Welding pipe inner welding rib height control method and system based on extrusion amount dynamic compensation

Abstract

The application discloses a kind of based on extrusion amount dynamic compensation's welded pipe inner weld rib height control method and system, it is related to high-frequency resistance welded pipe production technical field, it aims at solving the technical problem that existing inner weld rib height control scheme is caused by measurement point and execution point physical separation and leads to control instruction lag behind actual state change, cannot carry out advance prediction compensation, comprising: preset sampling frequency is continuously collected extrusion roller rear welded pipe inner wall's inner weld rib height time series data;Inner weld rib height time series data is differentially operated, at least extract first derivative feature, second derivative feature and third derivative feature;First derivative feature, second derivative feature and third derivative feature are input into preset prediction model;The application is by constructing based on inner weld rib height time series signal's multiple derivative feature extraction and advance prediction compensation mechanism, solved the problem that control instruction lag behind actual state change is caused by measurement point and execution point physical separation.

Description

Technical Field

[0001] This invention relates to the field of high-frequency resistance welded pipe production technology, and more specifically, to a method and system for controlling the height of weld beads inside welded pipes based on dynamic compensation of extrusion amount. Background Technology

[0002] In the production of high-frequency resistance welded (ERW) pipes, the inner weld bead (also known as the inner burr), an internal protrusion formed at the weld seam during the welding process, directly affects the zinc consumption level of subsequent galvanizing processes, the corrosion resistance of the pipe, and the internal fluid transport efficiency when its height is effectively controlled. Excessively high inner weld bead protrusions not only lead to uneven galvanized layer thickness and a significant increase in zinc consumption, but also make the uneven weld bead area more prone to corrosion, shortening the service life of the pipe. Therefore, controlling the height of the inner weld bead is a technical challenge in the welded pipe industry.

[0003] Currently, the control of the height of the inner weld bead mainly focuses on the following technical approaches: The first is post-processing removal, which involves removing or thinning the inner weld bead after it has formed through mechanical scraping or flattening. For example, a tool shank assembly can be used to remove burrs from the formed welded pipe online, or a flattening device can be used to physically roll the weld bead. While this approach can reduce the height of the inner weld bead, it increases process costs and can easily damage the pipe substrate and cause secondary defects such as indentations during the removal process. The second approach is static adjustment based on offline analysis. For example, metallographic samples are used to determine the correspondence between the extrusion deformation and weld parameters, and then the extrusion deformation is adjusted to obtain optimal weld parameters. Although this approach addresses the issue from the forming stage, its adjustment is based on offline metallographic testing, resulting in a long adjustment cycle and an inability to cope with real-time dynamic fluctuations during production. The third approach is a feedback control scheme based on the current measured deviation. This involves driving the actuator to compensate for the deviation of the inner weld bead height from the target value. While this scheme achieves online control, the physical distance between the measurement point and the extrusion roller's execution point creates a time lag between the deviation's occurrence, detection, and actuator response, causing the control command to always lag behind the actual state change. Therefore, this invention provides a method and system for controlling the inner weld bead height of welded pipes based on dynamic compensation of the extrusion amount. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for controlling the height of weld beads inside welded pipes based on dynamic compensation of extrusion amount, so as to solve the technical problem in the existing internal weld bead height control scheme that the control command lags behind the actual state change due to the physical separation of the measurement point and the execution point, and the inability to perform advanced prediction compensation.

[0005] To solve the above-mentioned technical problems, the present invention provides the following technical solution: a method for controlling the height of weld beads inside welded pipes based on dynamic compensation of extrusion amount, comprising: The time-series data of the height of the inner weld bead on the inner wall of the welded pipe behind the extrusion roller is continuously collected at a preset sampling frequency; Differential operation is performed on the time series data of the inner weld bead height to extract at least the first derivative feature, the second derivative feature and the third derivative feature, wherein the first derivative feature represents the instantaneous rate of change of the inner weld bead height, the second derivative feature represents the trend of the rate of change, and the third derivative feature represents the abrupt change characteristics of the signal. The first-order derivative features, second-order derivative features, and third-order derivative features are input into a preset prediction model. The prediction model predicts the height of the inner weld bead at a preset future time based on the different time scale changes represented by each derivative feature, and outputs the predicted value of the inner weld bead height. Calculate the predicted height deviation between the predicted value of the inner weld bead height and the preset target height value; Based on the predicted height deviation, the extrusion amount correction is calculated according to the preset extrusion amount-inner weld bead height mapping relationship, and the extrusion roller gap adjustment mechanism is driven to perform the extrusion amount compensation action before the future time is reached.

[0006] Preferably, the prediction model is a polynomial extrapolation model based on Taylor expansion, and the predicted value of the inner weld bead height is calculated by polynomial extrapolation based on the measured value of the inner weld bead height at the current time and the first, second, and third derivative features; or the prediction model is an artificial neural network model or a support vector regression model trained offline, and the first, second, and third derivative features are the model input features.

[0007] Preferably, the method further includes a multi-derivative feature frequency band differential compensation step: based on the time scale characteristics corresponding to the first-order derivative feature, second-order derivative feature, and third-order derivative feature, respectively, calculate the instantaneous micro-compensation component corresponding to the millisecond-level instantaneous fluctuation, the trend tracking compensation component corresponding to the second-level trend change, and the benchmark correction compensation component corresponding to the minute-level or hour-level gradual drift, and generate a comprehensive extrusion compensation command by weighting and superimposing the three types of compensation components.

[0008] Preferably, the method further includes an online adaptive correction step for prediction deviation: comparing the measured value of the inner weld bead height after the future time with the corresponding predicted value of the inner weld bead height, and calculating the prediction deviation; classifying the prediction deviation into systematic deviation and random deviation; updating the parameters of the prediction model online for the systematic deviation using the recursive least squares method; and temporarily correcting the compensation amount of the extrusion amount at the current time for the random deviation.

[0009] Preferably, it also includes an emergency compensation step for sudden interference: real-time monitoring of the absolute value of the third derivative feature; when the absolute value exceeds a preset threshold, it is determined to be a sudden interference event, triggering an emergency compensation mode; calculating the emergency extrusion correction amount based on the amplitude and direction of the third derivative feature; and simultaneously adjusting the high-frequency welding heat input parameters.

[0010] Preferably, the time-series data of the inner weld bead height is acquired by a vision inspection unit located behind the extrusion roller. The vision inspection unit is a structured light three-dimensional measurement device or a laser contour scanning device, which has high temperature protection and noise filtering functions.

[0011] Preferably, the extrusion roller gap adjustment mechanism is a hydraulically driven gap adjustment mechanism or an electric servo-driven gap adjustment mechanism, used to adjust the extrusion roller gap in response to the extrusion amount correction to change the extrusion amount.

[0012] A control system for the height of weld beads inside a welded pipe based on dynamic compensation of extrusion amount includes: The data acquisition module is used to continuously acquire the time-series data of the height of the inner weld bead on the inner wall of the welded pipe behind the extrusion roller at a preset sampling frequency; The feature extraction module is used to perform differential operations on the time-series data of the inner weld bead height, and to extract at least the first derivative feature, the second derivative feature, and the third derivative feature. The advanced prediction module is used to input the multi-derivative features into a preset prediction model, predict the height of the inner weld bead at a preset future time, and output the predicted value of the inner weld bead height. The deviation calculation module is used to calculate the predicted height deviation between the predicted value of the inner weld bead height and the preset target height value. The compensation execution module is used to calculate the extrusion amount correction based on the predicted height deviation, and drive the extrusion roller gap adjustment mechanism to perform the extrusion amount compensation action before the future time arrives.

[0013] An electronic device includes a memory and a processor, wherein the memory stores a computer program, and the processor executes the program to implement a method for controlling the height of weld beads inside welded pipes based on dynamic compensation of extrusion amount.

[0014] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements a method for controlling the height of weld beads inside welded pipes based on dynamic compensation of extrusion amount.

[0015] Compared with the prior art, the beneficial effects of the present invention are: 1. This invention constructs a multi-derivative feature extraction and advance prediction compensation mechanism based on the time-series signal of the inner weld bead height. First, it continuously collects time-series data of the inner weld bead height on the inner wall of the welded pipe behind the extrusion roller at a preset sampling frequency. Then, it performs differential operations on the collected time-series data to extract at least first-order, second-order, and third-order derivative features. After inputting these multi-derivative features into the prediction model, the model makes an advance prediction of the inner weld bead height at future times based on the different time-scale changes represented by each derivative feature and outputs the predicted value. The predicted value is then compared with a preset target height value to calculate the predicted height deviation. Finally, the extrusion amount correction is calculated based on this predicted height deviation, and the extrusion roller gap adjustment mechanism is driven to perform the extrusion amount compensation action before the future time arrives. This control logic transforms lagging control into predictive control, enabling the extrusion amount compensation command to be executed before the actual height deviation occurs, achieving compensation and improving the accuracy and consistency of inner weld bead height control.

[0016] 2. This invention further introduces a multi-derivative characteristic frequency band differentiated compensation mechanism on the basis of advanced predictive control. Based on the different time scale characteristics corresponding to the first-order derivative characteristics, second-order derivative characteristics, and third-order derivative characteristics, this invention calculates the instantaneous micro-compensation component, trend tracking compensation component, and benchmark correction compensation component respectively. The three compensation components are weighted and superimposed to generate a comprehensive squeezing amount compensation command. The calculation and superposition of the three components finally outputs a comprehensive squeezing amount compensation command that can both quickly respond to sudden disturbances and stably track long-term changes, avoiding the contradiction between fast response and high stability in the traditional single compensation strategy.

[0017] 3. This invention further constructs a dual-guarantee mechanism of emergency compensation for sudden interference and online adaptive correction of prediction deviation, based on frequency band differentiated compensation. On the one hand, this invention monitors the absolute value of the third derivative characteristic in real time. When the value exceeds a preset threshold, it is determined to be a sudden interference event and triggers the emergency compensation mode. The emergency extrusion amount correction is calculated based on the amplitude and direction of the third derivative characteristic, and the high-frequency welding heat input parameters are adjusted simultaneously. The emergency response is triggered before the measured value of the inner weld bead height has significantly shifted due to the interference, thus achieving early intervention. On the other hand, this invention compares the measured value of the inner weld bead height after the future arrival with the predicted value to calculate the prediction deviation. After online analysis of the prediction deviation, it distinguishes between systematic deviation and random deviation. For systematic deviation, the model parameters of the prediction model are updated online using the recursive least squares method. For random deviation, the extrusion amount compensation at the current moment is temporarily corrected. By separating and differentiating systematic and random biases, the predictive model can continuously self-optimize during long-term operation to adapt to gradual degradation processes such as equipment wear and sensor drift, while maintaining an immediate response capability to sudden external interference, preventing batch quality accidents caused by equipment aging or sudden unknown interference. Attached Figure Description

[0018] Figure 1 This is an overall flowchart of the method of the present invention; Figure 2 This is the logic diagram for multi-derivative feature extraction and advance prediction in this invention; Figure 3 This is a flowchart of the frequency band differentiation compensation process of the present invention; Figure 4 This is a flowchart of the online adaptive correction process for prediction bias of the present invention; Figure 5 This is a flowchart illustrating the experimental verification process of the present invention. Detailed Implementation

[0019] like Figures 1 to 4 As shown, this invention relates to a method for controlling the height of the inner weld bead in welded pipes based on dynamic compensation of extrusion amount. This method can be widely applied in automated production lines for various high-frequency resistance welded pipes, and is particularly suitable for the manufacturing processes of square and rectangular pipes, round pipes, and other products with strict requirements for the inner surface quality of the weld. The operating environment typically includes: an inner weld bead forming station, an extrusion device, a real-time height acquisition device, and an industrial controller or computer as the control core. The method includes: Step S1: During the inner weld bead forming process, the time sequence data of the inner weld bead height on the inner wall of the welded pipe behind the extrusion roller is continuously collected at a preset sampling frequency; The time-series data of the inner weld bead height refers to a series of numerical values ​​reflecting the height of the inner wall protrusion of the weld, continuously collected at fixed time intervals at a specific location behind the extrusion roller during the welding production process. This data is the fundamental information source for all subsequent calculations and control decisions.

[0020] Step S2: Perform differential operation on the collected time series data of the inner weld bead height, and extract at least the first derivative feature, the second derivative feature, and the third derivative feature; Among them, the first derivative feature refers to the physical quantity extracted by performing a first-order difference operation on the time series data of the inner weld bead height.

[0021] The second derivative feature refers to the physical quantity extracted by performing a further difference operation on the first derivative feature.

[0022] The third derivative feature refers to the physical quantity extracted by performing a further difference operation on the second derivative feature.

[0023] Step S3: Input the extracted first-order derivative features, second-order derivative features, and third-order derivative features into the prediction model. The prediction model will make an advance prediction of the height of the inner weld bead at future times based on the time scale change information represented by each derivative feature, and output the predicted value of the inner weld bead height. The prediction model refers to a mathematical model that runs in the controller or processor of this application. Its input is extracted multi-order derivative features, and its output is a predicted value of the weld bead height at a future time. This model can be a mathematical formula based on physical properties, or a machine learning model obtained by training on data.

[0024] The prediction model refers to a mathematical model that takes the first, second, and third derivative features extracted from the measured values ​​of the inner weld bead height at the current and historical times as input, and outputs the predicted value of the inner weld bead height at a preset future time. The prediction model can be a multinomial extrapolation model based on Taylor expansion, an offline-trained artificial neural network model, or a support vector regression model. The specific form and parameters of the model are determined through theoretical derivation or offline training before implementation, and can be updated online during operation by an adaptive correction step.

[0025] Step S4: Compare the predicted height of the inner weld bead with the preset target height value, and calculate the predicted height deviation; The target height refers to the upper limit control value of the inner weld bead height determined according to the welded pipe product specifications, subsequent galvanizing process requirements, and relevant standards such as GB / T3091. In other words, the control system aims to maintain the inner weld bead height at a target level. The target height is preset by process engineers based on specific product requirements and stored in the control system's parameter table.

[0026] Step S5: Based on the predicted height deviation, calculate the extrusion amount correction according to the preset extrusion amount-inner weld bead height mapping relationship, and drive the extrusion roller gap adjustment mechanism to perform the extrusion amount compensation action before the future time arrives.

[0027] The extrusion amount-inner weld bead height mapping relationship refers to a correspondence established in advance through process experiments and theoretical analysis, used to describe the quantitative correlation between the change in the extrusion amount of the extrusion roller and the final change in the height of the inner weld bead. This mapping relationship is the core conversion basis for calculating the extrusion amount correction.

[0028] The aforementioned extrusion amount-inner weld bead height mapping relationship refers to a mathematical model established in advance through process calibration experiments, describing the quantitative relationship between the change in extrusion amount of the extrusion roller and the change in the inner weld bead height. This mapping relationship is stored in the control system in the form of a functional expression, an interpolation table, or a fitted curve. The specific method for establishing this mapping relationship is as follows: before welded pipe production, under the target production speed and welding power conditions, the extrusion amount is changed by a preset step size, the corresponding steady-state inner weld bead height value is recorded, and the extrusion amount-inner weld bead height relationship curve is obtained through linear regression or polynomial fitting.

[0029] In this embodiment, the above steps work collaboratively. The entire control process begins with step S1, where the inner weld bead has just formed and has not yet fully cooled, by continuously sensing minute changes in height. Step S2 uses first, second, and third derivative features to capture the rate, acceleration, and abrupt change in height change, respectively. The prediction model in step S3 is no longer based solely on the current deviation but integrates these changes at different time scales to extrapolate the evolution of height at future prediction step lengths (prediction step length). This refers to the target time increment predicted by the prediction model from the current moment, with a reference value of 0.5s. This value is determined based on the production line operating speed of 25m / min and the distance from the measurement point to the center of the extrusion roller of 500mm (corresponding to the material transfer delay between the measurement point and the execution point). Steps S4 and S5, based on this future prediction deviation, calculate the amount of extrusion that needs to be changed to "compensate" for this impending deviation through a preset mapping relationship, and then drive the execution mechanism.

[0030] Taking a specific application scenario in a high-frequency straight seam welded pipe production line as an example: the steel pipe runs at a constant speed, and the measuring point is located behind the extrusion rollers. At time... Step S1 involves acquiring the current height value of the inner weld bead using a laser contour scanning device. Step S2 calculates the current first derivative based on the most recently collected consecutive height sequence. +0.1 mm / s (height is slowly increasing), second derivative The third derivative is +0.01 mm / s² (the rate of ascent itself is also accelerating). This is a small positive value close to 0 (acceleration changes smoothly). After receiving these features, the prediction model in step S3 predicts the future... At a certain moment, the height of the inner weld bead will exceed the preset target value. Step S4 immediately calculates this predicted height deviation, and step S5, based on this future deviation, calculates the height deviation at the current moment. An instruction is issued in advance to slightly reduce the gap between the extrusion rollers to increase the extrusion amount, thereby offsetting the height deviation before it actually occurs.

[0031] In one embodiment, regarding step S2 mentioned above, the first derivative feature characterizes the changing trend and rate of change of the inner weld bead height at the current moment, the second derivative feature characterizes the speed of change of this rate of change, and the third derivative feature characterizes the rate of change of the second derivative feature. These three derivative features are progressively more advanced: the first derivative provides the direct direction and speed of change, serving as the benchmark momentum term for predicting future positions; the second derivative reveals whether the change is strengthening or converging, providing a judgment on the sustainability of the trend; and the third derivative captures whether the process is smooth and whether new abrupt factors are about to intervene. Through this in-depth calculation of the essence of signal changes, a solid physical basis is provided for advanced prediction, avoiding trend misjudgments caused by relying solely on simple numerical extrapolation.

[0032] In one embodiment, for step S3 mentioned above, the prediction model is a polynomial extrapolation model based on Taylor expansion, which includes at least first-order, second-order, and third-order terms; the predicted value of the inner weld bead height is obtained by polynomial extrapolation calculation based on the measured value of the inner weld bead height at the current moment and the first-order derivative feature, second-order derivative feature, and third-order derivative feature.

[0033] Specifically, let the current time be... The sampling interval is The measured value of the inner weld bead height at the current moment is ,but: The first derivative feature The calculation formula is: ; The second derivative feature The calculation formula is: ; The third derivative characteristic The calculation formula is: ; Let the prediction step size be... Then the future moment Predicted value of inner weld bead height The formula for polynomial extrapolation is: In the formula: The current time indicates the latest acquisition time of the time-series data on the height of the inner weld bead; The sampling interval represents the time difference between two consecutive data collections of the inner weld bead height. For the current moment Measured value of the height of the inner weld bead; The previous sampling time Measured value of the height of the inner weld bead; For the current moment The first derivative characteristic (rate of change / velocity) is calculated by dividing the difference between the current measured altitude and the previous measured altitude by the sampling interval. For the current moment The second derivative characteristic (acceleration) is calculated by dividing the difference between the rate of change at the current moment and the rate of change at the previous moment by the sampling interval; To predict the step size, we represent the step size from the current time. The time increment for forward prediction; For the future The predicted value of the inner weld bead height is the output of the polynomial extrapolation model; For the current moment The third derivative characteristic (jerk) is calculated by dividing the difference between the acceleration at the current moment and the acceleration at the previous moment by the sampling interval.

[0034] In another embodiment, regarding step S3 mentioned above, the prediction model is an offline-trained artificial neural network model or support vector regression model. The first-order derivative features, second-order derivative features, and third-order derivative features serve as input features to the artificial neural network model or support vector regression model, which outputs the predicted value of the inner weld bead height. This approach provides an alternative data-driven prediction scheme. Compared to Taylor extrapolation models based on explicit mathematical formulas, artificial neural networks or support vector regression models can automatically learn the complex nonlinear dynamic characteristics of the inner weld bead height variation and the high-order coupling relationships between multi-order derivative features that are difficult to describe with fixed formulas from massive amounts of historical production data.

[0035] As a specific implementation, the artificial neural network model can be a fully connected feedforward network containing an input layer, several hidden layers, and an output layer. Its training data comes from sample pairs consisting of derivative feature sequences and corresponding true future height values ​​collected over a long period on the production line, covering different working conditions (such as different specifications, different materials, and different speeds). The support vector regression model can use the RBF kernel function to handle nonlinear mappings.

[0036] In one embodiment, the method further includes an online adaptive correction step for prediction deviation: comparing the measured value of the inner weld bead height after the future time with the predicted value of the inner weld bead height to calculate the prediction deviation; performing online analysis on the prediction deviation to classify the prediction deviation into systematic deviation and random deviation; updating the model parameters of the prediction model online using the recursive least squares method for the systematic deviation; and temporarily correcting the extrusion compensation amount at the current time for the random deviation.

[0037] This online adaptive correction step for prediction bias forms an external closed-loop optimization circuit. Its specific operational logic is as follows: when the prediction step size... After a period of time, step S1 naturally collected the time. True height value Compare this actual value with the previous value at time [time]. The prediction bias is obtained by subtracting the predicted value calculated by S3. At this point, an analyzer performs an online diagnostic of the bias. If the bias is consistently larger or smaller in one direction, it is judged as a systematic bias, indicating that the prediction model itself has inherent errors due to factors such as equipment wear and sensor drift. In this case, the recursive least squares method is activated to use the systematic error value to perform a small-step recursive correction on the parameters inside the model, prompting the model parameters to evolve in the direction of reducing errors. Conversely, if the bias appears suddenly, is short-lived, and has no clear direction, it is judged as a random bias, indicating that it is caused by unpredictable external interference such as voltage fluctuations or instantaneous impurities. In this case, the model itself is not corrected, but a temporary correction amount proportional to the random bias is directly added to the currently calculated extrusion compensation amount.

[0038] Specifically, let's assume a future time. The measured height of the inner weld bead after arrival is [value missing]. The predicted value of the inner weld bead height is Then the prediction bias The calculation formula is: In the formula: For future moments, i.e., predicting the step size The corresponding target time; For the future The actual height of the inner weld bead after arrival is acquired in real time by the vision inspection unit. For the future The predicted value of the inner weld bead height is obtained by calculation using a polynomial extrapolation model; Let the time series of prediction bias be... The time series of the prediction bias is analyzed online to determine the prediction bias. Decomposed into systematic bias and random bias : In the formula: For the future The prediction bias is the difference between the measured value and the predicted value. A positive value indicates that the prediction is too low, and a negative value indicates that the prediction is too high. This refers to the systematic bias component in the prediction bias, reflecting the persistent bias caused by model parameter aging or sensor zero-point drift. This refers to the random bias component in the prediction bias, reflecting transient biases caused by sudden external disturbances. Regarding the aforementioned systematic deviation The model parameters of the prediction model are updated using the recursive least squares method. The iterative calculation formula for the recursive least squares method is as follows: ; ; ; In the formula: for The model parameter estimation vector at time step includes the parameters to be identified in the Taylor extrapolation model (such as the weight coefficients corresponding to each derivative). This is the model parameter estimation vector updated by recursive least squares method; for The gain vector at time step controls the contribution of the new information (i.e., the current prediction bias) to the correction of the model parameters; for The covariance matrix at time t represents the uncertainty of the model parameter estimation; This is the covariance matrix updated using the recursive least squares method; for The regression vector at time step contains the input features of the Taylor extrapolation model at the current time step; The forgetting factor, with a value ranging from 0 to 1, is used to control the rate at which the weights of historical data decay during the recursive process. It is an identity matrix, and its dimensions are the same as those of the model parameter vector; For the random bias The temporary correction amount of the current squeeze compensation amount The calculation formula is: In the formula: This is a temporary correction amount for the current amount of extrusion compensation. This is the random deviation compensation coefficient, a preset positive number used to control the sensitivity of random deviation to the compensation amount correction.

[0039] In the formula system, the model parameter estimation vector Regression vector Gain vector Covariance matrix There is a recursive iterative relationship among the four elements. Input the model to calculate the predicted value. It is obtained by comparing the predicted value with the measured value. Depend on and Calculations show that Depend on and The product is corrected. Depend on and The four variables are recursively updated to form a complete online self-calibrating loop for the parameters.

[0040] As a specific implementation method, the distinction between systematic bias and random bias can be achieved by calculating the moving average of several predicted bias values ​​within the most recent sliding time window. If the moving average deviates from zero and exceeds a preset small threshold, the bias in the current period is considered to be systematic bias.

[0041] In one embodiment, the method further includes a multi-derivative feature frequency band differential compensation step: based on the different time scale characteristics corresponding to the first-order derivative feature, the second-order derivative feature, and the third-order derivative feature, an instantaneous micro-compensation component, a trend-following compensation component, and a benchmark correction compensation component are calculated respectively; the instantaneous micro-compensation component, the trend-following compensation component, and the benchmark correction compensation component are weighted and superimposed to generate a comprehensive compression amount compensation command; wherein, the first-order derivative feature is used to drive the instantaneous micro-compensation component corresponding to millisecond-level instantaneous fluctuations, the second-order derivative feature is used to drive the trend-following compensation component corresponding to second-level trend changes, and the third-order derivative feature is used to drive the benchmark correction compensation component corresponding to minute-level or hour-level gradual drift.

[0042] The internal logic of this step is to use the first, second, and third derivative features extracted by S2 as inputs to three parallel channels. In the first channel, the first derivative... It reflects a high instantaneous rate; any rapid fluctuation will be reflected in... The effect is immediately apparent, so it is multiplied by a gain factor. Generate instantaneous micro-compensation components for rapid hedging of transient disturbances. In the second channel, the second derivative... This reflects the acceleration of change, typically representing a macroscopic trend change lasting several seconds, and is therefore multiplied by a gain factor. Generate a trend-following compensation component to closely track and counteract this trend change. In the third channel, the third derivative... As a higher-order nonlinear characteristic, it is highly sensitive to extremely low-frequency baseline drift measured in minutes or hours caused by factors such as roller temperature rise and slow elastic deformation of the archway. Therefore, it is multiplied by a gain factor. Generates a reference correction compensation component for slowly but steadily correcting baseline drift. Finally, independent weighting coefficients are applied to these three compensation components, which are respectively for high-frequency, mid-frequency, and low-frequency interference sources. , , The results are then summed and combined to produce a final output.

[0043] Specifically, let the current time be... First derivative characteristics Second derivative characteristics and third derivative characteristics Input three parallel compensation component calculation units respectively: The instantaneous micro-compensation component The calculation formula is: In the formula: It is an instantaneous micro-compensation component used to cope with instantaneous fluctuations at the millisecond level, and its amplitude is proportional to the first derivative characteristic. The preset gain coefficient is the first derivative characteristic, used to map the rate of change signal to the amplitude of the instantaneous micro-compensation (the velocity gain is the scaling factor used to map instantaneous fluctuations to the instantaneous compensation intensity). The trend-following compensation component The calculation formula is: In the formula: The trend-following compensation component is used to address trend changes on a second-level scale, and its amplitude is proportional to the second derivative characteristic. The preset gain coefficient is the second derivative characteristic, used to map the acceleration signal to the amplitude of the trend tracking compensation (the acceleration gain is the scaling factor used to map the trend change rate to the tracking compensation intensity). The reference correction compensation component The calculation formula is: In the formula: As a reference correction compensation component, it is used to deal with slowly varying drift on the order of minutes or hours, and its amplitude is proportional to the third derivative characteristic. The preset gain coefficient for the third derivative characteristic is used to map the jerk signal to the amplitude of the reference correction compensation (the jerk gain is used to map the slowly varying drift rate to the scaling factor of the reference correction intensity). The comprehensive compression compensation command The weighted summation calculation formula is as follows: In the formula: The comprehensive extrusion compensation command is the weighted superposition result of the three compensation components, which is the final control command output to the extrusion roller gap adjustment mechanism. The preset weighting coefficients for the instantaneous micro-compensation components are used to adjust the weight of instantaneous fluctuation compensation in the entire integrated instruction. The preset weighting coefficients for the trend-following compensation component are used to adjust the weight of trend-following compensation in the entire integrated instruction. These are preset weighting coefficients for the reference correction compensation component, used to adjust the weight of the reference correction compensation in the entire integrated instruction.

[0044] As a specific implementation method, , , These are the preset weighting coefficients for the three compensation components, satisfying... The initial reference triplet is The three coefficients can be simultaneously fine-tuned around the initial reference value triplet, provided that the normalization constraint is met. The independent fine-tuning amplitude of each coefficient shall not exceed ±20% of its respective initial reference value. The feasible range is 0.48 to 0.72. The feasible range is 0.24 to 0.36. The feasible interval is from 0.08 to 0.12, and the sum of the three values ​​is always equal to 1. The intersection condition of this feasible region is equivalent to: , , When all three fall within the aforementioned interval, the normalization constraint automatically limits their degrees of freedom to 2.

[0045] In one embodiment, regarding step S1 mentioned above, a vision inspection unit located inside the welded pipe behind the extrusion roller continuously acquires the time-series data of the inner weld bead height. This vision inspection unit is either a structured light 3D measurement device or a laser contour scanning device. The vision inspection unit acquires the 3D contour data of the inner weld bead and extracts its height value in a high-temperature, steamy environment inside the welded pipe. This specifies the specific implementation method of the front-end sensing. Both structured light 3D measurement and laser contour scanning can acquire high-density point clouds or 2D contour lines of the inner weld bead surface contour in a non-contact manner at high frequencies. In harsh industrial environments such as high temperatures and steam, this type of optical measurement method has advantages over contact sensors such as mechanical probes, including faster response, zero wear, and accuracy unaffected by the physical deformation of the probe. To ensure stable operation in this environment, as a specific implementation, the vision inspection unit is typically encapsulated in a heat-resistant housing with air cooling and air curtain protection. The internal image processing board or backend processor uses specific filtering algorithms, such as morphological filtering or Gaussian smoothing, to suppress interference from water vapor scattering and ambient light noise from the raw contour data, accurately identifying the root plane and top tip of the weld bead, thereby calculating the precise height value of the inner weld bead. .

[0046] The visual inspection unit is an integrated optical measurement device consisting of a laser projector, an industrial camera, an optical lens, a high-temperature protective housing, a compressed air purging device, and a signal processing board. The laser projector projects line-structured light or coded structured light stripes onto the weld area inside the welded pipe. The industrial camera acquires the contour image of the weld area at a preset frame rate. The signal processing board or background processor performs filtering, contour extraction, and height calibration on the acquired image, outputting the height value of the inner weld bead. This device is fixedly installed inside the welded pipe behind the extrusion roller via a flange or bracket. The distance between the measurement point and the center of the extrusion roller is determined according to the production line layout, typically between 300mm and 800mm. The high-temperature protective housing adopts a double-layer water-cooled or air-cooled structure. The compressed air purging device forms an air curtain in front of the optical lens to prevent welding spatter and water vapor from adhering to the lens surface. The signal processing board incorporates morphological filtering or Gaussian smoothing filtering algorithms to suppress water vapor scattering and ambient light noise.

[0047] In one embodiment, the method further includes an emergency compensation step for sudden interference: real-time monitoring of the absolute value of the third derivative feature; when the absolute value of the third derivative feature exceeds a preset threshold, it is determined to be a sudden interference event; triggering an emergency compensation mode; calculating the emergency extrusion correction amount based on the amplitude and direction of the third derivative feature; and simultaneously adjusting the high-frequency welding heat input parameters.

[0048] This step's triggering mechanism relies on the third derivative. The ultra-sensitive characteristic to signal abrupt changes. When sudden situations occur, such as hard spots in the tube blank material or instantaneous instability in welding power, the height of the inner weld bead will show a "turning point" precursor. This symptom appears before the height value itself has significantly shifted, or even before the first and second derivatives are obvious, the third derivative... The absolute value will spike instantly and exceed the threshold. Once triggered, the system immediately enters an ultra-high priority emergency compensation mode. In this mode, compensation is no longer performed step-by-step according to frequency bands, but is directly based on... The amplitude and sign are determined by a higher emergency gain coefficient. A large-scale emergency extrusion correction is calculated and output as a control command, overriding the regular comprehensive compensation command. Simultaneously, the system recognizes that such abrupt changes are often the result of thermo-mechanical coupling imbalances, therefore it synchronously calculates the adjustment amount of the heat input power and coordinates the adjustment of the welding heat input.

[0049] Specifically, let the current time be... The third derivative is characterized by The preset threshold is The criteria for determining a sudden interference event are: When the aforementioned determination condition is met, the emergency compression amount is corrected. The calculation formula is: In the formula: It exhibits third-order derivative characteristics (accelerometer), and thus possesses a naturally high sensitivity to abrupt signals; The preset threshold is the trigger threshold for determining sudden interference events. Emergency compensation is initiated when the absolute value of the third derivative characteristic exceeds this value. This is the emergency extrusion correction amount, which replaces the regular compensation command as the execution command for the extrusion roll gap adjustment mechanism when a sudden disturbance event is triggered. (This is the emergency response gain, a proportional factor used to map the sudden change amplitude to the emergency correction amount; its value should be greater than...) ); and the emergency compression amount correction amount The response priority is higher than the comprehensive extrusion compensation command. ; An emergency compensation preset gain coefficient is used to map the abrupt change amplitude of the third derivative characteristic to the emergency compression amount correction amount; Simultaneously, the synchronous adjustment amount of high-frequency welding heat input parameters The calculation formula is: In the formula: This is the synchronous adjustment amount of the high-frequency welding heat input parameters, which synchronously adjusts the welding power when a sudden interference event is triggered; A preset gain coefficient is set for heat input adjustment, which is used to map the abrupt change amplitude of the third derivative characteristic to the heat input adjustment amount (the thermo-mechanical coupling gain is used as a scaling factor to map the abrupt change amplitude to the heat input adjustment amount, so as to realize the coordinated linkage of extrusion compensation and heat adjustment).

[0050] In the above formula system, and All triggered by the same signal The two drives are synchronized to achieve coordinated linkage between extrusion compensation and heat input regulation.

[0051] The preset threshold To determine the trigger threshold for sudden interference events, the unit is mm / s³. Its value is determined based on the statistical distribution of the third-order rate of change signal of the inner weld bead height under normal production conditions. The setting method is as follows: continuously collect no less than 1000 sets of third-order derivative characteristic data under normal production conditions, calculate the statistical distribution (mean and standard deviation) of their absolute values, and set the threshold as the mean plus 3 to 5 times the standard deviation; or record the third-order derivative characteristic values ​​of known sudden faults instantaneously before their occurrence in historical production data, and take the minimum value as the threshold. The threshold can be adjusted by process engineers according to the actual requirements of false alarm rate and false negative rate, such as under normal production conditions. The absolute values ​​have a mean of 2 to 5 (mm / s³) and a standard deviation of 1 to 3 (mm / s³). The value ranges from 5 to 20, with an initial reference value of 10, corresponding to the statistical threshold of the mean plus 3 times the standard deviation.

[0052] This step pushes the control response speed to the limits of physical feasibility. The third derivative characteristic has a naturally high sensitivity to abrupt signals, triggering an emergency compensation response even when the disturbance is still in its nascent stage. Synchronously adjusting the high-frequency welding heat input parameters to achieve coordinated extrusion compensation and heat source regulation is a fundamental adjustment based on a thermo-coupling model. It aims to mitigate disturbances by changing the welding heat input, avoiding the potential for insufficient response under extreme conditions with single extrusion compensation. This improves the robustness of the control system and effectively prevents batch scrap caused by sudden, unknown disturbances. As a specific implementation method, The preset threshold setting needs to balance the false alarm rate and the false negative rate. This can be achieved by recording and analyzing the data in the instant before various known sudden faults (such as contamination and sparking) occur in the historical production process. The value is determined scientifically by statistical distribution.

[0053] In one embodiment, for step S5 mentioned above, the extrusion roller gap adjustment mechanism is either a hydraulically driven extrusion roller gap adjustment mechanism or an electrically driven extrusion roller gap adjustment mechanism. This mechanism responds to the extrusion amount correction and adjusts the gap of the extrusion rollers before the future time arrives, thereby changing the extrusion amount on the inner weld bead. This feature clarifies the implementation method of the actuator. Hydraulic drive mechanisms offer extremely high response speed and powerful thrust, suitable for high-speed, high-pressure extrusion control of large-diameter, thick-walled tubes. Electrical drive mechanisms, such as servo motor-driven precision lead screws or ball screw mechanisms, offer higher position control accuracy and energy efficiency, suitable for precision forming of small and medium-diameter tubes. Regardless of the drive form used, its core function is to receive analog or digital pulse signals from the controller and, at the future time set in step S3 when the compensation command should take effect, adjust to the new target position, in accordance with the predictive control's advance timing requirements.

[0054] In one embodiment, this method is applied to the production process of high-frequency resistance welded square and rectangular tubes or high-frequency resistance welded round tubes. The inner weld bead is an internal burr formed at the weld seam on the inner wall of the welded tube during the high-frequency resistance welding process. This indicates the specific industrial application and core control objective of this technical solution. Whether it is the "R-corner" region or planar region of a square or rectangular tube, or the annular region of a round tube, the uniformity of the residual height of the inner weld bead (also known as the inner burr) is a key indicator for evaluating the internal quality and performance of the welded tube. This method can control this key quality indicator by adapting the gain coefficients and mapping relationships in the control links according to the process characteristics of different tube types.

[0055] It should be noted that, in some optional implementations, the preset gain coefficients in all the aforementioned formulas (including but not limited to) The specific values ​​of (etc.) need to be calibrated offline or self-tuned online based on the mechanical characteristics, material properties, production speed, and product specifications of the specific unit. The calibration process typically involves system identification techniques such as open-loop step response testing and closed-loop limit oscillation testing. Furthermore, the "extrusion amount-inner weld bead height mapping relationship" used in step S5 is not a constant function. In some implementations, it can be a variable parameter model that is dynamically corrected based on the wall thickness, material, yield strength, and current welding speed of the steel pipe to be welded. This mapping relationship can be stored in the controller memory as a table, allowing the system to look up the table in real time and perform linear interpolation. For the visual inspection unit in step S1, in another optional implementation, to overcome pipe wall vibration and electromagnetic interference, the acquired three-dimensional contour data is first passed through a bandpass filter to filter out low-frequency swaying introduced by mechanical vibration and high-frequency pulse noise introduced by the high-frequency electromagnetic field of welding before height value extraction. In the specific implementation of the online adaptive correction step for prediction deviation, the forgetting factor of the recursive least squares method... Typical values ​​for this parameter range from 0.95 to 0.99 to balance the agility and stability of model parameter updates. For emergency compensation steps in the event of sudden disturbances, a smooth transition algorithm is included to prevent abrupt changes in control commands, considering the duration of the emergency mode after triggering and how to smoothly switch back to the normal compensation mode.

[0056] As an alternative implementation, the gain coefficients involved in the prediction model and compensation control are based on the SI unit system, where the compression amount is in millimeters (mm), the time is in seconds (s), the power is in watts (W), and the height is in millimeters (mm). Prediction step size The value is fixed at 0.5s. This value is determined based on the production line running speed of 25m / min and the distance from the measuring point to the center of the extrusion roller of 500mm, which corresponds to the material transfer delay between the measuring point and the execution point.

[0057] The dimension is [s]. The effective stroke of the extrusion roller gap adjustment mechanism is 3.0mm to 7.0mm. The target control range for the height of the inner weld bead is 0.5mm to 2.0mm. The open-loop gain of the closed-loop system is set to 2.0 to 4.0, with a value range of 0.5 to 2.0. The initial reference value is 1.0. The dimensionless quantity is [s²], derived from With prediction step size Determined according to the matching relationship of Taylor expansion coefficients: ,Will , Substitute the initial reference value of 1.0. The initial reference value is 0.125, and the value range is from 0.05 to 0.25; The dimensionless quantity is [s³], from With prediction step size Determined according to the matching relationship of Taylor expansion coefficients: ,Will , Substituting the initial reference value of 0.125, The initial reference value is 0.021, and the range is from 0.005 to 0.05; the ratio of the three basic gain coefficients is as follows: This ratio is related to the Taylor expansion coefficient. : : exist Theoretical ratio of time Strict consistency, deviations originate from The rounding deviation rate is 0.96%.

[0058] Dimensions Its value is Five to fifteen times. Calculated using an initial reference value of 0.021. The initial reference value is 0.021 × 10 = 0.21, and the range is from 0.105 to 0.315. The dimension of this is [W·s³ / mm], and its value is determined by the ratio of the rated high-frequency welding power to the typical amplitude of the third-order rate of change. The rated high-frequency welding power is 300kW to 800kW, and the typical amplitude of the third-order rate of change ranges from 100 to 1000 (mm / s³). The value range is from 500 to 3000, with an initial reference value of 1500; It is dimensionless, and its value is less than the reciprocal of the open-loop gain of the closed-loop system (the reciprocal of the open-loop gain ranges from 0.25 to 0.50). The value ranges from 0.3 to 0.7, with an initial reference value of 0.5.

[0059] The initial values ​​of the aforementioned gain coefficients were obtained through offline calibration experiments. The method was as follows: under normal production line conditions, step disturbance signals of known amplitude were applied, the system response data was recorded, and the initial estimated values ​​of each gain coefficient were calculated based on the response curves. Subsequently, closed-loop fine-tuning was used to optimize the control effect to its optimal level. Each coefficient can be adjusted manually or automatically during operation based on the actual control effect.

[0060] This application also provides a weld bead height control system based on dynamic compensation of extrusion amount for welded pipes, including: The data acquisition module is used to continuously acquire the time-series data of the height of the inner weld bead on the inner wall of the welded pipe behind the extrusion roller at a preset sampling frequency; The feature extraction module is used to perform differential operations on the time-series data of the inner weld bead height, and to extract at least the first derivative feature, the second derivative feature, and the third derivative feature. The advanced prediction module is used to input the multi-derivative features into a preset prediction model, predict the height of the inner weld bead at a preset future time, and output the predicted value of the inner weld bead height. The deviation calculation module is used to calculate the predicted height deviation between the predicted value of the inner weld bead height and the preset target height value. The compensation execution module is used to calculate the extrusion amount correction based on the predicted height deviation, and drive the extrusion roller gap adjustment mechanism to perform the extrusion amount compensation action before the future time arrives.

[0061] This application also provides an electronic device, including a memory and a processor. The memory stores a computer program, and when the processor executes the program, it implements a method for controlling the height of weld beads inside welded pipes based on dynamic compensation of extrusion amount.

[0062] This application also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements a method for controlling the height of weld beads inside welded pipes based on dynamic compensation of extrusion amount.

[0063] like Figure 5 As shown in the embodiments of this application, experimental embodiments are also provided. To verify the effectiveness of the proposed method for controlling the height of weld beads inside welded pipes based on dynamic compensation of extrusion amount, a comparative verification experiment was conducted on a high-frequency resistance welded square and rectangular tube production line of a steel pipe manufacturing company. The experimental object was a square and rectangular tube with a specification of 200mm×200mm and a wall thickness of 6.0mm, made of Q355B low-alloy high-strength structural steel. The production line operating speed was 25m / min, and the high-frequency welding power was set to 480kW. Referring to the relevant requirements of GB / T3091 standard for the height of burrs inside welded pipes, the target height of the inner weld bead was set to 1.5mm.

[0064] During the experiment, the height data of the inner weld bead was continuously collected by a laser profile scanning device located behind the extrusion roller at a sampling frequency of 100Hz. The installation position of the laser profile scanning device was approximately 500mm from the center of the extrusion roller, with the specific installation position determined according to the production line layout. To avoid the influence of welding spatter and high temperatures on the optical lens, the sensor was equipped with a compressed air purging device. The data acquisition and control system was implemented based on a PLC platform, with the sampling cycle synchronized with the PLC's scanning cycle. All experimental data were exported from the production line's process control system and subjected to offline statistical analysis.

[0065] Each group of experiments continuously produced 100 standard-length steel pipes (each 12m in length), for a total of 400 steel pipes participating in the experiment.

[0066] The experiment was divided into four groups: Control group 1 (no control group): Production was carried out using a fixed extrusion roller gap, with the extrusion amount set to 5.0 mm, and no closed-loop or predictive control was applied.

[0067] Control Group 2 (Pure Feedback Control Group): The feedback control scheme based on the current measured deviation is adopted. The control algorithm adopts the PID control algorithm commonly used in industrial sites, and its parameters are tuned by the field engineer according to conventional methods.

[0068] Control Group 3 (Static Adjustment Group): Referring to the existing static adjustment scheme based on offline detection, the machine is stopped after every 50 steel pipes are produced for metallographic sampling, and the extrusion amount is adjusted based on the metallographic test results. This scheme refers to the production method of manual sampling and testing still used by some companies in the industry, and the sampling interval and testing method are implemented according to the company's current operating procedures.

[0069] Experimental group (method group of this invention): Using the method described in this invention, the first, second, and third derivative features of the time series data of the inner weld bead height were extracted, and advance prediction was performed using a polynomial extrapolation model based on Taylor expansion (prediction step size). Based on the production line speed and the distance from the measurement point to the extrusion roller (set to 0.5s), frequency band differential compensation, online adaptive correction of prediction deviation, and emergency compensation for sudden interference are implemented.

[0070] The statistical results of the measured data of the height of the inner weld bead of 100 steel pipes produced continuously by each group are shown in Table 1.

[0071] Table 1. Statistics on the control effect of weld bead height in each group ; As shown in Table 1, in Control Group 1 without any control measures, the average height of the inner weld bead was approximately 2.9 mm, with significant fluctuations and a pass rate of only about 50%. Control Group 2, using pure feedback control, showed improvement, with the average height decreasing to approximately 2.1 mm and the pass rate increasing to approximately 75%, but fluctuations still existed. Control Group 3, using a static adjustment scheme, showed limited improvement compared to the uncontrolled group. In the experimental group, after adopting the method of this invention, the average height of the inner weld bead decreased to approximately 1.4 mm, with a standard deviation of approximately 0.2 mm and a pass rate of approximately 98%, demonstrating the best performance among all comparative schemes. Compared to Control Group 2, the average height of the inner weld bead decreased by approximately 35%, and the standard deviation decreased by approximately 60%.

[0072] To verify the prediction accuracy of the polynomial extrapolation model based on multi-order derivative features proposed in this invention, 500 sets of data samples with consecutive sampling times were randomly selected from the production data of the experimental group, with future times (prediction step size) as the basis for the prediction. Based on the measured value of the inner weld bead height, the prediction errors of the prediction model of this invention (containing first-order, second-order and third-order derivative features) and the linear extrapolation model containing only first-order derivative features were calculated respectively. The statistical results are shown in Table 2.

[0073] Table 2 Comparison of prediction accuracy of different prediction models ; As shown in Table 2, the average absolute error of the polynomial extrapolation model used in this invention is approximately 0.09 mm, and the root mean square error is approximately 0.11 mm, both of which are superior to the linear extrapolation model containing only first-order derivative features. Statistical results show that the prediction accuracy is significantly improved after introducing second- and third-order derivative features, providing a reliable decision-making basis for the advanced dynamic compensation method of this invention.

[0074] To verify the effectiveness of the multi-derivative characteristic frequency band differentiated compensation strategy proposed in this invention, the control effect of the inner weld bead height was recorded in four modes: enabling only instant micro-compensation, enabling only trend tracking compensation, enabling only reference correction compensation, and the synergistic effect of the three compensation components. The results are shown in Table 3.

[0075] Table 3 Comparison of control effects under different compensation modes ; Table 3 shows that the three compensation components function against different frequency bands of interference sources: the instantaneous micro-compensation component mainly suppresses high-frequency fluctuations, the trend-following compensation component mainly reduces mid-frequency fluctuations, and the reference correction compensation component mainly eliminates low-frequency drift. When the three compensation components work synergistically, the overall standard deviation is approximately 0.20 mm, which is better than that of any single compensation mode, indicating that the frequency band differentiated compensation strategy can effectively cope with interference at multiple time scales.

[0076] During the experiment, for a batch of tube blanks with localized thickness fluctuations (thickness fluctuating from 6.0 mm to approximately 6.5 mm, lasting for about 0.8 s), the response of the inner weld bead height was recorded when the emergency compensation mode was enabled and disabled. The results are shown in Table 4.

[0077] Table 4 Comparison of Emergency Compensation Effects under Sudden Interference Conditions ; Table 4 shows that without the emergency compensation mode enabled, the peak height of the inner weld bead was approximately 2.5 mm, with an overshoot of approximately 1.0 mm. This section of the weld was deemed defective due to the excessive inner weld bead height. After enabling the emergency compensation mode, the peak height of the inner weld bead was approximately 1.8 mm, with an overshoot of approximately 0.3 mm. The recovery time was significantly shortened, and no defective products were generated. These results demonstrate that the emergency compensation mechanism based on third-order derivative characteristics proposed in this invention has good response capability to sudden disturbances.

[0078] To verify the long-term effect of the online adaptive correction step for prediction deviation proposed in this invention, the production data of the experimental group running continuously for 8 hours were statistically analyzed in hourly segments. The average absolute error of prediction for each time period was recorded under both modes of adaptive correction with and without. The results are shown in Table 5.

[0079] Table 5. Impact of Adaptive Correction on Long-Term Stability of Prediction Accuracy ; As shown in Table 5, without adaptive correction, the prediction error gradually increases with the duration of operation, from approximately 0.09 mm in the first hour to approximately 0.18 mm in the eighth hour. After enabling adaptive correction, the prediction error in each time period remains relatively stable at approximately 0.08–0.09 mm, without a significant increasing trend, indicating that the adaptive correction mechanism proposed in this invention has a positive effect on maintaining long-term operational accuracy.

[0080] The experimental results show that, using the method of this invention, the average height of the inner weld bead can be controlled at approximately 1.4 mm, a reduction of about 35% compared to pure feedback control, and the standard deviation is controlled at approximately 0.2 mm, a reduction of about 60% compared to pure feedback control. In a continuous 8-hour operation test, the mean absolute error of the prediction model stabilized at approximately 0.09 mm after enabling adaptive correction, while the error of the scheme without correction increased significantly within the same time period. In a comparative test simulating sudden disturbances, the peak overshoot and recovery time of the inner weld bead height were significantly better after enabling the emergency compensation mode than in the unenabled mode. The above experimental data preliminarily verify the effectiveness of the control system proposed in this invention. It should be noted that the data in the tables above are statistical results of measured data from 100 steel pipes in each group. Data from different batches in actual production may fluctuate to some extent, but the overall statistical trend is consistent with that listed in the tables.

[0081] The embodiments disclosed in this invention are preferred embodiments, but are not limited thereto. Those skilled in the art can easily understand the spirit of this invention based on the above embodiments and make different extensions and variations, but as long as they do not depart from the spirit of this invention, they are all within the protection scope of this invention.

Claims

1. A method for controlling the height of weld beads inside welded pipes based on dynamic compensation of extrusion amount, characterized in that, include: The time-series data of the height of the inner weld bead on the inner wall of the welded pipe behind the extrusion roller is continuously collected at a preset sampling frequency; Differential operation is performed on the time series data of the inner weld bead height to extract at least the first derivative feature, the second derivative feature and the third derivative feature, wherein the first derivative feature represents the instantaneous rate of change of the inner weld bead height, the second derivative feature represents the trend of the rate of change, and the third derivative feature represents the abrupt change characteristics of the signal. The first-order derivative features, second-order derivative features, and third-order derivative features are input into a preset prediction model. The prediction model predicts the height of the inner weld bead at a preset future time based on the different time scale changes represented by each derivative feature, and outputs the predicted value of the inner weld bead height. Calculate the predicted height deviation between the predicted value of the inner weld bead height and the preset target height value; Based on the predicted height deviation, the extrusion amount correction is calculated according to the preset extrusion amount-inner weld bead height mapping relationship, and the extrusion roller gap adjustment mechanism is driven to perform the extrusion amount compensation action before the future time is reached.

2. The method for controlling the height of weld beads inside welded pipes based on dynamic compensation of extrusion amount according to claim 1, characterized in that, The prediction model is a polynomial extrapolation model based on Taylor expansion. The predicted value of the inner weld bead height is calculated by polynomial extrapolation based on the measured value of the inner weld bead height at the current time and the first, second, and third derivative features. Alternatively, the prediction model is an artificial neural network model or a support vector regression model trained offline, with the first, second, and third derivative features serving as model input features.

3. The method for controlling the height of weld beads inside welded pipes based on dynamic compensation of extrusion amount according to claim 1, characterized in that, It also includes a multi-derivative feature frequency band differential compensation step: based on the time scale characteristics corresponding to the first-order derivative feature, second-order derivative feature, and third-order derivative feature, respectively, calculate the instantaneous micro-compensation component corresponding to the millisecond-level instantaneous fluctuation, the trend tracking compensation component corresponding to the second-level trend change, and the benchmark correction compensation component corresponding to the minute-level or hour-level gradual drift. The three types of compensation components are weighted and superimposed to generate a comprehensive extrusion compensation command.

4. The method for controlling the height of weld beads inside welded pipes based on dynamic compensation of extrusion amount according to claim 1, characterized in that, It also includes an online adaptive correction step for prediction deviation: comparing the measured value of the inner weld bead height after the future time with the corresponding predicted value of the inner weld bead height, and calculating the prediction deviation; classifying the prediction deviation into systematic deviation and random deviation, updating the parameters of the prediction model online for the systematic deviation using the recursive least squares method, and temporarily correcting the compensation amount of the extrusion amount at the current time for the random deviation.

5. The method for controlling the height of weld beads inside welded pipes based on dynamic compensation of extrusion amount according to claim 1, characterized in that, It also includes an emergency compensation step for sudden interference: real-time monitoring of the absolute value of the third derivative feature; when the absolute value exceeds a preset threshold, it is determined to be a sudden interference event, triggering the emergency compensation mode; calculating the emergency extrusion correction amount based on the amplitude and direction of the third derivative feature; and simultaneously adjusting the high-frequency welding heat input parameters.

6. The method for controlling the height of weld beads inside welded pipes based on dynamic compensation of extrusion amount according to claim 1, characterized in that, The timing data of the height of the inner weld bead is collected by a vision inspection unit located behind the extrusion roller. The vision inspection unit is a structured light three-dimensional measurement device or a laser contour scanning device, which has high temperature protection and noise filtering functions.

7. The method for controlling the height of weld beads inside welded pipes based on dynamic compensation of extrusion amount according to claim 1, characterized in that, The extrusion roller gap adjustment mechanism is a hydraulically driven gap adjustment mechanism or an electric servo-driven gap adjustment mechanism, used to adjust the extrusion roller gap in response to the extrusion amount correction to change the extrusion amount.

8. A control system for the height of weld beads inside a welded pipe based on dynamic compensation of extrusion amount, characterized in that, include: The data acquisition module is used to continuously acquire the time-series data of the height of the inner weld bead on the inner wall of the welded pipe behind the extrusion roller at a preset sampling frequency; The feature extraction module is used to perform differential operations on the time-series data of the inner weld bead height, and to extract at least the first derivative feature, the second derivative feature, and the third derivative feature. The advanced prediction module is used to input the multi-derivative features into a preset prediction model, predict the height of the inner weld bead at a preset future time, and output the predicted value of the inner weld bead height. The deviation calculation module is used to calculate the predicted height deviation between the predicted value of the inner weld bead height and the preset target height value. The compensation execution module is used to calculate the extrusion amount correction based on the predicted height deviation, and drive the extrusion roller gap adjustment mechanism to perform the extrusion amount compensation action before the future time arrives.

9. An electronic device, characterized in that, It includes a memory and a processor, wherein the memory stores a computer program, and the processor executes the program to implement the method for controlling the height of weld beads inside welded pipes based on dynamic compensation of extrusion amount as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method for controlling the height of weld beads inside welded pipes based on dynamic compensation of extrusion amount as described in any one of claims 1 to 7.

Images