Automatic welding and quality traceability method for prestressed pipe pile reinforcement cage

Abstract

The present application relates to prestressed pipe pile production technical field, disclose a prestressed pipe pile reinforcement cage automatic welding and quality traceability method, including before welding, obtain the material state information of the reinforcement cage to be welded, and generate dynamic initial welding parameter accordingly;In the welding process, the visual, molten pool sound spectrum and arc spectrum of the welding area are collected simultaneously, and the multi-modal process information is input into the welding quality feedforward prediction model, and the welding quality prediction value is calculated in real time;Based on the deviation of the prediction value and the target value, the welding parameters are real-time controlled;The whole process data is stored in association with the unique code of the reinforcement cage.The present application combines pre-compensation before welding, feedforward control during operation and post-holographic traceability, can actively adapt to the fluctuation of raw materials, realize the online avoidance of welding defects, and provide deep reproducible data support for quality diagnosis, significantly improve the stability and reliability of the welding quality.

Description

Technical Field

[0001] This invention relates to the field of prestressed pipe pile production technology, specifically to a method for automated welding and quality traceability of prestressed pipe pile reinforcement cages. Background Technology

[0002] As an important building foundation component, the core load-bearing structure of prestressed pipe piles is the internal steel cage. The welding quality of the steel cage determines the bending resistance, crack resistance, long-term reliability, and service life of the finished pipe pile.

[0003] In the existing production practice of prestressed pipe piles, the manufacturing process of the steel cage relies on manual welding, and the post-weld inspection relies on experience-based testing. The existing production model has technical defects. First, because the welding process depends on the individual skill level and operating condition of the welder, it is difficult to ensure the uniformity of the quality of all weld points. Quality problems such as weld point misalignment, insufficient weld height, or missed welds may occur, resulting in unstable product qualification rate and limiting the overall production efficiency.

[0004] Furthermore, the manual welding process makes it difficult to control welding parameters. Even slight fluctuations in parameters such as current, voltage, and welding speed during the welding process can directly affect the weld penetration, weld width, and internal metallographic structure, resulting in differences in mechanical properties between weld points and even between different batches of products, thus increasing the potential safety risks of the overall pipe pile structure.

[0005] To control quality, existing technologies typically employ a combination of post-weld sampling visual inspection and random mechanical testing. This approach is inherently a delayed and incomplete quality verification method. The nature of sampling inspection cannot cover all weld points on the rebar cage, posing a risk of overlooking potential quality issues. Furthermore, obtaining test results, such as those from mechanical tests, often requires a long timeframe. By the time a test report indicates the presence of non-conforming products, the batch has usually already been moved to subsequent processes such as concrete pouring. The rebar cages that have undergone subsequent concrete pouring then need to be scrapped, resulting in material waste and project delays.

[0006] Furthermore, the existing production methods lack a quality traceability system. Key information such as equipment status, operators, and actual process parameters during the welding process cannot be effectively recorded and stored. Therefore, once quality problems are found in the pipe piles in downstream processes or end applications, it is impossible to effectively trace back to the specific steel cage manufacturing process, making it difficult to locate and analyze the root cause of the problem, thus hindering the continuous improvement of the process and the enhancement of the quality management level. Summary of the Invention

[0007] To address the shortcomings of existing technologies, this invention provides an automated welding and quality traceability method for prestressed concrete pipe pile reinforcement cages. This method solves the problems of existing automated welding methods for prestressed concrete pipe pile reinforcement cages, which mainly rely on fixed process parameters and are not adaptable to fluctuations in the condition of raw materials; their quality control is mostly based on post-weld inspection, which cannot predict and avoid potential defects during the welding process and lacks proactive process control measures; at the same time, the quality traceability information has a single dimension, making it difficult to analyze the root causes of welding quality problems.

[0008] To achieve the above objectives, this invention provides an automated welding and quality traceability method for prestressed concrete pipe pile reinforcement cages, comprising the following steps:

[0009] S1. Before welding the steel cage begins, obtain the material state information of the steel cage to be welded, wherein the material state information includes material surface state information and material diameter information.

[0010] S2. Based on the material state information, compensate the preset reference welding parameters to generate dynamic initial welding parameters;

[0011] S3. Using the aforementioned dynamic initial welding parameters, during the welding process performed by the automated welding unit, the visual characteristics of the welding area, the acoustic spectrum characteristics of the molten pool, and the spectral characteristics of the arc are simultaneously collected to form multimodal process information.

[0012] S4. Input the multimodal process information into the welding quality feedforward prediction model and calculate the welding quality prediction value in real time.

[0013] S5. Based on the deviation between the predicted welding quality value and the preset quality target value, a control adjustment amount is generated, and the welding parameters are adjusted in real time.

[0014] S6. The material state information, multimodal process information, control adjustment amount, welding quality prediction value and the unique code of the steel cage are associated and stored to complete quality traceability.

[0015] Preferably, step S1, the step of obtaining the material state information of the reinforcing cage to be welded, specifically includes: obtaining the surface state information of the material of the reinforcing cage to be welded using an eddy current sensor; and obtaining the diameter information of the material of the reinforcing cage to be welded using a laser profilometer. The material state information consists of a material state vector. Description:

[0016] ;

[0017] in, This is the material state vector; This refers to the surface corrosion level parameter; This refers to the surface oil stain level parameter; These are the surface coating state parameters; This refers to the real-time diameter of the reinforcing bar. superscript here The transpose operation represents a matrix or vector. In mathematical notation, it is used to convert a row vector into a column vector, and it is the standard format for inputs in control theory and machine learning models.

[0018] Preferably, in step S2, the preset benchmark welding parameters are retrieved from the process database according to the specifications and model of the reinforcing cage; in step S5, the preset quality target value is a predicted score corresponding to the predetermined qualified weld quality grade.

[0019] Among them, the preset benchmark welding parameters are a set of process parameters, including welding current, arc voltage, wire feed speed, welding speed, shielding gas flow rate, etc. The best process parameters are established in the ideal state of the material of the steel cage to be welded, which is free of rust and oil and has standard dimensions. It is the baseline for the system to make adaptive adjustments.

[0020] Preferably, in step S3, the visual features include at least one of weld width, weld reinforcement height, molten pool area, and weld center offset; the molten pool acoustic spectrum features include at least one of short-time energy, short-time zero-crossing rate, spectral centroid, and spectral entropy; and the arc spectral features include arc plasma temperature and the intensity ratio of specific element spectral lines characterizing changes in the protective atmosphere.

[0021] Among them, weld width refers to the lateral dimension of the weld on the workpiece surface; weld reinforcement refers to the height of the filler metal that extends above the base metal surface; molten pool area refers to the area of ​​the liquid metal region on the surface of the workpiece in a molten state at the moment of welding; weld center offset refers to the positional deviation between the geometric center of the actual weld and the preset theoretical welding target point; short-time energy refers to the energy of the sound signal within a very short time window; short-time zero-crossing rate refers to the number of times the sound signal waveform crosses the zero level within a very short time window; spectral centroid is a more accurate frequency measure than zero-crossing rate; spectral entropy refers to the complexity or uncertainty of the sound signal spectrum.

[0022] Preferably, in step S3, the automated welding unit includes multiple welding torches; the welding process specifically involves: using the dynamic initial welding parameters, while the rebar cage is positioned, clamped, and rotated by the rebar cage positioning and clamping unit, the multiple welding torches simultaneously weld multiple welding points on the rebar cage; wherein, the dynamic initial welding parameters include at least one of dynamic initial welding current, dynamic initial welding voltage, and dynamic initial wire feed speed, and the dynamic initial welding parameters are generated by compensating the corresponding reference welding parameters according to the material state information.

[0023] Preferably, in step S4, the welding quality feedforward prediction model is constructed through offline supervised learning training based on historical process data and corresponding post-weld quality inspection results. The input of the welding quality feedforward prediction model further includes: electrical characteristics fed back from the welding power source in real time, wherein the electrical characteristics include the mean and variance of welding current and welding voltage.

[0024] The welding quality feedforward prediction model is built through offline supervised learning training. Before actual production begins, the system needs to accumulate a large amount of historical process data and corresponding post-weld quality inspection results.

[0025] Historical process data encompasses various process characteristics acquired in real time from multimodal sensing systems during controlled testing or early production phases.

[0026] Multimodal sensing systems include industrial cameras, acoustic sensors, spectrometers, and welding power sources.

[0027] These features constitute the model's input feature vector, including weld width, molten pool area, short-time energy, spectral centroid, arc plasma temperature, and intensity of specific element spectral lines.

[0028] Post-weld quality inspection results are used to guide model learning.

[0029] By using these historical process data as input and the corresponding post-weld quality inspection results as output, and training them with machine learning algorithms, the model learns and captures the complex, nonlinear mapping relationship between welding process characteristics and final welding quality.

[0030] Furthermore, in addition to the aforementioned multimodal process information, the input of the welding quality feedforward prediction model also specifically includes electrical characteristics fed back from the welding power source in real time. These electrical characteristics are key indicators reflecting the stability of the welding process and energy input, and have a direct impact on welding quality.

[0031] The average values ​​of welding current and welding voltage represent the core energy input level during the welding process. Values ​​that are too high or too low may lead to welding defects.

[0032] The variance of welding current and welding voltage reflects the stability of the welding process. In an ideal welding process, the fluctuations of current and voltage should be small. Large variance indicates problems such as arc instability, chaotic droplet transfer, and frequent short circuits. These can directly lead to quality defects such as poor weld formation, porosity, and increased spatter. By monitoring these variances in real time, the model can capture small disturbances in the welding process and thus accurately predict potential quality risks.

[0033] By combining multimodal process information with these key electrical characteristics as input to the model, the welding quality feedforward prediction model can understand the dynamic changes of the welding process from a more comprehensive and in-depth perspective, thereby improving its accuracy and robustness in predicting welding quality and providing a reliable basis for subsequent real-time control.

[0034] Preferably, in step S4, the welding quality feedforward prediction model includes a preset functional relationship that maps multimodal process information to welding quality prediction values, and its mathematical expression can be summarized as follows:

[0035] ;

[0036] in: This is the predicted output welding quality value; Represents the predictive model function; This represents the set of internal parameters of the model; It is a visual feature vector extracted from a visually inspected image; It is the acoustic spectrum feature vector extracted from the acoustic spectrum signal of the molten pool; It is a spectral feature vector extracted from the electric arc spectrum; It is an electrical feature vector extracted from the electrical signal fed back from the welding power source; These are the dynamic initial welding parameters under the current working conditions.

[0037] The step of calculating the predicted welding quality value in real time in S4 is as follows: the multimodal process information collected in real time is substituted into the functional relationship as the independent variable for calculation, and the calculation result is used as the predicted welding quality value.

[0038] Preferably, in step S5, the control adjustment amount is generated based on the deviation between the predicted welding quality value and the preset quality target value, as well as the rate of change of the deviation;

[0039] Step S5, the step of real-time adjustment of welding parameters, specifically includes:

[0040] The control adjustment is superimposed on the currently executed welding parameters to generate real-time operating parameters, and at least one of the welding current, welding voltage, shielding gas flow rate, or wire feed speed is adjusted through a digital welding power supply, flow controller, or servo wire feeder.

[0041] The real-time control of welding parameters involves superimposing the calculated control adjustment amount onto the currently executing welding parameters to generate new real-time operating parameters. This superposition process ensures the smoothness and continuity of control adjustment. The intelligent control unit sends the updated real-time operating parameter instructions to the underlying actuators of the automated welding unit via a standard industrial bus protocol. Specifically, this real-time control is achieved through one or more of the following methods:

[0042] The welding current or welding voltage can be adjusted using a digital welding power source.

[0043] The flow rate of protective gas delivered to the welding area is dynamically adjusted using an electronic mass flow controller.

[0044] The wire feeding speed is controlled by a servo motor-driven wire feeder.

[0045] In this way, the system applies control adjustments in real time to at least one of the welding current, welding voltage, shielding gas flow rate, or wire feed speed. This dynamic control process continues throughout the welding process, forming a closed-loop feedback control link with online quality prediction as its core. This effectively suppresses quality disturbances caused by factors such as arc fluctuations and changes in the local state of the workpiece.

[0046] Preferably, in step S5, the process of generating the control adjustment amount further includes: when the acoustic spectrum characteristics of the molten pool indicate a risk of welding defects, the arc spectrum characteristics are judged in conjunction; if the arc spectrum characteristics indicate an abnormal protective atmosphere, the control adjustment amount for adjusting the flow rate of the protective gas is generated.

[0047] Preferably, in step S6, the quality traceability includes: performing multi-dimensional dynamic reproduction of the welding process of any weld point based on the associated stored data, wherein the multi-dimensional dynamic reproduction synchronously displays the time series of visual images of the welding area, acoustic spectrum data of the molten pool, and arc spectrum data.

[0048] Preferably, the quality traceability step further includes: based on the associated stored data, performing multi-dimensional dynamic reproduction of the welding process of any specified weld point on the rebar cage. This reproduction function is realized through a visual traceability interface. When quality traceability is required, the user can retrieve its corresponding digital holographic file by scanning the QR code on the rebar cage or by entering the unique code of the rebar cage in the system, and select the specific weld point to be analyzed.

[0049] After the reproduction is initiated, the visual traceability interface synchronously displays the multi-dimensional information sequence of the solder joint during its formation process on a unified timeline.

[0050] Specifically, the interface simultaneously displays a sequence of visual monitoring images of the welding area, a time-spectrum diagram of the acoustic spectrum data of the molten pool, a curve showing the change in intensity of key element spectral lines in the arc spectrum, and real-time curves showing the change in various parameters, including welding current, voltage, wire feed speed, and quality prediction values.

[0051] The visual image sequence is a frame-by-frame video recording of the welding process, providing direct visual evidence for analyzing physical phenomena such as arc morphology, droplet transfer, molten pool behavior, and spatter. The time-spectrum diagram of the molten pool acoustic spectrum data, in the form of a two-dimensional waterfall plot, shows the changes in the frequency components of the acoustic signal over time, which can reveal specific acoustic events such as droplet short circuits and arc bursts. The change curve of the arc spectral data intuitively reflects the stability of the protective atmosphere in the welding area (such as the spectral intensity of nitrogen and oxygen elements) and the fluctuation of the arc temperature.

[0052] By precisely aligning and synchronously visualizing multi-dimensional data such as visual, acoustic, spectral, and electrical data, technicians can establish direct causal relationships between different physical phenomena. This multi-dimensional dynamic reproduction method provides objective and comprehensive data support for in-depth analysis of the root causes of welding quality problems and process optimization.

[0053] This invention provides an automated welding and quality traceability method for prestressed concrete pipe pile reinforcement cages. It has the following beneficial effects:

[0054] 1. This invention pre-scans the material condition information of the steel cage to be welded before welding to obtain real-time material surface condition and diameter information, and dynamically compensates the reference welding parameters according to the material condition information of the steel cage to be welded. This method enables the initial process parameters of welding to match the specific material condition of each section of the steel cage to be welded, eliminating the negative impact of uncertain factors such as raw material corrosion, oil stains or dimensional tolerances on welding quality from the source, and improving the stability and consistency of weld quality in mass production.

[0055] 2. During the welding process, this invention simultaneously collects multi-dimensional information such as visual data, molten pool acoustic spectrum, and arc spectrum, and uses a welding quality feedforward prediction model for real-time analysis. This allows the system to predict the trend of quality decline before physical defects are formed. Based on this prediction result, the system can perform closed-loop control in advance to eliminate potential quality problems, achieve proactive quality assurance in the process, and fundamentally reduce the defect rate.

[0056] 3. This invention associates and stores multi-dimensional data from the entire process before, during, and after welding, including material state, multi-modal process information, control adjustment amounts, and quality prediction values, with the unique code of the reinforcing cage. This provides static parameter traceability, enables traceability of product quality, and allows for multi-dimensional dynamic reproduction of the welding process of any weld point, including visual images, acoustic spectrum data, and spectral data. This provides data support for the root cause diagnosis of welding quality problems and the continuous optimization of process parameters. Attached Figure Description

[0057] Figure 1 This is a flowchart of the method of the present invention;

[0058] Figure 2 This is a logic diagram for generating digital holographic archives in this invention;

[0059] Figure 3 This is a flowchart of the welding process in this invention;

[0060] Figure 4 This is a real-time feedback control loop diagram for the welding process of the present invention. Detailed Implementation

[0061] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0062] Please see the appendix Figure 1 This invention provides an automated welding and quality traceability method for prestressed pipe pile reinforcement cages. The welding preparation stage of this method, that is, before welding is performed, includes the assessment of material condition and the adaptive generation of welding parameters.

[0063] Before welding of the reinforcing cage begins, a step is performed to obtain the material status information of the reinforcing cage to be welded. This step is achieved through a material pre-scanning unit, which is set on the feeding path of the production line, before the welding station. The material pre-scanning unit integrates an eddy current sensor and a high-frequency laser profilometer, which are used to continuously scan the material status information of the reinforcing cage to be welded.

[0064] Among them, the eddy current sensor is used to acquire the surface condition information of the material of the steel cage to be welded. When the eddy current sensor is working, the internal induction coil generates an alternating magnetic field on the surface of the steel bar. This magnetic field induces eddy currents on the surface of the steel cage to be welded. When there are changes in the state of the material surface of the steel cage to be welded, such as rust, oil stains or uneven coating, it will cause a local change in the conductivity or permeability of the material of the steel cage to be welded, which in turn causes changes in the amplitude and phase of the eddy current. By detecting the amount of change of the eddy current, a signal reflecting the surface condition of the steel cage to be welded can be obtained.

[0065] The high-frequency laser profilometer is used to obtain the material diameter information of the steel cage to be welded. When the high-frequency laser profilometer is working, the internal laser emitting device emits a laser beam to the material surface of the moving steel cage to be welded, and the laser receiving device receives the laser reflected from the material surface of the steel cage to be welded. The distance from the laser point to the sensor is calculated by the triangulation principle or the time-of-flight principle, thereby obtaining the profile and diameter data of the steel cage to be welded.

[0066] The acquired raw sensor signals are processed by signal processing circuits or software algorithms and quantified into specific material state parameters. For example, the changes in amplitude and phase of the electrical signal output by the eddy current sensor are mapped to predefined surface corrosion level parameters, surface oil contamination level parameters, and surface coating state parameters. Ultimately, at any given moment, the material state information of the reinforcing bar to be welded can be represented by a material state vector. Description:

[0067] ;

[0068] in, This is the material state vector; This refers to the surface corrosion level parameter; This refers to the surface oil stain level parameter; These are the surface coating state parameters; This refers to the real-time diameter of the reinforcing bar. superscript here The transpose operation represents a matrix or vector. In mathematical notation, it is used to convert a row vector into a column vector, and it is the standard format for inputs in control theory and machine learning models.

[0069] After acquiring the material state information, the process involves compensating the preset baseline welding parameters based on the material state information to generate dynamic initial welding parameters. This step is executed by the intelligent control unit.

[0070] First, the intelligent control unit retrieves a set of matching benchmark welding parameters from the local or cloud-based process database based on the specifications and model of the rebar cage to be produced. The baseline welding parameters are a set of process parameters that have been verified to be effective under ideal material conditions. Subsequently, the intelligent control unit will use the material state vector acquired in real time. As input, a pre-defined adaptive compensation model is substituted. The calculation is performed to determine a parameter adjustment amount. The adaptive compensation model can be a rule-based function or a regression model trained on historical data.

[0071] ;

[0072] in, For parameter adjustment amount; An adaptive compensation model; This is the material state vector.

[0073] Finally, the intelligent control unit will adjust the calculated parameters. With reference welding parameters The dynamic initial welding parameters are generated by superposition and then applied to this welding task. .

[0074] ;

[0075] The dynamic initial welding parameters It includes specific parameters such as dynamic initial welding current, dynamic initial welding voltage, and dynamic initial wire feed speed. These values ​​have been compensated for the actual material condition of the current reinforcing steel, providing more suitable initial process conditions for the subsequent welding process.

[0076] In the process of performing welding using the dynamically generated initial welding parameters, the method of this embodiment further includes the steps of real-time monitoring, prediction and control of the welding process.

[0077] While performing welding through an automated welding unit, visual features of the welding area, acoustic spectrum features of the molten pool, and arc spectrum features are simultaneously collected to form multimodal process information. This collection process is completed by a multimodal collaborative real-time quality monitoring unit.

[0078] The multimodal collaborative real-time quality monitoring unit is integrated into the working area of ​​the automated welding unit to ensure the time synchronization of all sensor data, thereby enabling accurate correlation of various physical phenomena occurring during the welding process.

[0079] The specific acquisition process includes visual feature acquisition. One or more high-speed, high-resolution industrial cameras are used to clearly capture images of the welding area. These cameras are mounted coaxially with the welding torch or at a specific viewing angle to ensure the field of view covers key areas such as the molten pool, weld formation, and arc root. The industrial cameras capture video streams and image frames of the welding process in real time. Image processing algorithms can then extract various visual features, including but not limited to:

[0080] Weld width: The transverse dimension of the molten pool, reflecting the degree of fusion;

[0081] Weld reinforcement: The height of the weld surface protruding from the base material, reflecting the amount of filler in the weld;

[0082] Molten pool area: The two-dimensional projected area of ​​the molten metal pool, which is closely related to the welding heat input and penetration depth;

[0083] Weld point center offset: The deviation between the actual weld point center and the preset welding trajectory, reflecting the accuracy of the welding position;

[0084] Arc morphology and stability: The combustion state of the electric arc is assessed by analyzing its brightness, shape, and jitter.

[0085] Spatter characteristics: The stability of droplet transfer, as well as the quantity and size of spatter, reflect the quality of the welding process;

[0086] The acoustic spectrum characteristics of the molten pool are acquired through one or more highly sensitive acoustic sensors. These sensors are installed close to the welding area, but effectively shielded from environmental noise and welding torch mechanical noise. The sensors are equipped with high-temperature and interference-resistant protective covers. The acoustic sensors capture the acoustic signals generated during the welding process in real time. These sounds are caused by physical events such as arc combustion, droplet transfer, metal solidification, and bubble collapse. By performing Fourier transform (FFT) or other time-frequency analyses on these raw acoustic signals, their acoustic spectrum characteristics can be obtained, including but not limited to:

[0087] Short-time energy: reflects the intensity of the sound signal over a short period of time, and is related to the instantaneous power and stability of the electric arc;

[0088] Short-time zero-crossing rate: The number of times an acoustic signal crosses zero in a short period of time, reflecting the frequency composition and noise level of the signal;

[0089] Spectral centroid: used to distinguish different types of droplet transition modes;

[0090] Spectral entropy: The complexity or disorder of the acoustic spectrum. A high entropy value may indicate instability or defects in the welding process.

[0091] The acquisition of arc spectral characteristics is achieved through a high-resolution spectrometer. Arc light is coupled into the spectrometer via a fiber optic probe, which is precisely aimed at the core region of the arc. Collimators and filters ensure the acquisition of representative arc plasma light. The spectrometer analyzes the emitted light in real time, decomposing it into spectra of different wavelengths and measuring the intensity of each wavelength. By analyzing this spectral data, various arc spectral characteristics can be extracted, including but not limited to:

[0092] Arc plasma temperature: The temperature of the arc plasma is calculated by analyzing the intensity ratio or broadening of spectral lines of specific elements, reflecting the welding heat input and penetration capability.

[0093] Characterizing the intensity ratio of specific elemental spectral lines in protective atmosphere changes: By monitoring the intensity and ratio changes of characteristic spectral lines of elements in the protective gas itself or in the air, the stability and purity of the protective gas flow rate and the presence of air entrapment in the welding area can be assessed in real time, which is crucial for preventing defects such as porosity.

[0094] Multimodal collaboration and real-time performance:

[0095] Time synchronization: All sensors (industrial cameras, acoustic sensors, spectrometers, and electrical characteristics fed back from the welding power source) are synchronized through a shared global clock or a high-precision time synchronization protocol, ensuring that all acquired data are aligned at any given time, thereby enabling accurate correlation between different modal information.

[0096] Data preprocessing: Before being transmitted to the welding quality feedforward prediction model, the collected raw data will undergo preliminary preprocessing (such as image denoising, acoustic spectrum denoising, spectral smoothing, etc.) to improve data quality and the accuracy of feature extraction.

[0097] Ultimately, these synchronized and preprocessed visual features, molten pool acoustic spectrum features, arc spectrum features, and real-time electrical features together constitute a rich and comprehensive multimodal process information.

[0098] To acquire visual features, high-definition industrial cameras and laser contour sensors are deployed in the welding area. The cameras acquire image sequences of the welding area in real time. Image processing algorithms, such as edge detection and threshold segmentation algorithms, are used to analyze the images to calculate geometric features such as weld width, weld reinforcement height, molten pool area, and weld center offset. These geometric features together constitute the visual feature vector. .

[0099] To obtain the acoustic spectrum characteristics of the molten pool, a high-sensitivity acoustic sensor is installed non-contactly near the welding torch, and sound insulation and noise reduction measures are taken. This sensor collects acoustic emission signals generated by the welding arc and the molten pool. The collected raw acoustic signals are processed by bandpass filtering to remove background noise from the production line environment. Subsequently, short-time Fourier transform (STFT) processing is performed to convert the one-dimensional time-domain acoustic signal into a two-dimensional time-spectrum diagram. Based on this time-spectrum diagram, acoustic spectrum characteristics that characterize the droplet transition behavior and the stable state of the molten pool, such as short-time energy, short-time zero-crossing rate, spectral centroid, and spectral entropy, are calculated. These acoustic spectrum characteristics collectively constitute the molten pool acoustic spectrum feature vector. .

[0100] To acquire the spectral characteristics of the electric arc, a fiber optic probe is integrated with the welding torch, and the other end is connected to a miniature spectrometer. This device acquires spectral data of the arc plasma emission in real time. Spectral analysis algorithms are used to identify spectral lines in the acquired data to locate characteristic spectral lines of specific elements such as iron (Fe), nitrogen (N), hydrogen (H), and oxygen (O). By calculating the intensity ratio of the spectral lines of specific elements to those of the matrix metal elements, the purity change of the protective atmosphere in the welding area is quantified. Simultaneously, the Boltzmann slope method is used to analyze the spectral line intensity and calculate the arc plasma temperature. These parameters together constitute the arc spectral characteristic vector. .

[0101] After acquiring the multimodal process information, it is input into a pre-defined welding quality feedforward prediction model to calculate the predicted welding quality value in real time. Before inputting the model, the visual feature vectors are collected synchronously. , Molten pool acoustic spectrum feature vector Arc spectrum eigenvectors And electrical characteristic vectors fed back in real time from digital welding power sources. The states are merged to form a comprehensive process state vector. .

[0102] FEED Prediction Model for Welding Quality It is built through offline training. The training process includes: collecting a large amount of historical welding process data, where each set of data contains a complete sequence of welding process state vectors and its corresponding post-weld quality inspection results (e.g., results obtained through non-destructive testing or mechanical performance testing), and using supervised learning algorithms (e.g., Long Short-Term Memory network LSTM) to train this data to obtain the model's internal parameter set. .

[0103] During the welding process, the intelligent control unit will collect the process state vector in real time. As a pre-trained prediction model The model takes the input as input and performs calculations based on a preset functional relationship, outputting a continuously changing predicted value for welding quality. .

[0104] This process can be described by the following formula:

[0105] ;

[0106] in, In order to be in Predicted welding quality at any given time; Represents the predictive model function; In order to be in The process state vector at any given time; This is the set of internal parameters of the model.

[0107] After obtaining the predicted welding quality value, a control adjustment amount is generated based on the deviation between the predicted value and the preset quality target value, and the welding parameters are adjusted in real time. The intelligent control unit continuously calculates the predicted welding quality value. With a preset quality target value Deviation between .

[0108] During the welding process, the system continuously calculates the predicted actual welding quality. And continuously adjust the welding parameters, with the aim of making Get as close as possible to and stabilize at On this value.

[0109] Based on this deviation and its rate of change, through a feedforward control function (For example, a PID control algorithm) calculates and generates a control adjustment. .

[0110] ;

[0111] in, This refers to the amount of control adjustment; It is a feedforward control function; This is a quality deviation; The rate of change of quality deviation; This is the time integral of the quality deviation.

[0112] Please see the appendix Figure 4 In one specific implementation, the process of generating the control adjustment amount also includes the linkage judgment of multimodal features. For example, when the molten pool acoustic spectrum feature vector... The parameters indicate the risk of porosity defects, leading to changes in the predicted value. During the descent, the intelligent control unit simultaneously determines the characteristic vector of the electric arc spectrum. If an abnormally high intensity of the characteristic spectral line of nitrogen (N), which characterizes air intrusion, is detected simultaneously, then the control function... It will generate a control adjustment that prioritizes increasing the flow rate of the protective gas.

[0113] Finally, the generated control adjustment amount These parameters are superimposed on the currently executed welding parameters to generate real-time operating parameters. Furthermore, by means of actuators such as digital welding power supplies, flow controllers, or servo wire feeders, at least one of the following can be adjusted: welding current, welding voltage, shielding gas flow rate, or wire feed speed, in order to suppress the occurrence of quality fluctuations.

[0114] The current deviation, the cumulative deviation, and the trend of deviation change are all input into the control function. In this process, the system can generate a fast and stable control adjustment value. This enables real-time, proactive, and high-precision control of welding quality.

[0115] Please see the appendix Figure 2 and attached Figure 3 In one embodiment of the present invention, the method further includes the step of archiving the entire process data to achieve quality traceability after the welding process is completed.

[0116] This step associates and stores the material state information, multimodal process information, generated control adjustment quantities, and calculated welding quality prediction values ​​obtained in the previous steps with the unique code of the rebar cage. Before each rebar cage is produced, the intelligent control unit generates a unique identification code for it and uses equipment such as a laser marking machine to solidify the identification code on the main reinforcement of the rebar cage.

[0117] Throughout the welding process, all collected or generated data is tagged with a precise timestamp and bound to a unique identifier. This ensures that all data, from the state of the raw materials to the final welded product, points to a unique physical entity. Based on this association mechanism, the system constructs an independent digital holographic file for each rebar cage. .

[0118] The data structure of this file is as follows:

[0119] ;

[0120] in: It is a digital holographic archive; It is the unique code of the steel cage; It is a pre-welding data set, which includes the specifications and model of the rebar cage, operator information, equipment status records, and complete material status information obtained by the material pre-scanning unit; It is a welding process data set, which is a set of timestamped multivariate time series data containing real-time operating parameters throughout the welding process. Comprehensive process state vector and continuous welding quality prediction values ; It is a post-weld data set, which includes offline sampling non-destructive testing reports or mechanical property test results, product flow information, and the final finished product inspection report.

[0121] Digital holographic archives are simultaneously stored on local servers and in cloud databases to build the data foundation for quality traceability.

[0122] The quality traceability steps in this embodiment of the invention further include multi-dimensional dynamic reproduction of the welding process of any weld point based on the associated stored data. When quality traceability is required, the user can retrieve the corresponding digital holographic file by scanning the QR code on the steel cage or by entering its unique code in the system.

[0123] The system provides a visual traceability interface that can synchronously replay multi-dimensional information of a specified weld point during the welding process on a unified timeline. Specifically, the interface synchronously displays the visual monitoring image sequence of the welding area, the time-spectrum diagram of the acoustic spectrum signal of the molten pool, the change curves of the intensity of the spectral lines of key elements in the arc spectrum, and the real-time change curves of various parameters including welding current, voltage, wire feed speed, and quality prediction values. This method of reproducing and synchronously visualizing visual, acoustic, spectral, and electrical data in a time-aligned manner provides objective and comprehensive data for technicians to diagnose quality problems and perform process optimization analysis.

Claims

1. A method for automated welding and quality traceability of prestressed concrete pipe pile reinforcement cages, characterized in that, Includes the following steps: S1. Before welding the steel cage begins, obtain the material state information of the steel cage to be welded, wherein the material state information includes material surface state information and material diameter information. S2. Based on the material state information, compensate the preset reference welding parameters to generate dynamic initial welding parameters; S3. Using the aforementioned dynamic initial welding parameters, welding is performed by an automated welding unit. During the welding process, visual features of the welding area, acoustic spectrum features of the molten pool, and arc spectrum features are simultaneously collected to form multimodal process information. S4. Input the multimodal process information into the welding quality feedforward prediction model and calculate the welding quality prediction value in real time. S5. Based on the deviation between the predicted welding quality value and the preset quality target value, a control adjustment amount is generated, and the welding parameters are adjusted in real time. S6. The material state information, multimodal process information, control adjustment amount, welding quality prediction value and the unique code of the steel cage are associated and stored to complete quality traceability.

2. The method for automated welding and quality traceability of prestressed concrete pipe pile reinforcement cages according to claim 1, characterized in that, In step S1, the step of obtaining the material state information of the steel cage to be welded specifically includes: The surface condition information of the steel cage to be welded is obtained by using an eddy current sensor. The material diameter information of the steel cage to be welded is obtained by using a laser profilometer.

3. The method for automated welding and quality traceability of prestressed concrete pipe pile reinforcement cages according to claim 1, characterized in that, In step S2, the preset benchmark welding parameters are retrieved from the process database according to the specifications and model of the reinforcing cage; in step S5, the preset quality target value is a predicted score corresponding to the predetermined qualified weld quality grade.

4. The method for automated welding and quality traceability of prestressed concrete pipe pile reinforcement cages according to claim 1, characterized in that, In step S3, the visual features include at least one of weld width, weld reinforcement height, molten pool area, and weld center offset; the molten pool acoustic spectrum features include at least one of short-time energy, short-time zero-crossing rate, spectral centroid, and spectral entropy; and the arc spectral features include arc plasma temperature and the intensity ratio of specific element spectral lines characterizing changes in the protective atmosphere.

5. The automated welding and quality traceability method for prestressed concrete pipe pile reinforcement cages according to claim 1, characterized in that, In step S3, the automated welding unit includes multiple welding torches; the specific steps for performing welding are as follows: Using the dynamic initial welding parameters, while the steel cage is positioned, clamped and rotated by the steel cage positioning and clamping unit, the multiple welding torches simultaneously weld multiple welding points on the steel cage. The dynamic initial welding parameters include at least one of dynamic initial welding current, dynamic initial welding voltage, and dynamic initial wire feed speed, and the dynamic initial welding parameters are generated by compensating the corresponding reference welding parameters based on the material state information.

6. The method for automated welding and quality traceability of prestressed concrete pipe pile reinforcement cages according to claim 1, characterized in that, In step S4, the welding quality feedforward prediction model is constructed through offline supervised learning training based on historical process data and corresponding post-weld quality inspection results. The inputs to the welding quality feedforward prediction model also include: The electrical characteristics fed back from the welding power source in real time include the mean and variance of the welding current and welding voltage.

7. The method for automated welding and quality traceability of prestressed concrete pipe pile reinforcement cages according to claim 1, characterized in that, In step S4, the welding quality feedforward prediction model includes a preset functional relationship that can map multimodal process information to welding quality prediction values; In step S4, the step of obtaining the predicted welding quality value in real time is specifically as follows: The multimodal process information collected in real time is substituted into the functional relationship as independent variables for calculation, and the calculation result is used as the predicted value of welding quality.

8. The method for automated welding and quality traceability of prestressed concrete pipe pile reinforcement cages according to claim 1, characterized in that, In step S5, the control adjustment amount is generated based on the deviation between the predicted welding quality value and the preset quality target value, as well as the rate of change of the deviation; Step S5, the step of real-time adjustment of welding parameters, specifically includes: The control adjustment is superimposed on the currently executed welding parameters to generate real-time operating parameters, and at least one of the welding current, welding voltage, shielding gas flow rate, or wire feed speed is adjusted through a digital welding power supply, flow controller, or servo wire feeder.

9. The automated welding and quality traceability method for prestressed concrete pipe pile reinforcement cages according to claim 1, characterized in that, In step S5, the step of generating the control adjustment amount further includes: When the acoustic spectrum characteristics of the molten pool indicate a risk of welding defects, the arc spectrum characteristics are judged in conjunction with the process. If the arc spectrum characteristics indicate an abnormal protective atmosphere, the control adjustment amount for adjusting the protective gas flow rate is generated.

10. The method for automated welding and quality traceability of prestressed concrete pipe pile reinforcement cages according to claim 1, characterized in that, In step S6, the quality traceability includes: Based on the associated stored data, the welding process of any weld point is dynamically reproduced in multiple dimensions. The multidimensional dynamic reproduction synchronously displays the visual image of the welding area, the acoustic spectrum data of the molten pool, and the time series of the arc spectrum data.

Images