Piping assembly welding full life cycle data traceability control workstation and control method

Abstract

The application provides a pipeline assembly welding full life cycle data tracing control workstation and a control method, belongs to the technical field of tracing control, and establishes a full life cycle digital file from raw materials to welded joint finished products, conforms to the digital tracing requirements of HAF003 and ASME BPVC Section III for nuclear power construction quality assurance. In-service inspection reliability is enhanced: a high-precision initial manufacturing state reference line is provided. According to the calculation, the proportion of unknown signals in the in-service inspection is increased to manufacturing marks. The welding quality evaluation is changed from post-detection to process pre-control, and the defect prediction model based on cost-sensitive learning fully considers the engineering requirement of zero tolerance for missed detection in the nuclear power industry. The full-cycle data closed-loop management realizes the seamless transfer of construction period quality control data to the in-service inspection in the operation period.

Description

Technical Field

[0001] This invention belongs to the field of traceability control technology, specifically relating to a data traceability control workstation and control method for the entire life cycle of pipeline component welding. Background Technology

[0002] In fields with extremely high safety requirements, such as nuclear power and petrochemicals, the weld quality of thick-walled pipelines (such as main steam pipelines and coolant pipelines) directly affects the integrity of the primary circuit pressure boundary. Existing technologies face the following significant bottlenecks in data management and full-cycle traceability:

[0003] The problems of data silos and fragmentation exist. Currently, pre-weld assembly measurements, in-weld process monitoring, and post-weld non-destructive testing (NDT) are typically performed independently by different equipment. Pre-weld misalignment data cannot guide in-weld heat input compensation, and in-weld current and voltage fluctuations cannot be correlated to the final radiographic (RT) film display. Manufacturing data from the construction phase cannot provide an initial baseline for in-service inspection (ISI) during operation; in-service inspection lacks high-precision baselines, and pre-service inspection (PSI) before nuclear power plant operation largely relies on ultrasonic / radiological scanning, recording only macroscopic geometric dimensions. Due to the lack of precise coordinate records of the interlayer filling morphology and heat input history of welds during the construction period, it is difficult to determine whether minor signal changes detected during in-service inspections are benign geometric reflections left over from the original manufacturing process or defects that arise during service. Although existing technologies can ensure the weld appearance is qualified, risk control for internal non-fusion and interlayer inclusions still relies on welder experience and post-weld non-destructive testing. There is a lack of real-time quantitative analysis of the correlation between groove filling volume deviation, thermal deformation trend and NDT defect probability during the welding process. There is also a lack of a closed-loop condition monitoring system throughout the entire life cycle. The thermal cycle history (temperature field distribution, interlayer cooling time) data during the welding process is not structured and bound to spatial coordinates. Once weld anomalies occur during the operation of the nuclear power plant, it is difficult to trace back and it is impossible to achieve reverse tracing from in-service inspection signal anomalies to welding thermal process data during the construction period. Summary of the Invention

[0004] This invention aims to provide an intelligent welding system that can provide digital support for the entire lifecycle management of nuclear power plants. Specific objectives include:

[0005] A digital twin baseline for welds is constructed, precisely binding the pre-weld assembly geometry, interpass filling profile, welding electrical parameters, and environmental data according to spatial coordinates to form an initial manufacturing data archive that is completely mapped to the physical weld. A correlation mechanism between process parameters and non-destructive testing is established. By establishing a correlation model between the filling state feature vector and typical welding defects (lack of fusion, porosity), early warning of potential NDT risk areas during the welding process is achieved. Full-process status monitoring and data closure are realized, breaking down the data barriers between pre-weld, during-weld, and post-weld stages, so that the signal interpretation during in-service inspection can be traced back to the specific welding pass, interpass temperature, and real-time visual profile data during the construction period.

[0006] The present invention employs the following technical solution.

[0007] A method for data traceability and control of the entire lifecycle of pipe assembly welding includes:

[0008] Step 1: Collect and digitize baseline data for pipe assembly welding;

[0009] Step 2: Real-time recording and quality evaluation of multi-dimensional status data during the welding of pipe components;

[0010] Step 3: Perform post-weld digital twin model construction of pipeline components and generate in-service inspection benchmark reports.

[0011] Furthermore, step 1 specifically includes:

[0012] Step 1-1: Weld circumferential development mapping method based on the characteristic axis of the bevel root;

[0013] Step 1-2: Execute the method for generating a unique digital identity code for the weld based on structural checksum;

[0014] Steps 1-3: Form a binding storage structure between multidimensional data and spatial grid.

[0015] Furthermore, step 1-1 specifically includes:

[0016] Feature anchor point definition is performed, that is, during the pre-welding assembly scanning stage, the multimodal vision perception system scans along 360° to obtain the full-circumference misalignment distribution curve. and gap distribution curve The data is then transmitted to the intelligent control system, which automatically identifies it. global maximum point This point is defined as the circumferential zero-position anchor point of the weld. =0°;

[0017] Define the axial reference plane, that is, take the least squares fitting plane of the root of the groove in the axial direction as the axial reference plane Z=0, and the direction of the weld axis is the positive direction of the Z axis;

[0018] The unfolding mapping rule involves unfolding the circumferential weld into a two-dimensional plane starting at θ=0° and proceeding counterclockwise. The internal space mesh mapping formula is: Grid(i,j)=Point(Z_i,θ_j), where Z_i is the axial depth from the reference plane, and θ_j is the circumferential angle relative to the feature anchor point.

[0019] Furthermore, steps 1-2 specifically include:

[0020] The weld UID, which serves as the unique digital identification code for welds, consists of four segments, in the format of unit number-system code-weld classification number-check code.

[0021] Furthermore, step 2 specifically includes:

[0022] Step 2-1: Perform structured extraction of multidimensional process feature vectors;

[0023] Step 2-2: Construct a defect probability prediction model based on cost-sensitive weighted random forest;

[0024] Steps 2-3: Defect prediction interpretability output based on SHAP values.

[0025] Furthermore, step 2-1 specifically includes:

[0026] For each Grid(i,j) grid cell, the intelligent control system extracts the feature vector F(i,j) as shown below:

[0027] Geometric features include the rate of change of interlayer fusion angle dα / dθ, the fill fullness η=V_actual / V_theory, and the undercut depth of the bevel sidewall, where α is the interlayer fusion angle, θ is the circumferential angle of the weld, V_actual is the actual fill volume, and V_theory is the theoretical fill volume.

[0028] The process characteristics include the instantaneous heat input correction coefficient k_heat=(U×I×60) / (v×1000) / HI_nominal and the arc voltage variation coefficient CV_U=σ_U / μ_U, where U is the instantaneous arc voltage, I is the instantaneous welding current, v is the welding travel speed, HI_nominal is the preset rated standard heat input, μ_U is the mean value of the arc voltage within the grid time period, and σ_U is the standard deviation of the arc voltage.

[0029] The temporal characteristic is the interlayer cooling time of the grid.

[0030] Furthermore, step 2-2 specifically includes:

[0031] To construct the training dataset, we accumulate historical welding data of pipes of the same material and specifications. After each weld is inspected by RT and phased array UT, certified inspectors mark the location and type of defective mesh cells. The marked defective mesh cells are used as positive samples, and at least 5 times the number of defect-free mesh cells are randomly selected as negative samples.

[0032] Construct a cost-sensitive weight matrix, that is, in view of the nuclear power industry's zero tolerance requirement for missed detection, construct a weighted matrix, and assign the penalty weight for missed detection to 5-10 times that for false positives;

[0033] The model then outputs three types of results for each grid cell: normal, attention, and high risk. When more than three consecutive grid cells in a certain area are judged as high risk, the system triggers welding parameter intervention or post-weld key inspection prompts.

[0034] Furthermore, steps 2-3 specifically include:

[0035] For each grid cell identified as high-risk, output the SHAP contribution values ​​of each feature in sorted order.

[0036] Furthermore, step 3 specifically includes:

[0037] Step 3-1: Acquire and register the interlayer contour curve sequence;

[0038] Step 3-2: Perform the method for extracting the interface lines of the internal layers of the weld;

[0039] Step 3-3: Record and store the topology of the layer layout;

[0040] Steps 3-4: Map and visualize the digital twin model.

[0041] Furthermore, step 3-1 specifically includes:

[0042] After each layer and each weld pass is completed, the robot carries a laser profile sensor to scan along the circumference of the weld and obtain the two-dimensional profile point set P_k={(r_i,z_i)} of the current remaining bevel, where k is the scan number, r_i is the radial coordinate, and z_i is the axial coordinate. All scan data are uniformly registered to the weld feature coordinate system.

[0043] Furthermore, step 3-2 specifically includes:

[0044] The system extracts the interface line of the new weld bead by calculating the difference between two adjacent scan profiles:

[0045] Weld contour extraction is performed, that is, the radial difference Δr(θ) between the contour P_k of the k-th scan and the contour P_{k-1} of the (k-1)-th scan at the same circumferential angle is Δr(θ) = r_k(θ) - r_{k-1}(θ). If Δr(θ) > 0, then the contour increment corresponds to the cross section of the new weld, where r_k(θ) is the radial distance at the circumferential θ position of the k-th scan, and r_{k-1}(θ) is the radial distance at the circumferential θ position of the (k-1)-th scan.

[0046] Weld line identification is performed, and points of curvature abrupt change and color or reflectivity anomalies are identified on the P_{k-1} contour as key points of the weld line interface.

[0047] A pipeline assembly welding full lifecycle data traceability and control workstation includes:

[0048] An automatic assembly welding positioning system, which includes a base, a three-axis servo slide, and a three-jaw self-centering chuck connected sequentially from bottom to top;

[0049] The center frame supports the trolley, which includes a four-jaw self-centering center frame, a servo-driven moving trolley, and a linear guide pair. The linear guide pair is mounted on the base, the servo-driven moving trolley is movably mounted on the linear guide pair, and the four-jaw self-centering center frame is mounted on the servo-driven moving trolley.

[0050] A six-axis welding robot, which uses a six-axis articulated robot;

[0051] A multimodal visual perception system, comprising a laser vision sensor, an auxiliary laser ranging sensor, and a temperature sensing module;

[0052] The intelligent control system includes a distributed control architecture consisting of a host computer and a slave computer, and is equipped with an industrial touch screen human-machine interface.

[0053] Furthermore, the three-jaw self-centering chuck is equipped with detachable quick-change clamping modules, each of which is connected via a standardized quick-change interface.

[0054] Furthermore, the four-jaw self-centering center frame is synchronously driven by a servo motor through a lead screw and a bevel gear transmission mechanism connected thereto, thereby achieving self-centering clamping of the geometric center of the tube.

[0055] Furthermore, a laser vision sensor is mounted at the end of the six-axis articulated robot, with a built-in high-speed CMOS camera and a blue or green laser.

[0056] The auxiliary laser rangefinder is fixedly installed next to the automatic assembly welding positioning system;

[0057] The temperature sensing module is a non-contact infrared thermometer or a contact thermocouple, used to monitor the interlayer temperature.

[0058] The beneficial effects of the present invention are as follows: Compared with the prior art, the technical effects of the present invention include:

[0059] This invention establishes a digital archive covering the entire lifecycle from raw materials to finished welds, meeting the digital traceability requirements for nuclear power plant construction quality assurance under HAF003 and ASME BPVC Section III. Enhanced reliability of in-service inspections: It provides a high-precision baseline for the initial manufacturing state. A unique feature anchoring method ensures repeatable alignment of the coordinate systems during construction and operation. Calculations show that this technology can increase the proportion of unclear signals identified as manufacturing traces during in-service inspections, effectively reducing the risk of unplanned shutdowns due to misjudgments. Welding quality evaluation shifts from post-inspection to process pre-control. The defect prediction model based on cost-sensitive learning fully considers the nuclear power industry's zero-tolerance engineering requirements for missed inspections. The model's prediction results are interpretable through the SHAP method, supporting manual review decisions. Closed-loop management of full-cycle data enables seamless transfer of construction-phase quality control data to in-service inspections during operation. The weld digital twin model can be imported into the nuclear power plant's PLM system, providing a data foundation for life extension assessments and digital overhauls. Attached Figure Description

[0060] Figure 1 This is a schematic diagram of the structure of the pipeline component welding full life cycle data traceability and control workstation in this invention;

[0061] Figure 2 This is a schematic diagram of the structure of the central frame supporting the trolley in this invention;

[0062] Figure 3 This is a schematic diagram of the automatic assembly welding positioning system in this invention;

[0063] Figure 4 This is a schematic diagram of the structure of the six-axis welding robot in this invention;

[0064] Figure 5 This is a schematic diagram of the structure of the industrial touch screen human-machine interface in this invention;

[0065] Figure 6 This is a schematic diagram of the structure of the multimodal visual perception system in this invention;

[0066] Figure 7 This is the overall flowchart of the pipeline component welding full life cycle data traceability and control method described in this invention. Detailed Implementation

[0067] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of this invention. The embodiments described in this application are merely some embodiments of this invention, and not all embodiments. Based on the spirit of this invention, any other embodiments obtained by those skilled in the art without creative effort are within the protection scope of this invention.

[0068] like Figure 7 As shown, the present invention provides a method for data traceability and control of the entire lifecycle of welding of pipe components, comprising:

[0069] Step 1: Collect and digitize baseline data for pipe assembly welding;

[0070] The core problem addressed in Step 1 is how to establish a unique, precisely traceable, and anti-counterfeiting digital reference file for each weld, with a complete lifecycle, accurate spatial coordinates, and data integrity verification capabilities. This method differs from common spatial coordinate recording methods, specifically in the following three technical aspects, which are the preferred but non-limiting embodiments of this invention. Step 1 specifically includes:

[0071] Step 1-1: Weld circumferential development mapping method based on the characteristic axis of the bevel root;

[0072] In a preferred but non-limiting embodiment of the present invention, step 1-1 specifically includes:

[0073] Traditional methods typically use the positioner's rotation angle or the robot's TCP path points as spatial references. These external references are difficult to accurately reproduce after long-term operation or equipment replacement. This invention proposes a spatial positioning method based on feature anchor points of the weld's own geometric characteristics:

[0074] Feature anchor point definition is performed, that is, during the pre-welding assembly scanning stage, the multimodal vision perception system scans along 360° to obtain the full-circumference misalignment distribution curve. and gap distribution curve The data is then transmitted to the intelligent control system, which automatically identifies it. global maximum point (i.e., the location of the maximum misalignment), and define this point as the circumferential zero-position anchor point of the weld. =0°;

[0075] It should be noted that the multimodal vision perception system acquires the full-circumference misalignment distribution curve by scanning along 360°. and gap distribution curve The methods specifically include:

[0076] Straight pipes / elbows are clamped by a three-jaw self-centering chuck, and a four-jaw self-centering center frame provides auxiliary coaxial support.

[0077] The auxiliary laser rangefinder quickly reads the axial coordinates of the two pipe end faces, and the three-axis servo slide is finely adjusted to complete the initial alignment of the bevels;

[0078] After the multimodal visual perception system determines that there are no significant deviations in the assembly, it initiates a 360° full-circle laser scanning program; the automatic assembly welding positioning system drives the pipeline to rotate 360° at a uniform speed, and the laser continuously acquires the profile of the bevel section.

[0079] The six-axis welding robot stops the laser profilometer at the axial reference plane of the bevel (the least squares fitting plane Z=0 at the root of the bevel).

[0080] The servo-driven pipe of the automatic assembly welding positioning system rotates continuously from 0° to 360° around the circumference, with a circumferential sampling interval of 1° (corresponding to an arc length of approximately 2.6 mm for a DN300 pipe).

[0081] For every 1° rotation, a blue or green laser projects a linear laser beam that passes vertically through the bevel. A high-speed CMOS camera captures the laser stripe deformation and outputs a complete two-dimensional profile point set of the bevel at the current circumferential angle θ, including the radial coordinates r of the left outer wall of the parent material, the root of the bevel, and the right outer wall of the parent material.

[0082] The automatic assembly welding positioning system transmits the current precise circumferential angle in real time. Each frame's contour data is bound to a pair of data pairs. The coordinates of the contour point cloud are P(r,z)).

[0083] single angle Next, the corresponding algorithm is applied to calculate the offset amount at the current position. bevel gap :

[0084] Based on laser triangulation to analyze contour coordinates, geometric parameters are calculated frame by frame:

[0085] (1) Bevel gap Solve

[0086] From the present In the corresponding contour point cloud, identify the feature points at the root edge of the bevel on both the left and right sides of the bevel.

[0087] Extracting the radial coordinate difference between two points gives the bevel root clearance at that circumferential angle:

[0088] =r 左根部 -r 右根部 ,in Circumferential angle The gap at the root of the bevel at the location, r 左根部r is the radial radius of the bevel root of the left-side pipe fitting. 右根部 The radial radius of the bevel root of the pipe fitting on the right side;

[0089] (2) Solving for the offset quantity Δ(θ)

[0090] Extract the reference points on the outer wall of the left and right sections of the pipe base material respectively;

[0091] Calculate the radial height difference between the outer walls of the two parent materials to obtain the misalignment amount at this angle:

[0092] =|r 左母材 -r 右母材 |, among which Circumferential angle Angular offset measurement at the location, r 左母材 For the same circumferential angle Below, the radial radius r of the outer wall of the base material of the left-side connector. 右母材 For the same circumferential angle Below, the radial radius of the outer wall of the base material of the right-side connector;

[0093] The subpixel contour optimization algorithm is applied to filter out interference from pipe surface oxidation and minor scratches, ensuring a measurement accuracy of ±0.05mm.

[0094] After summarizing the data for the entire week and constructing a discrete sampling sequence pipeline that has rotated a full circle (0°~360°), the system obtains three sets of discrete arrays:

[0095] Circumferential angle sequence: =0 。 ,1 。 ,2 。 ,...,359 。 ;

[0096] Corresponding staggered edge sequence: , ,..., ;

[0097] Corresponding gap sequence: , ,..., ;

[0098] Calibrate the circumferential zero anchor point θ=0° to unify the curve coordinate reference:

[0099] In all Global maximum point retrieved from array (i.e., the location of the maximum misalignment), and define this point as the circumferential zero-position anchor point of the weld. =0°;

[0100] Perform coordinate translation and remapping on all angle data, that is, set the original maximum value point to 0°, and rearrange the remaining angles counterclockwise to eliminate the coordinate system deviation caused by the initial rotation zero point of the positioner.

[0101] After realignment, both sets of data are unified with the feature anchor point θ=0° as the starting point, forming standardized and reproducible full-circle data.

[0102] Interpolation and smoothing are performed to generate a continuous distribution curve:

[0103] Piecewise linear interpolation / spline smoothing is performed on discrete sampling points at 1° intervals to eliminate data spikes caused by minor rotational jitter.

[0104] Generate two continuous two-dimensional curves:

[0105] Misalignment distribution curve: the horizontal axis is the circumferential angle θ (0°~360°), and the vertical axis is the misalignment value Δ;

[0106] Gap distribution curve: the horizontal axis represents the circumferential angle θ (0°~360°), and the vertical axis represents the bevel gap Gap;

[0107] The curves are bound to the weld UID and L3-level (Z,θ) spatial grid and stored in the database as the initial geometric reference for the weld digital twin.

[0108] Define the axial reference plane, that is, take the least squares fitting plane of the root of the groove in the axial direction as the axial reference plane Z=0, and the direction of the weld axis is the positive direction of the Z axis;

[0109] The unfolding mapping rule involves unfolding the circumferential weld into a two-dimensional plane starting at θ=0° and proceeding counterclockwise. The internal space mesh mapping formula is: Grid(i,j)=Point(Z_i,θ_j), where Z_i is the axial depth from the reference plane, and θ_j is the circumferential angle relative to the feature anchor point.

[0110] Furthermore, in step 1-1, since the maximum misalignment is an objective indicator of weld geometric discontinuity, structured light vision (such as a laser profilometer) can accurately identify profile jumps of ±0.05mm. This feature can still be reproduced through weld appearance geometric measurement after welding is completed, ensuring that the spatial coordinate system can be repeatedly aligned between the construction period and the in-service inspection period.

[0111] Step 1-2: Execute the method for generating a unique digital identity code (weld UID) for the weld based on structural checksum;

[0112] In a preferred but non-limiting embodiment of the present invention, steps 1-2 specifically include:

[0113] This invention proposes a weld seam UID encoding system with self-verification and self-description capabilities, which differs from simple serial numbers or random codes:

[0114] Encoding Structure: The weld UID, which serves as the unique digital identification code for welds, consists of four segments, in the format of unit number-system code-weld classification number-check code.

[0115] Example as follows:

[0116] Unit number (3 digits): Nuclear power unit number, such as R01.

[0117] System code (4 digits): Identifier of nuclear island system, such as RCP0 (reactor coolant system).

[0118] Weld classification number (5 digits): includes pipe section number, weld type identifier (B=butt weld, S=socket weld, F=flange weld) and sequence number.

[0119] Checksum (4 digits): A hexadecimal checksum generated using the CRC-16 algorithm based on the first three segments of the string, ensuring the integrity of the encoded data during transmission and storage.

[0120] As R01-RCP0-B0012-A3F2 indicates, this represents the 12th butt weld of the reactor coolant system of Unit 1, with check code A3F2.

[0121] CRC-16 verification has been widely used in nuclear power digital instrumentation and control systems (such as the IEC 61508 functional safety standard), and can detect all single-bit errors and more than 99.9% of multi-bit errors.

[0122] Steps 1-3: Form a binding storage structure between multidimensional data and spatial grid.

[0123] Each weld seam UID corresponds to a hierarchical data storage structure, as shown in Table 1:

[0124] Table 1

[0125]

[0126] This structure is typically positioned with millimeter-level accuracy based on signal analysis from ultrasonic testing (UT) at nuclear power plants. A 1° circumferential resolution corresponds to approximately 2.6 mm of arc length on a DN300 pipe, and combined with a 1 mm axial resolution, it is sufficient to accurately locate the spatial position of minute signals while maintaining a manageable amount of stored data.

[0127] Specifically, step 1 involves a robot equipped with a laser vision sensor performing a 360° scan of the irregularly shaped pipe fittings (flanges / elbows) and the straight pipe end faces. This generates end face ellipse fitting parameters, a full-circumferential misalignment distribution curve Δ(θ), and an initial gap distribution curve Gap(θ). The system automatically identifies the global maximum point θ_max of Δ(θ) and defines it as the weld circumferential zero-position anchor point θ=0°. Based on unit information, system code, and weld type, the system generates a unique weld UID and CRC-16 checksum. A weld file is established, recording the initial geometric deviation vector [ΔX(θ), ΔY(θ), Gap(θ)], and storing it in an L1 / L2 / L3 hierarchical data structure. Here, ΔX(θ) represents the coordinate deviation in the X direction of the corresponding measuring points on the left and right pipe fittings at the circumferential angle θ; ΔY(θ) represents the coordinate deviation in the Y direction of the corresponding measuring points on the left and right pipe fittings at the circumferential angle θ. This data serves as the benchmark for determining geometric stress concentration during in-service inspections. If the original misalignment at a certain point reaches the standard upper limit of 0.5mm, the background noise threshold of the ultrasonic signal at that point will be adjusted accordingly during operation.

[0128] Step 2: Real-time recording and quality evaluation of multi-dimensional status data during the welding of pipe components;

[0129] In a preferred but non-limiting embodiment of the present invention, step 2 specifically includes:

[0130] Step 2 addresses the core issue of establishing an interpretable mapping between welding process features (visual contours, electrical parameters) and post-weld nondestructive testing results, and enabling real-time risk warnings during the welding process. This method possesses unique features that distinguish it from general machine learning, specifically including:

[0131] Step 2-1: Perform structured extraction of multidimensional process feature vectors;

[0132] In a preferred but non-limiting embodiment of the present invention, step 2-1 specifically includes:

[0133] For each Grid(i,j) grid cell, the intelligent control system extracts the feature vector F(i,j) as shown below:

[0134] Geometric features include the rate of change of interlayer fusion angle dα / dθ, fill fullness η=V_actual / V_theory, and undercut depth of the bevel sidewall, where α is the interlayer fusion angle, θ is the circumferential angle of the weld, V_actual is the actual fill volume (the actual deposited metal volume of a single mesh), and V_theory is the theoretical fill volume (the theoretical fill metal volume required by a single mesh element).

[0135] The process characteristics include the instantaneous heat input correction coefficient k_heat=(U×I×60) / (v×1000) / HI_nominal and the arc voltage variation coefficient CV_U=σ_U / μ_U (calculated within the welding time window corresponding to the grid), where U is the instantaneous arc voltage, I is the instantaneous welding current, v is the welding travel speed, HI_nominal is the preset rated standard heat input, μ_U is the mean value of the arc voltage within the grid time period, and σ_U is the standard deviation of the arc voltage;

[0136] The temporal characteristic is the interlayer cooling time of the grid (the time interval between the previous grid position).

[0137] Step 2-2: Construct a defect probability prediction model based on cost-sensitive weighted random forest;

[0138] In a preferred but non-limiting embodiment of the present invention, step 2-2 specifically includes:

[0139] This invention employs a cost-sensitive weighted random forest algorithm to construct a prediction model. Its core design, which distinguishes it from general methods, lies in:

[0140] To construct the training dataset, we accumulate historical welding data of pipes of the same material and specifications. After each weld is inspected by RT and phased array UT, certified inspectors mark the location and type of defect mesh cells (circular defect / strip defect / lack of fusion / crack). The marked defect mesh cells are used as positive samples, and at least 5 times the number of defect-free mesh cells are randomly selected as negative samples.

[0141] A cost-sensitive weighting matrix is ​​constructed, specifically a weighted matrix designed to meet the nuclear power industry's zero-tolerance requirement for missed detections. The penalty weight for a missed detection (predicting a defective cell as non-defective) is 5-10 times that for a false alarm (predicting a non-defective cell as defective). Its mathematical expression is as follows:

[0142] Cost_Matrix=[[0,C_FP],[C_FN,0]];

[0143] Where Cost_Matrix is ​​the matching cost matrix, C_FN=λ×C_FP, λ takes the value of 5-10 (which can be adjusted according to the nuclear safety level), C_FN is the false negative cost, and C_FP is the false positive cost.

[0144] The model then outputs three types of results for each grid cell: normal (defect probability < 0.2), attention (0.2-0.5), and high risk (> 0.5). When more than three consecutive grid cells in a certain area are judged as high risk, the system triggers welding parameter intervention or post-weld key inspection prompts.

[0145] In step 2-2, cost-sensitive learning is a classic solution to the imbalanced classification problem (Reference: Elkan, C. "The Foundations of Cost-Sensitive Learning", IJCAI 2001). In the field of nuclear power nondestructive testing, the acceptance of missed detections is extremely low (ASME BPVC Sec.III requires 100% volumetric inspection of critical welds), therefore, the design of high penalty weights for missed detections has a clear engineering basis.

[0146] Steps 2-3: Defect prediction interpretability output based on SHAP values.

[0147] In a preferred but non-limiting embodiment of the present invention, steps 2-3 specifically include:

[0148] This invention also introduces the SHAP method to quantify and interpret the model prediction results, providing decision support for nuclear power quality engineers:

[0149] For each grid cell identified as high-risk, output the SHAP contribution values ​​of each feature in sorted order.

[0150] For example, the mesh (θ=185°, Z=3mm) is judged as high risk, with the main contributing characteristics being: sidewall fusion angle <20° (SHAP=+0.35) and fill rate <0.85 (SHAP=+0.28).

[0151] Interpretable outputs enable quality inspectors to understand why the model makes predictions, supporting human review decisions based on ASME standards.

[0152] Specifically, step 2 includes:

[0153] During the root pass welding stage, the vision system tracks the changes in the bevel gap at the front end of the molten pool in real time and records the gap-heat input response curve.

[0154] Interlayer scanning and fill evaluation phase:

[0155] After each weld bead cools, the remaining bevel profile is scanned with a laser.

[0156] The system extracts a multidimensional feature vector F(i,j) (including fusion angle, fill rate, thermal input coefficient, etc.).

[0157] Each grid cell is evaluated using a cost-sensitive random forest model. If a cell is deemed high-risk, it is labeled with a quality concern tag and parameter intervention suggestions are triggered.

[0158] Generate a data file of the interlayer contour evolution time series.

[0159] Step 3: Perform post-weld digital twin model construction of the piping assembly and generate an in-service inspection baseline report. In a preferred but non-limiting embodiment of the present invention, step 3 specifically includes:

[0160] Step 3 addresses the core issue of how to reconstruct the discrete laser profile scanning data collected during multi-layer, multi-pass welding into a visible, queryable, and traceable three-dimensional digital twin model of the weld interior. The specific method is as follows:

[0161] Step 3-1: Acquire and register the interlayer contour curve sequence;

[0162] In a preferred but non-limiting embodiment of the present invention, step 3-1 specifically includes:

[0163] After each layer and each weld pass is completed, the robot carries a laser profile sensor to scan along the circumference of the weld seam and obtain the two-dimensional profile point set P_k={(r_i,z_i)} of the current remaining bevel, where k is the scan sequence number (corresponding to the state after the completion of a specific layer and pass), r_i is the radial coordinate, and z_i is the axial coordinate. All scan data are uniformly registered to the weld seam feature coordinate system (anchor point θ=0°, reference plane Z=0).

[0164] Step 3-2: Perform the method for extracting the interface lines of the internal layers of the weld;

[0165] In a preferred but non-limiting embodiment of the present invention, step 3-2 specifically includes:

[0166] The system extracts the interface line of the new weld bead by calculating the difference between two adjacent scan profiles:

[0167] Weld contour extraction is performed, that is, the radial difference Δr(θ) between the contour P_k of the k-th scan and the contour P_{k-1} of the (k-1)-th scan at the same circumferential angle is Δr(θ) = r_k(θ) - r_{k-1}(θ). If Δr(θ) > 0 (filling), then the contour increment corresponds to the cross section of the new weld, where r_k(θ) is the radial distance at the circumferential θ position of the k-th scan, and r_{k-1}(θ) is the radial distance at the circumferential θ position of the (k-1)-th scan.

[0168] Weld line identification is performed, and points of curvature abrupt change and color or reflectivity anomalies are identified on the P_{k-1} profile (due to the difference in optical properties between the remelted area and the original bevel surface) as key points of the weld line interface.

[0169] In step 3-2, laser triangulation has been successfully applied in weld contour inspection, with a Z-axis measurement accuracy of ±0.02 mm (Reference: Huang, W. & Kovacevic, R., "A laser-based vision system for weldquality inspection", Sensors, 2011). The change in laser reflectivity due to surface oxidation in the weld heat-affected zone can serve as an auxiliary feature for fusion line identification.

[0170] Step 3-3: Record and store the topology of the layer layout;

[0171] It should be noted that step 3-3 specifically includes:

[0172] The system establishes a topology record for each weld pass, storing the following information:

[0173] Weld markings: Layer number L, pass number P.

[0174] Weld types: root pass, fill pass, cover pass.

[0175] Execution order: Global welding sequence number.

[0176] Spatial parameters: weld centerline trajectory (curve equation on the unfolded plane), weld width, and weld height.

[0177] Welding parameters: average current, voltage, wire feed speed, welding speed, and oscillation parameters used in this welding pass.

[0178] Parent weld pointer: If there are multiple layers and multiple passes, record the pass number of the previous layer weld that the next layer weld relies on, forming a parent-child relationship topology between layers.

[0179] Steps 3-4: Map and visualize the digital twin model.

[0180] It should be noted that steps 3-4 specifically include:

[0181] The system integrates data from all layers to construct an interactive 3D digital twin of the weld seam.

[0182] Geometric reconstruction: All layer contour point sets are fitted with NURBS surfaces to generate a smooth 3D surface inside the weld.

[0183] Attribute mapping: The welding parameters of each layer area are rendered to the surface of the twin model in a color mapping manner (e.g., areas with high heat input are displayed in warm colors, and normal areas are displayed in cool colors).

[0184] Query interface: When a user clicks on any location in the twin model, the system returns the layer number, welding parameters, interpass temperature, and quality status assessment results for that point.

[0185] Output format: Supports exporting to STEP AP242 format (a 3D model standard containing manufacturing information), which can be directly imported into nuclear power plant PLM / lifecycle management systems.

[0186] Specifically, step 3 includes:

[0187] Weld reconstruction: The system summarizes all layer contour scanning data, extracts layer interface lines, establishes layer topology, and generates a digital twin model of the weld interior through NURBS surface fitting.

[0188] It should be noted that the data traceability and control method for the entire lifecycle of pipeline component welding also includes: data traceability services for in-service inspections during the operational phase, one specific embodiment of which is as follows:

[0189] Scenario: Five years after the nuclear power plant began operation, an in-service inspection ultrasonic probe detected a suspected signal at a position 112° at the root of the weld.

[0190] This invention traceability process:

[0191] Input coordinates Weld_ID_001, θ=112°, Layer=Root.

[0192] The system retrieved construction period data, showing that the original gap at this point was 2.1mm, slightly larger than the average; the heat input was adjusted by +5% during welding; and the interlayer scan showed full filling.

[0193] The conclusion supports the conclusion that the system analysis indicates the signal most likely originated from a sudden change in the root height morphology during original manufacturing (due to gap variations), rather than service-related cracks. The data supports eliminating the need for maintenance decisions, significantly reducing nuclear power plant operation and maintenance costs.

[0194] like Figures 1 to 6 As shown, the pipeline assembly welding full life cycle data traceability and control workstation of the present invention includes:

[0195] Automatic assembly welding positioning system 1, which includes a high-rigidity base, an X / Y / Z three-axis servo slide and a modular three-jaw self-centering chuck connected sequentially from bottom to top;

[0196] In a preferred but non-limiting embodiment of the present invention, the modular three-jaw self-centering chuck is equipped with a detachable quick-change clamping module, which has a dedicated structure designed for different workpieces such as flanges, elbows, and straight pipes. Each detachable quick-change clamping module is connected through a standardized quick-change interface.

[0197] The center frame supports the trolley 2, which includes a four-jaw self-centering center frame, a servo-driven moving trolley, and a linear guide pair. The linear guide pair is mounted on a high-rigidity base, the servo-driven moving trolley is movably mounted on the linear guide pair, and the four-jaw self-centering center frame is mounted on the servo-driven moving trolley.

[0198] In a preferred but non-limiting embodiment of the present invention, the four-jaw self-centering center frame is synchronously driven by a servo motor through a precision lead screw and a bevel gear transmission mechanism connected thereto, so as to achieve self-centering clamping of the geometric center of the pipe, which is suitable for pipe diameter range of Φ60.3mm-325mm.

[0199] The six-axis welding robot 3 is a high-rigidity industrial-grade six-axis articulated robot with a repeatability of ±0.05mm and a laser vision sensor installed at the end.

[0200] The multimodal visual perception system 4 includes a laser vision sensor, an auxiliary laser ranging sensor, and a temperature sensing module.

[0201] In a preferred but non-limiting embodiment of the present invention, the laser vision sensor is mounted on the end effector of a high-rigidity industrial-grade six-axis articulated robot, and has a built-in high-speed CMOS camera and a blue or green laser with a sampling frequency ≥50Hz.

[0202] The auxiliary laser rangefinder is fixedly installed next to the automatic assembly welding positioning system to quickly determine the approximate position of the pipe end face, with a measurement accuracy of ±1mm.

[0203] The temperature sensing module is a non-contact infrared thermometer or a contact thermocouple, used to monitor the interlayer temperature.

[0204] The intelligent control system includes a distributed control architecture consisting of a host computer (VRCPS) and a slave computer (PLC), equipped with an industrial touch screen human-machine interface, a multimodal vision perception system, a six-axis welding robot, an automatic assembly welding positioning system, a center frame support trolley, and the host computer, all of which are connected to the slave computer.

[0205] The method of the present invention is further illustrated by taking the welding of DN300 stainless steel main pipes and elbows used in nuclear power plants as an example:

[0206] (1) Initialization and filing: The operator enters the workpiece number RCP-PIPE-09 into the HMI. The system automatically generates the weld UID CPR-09-W01 and associates it with the material library (316L) and the qualified process specification (WPS).

[0207] (2) Pre-welding data acquisition: Visual scanning confirmed that the misalignment between the elbow bevel and the straight pipe fluctuated within the range of 0.2-0.4mm. Data recording: Pre_Weld_Alignment_Map.csv was stored in the database.

[0208] (3) Filler Weld Data Tracking: When welding to the 2nd pass of the 5th layer, the vision system scan revealed that the remaining bevel shape had asymmetrical shrinkage due to thermal deformation (deeper on the left side and shallower on the right side). The system adjusted the swing center offset by +1.2mm within milliseconds. Key Record: Database record event Event_ID:2215,Layer5_Pass2,Offset:+1.2mm,Cause:Thermal_Deformation.

[0209] (4) NDT Mapping and Delivery: After welding, phased array ultrasonic testing (PAUT) was performed on the weld. A tiny wavy pattern less than 0.5 mm was found at the 270° circumferential position. By querying the system data link, it was confirmed that this position was exactly the recorded interlayer thermal deformation compensation position, and the fusion line tortuosity caused the sound beam reflection. The system output report stated: "This display is a recordable manufacturing instruction during the construction period, not an in-service defect."

[0210] (5) Full lifecycle handover: The workstation will hand over the encrypted data package containing manufacturing geometric reference, thermal history data and compensation event log to the owner's operation department as a permanent electronic identity card for the weld.

[0211] It should be recognized that embodiments of the present invention may be implemented or carried out by computer hardware, a combination of hardware and software, or by computer instructions stored in a non-transitory computer-readable storage medium.

[0212] The method can be implemented using standard programming techniques, including a non-transitory computer-readable storage medium configured with a computer program in the computer program, wherein the storage medium is configured such that the computer operates in a specific and predefined manner.

[0213] Each program can be implemented in a high-level procedural or object-oriented programming language to communicate with the computer system; however, if required, the program can be implemented in assembly or machine language.

[0214] In any case, the language can be either compiled or interpreted.

[0215] Furthermore, for this purpose, the program can run on a programmed application-specific integrated circuit.

[0216] The processes described herein (or variations and / or combinations thereof) can be executed under the control of one or more computer systems configured with executable instructions, and can be implemented by hardware or a combination thereof as code (e.g., executable instructions, one or more computer programs, or one or more applications) that commonly executes on one or more processors. The computer program includes a plurality of instructions executable by one or more processors.

[0217] Furthermore, the method can be implemented in any suitable computing platform, including but not limited to personal computers, minicomputers, mainframes, workstations, networked or distributed computing environments, standalone or integrated computer platforms, or in communication with charged particle tools or other imaging devices.

[0218] Various aspects of the present invention can be implemented in machine-readable code stored on a non-transitory storage medium or device, whether portable or integrated into a computing platform, such as a hard disk, optical read and / or write storage medium, RAM, ROM, etc., such that it can be read by a programmable computer, and when the storage medium or device is read by the computer, it can be used to configure and operate the computer to perform the processes described herein.

[0219] Furthermore, machine-readable code, or parts thereof, can be transmitted via wired or wireless networks.

[0220] When such media includes instructions or programs that combine with a microprocessor or other data processor to implement the steps described above, the invention described herein includes these and other different types of non-transitory computer-readable storage media.

[0221] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for data traceability and control throughout the entire lifecycle of pipe assembly welding, characterized in that, include: Step 1: Collect and digitize baseline data for pipe assembly welding; Step 2: Real-time recording and quality evaluation of multi-dimensional status data during the welding of pipe components; Step 3: Execute the post-weld digital twin model construction of the piping assembly and generate the in-service inspection baseline report; Step 1 specifically includes: Step 1-1: Weld circumferential development mapping method based on the characteristic axis of the bevel root; Step 1-2: Execute the method for generating a unique digital identity code for the weld based on structural checksum; Steps 1-3: Form a binding storage structure between multidimensional data and spatial grid; Step 1-1 specifically includes: Feature anchor point definition is performed, that is, during the pre-welding assembly scanning stage, the multimodal vision perception system scans along 360° to obtain the full-circumference misalignment distribution curve. and gap distribution curve The data is then transmitted to the intelligent control system, which automatically identifies it. global maximum point This point is defined as the circumferential zero-position anchor point of the weld. =0°; Define the axial reference plane, that is, take the least squares fitting plane of the root of the groove in the axial direction as the axial reference plane Z=0, and the direction of the weld axis is the positive direction of the Z axis; The expansion mapping rule is to expand the circumferential weld into a two-dimensional plane starting from θ=0° and proceeding counterclockwise. The internal space grid mapping formula is: Grid(i,j)=Point(Z_i,θ_j), where Z_i is the axial depth from the reference plane and θ_j is the circumferential angle relative to the feature anchor point.

2. The method for data traceability and control of the entire lifecycle of pipeline component welding according to claim 1, characterized in that, Steps 1-2 specifically include: The weld UID, which serves as the unique digital identification code for welds, consists of four segments, in the format of unit number-system code-weld classification number-check code.

3. The method for data traceability and control of the entire lifecycle of pipeline component welding according to claim 2, characterized in that, Step 2 specifically includes: Step 2-1: Perform structured extraction of multidimensional process feature vectors; Step 2-2: Construct a defect probability prediction model based on cost-sensitive weighted random forest; Steps 2-3: Defect prediction interpretability output based on SHAP values.

4. The method for data traceability and control of the entire lifecycle of pipeline component welding according to claim 3, characterized in that, Step 2-1 specifically includes: For each Grid(i,j) grid cell, the intelligent control system extracts the feature vector F(i,j) as shown below: Geometric features include the rate of change of interlayer fusion angle dα / dθ, the fill fullness η=V_actual / V_theory, and the undercut depth of the bevel sidewall, where α is the interlayer fusion angle, θ is the circumferential angle of the weld, V_actual is the actual fill volume, and V_theory is the theoretical fill volume. The process characteristics include the instantaneous heat input correction coefficient k_heat=(U×I×60) / (v×1000) / HI_nominal and the arc voltage variation coefficient CV_U=σ_U / μ_U, where U is the instantaneous arc voltage, I is the instantaneous welding current, v is the welding travel speed, HI_nominal is the preset rated standard heat input, μ_U is the mean value of the arc voltage within the grid time period, and σ_U is the standard deviation of the arc voltage. The temporal characteristic is the interlayer cooling time of the grid; Step 2-2 specifically includes: To construct the training dataset, we accumulate historical welding data of pipes of the same material and specifications. After each weld is inspected by RT and phased array UT, certified inspectors mark the location and type of defective mesh cells. The marked defective mesh cells are used as positive samples, and at least 5 times the number of defect-free mesh cells are randomly selected as negative samples. Construct a cost-sensitive weight matrix, that is, in view of the nuclear power industry's zero tolerance requirement for missed detection, construct a weighted matrix, and assign the penalty weight for missed detection to 5-10 times that for false positives; The model then outputs three types of results for each grid cell: normal, attention, and high risk. When more than three consecutive grid cells in a certain area are judged as high risk, the system triggers welding parameter intervention or post-weld key inspection prompts. Steps 2-3 specifically include: For each grid cell identified as high-risk, output the SHAP contribution values ​​of each feature in sorted order.

5. The method for data traceability and control of the entire lifecycle of pipe assembly welding according to claim 4, characterized in that, Step 3 specifically includes: Step 3-1: Acquire and register the interlayer contour curve sequence; Step 3-2: Perform the method for extracting the interface lines of the internal layers of the weld; Step 3-3: Record and store the topology of the layer layout; Steps 3-4: Map and visualize the digital twin model; Step 3-1 specifically includes: After each layer and each weld pass is completed, the robot carries a laser contour sensor to scan along the circumference of the weld and obtain the two-dimensional contour point set P_k={(r_i,z_i)} of the current remaining bevel, where k is the scan number, r_i is the radial coordinate, and z_i is the axial coordinate. All scan data are uniformly registered to the weld feature coordinate system. Step 3-2 specifically includes: The system extracts the interface line of the new weld bead by calculating the difference between two adjacent scan profiles: Weld contour extraction is performed, that is, the radial difference Δr(θ) between the contour P_k of the k-th scan and the contour P_{k-1} of the (k-1)-th scan at the same circumferential angle is Δr(θ) = r_k(θ) - r_{k-1}(θ). If Δr(θ) > 0, then the contour increment corresponds to the cross section of the new weld, where r_k(θ) is the radial distance at the circumferential θ position of the k-th scan, and r_{k-1}(θ) is the radial distance at the circumferential θ position of the (k-1)-th scan. Weld line identification is performed, and points of curvature abrupt change and color or reflectivity anomalies are identified on the P_{k-1} contour as key points of the weld line interface.

6. A data traceability and control workstation for the entire lifecycle of pipe assembly welding, characterized in that, include: An automatic assembly welding positioning system, which includes a base, a three-axis servo slide, and a three-jaw self-centering chuck connected sequentially from bottom to top; The center frame supports the trolley, which includes a four-jaw self-centering center frame, a servo-driven moving trolley, and a linear guide pair. The linear guide pair is mounted on the base, the servo-driven moving trolley is movably mounted on the linear guide pair, and the four-jaw self-centering center frame is mounted on the servo-driven moving trolley. A six-axis welding robot, which uses a six-axis articulated robot; A multimodal visual perception system, comprising a laser vision sensor, an auxiliary laser ranging sensor, and a temperature sensing module; The intelligent control system includes a distributed control architecture consisting of a host computer and a slave computer, and is equipped with an industrial touch screen human-machine interface.

7. The pipeline component welding full life cycle data traceability and control workstation according to claim 6, characterized in that, The three-jaw self-centering chuck is equipped with detachable quick-change clamping modules, which are connected via standardized quick-change interfaces.

8. The pipeline component welding full life cycle data traceability and control workstation according to claim 7, characterized in that, The four-jaw self-centering center frame is synchronously driven by a servo motor through a lead screw and a bevel gear transmission mechanism connected to it, so as to achieve self-centering clamping of the geometric center of the tube; The laser vision sensor is mounted on the end of the six-axis articulated robot and has a built-in high-speed CMOS camera and a blue or green laser. The auxiliary laser rangefinder is fixedly installed next to the automatic assembly welding positioning system; The temperature sensing module is a non-contact infrared thermometer or a contact thermocouple, used to monitor the interlayer temperature.

Images