life prediction provided in an embodiment of the present application;

[0016] Figure 2A schematic structural diagram of a weld segment detection system based on weld life prediction provided in an embodiment of the present application. DETAILED DESCRIPTION

[0017] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below.

[0018] In order to make the purpose, technical solutions and advantages of this application clearer, the application will be further described in detail below with reference to the accompanying drawings. The described embodiments should not be regarded as limiting this application. All other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0019] In the following description, reference is made to “some embodiments”, which describes a subset of all possible embodiments, but it will be understood that “some embodiments” may be the same subset or different subsets of all possible embodiments and may be combined with each other without conflict, and the terms “first\second” involved are merely used to distinguish similar objects and do not represent a specific ordering of the objects. The terms “including” and “having” and any variations are intended to cover non-exclusive inclusions. 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 clearly listed, but may include other steps or modules that are not clearly listed or that are inherent to these processes, methods, products, or devices. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which this application belongs. The terms used herein are for the purpose of describing the embodiments of this application only.

[0020] The embodiment of the present application provides a weld segmentation detection method based on weld life prediction, and the method is applied to a weld segmentation detection system based on weld life prediction, such as Figure 1 As shown, the method includes:

[0021] Step A100, defining weld segmentation rules; in one possible implementation, step A100 further includes step A110, traversing the target weld to collect data, determining the weld starting point based on the weld data set, and establishing a one-dimensional X-axis coordinate system along the weld direction with the weld starting point as the origin; executing step A120, performing adjacent overlap analysis according to the one-dimensional X-axis coordinate system, and setting the adjacent overlap ratio; executing step A130, performing segmentation preset based on the adjacent overlap ratio, automatically encrypting the segment length according to the preset result, and defining the weld segmentation rules.

[0022] First, a laser scanner, a visual camera, or an ultrasonic probe and other sensors can be used to perform high-precision scanning along the weld path to collect parameters such as weld surface morphology, width, and depth to obtain a weld dataset. At the same time, the weld dataset is used to locate the weld starting point according to weld characteristics such as width mutation, depth increase, and color / texture change to obtain the weld starting point. The weld starting point is further defined as the origin along the weld direction, and the positive direction in the one-dimensional X-axis is the welding direction, thereby determining the one-dimensional X-axis coordinate system. The overlapping area length on the X-axis is used as the numerator and the adjacent weld segments are used as the denominator to calculate the adjacent overlap of the one-dimensional X-axis coordinate system to determine the overlap ratio. The adjacent overlap ratio of the weld segments is set according to the overlap ratio and the welding process requirements. The adjacent overlap ratio can be 20%-30%. Furthermore, based on the adjacent overlap ratio, For example, segmentation preset refers to setting the initial segment length according to the complexity of the weld feature, such as uniform weld: 100mm / segment; complex weld: 50mm / segment. If the weld width change rate exceeds 5% / 10mm, the segment length is shortened, and then the segment length is automatically encrypted according to the preset result, which means that for the defect clustering area, the segment length is automatically encrypted to improve the detection accuracy, that is, if the overlap ratio is <20%, the subsequent segment length is shortened (such as reducing it by 10%), and if the overlap ratio is >30%, the subsequent segment length is extended (such as increasing it by 10%). Finally, the weld segmentation rule is defined, which can include the minimum segment length: not less than the sensor sampling resolution, the maximum segment length: not exceeding the process allowable range, as well as the segment number, starting point coordinates, end point coordinates, segment length, and overlap ratio to achieve automatic segmentation and rule definition of the weld.

[0023] Execute step A200 to collect and obtain heterogeneous data from different detection devices for registration to obtain a registration data set; in one possible implementation, step A200 further includes step A210 to collect and obtain heterogeneous data from different detection devices, perform unstructured detection based on the weld data set, and extract multiple weld key points; execute step A220 to map the multiple weld key points to the one-dimensional X-axis coordinate system to obtain multiple coordinate groups; execute step A230 to normalize the multiple coordinate groups to obtain normalized coordinate values; execute step A240 to add the normalized coordinate values ​​to the registration data set.

[0024] In order to integrate the heterogeneous data of multi-source detection equipment and form a unified data set, unstructured detection is performed by accessing the original data of different detection equipment (such as ultrasonic flaw detectors, X-ray imagers, and magnetic particle detectors). This means that deep learning can be performed on image data to detect weld defects (such as cracks and pores), point cloud segmentation can be applied to point cloud data to extract weld contour features, time-frequency analysis (such as wavelet transform) can be used on signal data to identify abnormal waveforms, image key points can be determined by corner detection or feature point matching, point cloud key points can be extracted based on curvature analysis or normal vector changes, and signal key points can be located by peak detection, thereby determining multiple key points of the crack. The multiple weld key points determined can include the crack starting point, end point, bifurcation point, and defect center. Furthermore, the weld length direction is used as the X-axis. , the vertical direction (such as depth and width) is the auxiliary dimension (Y and Z axis), and the multiple weld key points are mapped to the one-dimensional X-axis coordinate system, which means that the three-dimensional coordinates can be projected along the length direction of the weld to map the three-dimensional coordinates to one-dimensional coordinates. If there is a branch weld, an offset is added according to the branch number. At the same time, each weld is segmented to generate a key point coordinate group, and one-dimensional extraction is performed based on the key point coordinate group to obtain multiple one-dimensional X coordinate groups as the final coordinate group. Furthermore, the multiple coordinate groups are normalized, which means that the X coordinate group of each weld segment is normalized to generate normalized coordinate values, and then the normalized coordinate values ​​are associated with the weld segment identification and the key point type to generate a registration data set, thereby realizing standardized processing of weld inspection data and providing high-quality input data for subsequent weld risk assessment and life prediction.

[0025] Execute step A300, extract features of the registration data set according to the weld segmentation rule, construct a feature framework, construct a weld life prediction model based on the feature framework, and obtain multiple remaining life prediction results; in one possible implementation, step A300 further includes step A310, match the normalized coordinate values ​​according to the weld segmentation rule to generate segmented coordinate matching results; execute step A320, extract segmentation basic attributes, search the segmented coordinate matching results according to the segmentation basic attributes, and determine defect attribute data; execute step A330, use the defect attribute data as an index to search for welding process parameters, and determine associated process parameters; execute step A340, perform feature analysis and reorganization on the multiple weld key points according to the associated process parameters to construct the feature framework.

[0026] Associating normalized coordinate values ​​with weld segmentation rules means mapping normalized coordinate values ​​to corresponding segmentation intervals according to preset weld segmentation rules (such as segment length, overlap ratio, and feature point distribution), and outputting a segmentation coordinate matching table. It can generate segmentation coordinate matching results containing fields such as segment number, normalized coordinate value range, and original coordinate value range. At the same time, it can extract key attributes from the segmentation coordinate matching results, which may include segment length, thickness, weld type, and material parameters, and retrieve the segmentation coordinate matching results according to the segmentation basic attributes. It means querying the defect database (such as cracks, pores, and unfusion) based on the segmentation basic attributes to obtain defect information including defect type, size, density, and relative position. The defect attribute data is further used as an index to query the process parameter database (such as welding current, voltage, speed, and shielding gas flow rate) to obtain the associated process parameters including welding current, post-weld heat treatment conditions, and service environment load spectrum. Furthermore, the associated process parameters are analyzed in correlation with the key points of the weld (such as starting point, inflection point, and defect point). The feature can be constructed by taking the characteristic indicators, that is, the welding quality score at the key point (such as uniformity of penetration and width of heat-affected zone), and the process parameter-defect correlation strength (such as the probability of cracks caused by high current), to determine the feature framework, thereby realizing the integrated processing of weld detection, defect location, process optimization, and feature modeling, and providing technical support for the intelligent welding system.

[0027] In one possible implementation, step A300 further includes step A350, performing fracture mechanics calculations based on the registered data set to obtain multiple initial life prediction values ​​for multiple segments; executing step A360, using a data-driven algorithm to train the feature framework and construct a weld life prediction model; first, extracting key parameters from the registered data set to obtain weld geometric characteristics (such as crack length, depth, shape), material properties (such as yield strength, fracture toughness), and load conditions (such as stress intensity factor KI, cyclic loading frequency), and then using the Paris formula to perform fracture mechanics calculations. The fracture mechanics calculation is performed by taking the crack length as the numerator and the number of cycles as the denominator, and calculating according to the crack material constant and the stress intensity factor amplitude. The number of cycles required for the crack to extend to the critical size is calculated by integration to obtain the initial life prediction value of each segment, thereby determining multiple initial life prediction values ​​for multiple segments, and then using a data-driven algorithm to train the feature framework, which means that an LSTM model can be selected to process time series data (such as a service environment load spectrum) to predict the remaining life attenuation curve. The model hyperparameters are optimized through cross-validation to ensure generalization ability.

[0028] In one possible implementation, step A360 further includes step A361, performing structured analysis according to the feature framework, extracting a structured data set, and using the structured data set as a training data set; executing step A362, performing time series processing on the training data set using the LSTM model, and constructing a remaining life decay curve; executing step A363, performing reverse cross-validation on the training data set based on the remaining life decay curve, and constructing the weld life prediction model according to the verification results.

[0029] Extracting a structured data set from an unstructured feature framework refers to mapping the multimodal data in the feature framework (such as weld geometry, process parameters, and defect attributes) to a structured table, then generating time series data for each weld segment, and recording feature evolution by timestamp (such as the change in crack extension length over time). The time series data is further processed using the LSTM model. The time series features (such as crack length and stress intensity factor) can be received through the input layer of the LSTM model, and multi-layer LSTM units (such as 2 layers, 128 units per layer) are used through the hidden layer of the LSTM model to capture the temporal dependency. Finally, the remaining life value or the time for the crack to extend to the critical size is predicted through the output layer of the LSTM model. The time series data is then divided into a training set (the first 80%) and a validation set (the last 20%) in chronological order. For each weld segment, the time series features are input into the LSTM model, and a curve of the remaining life prediction value changing over time is output to obtain a remaining life decay curve.

[0030] Further reverse cross-validation of the training data set based on the remaining life decay curve refers to extracting key time points (such as the time when the crack expands to 50% of the critical size) from the life decay curve, using these time points as "pseudo-labels", reversely verifying the prediction accuracy of the training data set, calculating verification indicators (such as MSE, MAE), evaluating the prediction performance of the model in different time periods, and constructing a closed-loop framework of "feature extraction-time series modeling-reverse verification-model update" to obtain the final weld life prediction model, thereby realizing the intelligent and dynamic prediction of weld life.

[0031] Execute step A370 to synchronize the registered data set to the weld life prediction model for analysis to obtain multiple life prediction values ​​for multiple segments; execute step A380 to weightedly fuse the multiple initial life prediction values ​​with the multiple life prediction values ​​to generate the multiple remaining life prediction results.

[0032] The registration data set (such as weld segment coordinates and material properties) is mapped to the feature framework to generate standardized input data, the synchronized data is input into the trained life prediction model, and the life prediction value of each segment is output. Finally, the multiple life prediction values ​​predicted by the model are compared with the multiple initial life prediction values ​​calculated by fracture mechanics to analyze the source of differences (such as model generalization ability and simplification of mechanical assumptions). The multiple initial life prediction values ​​are further weighted and fused with the multiple life prediction values. By defining weighting coefficients, the weighting coefficients of the multiple life prediction values ​​are multiplied by the multiple life prediction values, and then the values ​​are added to the multiplication of the weighting coefficients of the multiple initial life prediction values ​​with the multiple initial life prediction values ​​to generate the remaining life prediction results of each segment, thereby realizing the precision and intelligence of weld life prediction and providing technical support for industrial equipment maintenance and safety assessment.

[0033] Execute step A400 to prioritize the welds by segment according to the plurality of remaining life prediction results and generate a detection decision report.

[0034] In one possible implementation, step A400 further includes step A410, performing weld risk analysis based on the multiple remaining life prediction results, and defining multiple weld segment risk levels; executing step A420, retrieving environmental risk factors, superimposing the multiple weld segment risk levels and the environmental risk factors to obtain a comprehensive risk score; executing step A430, sorting the weld segments according to the comprehensive risk score, and formulating a priority sorting result; executing step A440, performing trend analysis on the priority sorting results and historical inspection data, outputting a long-term maintenance strategy, and adding the long-term maintenance strategy to the inspection decision report.

[0035] Based on the multiple remaining life prediction results, weld risk analysis is performed, which can be divided into high risk, that is, remaining life <1 year (immediate repair is required), medium risk, that is, 1 year ≤ remaining life ≤3 years (planned repair is required), and low risk, that is, remaining life >3 years (routine monitoring). For each weld segment, its remaining life prediction value is mapped to the corresponding risk level, thereby defining multiple weld segment risk levels, and further extracting environmental risk factors from the environmental monitoring system, which may include corrosion factors (such as chloride ion concentration, humidity), temperature factors (such as extreme high temperature / low temperature exposure time), load factors (such as vibration frequency, stress concentration factor), and then superimposing the multiple weld segment risk levels and the environmental risk factors. This means that for each weld segment, a comprehensive risk score is obtained by multiplying the superimposed risk level and the environmental factor, and the weld segment is classified according to the comprehensive risk score. The segments are sorted in descending order, and priority sorting results are formulated. Combined with historical detection data (such as crack growth rate and repair records in the past three years), the risk change trend is analyzed. That is, if the score of a segment continues to rise, it is marked as "trend deterioration"; if the score is stable or decreases, it is marked as "risk controllable". The trend analysis results are determined, and finally a long-term maintenance strategy can be formulated based on the priority sorting results and trend analysis results. The long-term maintenance strategy includes high priority + trend deterioration, that is, immediate repair, arrangement of emergency resources, high priority + controllable risk, that is, planned repair, setting a repair window period, medium priority, that is, strengthening monitoring, shortening the detection cycle, low priority, that is, routine monitoring, maintaining the existing detection plan, and further embedding the long-term maintenance strategy into the report to clarify the time nodes and resource requirements. The detection decision report provides management with a visual decision-making basis and supports resource optimization allocation.

[0036] The embodiments of the present application solve the technical problems of inconsistent segmentation standards, low life prediction accuracy and difficulty in integrating multi-source data caused by the heterogeneity of weld inspection data. It achieves the technical effect of improving prediction accuracy by unifying segmentation rules, integrating physical and data models, and generating priority inspection reports to achieve efficient maintenance and accident risk prevention and control.

[0037] In the above, refer to Figure 1 The weld segment detection method based on weld life prediction according to the embodiment of the present application is described in detail. Figure 2 A weld segmentation detection system based on weld life prediction according to an embodiment of the present application is described.

[0038] The weld segmentation inspection system based on weld life prediction according to the embodiments of the present application solves the technical problems of inconsistent segmentation standards, low life prediction accuracy, and difficulty integrating multi-source data due to the heterogeneity of weld inspection data. This system achieves the technical effect of improving prediction accuracy by unifying segmentation rules and integrating physical and data models, and generating priority inspection reports, thereby achieving efficient maintenance and accident risk prevention. The weld segmentation inspection system based on weld life prediction includes: a rule definition module 10, a data registration module 20, a feature extraction module 30, and a segment sorting module 40.

[0039] A rule definition module 10 is used to define weld segmentation rules; a data registration module 20 is used to collect and obtain heterogeneous data from different detection equipment for registration to obtain a registration data set; a feature extraction module 30 is used to extract features from the registration data set according to the weld segmentation rules, construct a feature framework, and construct a weld life prediction model based on the feature framework to obtain multiple remaining life prediction results; a segmentation sorting module 40 is used to sort the welds by segment priority according to the multiple remaining life prediction results and generate a detection decision report.

[0040] The specific configuration of the rule definition module 10 will be described in detail below. As described above, the process of defining weld segmentation rules can further include: traversing the target weld to collect data, determining the weld starting point based on the weld data set, establishing a one-dimensional X-axis coordinate system along the weld direction with the weld starting point as the origin; performing adjacent overlap analysis according to the one-dimensional X-axis coordinate system and setting the adjacent overlap ratio; performing segmentation presets based on the adjacent overlap ratio, automatically increasing the segment length based on the preset results, and defining the weld segmentation rules.

[0041] The specific configuration of the data registration module 20 will be described in detail below. As described above, heterogeneous data from different testing devices is collected and registered to obtain a registered dataset. The data registration module 20 may further include: collecting heterogeneous data from different testing devices, performing unstructured inspection based on the weld dataset, and extracting multiple weld key points; mapping the multiple weld key points to the one-dimensional X-axis coordinate system to obtain multiple coordinate groups; normalizing the multiple coordinate groups to obtain normalized coordinate values; and adding the normalized coordinate values ​​to the registered dataset.

[0042] The specific configuration of the feature extraction module 30 will be described in detail below. As described above, the feature extraction module 30 performs feature extraction on the registration dataset according to the weld segmentation rule to construct a feature framework. The feature extraction module 30 may further include: matching the normalized coordinate values ​​according to the weld segmentation rule to generate segment coordinate matching results; extracting segment basic attributes, searching the segment coordinate matching results according to the segment basic attributes to determine defect attribute data; using the defect attribute data as an index to search for welding process parameters to determine associated process parameters; and performing feature analysis and reorganization on the multiple weld key points according to the associated process parameters to construct the feature framework.

[0043] The specific configuration of the feature extraction module 30 will be described in detail below. As described above, the weld life prediction model is constructed based on the feature framework to obtain multiple remaining life prediction results. The feature extraction module 30 may further include: performing fracture mechanics calculations based on the registered dataset to obtain multiple initial life prediction values ​​for multiple segments; training the feature framework using a data-driven algorithm to construct the weld life prediction model; synchronizing the registered dataset with the weld life prediction model for analysis to obtain multiple life prediction values ​​for multiple segments; and performing a weighted fusion of the multiple initial life prediction values ​​with the multiple life prediction values ​​to generate the multiple remaining life prediction results.

[0044] The specific configuration of feature extraction module 30 will be described in detail below. As described above, the feature framework is trained using a data-driven algorithm to construct a weld life prediction model. The feature extraction module 30 may further include: performing structured analysis based on the feature framework to extract a structured dataset, using the structured dataset as a training dataset; performing time series processing on the training dataset using an LSTM model to construct a remaining life decay curve; and performing reverse cross-validation on the training dataset based on the remaining life decay curve, and constructing the weld life prediction model based on the validation results.

[0045] The specific configuration of the segmentation ranking module 40 will be described in detail below. As described above, the welds are prioritized by segmentation based on the multiple remaining life prediction results and an inspection decision report is generated. The segmentation ranking module 40 may further include: performing a weld risk analysis based on the multiple remaining life prediction results to define multiple weld segmentation risk levels; retrieving environmental risk factors and superimposing the multiple weld segmentation risk levels with the environmental risk factors to obtain a comprehensive risk score; sorting the weld segments according to the comprehensive risk score to formulate a priority ranking result; performing a trend analysis on the priority ranking result and historical inspection data to output a long-term maintenance strategy, and adding the long-term maintenance strategy to the inspection decision report.

[0046] The weld segmentation detection system based on weld life prediction provided in the embodiment of the present application can execute the weld segmentation detection method based on weld life prediction provided in any embodiment of the present application, and has the corresponding functional modules and beneficial effects of the execution method.

[0047] Although the present application makes various references to certain modules in the system according to the embodiments of the present application, any number of different modules may be used and run on the user terminal and / or server, and the various units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of the functional units are only for the convenience of distinguishing each other and are not used to limit the scope of protection of the present application.

[0048] The above specific embodiments do not constitute a limitation on the scope of protection of this application. Those skilled in the art should understand that various modifications, combinations, and substitutions may be made based on design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application shall be included within the scope of protection of this application.

Claims

1. A weld segmentation detection method based on weld life prediction, characterized in that: The method comprises: Define weld segmentation rules; Collect and obtain heterogeneous data from different detection devices for registration to obtain a registration data set; Extracting features from the registration data set according to the weld segmentation rule, constructing a feature framework, and constructing a weld life prediction model based on the feature framework to obtain multiple remaining life prediction results; The welds are prioritized by segment according to the multiple remaining life prediction results, and an inspection decision report is generated.

2. The weld segment detection method based on weld life prediction according to claim 1, characterized in that: The process of defining weld segmentation rules by: Traverse the target weld to collect data, determine the weld starting point according to the weld data set, and establish a one-dimensional X-axis coordinate system along the weld direction with the weld starting point as the origin; Perform adjacent overlap analysis according to the one-dimensional X-axis coordinate system and set an adjacent overlap ratio; The segmentation is preset based on the adjacent overlapping ratio, the segmentation length is automatically encrypted according to the preset result, and the weld segmentation rule is defined.

3. The weld segment detection method based on weld life prediction according to claim 2, characterized in that: Collect and obtain heterogeneous data from different detection devices for registration to obtain a registration dataset. The methods include: Collect heterogeneous data from different inspection equipment, perform unstructured inspection based on weld datasets, and extract multiple weld key points; Mapping the plurality of weld key points to the one-dimensional X-axis coordinate system to obtain a plurality of coordinate groups; Normalizing the multiple coordinate groups to obtain normalized coordinate values; The normalized coordinate values ​​are added to the registration dataset.

4. The weld segment detection method based on weld life prediction according to claim 3, characterized in that: Extracting features from the registration data set according to the weld segmentation rule to construct a feature framework, the method comprising: Matching the normalized coordinate values ​​according to the weld segmentation rule to generate a segmented coordinate matching result; Extracting segment basic attributes, searching the segment coordinate matching results according to the segment basic attributes, and determining defect attribute data; Using the defect attribute data as an index to search for welding process parameters and determine associated process parameters; The plurality of weld key points are subjected to feature analysis and reorganization according to the associated process parameters to construct the feature framework.

5. The weld segment detection method based on weld life prediction according to claim 1, characterized in that: A weld life prediction model is constructed based on the characteristic framework to obtain multiple remaining life prediction results, the method comprising: performing fracture mechanics calculations based on the registered data set to obtain a plurality of initial life prediction values ​​for a plurality of segments; Using a data-driven algorithm to train the feature framework and construct a weld life prediction model; Synchronizing the registered data set to the weld life prediction model for analysis to obtain a plurality of life prediction values ​​for a plurality of segments; The multiple initial life prediction values ​​are weightedly fused with the multiple life prediction values ​​to generate the multiple remaining life prediction results.

6. The weld segment detection method based on weld life prediction according to claim 5, characterized in that: The feature framework is trained using a data-driven algorithm to construct a weld life prediction model, the method comprising: Performing structured analysis based on the feature framework, extracting a structured data set, and using the structured data set as a training data set; Using the LSTM model to perform time series processing on the training data set to construct a remaining life decay curve; The training data set is subjected to reverse cross validation based on the remaining life decay curve, and the weld life prediction model is constructed according to the validation results.

7. The weld segment detection method based on weld life prediction according to claim 1, characterized in that: Segmented priority sorting of welds is performed according to the plurality of remaining life prediction results, and a detection decision report is generated, the method comprising: Performing weld risk analysis based on the multiple remaining life prediction results to define multiple weld segment risk levels; Retrieving environmental risk factors, and superimposing the plurality of weld segment risk levels with the environmental risk factors to obtain a comprehensive risk score; Sort the weld segments according to the comprehensive risk score and formulate a priority ranking result; Perform trend analysis on the priority ranking results and historical detection data, output a long-term maintenance strategy, and add the long-term maintenance strategy to the detection decision report.

8. The weld segment detection system based on weld life prediction is characterized by: The system is used to implement the weld segment detection method based on weld life prediction according to any one of claims 1 to 7, comprising: Rule definition module, used to define weld segmentation rules; The data registration module is used to collect and register heterogeneous data from different detection devices to obtain a registration data set; a feature extraction module, configured to extract features from the registration data set according to the weld segmentation rule, construct a feature framework, and construct a weld life prediction model based on the feature framework to obtain multiple remaining life prediction results; The segmented sorting module is used to sort the welds by segmented priority according to the multiple remaining life prediction results and generate a detection decision report.

Images