Intelligent laser-based metal square tube forming production line control system and method

Abstract

This application provides a control system and method for a metal square tube forming production line based on intelligent laser technology, belonging to the field of metal processing and intelligent control technology. The system includes an intelligent uncoiling and leveling module, an adaptive laser cutting and separating plate module, a flexible roller forming module, an intelligent laser welding and joining module, a multi-dimensional weld laser detection module, and an edge computing server connected in sequence. The adaptive laser cutting and separating plate module integrates a deep neural network and a fuzzy logic compensation controller, dynamically generating and compensating for optimal cutting parameters in real time based on the plate material, thickness, laser aging coefficient, and production cycle requirements. The multi-dimensional weld laser detection module integrates a data fusion processor, jointly judging internal defects and surface defects under the same weld coordinates, and generating welding process optimization suggestions through association rule mining, which are then fed back to the intelligent laser welding and joining module. This invention enables flexible manufacturing of metal square tubes, improves precision and efficiency, and reduces energy consumption and scrap rate.

Description

Technical Field

[0001] This application belongs to the field of metal processing and intelligent control technology, specifically relating to a control system and method for a metal square tube forming production line based on intelligent laser. Background Technology

[0002] Metal square tubes are key structural materials in construction, machinery, and automotive industries, and their production efficiency and quality directly affect the costs and safety of downstream industries. With the accelerated transformation of the manufacturing industry towards intelligent and energy-saving technologies, the market is placing unprecedented demands on the flexibility, precision machining capabilities, and environmental compliance of metal square tube production lines. However, existing metal square tube forming production lines or processing methods still suffer from the following technical deficiencies: Firstly, the raw material adaptation and cutting processes are rudimentary and lack flexibility. Traditional production lines are mostly designed for single-specification tube shapes, requiring manual replacement of rollers and adjustment of cutters when switching products, which is time-consuming and inefficient. Even when some production lines introduce laser cutting, the cutting parameters (such as power, speed, and focus) are mostly fixed values ​​or only simple segmented settings, lacking the ability to adaptively perceive and dynamically adjust to fluctuations in the thickness of incoming sheet materials, material differences, and equipment conditions (such as laser aging). This ultimately leads to unstable cutting accuracy, easy generation of burrs and excessively large heat-affected zones, affecting the quality of subsequent forming and welding, and making it impossible to achieve efficient and flexible switching between multiple categories and small batch orders. Secondly, the weld inspection methods are limited and lack multi-dimensional and quantitative defect identification and traceability capabilities. Weld quality is a decisive factor in the strength of square tubes. Existing technologies mostly rely on manual visual inspection, online eddy current testing, or single X-ray inspection. These methods have significant drawbacks: manual inspection is highly subjective, prone to missed defects, and inefficient; eddy current testing struggles to detect deep internal defects; while X-rays can detect flaws, the equipment is expensive, poses radiation risks, and cannot simultaneously acquire surface morphology information. More importantly, existing technologies only detect defects, lacking cross-module, multi-dimensional data correlation and mining between defect characteristics (such as type, size, and location) and process parameters during welding (such as gap, temperature, and power fluctuations). This prevents the root cause tracing of defects and proactive process optimization, resulting in persistently high scrap rates.

[0003] Therefore, there is an urgent need for a control method for metal square tube forming production lines that can achieve highly flexible adaptive processing and has multi-dimensional weld seam precision detection and intelligent traceability capabilities. Summary of the Invention

[0004] In a first aspect, embodiments of this application provide a control system for a metal square tube forming production line based on intelligent laser, including an intelligent uncoiling and leveling module, an adaptive laser cutting and separating plate module, a flexible roller forming module, an intelligent laser welding seam module, a multi-dimensional weld seam laser detection module, and an edge computing server that is communicatively connected to each of the above modules. The edge computing server acquires the sensor data and operating status of each module in real time through a dual-mode communication network, and uses the built-in intelligent MES subsystem to issue control commands to each module to perform closed-loop collaborative control of the production line. The adaptive laser cutting and separating plate module integrates a pre-trained deep neural network model and a fuzzy logic compensation controller, which dynamically calculates and adjusts the optimal cutting parameters based on the real-time physical parameters of the incoming plate and the equipment status. The multidimensional weld laser inspection module integrates a laser ultrasonic inspection component, a vision imaging subsystem, and a data fusion processor. It is used to quantitatively identify internal and surface defects in the weld in three-dimensional space, and to perform correlation analysis with the process parameters of the intelligent laser welding joint module to optimize the welding quality in a closed loop.

[0005] Furthermore, the adaptive laser cutting and separating sheet metal module includes a sheet metal thickness detection sensor, a vision positioning subsystem, an edge controller, and several sets of fiber laser cutting heads; The sheet thickness detection sensor is used to detect the material coefficient and actual thickness of the incoming sheet in real time and output the data to the edge controller. The visual positioning subsystem is used to acquire images of the cut edges in real time during the cutting process, analyze and generate cut quality feedback values, and output them to the edge controller. A pre-trained deep neural network model and a fuzzy logic compensation controller are deployed in the edge controller; The deep neural network model receives the material coefficient, actual plate thickness, cut quality feedback value, production cycle requirements, and laser aging coefficient as input, and outputs the initial optimal laser power, optimal cutting speed, and optimal focal position offset after calculation. The fuzzy logic compensation controller calculates the power compensation amount in real time based on the deviation between the cut quality feedback value and the preset standard value, and adds it to the optimal laser power to generate the final laser power control command. Each fiber laser cutting head receives the final laser power control command, optimal cutting speed command, and optimal focus position offset command issued by the edge controller, and performs the cutting action on the plate.

[0006] Furthermore, the laser ultrasonic detection component is used to emit pulsed lasers into the weld and receive the reflected ultrasonic echo signals. Based on the time difference between emission and reception and the echo amplitude, it quantitatively calculates the depth, equivalent size, and position coordinates of the internal defects in the weld, and outputs them as internal defect data to the data fusion processor. The visual imaging subsystem is used to synchronously acquire images of the weld surface, identify the contours, areas and pixel coordinates of surface defects in the images through a semantic segmentation network, and output them as surface defect data to the data fusion processor. The data fusion processor receives internal defect data and surface defect data, establishes a spatial synchronization mapping matrix, converts the pixel coordinates into position coordinates along the weld length direction, and performs joint judgment logic on internal defects and surface defects under the same position coordinates, outputting a comprehensive weld rating result; at the same time, it performs association rule mining on the identified defect features and externally input welding process parameters, generates process optimization suggestions, and outputs them to the edge computing server.

[0007] Secondly, embodiments of this application also provide a control method for a metal square tube forming production line based on intelligent laser, which applies the intelligent laser-based metal square tube forming production line control system described in the first aspect, and includes the following steps: S1. The finished steel coil is uncoiled and leveled using an intelligent uncoiling and leveling module to obtain a flat sheet material; S2. The adaptive laser cutting separation plate module uses a multi-source data fusion decision subsystem to dynamically generate optimal cutting parameters and perform high-precision cutting on the flat plate to obtain at least one set of blanks that meet the preset width. S3. Using a flexible roller forming module, the roller parameters are automatically adjusted according to the target square tube specifications to continuously roll the blank to form an open square tube prototype; S4. Using an intelligent laser welding seam module, laser welding is performed on the seam of the square tube prototype to form a closed square tube; S5. Using the multi-dimensional weld laser detection module, the internal and surface defects of the weld are quantitatively identified in three dimensions by using a multi-level data fusion method of laser ultrasound and vision. The data fusion processor executes the joint judgment logic to output the comprehensive rating result of the weld. At the same time, the defect features and welding process parameters are correlated by rule mining to generate process optimization suggestions, which are then fed back to step S4.

[0008] Furthermore, the specific steps of step S2 are as follows: S21. The material coefficient and actual thickness of the flat plate are detected in real time by the plate thickness detection sensor, and the aging coefficient of the current laser is obtained. The material coefficient, actual plate thickness and laser aging coefficient are output to the edge controller. S22. The edge controller inputs the received material coefficient, actual plate thickness, laser aging coefficient, and preset production cycle requirements into a pre-trained deep neural network model. After calculation, it outputs the initial optimal cutting parameters, including the optimal laser power. Optimal cutting speed and optimal focus position offset ; S23. The fiber laser cutting head is started to cut the plate according to the initial optimal cutting parameters. At the same time, the visual positioning subsystem collects the cut edge image in real time, analyzes and generates the cut quality feedback value, and outputs it to the edge controller. S24. The fuzzy logic compensation controller in the edge controller calculates the power compensation amount in real time based on the deviation between the cut quality feedback value and the preset standard value, and adds it to the initial optimal laser power to generate the final laser power control command. S25. The edge controller sends the final laser power control command, the optimal cutting speed command, and the optimal focal position offset command to the fiber laser cutting head to execute the cutting action and obtain at least one set of blanks that meet the preset width.

[0009] Furthermore, the training process of the deep neural network model in step S22 is as follows: Historical cutting process data is collected as training samples, and each training sample includes an input feature vector. and the corresponding optimal cutting parameter labels Use training samples to build a training dataset; in, Material coefficient, This refers to the actual thickness of the board material. This refers to the laser aging factor. To meet production cycle requirements; The deep neural network model is trained under supervision using the training dataset. The mean squared error is used as the loss function. The weights and biases of the model are iteratively updated through the backpropagation algorithm until the loss function converges, and the trained deep neural network model is obtained. Deploy the trained deep neural network model to the edge controller; When running online, the edge controller will acquire the material coefficients in real time. Actual thickness of the board Laser aging coefficient Production cycle time requirements Constructing the input feature vector The input is fed into the trained deep neural network model; After forward computation, the deep neural network model outputs the optimal cutting parameter labels. These are respectively used as the initial optimal laser power, optimal cutting speed, and optimal focal position offset.

[0010] Furthermore, the specific steps in step S24 are as follows: S241. Obtain the incision quality feedback value and calculate the deviation between the incision quality feedback value and the preset standard value. ; S242. Based on the aforementioned deviation The proportional compensation component is calculated using the following formula. :

[0011] in, This is a preset proportional coefficient; S243. Based on the aforementioned deviation The integral compensation component is calculated using the following formula. :

[0012] in, The preset integral coefficient; S244. The proportional compensation component With integral compensation component Add them together to get the power compensation amount. ; S245. Adjust power compensation amount Superimposed to the initial optimal laser power The final laser power control command is obtained. .

[0013] Furthermore, the specific steps of step S3 are as follows: S31. The edge computing server sends roller adjustment instructions to the flexible roller forming module according to the target square tube specifications. S32. The flexible roller forming module automatically adjusts the parameters of its constituent forming rollers according to the roller adjustment command to continuously roll the blank to form an open square tube prototype; the parameters of the forming rollers include spacing, angle and rotation speed; S33. Through the shape detection function of the flexible roller forming module, the cross-sectional shape data of the square tube prototype is collected in real time, the actual cross-sectional outline is generated through 3D modeling, and fed back to the edge computing server; S34. The edge computing server compares the actual cross-sectional profile with the preset specifications. If there is a deviation, it sends a correction command to the flexible roller forming module to adjust the parameters of the forming rollers until the side length deviation of the square tube prototype meets the preset requirements.

[0014] Furthermore, the specific steps of step S4 are as follows: S41. Through the visual tracking function of the intelligent laser welding seam module, the position of the seam of the square tube prototype is captured in real time, the compensation offset during the operation of the square tube is automatically calculated, and sent to the edge computing server. S42. By using the weld gap detection function of the intelligent laser welding joint module, the gap size at the joint is monitored and output as gap data to the edge computing server; S43. The edge computing server adjusts the power and welding speed of the main welding laser in the intelligent laser welding seam module according to the gap data, and starts the main welding laser to weld the seam. S44. Through the temperature monitoring function of the intelligent laser welding seam module, the temperature of the weld area is monitored in real time, and the temperature data is fed back to the edge computing server to control the welding process based on the temperature data, thereby avoiding overheating and deformation; S45. If the weld inspection reveals incomplete penetration defects, the edge computing server activates the repair laser in the intelligent laser welding seam module to perform targeted repair welding, forming a closed square tube.

[0015] Furthermore, the specific steps of step S5 are as follows: S51. The laser ultrasonic detection component emits a pulsed laser to the weld and receives the reflected ultrasonic echo signal. Based on the time difference between emission and reception and the echo amplitude, the depth, equivalent size and position coordinates of the internal defects in the weld are quantitatively calculated, and the calculation results are output as internal defect data to the data fusion processor. S52. Simultaneously acquire weld surface images through the visual imaging subsystem, identify the contour, area, and pixel coordinates of surface defects in the image through the semantic segmentation network, and output the identification results as surface defect data to the data fusion processor; S53. The data fusion processor receives internal defect data and surface defect data, establishes a spatial synchronization mapping matrix, and converts the pixel coordinates of the surface defects into position coordinates along the weld length direction; S54. The data fusion processor performs joint judgment logic on internal defects and surface defects at the same location coordinates and outputs the comprehensive weld rating result; The S55 data fusion processor performs association rule mining on the identified defect features and welding process parameters, generates process optimization suggestions, and outputs them to the edge computing server.

[0016] Furthermore, the specific steps of step S51 are as follows: S511. A pulsed laser is emitted toward the weld seam through the laser ultrasonic detection component, and a timer is started to record the emission time; S512. Receive the ultrasonic echo signal reflected from internal defects in the weld, record the moment the echo is received, and calculate the time difference between the transmission and reception times. ; S513. Calculate the depth of internal defects according to the following formula. :

[0017] in, The propagation speed of ultrasound in the metal square tube base material; S514. Extracting the amplitude of the ultrasonic echo signal reflected from the defect. The equivalent magnitude of the internal defect is calculated using the following formula. :

[0018] in, and These are pre-calibrated coefficients; S515. Record the position coordinates of the internal defect along the weld direction and the defect depth. Equivalent size The location coordinates are output as internal defect data to the data fusion processor; The specific steps of step S54 are as follows: S541. The data fusion processor iterates through the coordinates of each position on the weld, and determines the equivalent size of the internal defect at the corresponding coordinate. If the area of ​​the surface defect is greater than the preset first threshold, or the area of ​​the surface defect is greater than the preset second threshold, then the coordinates of this location are determined to be a defect point. S542. Based on the number and severity of defects on the entire weld, output the overall weld rating as qualified, repairable, or scrap.

[0019] As can be seen from the above technical solutions, this application has the following advantages: The intelligent laser-based metal square tube forming production line control system and method provided in this application achieve intelligent, flexible, and unmanned production throughout the entire metal square tube forming process by constructing a fully closed-loop production line system that includes intelligent uncoiling, adaptive cutting, flexible forming, intelligent welding, and multi-dimensional detection. This solves the problems of isolated equipment, excessive manual intervention, and difficult production changeover in traditional production lines. Through the dual-mode communication network between the edge computing server and each module, and the built-in intelligent MES subsystem, millisecond-level interaction and global collaborative control of production data are achieved, improving production scheduling efficiency and production line stability. The multi-source data fusion decision subsystem in the adaptive laser cutting and separating plate module dynamically optimizes cutting parameters based on the thickness, material, and equipment aging status of the incoming material, thereby ensuring the consistency of cutting quality for different batches of plates. The laser ultrasonic and visual data fusion method in the multi-dimensional weld laser detection module enables three-dimensional quantitative identification and joint judgment of internal and surface defects in welds, solving the problem of traditional detection methods being singular and unable to accurately quantify. The integration of laser cleaning and AI visual appearance inspection achieves environmentally friendly surface treatment and fully automated dimensional inspection, reducing labor costs and error rates. Attached Figure Description

[0020] To more clearly illustrate the technical solution of this application, the accompanying drawings used in the description will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 This is a schematic diagram of the intelligent laser-based metal square tube forming production line control system of the present invention.

[0022] Figure 2 This is a schematic diagram of the process for controlling the metal square tube forming production line based on intelligent laser according to the present invention. Detailed Implementation

[0023] The various embodiments of this disclosure will be described more fully in the following detailed description of the intelligent laser-based metal square tube forming production line control system. This disclosure may have various embodiments, and adjustments and changes may be made therein. However, it should be understood that there is no intention to limit the various embodiments of this disclosure to the specific embodiments disclosed herein, but rather this disclosure should be understood to cover all adjustments, equivalents, and / or alternatives falling within the spirit and scope of the various embodiments of this disclosure.

[0024] This embodiment provides a control system for a metal square tube forming production line based on intelligent laser. Based on intelligent laser and edge computing, it realizes flexible production of multiple specifications and multi-dimensional full inspection of welds, which greatly improves accuracy and efficiency and reduces energy consumption and labor costs.

[0025] The technical solutions of 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.

[0026] Please see Figure 1 The diagram shown is a schematic of a metal square tube forming production line control system based on intelligent laser in a specific embodiment. The system includes an intelligent uncoiling and leveling module, an adaptive laser cutting and separating plate module, a flexible roller forming module, an intelligent laser welding and joining module, a multi-dimensional weld laser detection module, and an edge computing server that is communicatively connected to each of the above modules. The edge computing server acquires the sensor data and operating status of each module in real time through a dual-mode communication network, and uses the built-in intelligent MES subsystem to issue control commands to each module to perform closed-loop collaborative control of the production line. The adaptive laser cutting and separating plate module integrates a pre-trained deep neural network model and a fuzzy logic compensation controller, which dynamically calculates and adjusts the optimal cutting parameters based on the real-time physical parameters of the incoming plate and the equipment status. The multidimensional weld laser inspection module integrates a laser ultrasonic inspection component, a visual imaging subsystem, and a data fusion processor. It is used to quantitatively identify internal and surface defects in welds in three-dimensional space, and to perform correlation analysis with the process parameters of the intelligent laser welding joint module to optimize the welding quality in a closed loop. It should be noted that the intelligent uncoiling and leveling module can automatically identify the material and diameter of the steel coil, and dynamically adjust the pressure and spacing of the leveling rollers through the PID algorithm, which solves the problem of poor leveling effect caused by the hardness difference of different batches of steel coils, and ensures the flatness of the output sheet. The adaptive laser cutting separation module can dynamically generate optimal cutting parameters based on the real-time plate thickness, material coefficient, and laser aging degree. Compared with traditional fixed parameter cutting, it can adapt to material fluctuations, ensure consistent cut quality, reduce slag and heat-affected zone width, and improve billet precision. The intelligent laser welding seam module results in a small heat-affected zone, high welding speed, and high weld strength. Combined with visual tracking, it can automatically compensate for weld misalignment, ensuring the stability of the welding process and the quality of weld formation, and realizing high-energy-density precision welding. The multidimensional weld laser inspection module breaks through the limitations of single-dimensional inspection, and can simultaneously detect internal and surface defects in welds and perform three-dimensional quantitative analysis, thus avoiding the quality risks brought about by random inspection. Edge computing servers acquire data from the entire production line with low latency through a dual-mode communication network, and use the intelligent MES subsystem for global scheduling and process optimization, connecting previously isolated devices into a whole, thus realizing transparency and intelligence in the production process.

[0027] This embodiment overcomes the problems of insufficient flexibility and difficulty in detecting welding defects in traditional production lines by using intelligent laser and edge computing closed-loop control, and realizes high-precision, full-process inspection and intelligent management of metal square tube forming.

[0028] Furthermore, as a refinement and extension of the specific implementation of the above embodiments, in order to fully illustrate the specific implementation process in this embodiment, another intelligent laser-based metal square tube forming production line control system is provided. The system includes an intelligent uncoiling and leveling module, an adaptive laser cutting and separating plate module, a flexible roller forming module, an intelligent laser welding seam module, a multi-dimensional weld seam laser detection module, and an edge computing server that is communicatively connected to each of the above modules. The edge computing server acquires the sensor data and operating status of each module in real time through a dual-mode communication network, and uses the built-in intelligent MES subsystem to issue control commands to each module to perform closed-loop collaborative control of the production line. It should be noted that the dual-mode communication network includes a 5G communication link and a MODBUS TCP communication link; Among them, the 5G communication link is used as the primary communication method to realize high-speed data transmission between the edge computing server and various modules, with a transmission rate of ≥1Gbps and a communication latency of ≤10ms; The MODBUS TCP communication link serves as a backup communication method, automatically switching when the 5G communication link fails or the signal is unstable, ensuring the continuous transmission of control commands and sensor data. In this embodiment, the dual-mode communication network adopts a dual-redundancy design of 5G communication and MODBUS TCP communication; the 5G communication module is used to transmit high-volume real-time data, such as visual positioning images, laser ultrasonic echo signals, AI visual inspection images, etc.; MODBUS TCP communication is used to transmit low-volume but high-reliability data such as equipment status parameters, control commands, and process parameters; when the 5G signal strength is lower than a preset threshold (e.g., -95dBm), the edge computing server automatically switches the data transmission to the MODBUS TCP link, with a switching time of ≤50ms, to ensure uninterrupted operation of the production line; The adaptive laser cutting and separating plate module integrates a pre-trained deep neural network model and a fuzzy logic compensation controller, which dynamically calculates and adjusts the optimal cutting parameters based on the real-time physical parameters of the incoming plate and the equipment status. The multidimensional weld laser inspection module integrates a laser ultrasonic inspection component, a visual imaging subsystem, and a data fusion processor. It is used to quantitatively identify internal and surface defects in welds in three-dimensional space, and to perform correlation analysis with the process parameters of the intelligent laser welding joint module to optimize the welding quality in a closed loop. The adaptive laser cutting and separating sheet metal module includes a sheet metal thickness detection sensor, a vision positioning subsystem, an edge controller, and several sets of fiber laser cutting heads; The sheet thickness detection sensor is used to detect the material coefficient and actual thickness of the incoming sheet in real time and output the data to the edge controller. The visual positioning subsystem is used to acquire images of the cut edges in real time during the cutting process, analyze and generate cut quality feedback values, and output them to the edge controller. A pre-trained deep neural network model and a fuzzy logic compensation controller are deployed in the edge controller; The deep neural network model receives the material coefficient, actual plate thickness, cut quality feedback value, production cycle requirements, and laser aging coefficient as input, and outputs the initial optimal laser power, optimal cutting speed, and optimal focal position offset after calculation. The fuzzy logic compensation controller calculates the power compensation amount in real time based on the deviation between the cut quality feedback value and the preset standard value, and adds it to the optimal laser power to generate the final laser power control command. Each fiber laser cutting head receives the final laser power control command, the optimal cutting speed command, and the optimal focus position offset command issued by the edge controller, and performs the cutting action on the plate. The laser ultrasonic testing component is used to emit pulsed lasers into the weld and receive the reflected ultrasonic echo signals. Based on the time difference between emission and reception and the echo amplitude, it quantitatively calculates the depth, equivalent size, and position coordinates of internal defects in the weld, and outputs them as internal defect data to the data fusion processor. The visual imaging subsystem is used to synchronously acquire images of the weld surface, identify the contours, areas and pixel coordinates of surface defects in the images through a semantic segmentation network, and output them as surface defect data to the data fusion processor. The data fusion processor receives internal defect data and surface defect data, establishes a spatial synchronization mapping matrix, converts the pixel coordinates into position coordinates along the weld length direction, and performs joint judgment logic on internal defects and surface defects under the same position coordinates, outputting a comprehensive weld rating result; at the same time, it performs association rule mining on the identified defect features and externally input welding process parameters, generates process optimization suggestions, and outputs them to the edge computing server. The intelligent uncoiling and leveling module includes an intelligent uncoiling machine, a servo-controlled leveling machine, a material identification sensor, an infrared distance sensor, and multiple sets of intelligent adjusting leveling rollers; Material identification sensors are used to identify the material type of the steel coil and output the result to the edge computing server; Infrared ranging sensors are used to monitor changes in the diameter of steel coils in real time and output the results to an edge computing server. The edge computing server automatically matches the unwinding speed and leveling scheme based on the material type and diameter variation data, generates leveling parameters, and sends them to the servo-controlled leveling machine. The servo-controlled leveling machine dynamically adjusts the pressure of each intelligent leveling roller and the plate spacing according to the leveling parameters through a PID algorithm to level the steel coil and obtain a flat plate. The flexible roller forming module includes several forming rollers, a servo adjustment subsystem, and a shape detection sensor; The edge computing server issues roller adjustment commands based on the target square tube specifications. The servo adjustment subsystem receives roller adjustment commands and automatically adjusts the spacing, angle, and speed of each group of rollers; Shape detection sensors are used to collect cross-sectional shape data in real time during the sheet forming process, generate the actual cross-sectional outline of the square tube prototype through 3D modeling, and feed it back to the edge computing server; The edge computing server compares the actual cross-sectional profile with the preset specifications. If there is a deviation, it sends a correction command to the servo adjustment subsystem to adjust the parameters of the forming rollers until the side length deviation of the square tube prototype meets the preset requirements. The intelligent laser welding seam module includes a dual-fiber laser, a vision tracking subsystem, a weld gap detection sensor, and a temperature monitoring sensor. The visual tracking subsystem is used to capture the position of the joint of the square tube prototype in real time, automatically calculate the compensation offset during the operation of the square tube, and output it to the edge computing server. The weld gap detection sensor is used to monitor the gap size at the joint and output the gap size data to the edge computing server. The edge computing server adjusts the power and welding speed of the main welding laser in the dual-fiber laser system based on the gap size data. Temperature monitoring sensors are used to monitor the temperature of the weld area in real time and feed the temperature data back to the edge computing server to avoid overheating and deformation of the square tube. The system also includes a laser cleaning module, an AI visual appearance and size detection module, an intelligent laser length cutting module, and an autonomous navigation AGV handling module; The laser cleaning module is positioned between the intelligent laser welding seam module and the multi-dimensional weld seam laser inspection module. The AI ​​visual appearance and size detection module, the intelligent laser length cutting module, and the autonomous navigation AGV handling module are sequentially set after the multi-dimensional weld laser detection module. The laser cleaning module is used to perform pulsed laser cleaning on the surface of the welded closed square tube to remove oxide layer, oil and welding slag impurities. The AI ​​visual appearance and size inspection module is set after the multi-dimensional weld laser inspection module and is used to conduct a comprehensive inspection of the appearance, size and surface quality of square tubes that have passed the weld inspection. The intelligent laser length-cutting module is set after the AI ​​visual appearance size detection module and is used to cut qualified square tubes with high precision according to the preset product length. The autonomous AGV handling module is set after the intelligent laser length cutting module and is used to transport the cut square tubes to designated locations for sorting and storage. The laser cleaning module includes an adjustable focal length laser head, a surface cleanliness detection sensor, and a dust purification subsystem. The edge computing server automatically matches laser cleaning parameters based on the type of contaminants on the surface of the square tube and sends them to the corresponding laser head; The laser head performs non-contact pulsed laser cleaning on the surface of the square tube according to the cleaning parameters; The surface cleanliness detection sensor uses infrared spectroscopy analysis to detect the amount of residual oil on the surface after cleaning in real time and outputs the data to the edge computing server. If the edge computing server determines that the amount of residual oil on the surface exceeds a preset threshold, it will automatically initiate a secondary cleaning process. The dust purification subsystem is used to collect dust generated during the cleaning process and filter it according to a preset precision (e.g., 0.3 microns). After cleaning and achieving the required cleanliness level, the square tube is transported to the multi-dimensional weld laser inspection module. The AI ​​visual appearance size detection module includes several sets of CCD cameras with a resolution higher than a preset threshold, several sets of laser rangefinders, and an AI image analysis subsystem. The CCD camera is used to acquire images of the square tube's appearance from different angles and output them to the AI ​​image analysis subsystem; The laser rangefinder is used to measure the side length, diagonal and straightness data of the square tube, and outputs the data as size data to the AI ​​image analysis subsystem. The AI ​​image analysis subsystem uses deep learning algorithms to identify dimensional deviations and surface defects, generates detection results, and synchronizes them to the edge computing server. If the edge computing server determines that the square tube is unqualified, it sends a marking and diversion instruction to the subsequent modules. The qualified square tube then enters the intelligent laser length cutting module. The intelligent laser length-cutting module includes a Q-switched pulse laser, a laser positioning subsystem, and a length measurement sensor; The length measurement sensor is used to monitor the running length of the square tube in real time, with a measurement accuracy of ±0.03mm, and outputs the length data to the edge computing server; When the length data reaches the preset trimming length, the edge computing server sends a trimming command to the Q-switched pulse laser. The laser positioning subsystem is used to ensure the accuracy of the cutting position; The Q-switched pulsed laser cuts the square tube according to the cutting command, with a cutting accuracy of ±0.04mm. The waste generated during cutting automatically falls into the waste collection device. The autonomous navigation AGV handling module includes a laser SLAM autonomous navigation subsystem, a vision recognition subsystem, a robotic arm, and a weight sensor; The visual recognition subsystem is used to identify the specification markings and inspection result marks on the square tubes and output the recognition information to the edge computing server; The weight sensor is used to detect the weight of the square tube and output the weight data to the edge computing server. The edge computing server issues storage instructions and transportation routes based on the identification information and weight data; The laser SLAM autonomous navigation subsystem automatically navigates according to the transport path and dynamically plans obstacle avoidance routes. The robotic arm automatically adjusts its gripping force according to the size of the square tube, moves the square tube to the designated storage area, such as the corresponding position in the qualified or unqualified product area, and feeds the handling data back to the edge computing server in real time.

[0029] like Figure 2 As shown, the following are embodiments of the intelligent laser-based metal square tube forming production line control method provided in this disclosure. This method and the intelligent laser-based metal square tube forming production line control system of the above embodiments belong to the same inventive concept. For details not described in detail in the embodiments of the intelligent laser-based metal square tube forming production line control method, please refer to the embodiments of the intelligent laser-based metal square tube forming production line control system described above.

[0030] The method includes the following steps: S1. The finished steel coil is uncoiled and leveled using an intelligent uncoiling and leveling module to obtain a flat sheet material; It should be noted that this step involves a leveling scheme, which eliminates the internal stress and original curvature of the steel coil, preventing subsequent cutting errors or forming distortions caused by uneven raw materials, and ensuring the smooth progress of the production process. S2. The adaptive laser cutting separation plate module uses a multi-source data fusion decision subsystem to dynamically generate optimal cutting parameters and perform high-precision cutting on the flat plate to obtain at least one set of blanks that meet the preset width. It should be noted that this step uses multi-source data fusion decision-making to dynamically adjust the cutting power and speed, ensuring a smooth and burr-free cut, and can be flexibly adjusted according to production cycle requirements, providing the forming process with dimensionally accurate strip steel billets and ensuring the accuracy of the billet dimensions. S3. Using a flexible roller forming module, the roller parameters are automatically adjusted according to the target square tube specifications to continuously roll the blank to form an open square tube prototype; It should be noted that this step automatically adjusts the roller parameters to continuously roll the flat plate into an open square tube prototype, and the cross-sectional shape data during the forming process is fed back and corrected in real time, ensuring the geometric accuracy of the prototype and providing a good assembly basis for laser welding. S4. Using an intelligent laser welding seam module, laser welding is performed on the seam of the square tube prototype to form a closed square tube; It should be noted that this step uses the high precision and high stability of laser welding to weld the open prototype into a closed square tube. The weld has high strength and small deformation, which ensures the mechanical properties of the square tube. S5. Using the multi-dimensional weld laser detection module, the internal and surface defects of the weld are quantitatively identified in three dimensions by using laser ultrasound and vision multi-level data fusion. The data fusion processor executes the joint judgment logic to output the comprehensive rating result of the weld. At the same time, the defect features and welding process parameters are correlated by rule mining to generate process optimization suggestions, which are then fed back to step S4. It should be noted that this step not only performs a full inspection of the weld, but also performs correlation analysis between the detected defect features and the welding parameters at that time, generating optimization suggestions and feeding them back to the intelligent laser welding seam module, which can continuously improve the welding yield and realize real-time quality monitoring and process evolution.

[0031] This embodiment uses laser ultrasound and vision fusion to achieve three-dimensional quantitative detection of welds, combined with deep neural network adaptive cutting, which solves the problems of difficult production line changeover, missed quality inspections and high energy consumption, and improves the level of intelligent production.

[0032] Furthermore, as a refinement and extension of the specific implementation of the above embodiments, in order to fully illustrate the specific implementation process in this embodiment, another control method for a metal square tube forming production line based on intelligent laser is provided. This method includes the following steps: S1. The finished steel coil is uncoiled and leveled using an intelligent uncoiling and leveling module to obtain a flat sheet material; The specific steps of step S1 are as follows: S11. The intelligent uncoiling and leveling module identifies the material parameters of the steel coil and monitors the change in the diameter of the steel coil in real time, and uploads the material parameters and diameter change data to the edge computing server; S12. The edge computing server automatically matches the unwinding speed and leveling scheme based on the material parameters and diameter change data, generates leveling parameters, and sends them to the intelligent unwinding and leveling module; S13. The intelligent uncoiling and leveling module dynamically adjusts the leveling pressure and plate spacing according to the leveling parameters through a PID algorithm to level the steel coil and obtain a flat plate. S2. The adaptive laser cutting separation plate module uses a multi-source data fusion decision subsystem to dynamically generate optimal cutting parameters and perform high-precision cutting on the flat plate to obtain at least one set of blanks that meet the preset width. The specific steps of step S2 are as follows: S21. The material coefficient and actual thickness of the flat plate are detected in real time by the plate thickness detection sensor, and the aging coefficient of the current laser is obtained. The material coefficient, actual plate thickness and laser aging coefficient are output to the edge controller. S22. The edge controller inputs the received material coefficient, actual plate thickness, laser aging coefficient, and preset production cycle requirements into a pre-trained deep neural network model. After calculation, it outputs the initial optimal cutting parameters, including the optimal laser power. Optimal cutting speed and optimal focus position offset ; In step S22, the deep neural network model adopts a three-layer structure, including an input layer, a hidden layer, and an output layer; The input layer contains four neurons, each corresponding to a material coefficient. Actual thickness of the board Production cycle time requirements and laser aging coefficient Four input features; The hidden layer contains a number of neurons (e.g., 32) and uses the ReLU activation function; The output layer contains three neurons, each corresponding to the optimal laser power. Optimal cutting speed Optimal focal position offset ; The training process of the deep neural network model in step S22 is as follows: Historical cutting process data is collected as training samples, and each training sample includes an input feature vector. and the corresponding optimal cutting parameter labels Use training samples to build a training dataset; in, Material coefficient, This refers to the actual thickness of the board material. This refers to the laser aging factor. To meet production cycle requirements; The deep neural network model is trained under supervision using the training dataset. The mean squared error is used as the loss function. The weights and biases of the model are iteratively updated through the backpropagation algorithm until the loss function converges, and the trained deep neural network model is obtained. Furthermore, to improve the prediction accuracy of deep neural network models for laser aging attenuation, the laser aging coefficient... The method of obtaining it is: Edge computing servers record the cumulative operating time of fiber lasers. And regularly perform power calibration procedures to measure the current actual output power. With nominal power The ratio of the aging coefficient to the aging coefficient is defined as follows: ; in, This refers to the rated lifespan of the laser. The attenuation factor is obtained by fitting historical attenuation curves, and its value ranges from 0.05 to 0.15. The aging coefficients in the training dataset cover the entire range from 0.85 (near the end of life) to 1.00 (brand new laser), with a step size of 0.01 and a total of 16 levels, ensuring that the model has the ability to generalize to parameter drift throughout the entire life cycle of the laser. Furthermore, after the deep neural network model is deployed, the edge computing server automatically extracts the 100 most recent sets of process data with the best actual cutting effect from the production database every 1000 cutting tasks completed, using the criteria that the burr height of the cut is ≤0.02mm and the width of the heat-affected zone is ≤0.08mm, as new training samples to incrementally learn the model. The model weights are updated every 24 hours to enable the model to continuously adapt to equipment wear and environmental changes. Deploy the trained deep neural network model to the edge controller; When running online, the edge controller will acquire the material coefficients in real time. Actual thickness of the board Laser aging coefficient Production cycle time requirements Constructing the input feature vector The input is fed into the trained deep neural network model; After forward computation, the deep neural network model outputs the optimal cutting parameter labels. These are respectively used as the initial optimal laser power, optimal cutting speed, and optimal focal position offset; S23. The fiber laser cutting head is started to cut the plate according to the initial optimal cutting parameters. At the same time, the visual positioning subsystem collects the cut edge image in real time, analyzes and generates the cut quality feedback value, and outputs it to the edge controller. S24. The fuzzy logic compensation controller in the edge controller calculates the power compensation amount in real time based on the deviation between the cut quality feedback value and the preset standard value, and adds it to the initial optimal laser power to generate the final laser power control command. The specific steps in step S24 are as follows: S241. Obtain the incision quality feedback value and calculate the deviation between the incision quality feedback value and the preset standard value. ; S242. Based on the aforementioned deviation The proportional compensation component is calculated using the following formula. :

[0033] in, This is a preset proportional coefficient; S243. Based on the aforementioned deviation The integral compensation component is calculated using the following formula. :

[0034] in, The preset integral coefficient; S244. The proportional compensation component With integral compensation component Add them together to get the power compensation amount. ; S245. Adjust power compensation amount Superimposed to the initial optimal laser power The final laser power control command is obtained. ; Furthermore, in the fuzzy logic compensation controller, the proportional coefficient and integral coefficient It is not a fixed constant, but rather depends on the deviation of the current cut quality. The size is adjusted adaptively in segments; specifically: when When the value is mm, it indicates a large deviation in cut quality, and the system is in the coarse adjustment stage. Setting... , To quickly eliminate deviations; when At that time, the system was in the fine-tuning stage, and settings were being implemented. , To avoid overshooting; when When mm, the system is in the fine-tuning and holding phase, setting , Only minor compensation is made to maintain stability; The above segmented thresholds and coefficient values ​​were obtained by conducting 500 cutting tests on plates of different thicknesses (e.g., 0.3-12mm) and different materials (e.g., Q235 steel, Q355 steel, and 304 stainless steel) with the goal of minimizing the standard deviation of cut quality. This segmented adaptive strategy resolves the contradiction between response speed and stability under a wide operating range with fixed PID coefficients. S25. The edge controller sends the final laser power control command, the optimal cutting speed command, and the optimal focus position offset command to the fiber laser cutting head to execute the cutting action and obtain at least one set of blanks that meet the preset width. S3. Using a flexible roller forming module, the roller parameters are automatically adjusted according to the target square tube specifications to continuously roll the blank to form an open square tube prototype; The specific steps of step S3 are as follows: S31. The edge computing server sends roller adjustment instructions to the flexible roller forming module according to the target square tube specifications. S32. The flexible roller forming module automatically adjusts the parameters of its constituent forming rollers according to the roller adjustment command to continuously roll the blank to form an open square tube prototype; the parameters of the forming rollers include spacing, angle and rotation speed; S33. Through the shape detection function of the flexible roller forming module, the cross-sectional shape data of the square tube prototype is collected in real time, the actual cross-sectional outline is generated through 3D modeling, and fed back to the edge computing server; S34. The edge computing server compares the actual cross-sectional profile with the preset specifications. If there is a deviation, it sends a correction command to the flexible roller forming module to adjust the parameters of the forming roller until the side length deviation of the square tube prototype meets the preset requirements. S4. Using an intelligent laser welding seam module, laser welding is performed on the seam of the square tube prototype to form a closed square tube; The specific steps of step S4 are as follows: S41. Through the visual tracking function of the intelligent laser welding seam module, the position of the seam of the square tube prototype is captured in real time, the compensation offset during the operation of the square tube is automatically calculated, and sent to the edge computing server. S42. By using the weld gap detection function of the intelligent laser welding joint module, the gap size at the joint is monitored and output as gap data to the edge computing server; S43. The edge computing server adjusts the power and welding speed of the main welding laser in the intelligent laser welding seam module according to the gap data, and starts the main welding laser to weld the seam. S44. Through the temperature monitoring function of the intelligent laser welding seam module, the temperature of the weld area is monitored in real time, and the temperature data is fed back to the edge computing server to control the welding process based on the temperature data, thereby avoiding overheating and deformation; S45. If the weld inspection reveals incomplete penetration defects, the edge computing server activates the repair laser in the intelligent laser welding seam module to perform targeted repair welding, forming a closed square tube. S5. Using the multi-dimensional weld laser detection module, the internal and surface defects of the weld are quantitatively identified in three dimensions by using laser ultrasound and vision multi-level data fusion. The data fusion processor executes the joint judgment logic to output the comprehensive rating result of the weld. At the same time, the defect features and welding process parameters are correlated by rule mining to generate process optimization suggestions, which are then fed back to step S4. The specific steps of step S5 are as follows: S51. The laser ultrasonic detection component emits a pulsed laser to the weld and receives the reflected ultrasonic echo signal. Based on the time difference between emission and reception and the echo amplitude, the depth, equivalent size and position coordinates of the internal defects in the weld are quantitatively calculated, and the calculation results are output as internal defect data to the data fusion processor. The specific steps of step S51 are as follows: S511. A pulsed laser is emitted toward the weld seam through the laser ultrasonic detection component, and a timer is started to record the emission time; S512. Receive the ultrasonic echo signal reflected from internal defects in the weld, record the reception time of the received echo, and calculate the time difference between the transmission and reception times. ; S513. Calculate the depth of internal defects according to the following formula. :

[0035] in, The propagation speed of ultrasound in the metal square tube base material; S514. Extracting the amplitude of the ultrasonic echo signal reflected from the defect. The equivalent magnitude of the internal defect is calculated using the following formula. :

[0036] in, and These are pre-calibrated coefficients; S515. Record the position coordinates of the internal defect along the weld direction and the defect depth. Equivalent size The location coordinates are output as internal defect data to the data fusion processor; Furthermore, to improve the signal-to-noise ratio of the ultrasonic echo signal, the pulsed laser emitted in step S511 uses an coded pulse sequence instead of a single pulse. The coded pulse sequence contains three sub-pulses with the same pulse width (e.g., 10 ns) and an interval of 5 μs. The received ultrasonic echo signal is processed by a matched filter for pulse compression, which improves the signal-to-noise ratio by about 9 dB, thereby enabling the detection of micro-defects with a minimum equivalent size of 0.05 mm, which is better than the detection limit of 0.08 mm for conventional single-pulse laser ultrasound. In addition, the pre-calibrated coefficients mentioned in step S514 and The method for determining it is as follows: Standard test blocks of Q355 steel containing artificial defects (such as flat-bottomed holes) were prepared with defect diameters of 0.05 mm, 0.08 mm, 0.10 mm, 0.15 mm, and 0.20 mm, and depths of 1 mm, 2 mm, 3 mm, 5 mm, and 8 mm, respectively. Each defect was tested using a laser ultrasonic testing kit, and the echo amplitude was recorded. The least squares method was used to obtain linear fitting. mm / mV, mm; Different base materials need to be specified separately. and The values ​​are stored in the material parameter library of the edge computing server; S52. Simultaneously acquire weld surface images through the visual imaging subsystem, identify the contour, area, and pixel coordinates of surface defects in the image through the semantic segmentation network, and output the identification results as surface defect data to the data fusion processor; S53. The data fusion processor receives internal defect data and surface defect data, establishes a spatial synchronization mapping matrix, and converts the pixel coordinates of the surface defects into position coordinates along the weld length direction; S54. The data fusion processor performs joint judgment logic on internal defects and surface defects at the same location coordinates and outputs the comprehensive weld rating result; The specific steps of step S54 are as follows: S541. The data fusion processor iterates through the coordinates of each position on the weld, and determines the equivalent size of the internal defect at the corresponding coordinate. If the area of ​​the surface defect is greater than a preset first threshold (e.g., 0.08 mm), or the area of ​​the surface defect is greater than a preset second threshold, then the coordinates of this location are determined to be a defect point. S542. Based on the number and severity of defects on the entire weld, output the overall weld rating as qualified, repairable, or scrap. Furthermore, in the joint determination logic, for welds determined to be repairable, the data fusion processor outputs the specific repair location coordinates. The system identifies the defect type (e.g., internal porosity, surface depressions, etc.) and automatically generates a repair welding path. The edge computing server sends the repair welding path to the repair welding laser of the intelligent laser welding seam module. The repair welding laser performs point remelting repair at a speed of 0.5 m / min along the weld direction with a power 20% lower than that of the main welding laser. After the repair is completed, the multi-dimensional weld laser detection module automatically re-inspects the repaired area. If the re-inspection is still unqualified, it is marked as scrap. The association rule mining uses the Apriori algorithm and sets a minimum support. minimum confidence The target rule for mining is in the form of: {welding parameter conditions} → {defect type}; for example, by mining the welding data and inspection data of 2000 consecutive square tubes, it was found that the rule "weld gap ∈ [0.25mm, 0.35mm] and welding speed ∈ [3.5m / min, 4.0m / min]" → "porosity defect" has a confidence level of 0.92 and a support level of 0.35. Based on this rule, when the weld gap is detected to fall into the above range in real time, the edge computing server automatically reduces the welding speed to 3.0 m / min and increases the laser power by 5%, thereby reducing the porosity defect rate from 12% to 2.3%. All the strongly correlated rules discovered are stored in the process knowledge base of the edge computing server and are updated regularly (e.g., weekly) based on new production data to achieve continuous self-evolution of the welding process. The S55 data fusion processor performs association rule mining on the identified defect features and welding process parameters, generates process optimization suggestions, and outputs them to the edge computing server. Step S5 is followed by the following steps: S6. The surface of the welded square tube is cleaned by pulsed laser using a laser cleaning module; S7. The appearance dimensions and surface quality of the square tube are detected by the AI ​​visual appearance and size detection module; S8. Cut the qualified square tubes according to the preset length using the intelligent laser length cutting module, and transport the cut square tubes to the designated location for classification and storage using the autonomous navigation AGV handling module.

[0037] For example, taking the production of Q355 steel square tubes with a side length of 80mm × 80mm and a length of 9m as an example, the following steps are performed: First, select the raw materials and preset the parameters: Q355 steel coils with a diameter of 1200mm, a thickness of 5mm, and a width of 2.0m were selected. The product parameters are entered through the intelligent MES subsystem of the edge computing server as 80mm×80mm on each side and 9m in length. The quality standards are set as follows: internal defect equivalent ≤ 0.08 mm, surface defect area ≤ 0.5 mm², and dimensional deviation ≤ ±0.02 mm. The production plan is set at 500 pieces per day.

[0038] Next are the running parameters and results of each module: The intelligent uncoiling and leveling module uses a material identification sensor to identify the steel coil as Q355 steel, an infrared ranging sensor to monitor the initial diameter of 1200mm, an edge computing server to match the leveling scheme, and a servo-controlled leveling machine to adjust the leveling roller pressure using a PID algorithm, resulting in an output plate flatness of ≤0.05mm / m. The adaptive laser cutting and separating plate module's plate thickness detection sensor measured a plate thickness of 5.02mm, with a laser aging coefficient of 0.92 (cumulative working time of 800 hours) and a production cycle requirement of 6 pieces / minute. The deep neural network model outputs initial cutting parameters: optimal laser power of 2600W, optimal cutting speed of 6m / min, and optimal focal point position offset of -2.05mm. During the cutting process, visual positioning feedback indicates a burr height of 0.02mm, and the fuzzy logic compensation controller calculates the power compensation amount. =+80W, the final laser power is 2680W; the width of the blank after cutting is 320mm, the dimensional deviation is ±0.03mm, the width of the heat-affected zone is 0.08mm, and the cut is burr-free; The flexible roller forming module receives a specification of 80mm side length from the edge computing server, automatically adjusts the roller spacing to 80mm, performs real-time 3D modeling through shape detection sensors to obtain the actual cross-sectional outline of the square tube prototype, and feeds it back to the edge computing server. The edge computing server compares the actual cross-sectional outline with the preset specification. If there is a deviation, it issues a correction command to adjust the roller spacing until the side length deviation of the square tube prototype meets the preset requirements. Finally, the side length deviation is controlled within ±0.05mm. The intelligent laser welding seam module calculates a compensation offset of 0.015mm through the visual tracking subsystem, obtains a weld gap of 0.3mm through the weld gap detection sensor, and feeds this information back to the edge computing server. The edge computing server adjusts the main welding power to 4200W and the welding speed to 4m / min. The temperature monitoring sensor shows that the temperature in the weld area is 650±20℃, and no overheating alarm is triggered. The repair welding laser is not activated. The weld strength reaches 96% of the base material, and the surface roughness Ra is 2.8μm. The multidimensional weld laser inspection module uses a laser ultrasonic inspection component to determine that there are no defects ≥0.08mm inside the weld, and a visual imaging subsystem to determine that there are no defects ≥0.5mm² on the weld surface. Together, they are judged as qualified products. The data fusion processor performs association rule mining to discover the rule that "gap of 0.3-0.35mm and welding power need to be increased by 5%", which is automatically recorded to the process knowledge base.

[0039] In subsequent modules, after laser cleaning by the laser cleaning module, the residual oil on the surface of the square tube is 0.003g / m²; the AI ​​visual appearance inspection module determines that the side length of the square tube is 80.01mm and the straightness is 0.02mm / m, which is qualified; the intelligent laser length cutting module determines the accuracy to be ±0.03mm and the perpendicularity of the cutting surface to be 0.015mm / m; the autonomous navigation AGV handling module transports the square tube to the qualified product area.

[0040] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0041] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A control system for a metal square tube forming production line based on intelligent laser, characterized in that, It includes a smart uncoiling and leveling module, an adaptive laser cutting and separating plate module, a flexible roller forming module, a smart laser welding and joining module, a multi-dimensional weld laser detection module, and an edge computing server that is connected in sequence to each of the above modules. The edge computing server acquires the sensor data and operating status of each module in real time through a dual-mode communication network, and uses the built-in intelligent MES subsystem to issue control commands to each module to perform closed-loop collaborative control of the production line. The adaptive laser cutting and separating plate module integrates a pre-trained deep neural network model and a fuzzy logic compensation controller, which dynamically calculates and adjusts the optimal cutting parameters based on the real-time physical parameters of the incoming plate and the equipment status. The multidimensional weld laser inspection module integrates a laser ultrasonic inspection component, a vision imaging subsystem, and a data fusion processor. It is used to quantitatively identify internal and surface defects in the weld in three-dimensional space, and to perform correlation analysis with the process parameters of the intelligent laser welding joint module to optimize the welding quality in a closed loop.

2. The intelligent laser-based metal square tube forming production line control system according to claim 1, characterized in that, The adaptive laser cutting and separating sheet metal module includes a sheet metal thickness detection sensor, a vision positioning subsystem, an edge controller, and several sets of fiber laser cutting heads; The sheet thickness detection sensor is used to detect the material coefficient and actual thickness of the incoming sheet in real time and output the data to the edge controller. The visual positioning subsystem is used to acquire images of the cut edges in real time during the cutting process, analyze and generate cut quality feedback values, and output them to the edge controller. A pre-trained deep neural network model and a fuzzy logic compensation controller are deployed in the edge controller; The deep neural network model receives the material coefficient, actual plate thickness, cut quality feedback value, production cycle requirements, and laser aging coefficient as input, and outputs the initial optimal laser power, optimal cutting speed, and optimal focal position offset after calculation. The fuzzy logic compensation controller calculates the power compensation amount in real time based on the deviation between the cut quality feedback value and the preset standard value, and adds it to the optimal laser power to generate the final laser power control command. Each fiber laser cutting head receives the final laser power control command, optimal cutting speed command, and optimal focus position offset command issued by the edge controller, and performs the cutting action on the plate.

3. The intelligent laser-based metal square tube forming production line control system according to claim 1, characterized in that, The laser ultrasonic testing component is used to emit pulsed lasers into the weld and receive the reflected ultrasonic echo signals. Based on the time difference between emission and reception and the echo amplitude, it quantitatively calculates the depth, equivalent size, and position coordinates of internal defects in the weld, and outputs them as internal defect data to the data fusion processor. The visual imaging subsystem is used to synchronously acquire images of the weld surface, identify the contours, areas and pixel coordinates of surface defects in the images through a semantic segmentation network, and output them as surface defect data to the data fusion processor. The data fusion processor receives internal defect data and surface defect data, establishes a spatial synchronization mapping matrix, converts the pixel coordinates into position coordinates along the weld length direction, and performs joint judgment logic on internal defects and surface defects under the same position coordinates, outputting a comprehensive weld rating result; at the same time, it performs association rule mining on the identified defect features and externally input welding process parameters, generates process optimization suggestions, and outputs them to the edge computing server.

4. A control method for a metal square tube forming production line based on intelligent laser, using the intelligent laser-based metal square tube forming production line control system according to any one of claims 1-3, characterized in that, Includes the following steps: S1. The finished steel coil is uncoiled and leveled using an intelligent uncoiling and leveling module to obtain a flat sheet material; S2. The adaptive laser cutting separation plate module uses a multi-source data fusion decision subsystem to dynamically generate optimal cutting parameters and perform high-precision cutting on the flat plate to obtain at least one set of blanks that meet the preset width. S3. Using a flexible roller forming module, the roller parameters are automatically adjusted according to the target square tube specifications to continuously roll the blank to form an open square tube prototype; S4. Using an intelligent laser welding seam module, laser welding is performed on the seam of the square tube prototype to form a closed square tube; S5. Using the multi-dimensional weld laser detection module, the internal and surface defects of the weld are quantitatively identified in three dimensions by using a multi-level data fusion method of laser ultrasound and vision. The data fusion processor executes the joint judgment logic to output the comprehensive rating result of the weld. At the same time, the defect features and welding process parameters are correlated by rule mining to generate process optimization suggestions, which are then fed back to step S4.

5. The method for controlling a metal square tube forming production line based on intelligent laser according to claim 4, characterized in that, The specific steps of step S2 are as follows: S21. The material coefficient and actual thickness of the flat plate are detected in real time by the plate thickness detection sensor, and the aging coefficient of the current laser is obtained. The material coefficient, actual plate thickness and laser aging coefficient are output to the edge controller. S22. The edge controller inputs the received material coefficient, actual plate thickness, laser aging coefficient, and preset production cycle requirements into a pre-trained deep neural network model. After calculation, it outputs the initial optimal cutting parameters, including the optimal laser power. Optimal cutting speed and optimal focus position offset ; S23. The fiber laser cutting head is started to cut the plate according to the initial optimal cutting parameters. At the same time, the visual positioning subsystem collects the cut edge image in real time, analyzes and generates the cut quality feedback value, and outputs it to the edge controller. S24. The fuzzy logic compensation controller in the edge controller calculates the power compensation amount in real time based on the deviation between the cut quality feedback value and the preset standard value, and adds it to the initial optimal laser power to generate the final laser power control command. S25. The edge controller sends the final laser power control command, the optimal cutting speed command, and the optimal focal position offset command to the fiber laser cutting head to execute the cutting action and obtain at least one set of blanks that meet the preset width.

6. The method for controlling a metal square tube forming production line based on intelligent laser according to claim 5, characterized in that, The training process of the deep neural network model in step S22 is as follows: Historical cutting process data is collected as training samples, and each training sample includes an input feature vector. and the corresponding optimal cutting parameter labels Use training samples to build a training dataset; in, Material coefficient, This refers to the actual thickness of the board material. This refers to the laser aging factor. To meet production cycle requirements; The deep neural network model is trained under supervision using the training dataset. The mean squared error is used as the loss function. The weights and biases of the model are iteratively updated through the backpropagation algorithm until the loss function converges, and the trained deep neural network model is obtained. Deploy the trained deep neural network model to the edge controller; When running online, the edge controller will acquire the material coefficients in real time. Actual thickness of the board Laser aging coefficient Production cycle time requirements Constructing the input feature vector The input is fed into the trained deep neural network model; After forward computation, the deep neural network model outputs the optimal cutting parameter labels. These are respectively used as the initial optimal laser power, optimal cutting speed, and optimal focal position offset; The specific steps in step S24 are as follows: S241. Obtain the incision quality feedback value and calculate the deviation between the incision quality feedback value and the preset standard value. ; S242. Based on the aforementioned deviation The proportional compensation component is calculated using the following formula. : in, This is a preset proportionality coefficient; S243. Based on the aforementioned deviation The integral compensation component is calculated using the following formula. : in, The preset integral coefficient; S244. The proportional compensation component With integral compensation component Add them together to get the power compensation amount. ; S245. Adjust power compensation amount Superimposed to the initial optimal laser power The final laser power control command is obtained. .

7. The method for controlling a metal square tube forming production line based on intelligent laser according to claim 5, characterized in that, The specific steps of step S3 are as follows: S31. The edge computing server sends roller adjustment instructions to the flexible roller forming module according to the target square tube specifications. S32. The flexible roller forming module automatically adjusts the parameters of its constituent forming rollers according to the roller adjustment command to continuously roll the blank to form an open square tube prototype; the parameters of the forming rollers include spacing, angle and rotation speed; S33. Through the shape detection function of the flexible roller forming module, the cross-sectional shape data of the square tube prototype is collected in real time, the actual cross-sectional outline is generated through 3D modeling, and fed back to the edge computing server; S34. The edge computing server compares the actual cross-sectional profile with the preset specifications. If there is a deviation, it sends a correction command to the flexible roller forming module to adjust the parameters of the forming rollers until the side length deviation of the square tube prototype meets the preset requirements.

8. The method for controlling a metal square tube forming production line based on intelligent laser according to claim 5, characterized in that, The specific steps of step S4 are as follows: S41. Through the visual tracking function of the intelligent laser welding seam module, the position of the seam of the square tube prototype is captured in real time, the compensation offset during the operation of the square tube is automatically calculated, and sent to the edge computing server. S42. By using the weld gap detection function of the intelligent laser welding joint module, the gap size at the joint is monitored and output as gap data to the edge computing server; S43. The edge computing server adjusts the power and welding speed of the main welding laser in the intelligent laser welding seam module according to the gap data, and starts the main welding laser to weld the seam. S44. Through the temperature monitoring function of the intelligent laser welding seam module, the temperature of the weld area is monitored in real time, and the temperature data is fed back to the edge computing server to control the welding process based on the temperature data, thereby avoiding overheating and deformation; S45. If the weld inspection reveals incomplete penetration defects, the edge computing server activates the repair laser in the intelligent laser welding seam module to perform targeted repair welding, forming a closed square tube.

9. The control method for a metal square tube forming production line based on intelligent laser according to claim 4, characterized in that, The specific steps of step S5 are as follows: S51. The laser ultrasonic detection component emits a pulsed laser to the weld and receives the reflected ultrasonic echo signal. Based on the time difference between emission and reception and the echo amplitude, the depth, equivalent size and position coordinates of the internal defects in the weld are quantitatively calculated, and the calculation results are output as internal defect data to the data fusion processor. S52. Simultaneously acquire weld surface images through the visual imaging subsystem, identify the contour, area, and pixel coordinates of surface defects in the image through the semantic segmentation network, and output the identification results as surface defect data to the data fusion processor; S53. The data fusion processor receives internal defect data and surface defect data, establishes a spatial synchronization mapping matrix, and converts the pixel coordinates of the surface defects into position coordinates along the weld length direction; S54. The data fusion processor performs joint judgment logic on internal defects and surface defects at the same location coordinates and outputs the comprehensive weld rating result; The S55 data fusion processor performs association rule mining on the identified defect features and welding process parameters, generates process optimization suggestions, and outputs them to the edge computing server.

10. The method for controlling a metal square tube forming production line based on intelligent laser according to claim 9, characterized in that, The specific steps of step S51 are as follows: S511. A pulsed laser is emitted toward the weld seam through the laser ultrasonic detection component, and a timer is started to record the emission time; S512. Receive the ultrasonic echo signal reflected from internal defects in the weld, record the reception time of the received echo, and calculate the time difference between the transmission and reception times. ; S513. Calculate the depth of internal defects according to the following formula. : in, The propagation speed of ultrasound in the metal square tube base material; S514. Extracting the amplitude of the ultrasonic echo signal reflected from the defect. The equivalent magnitude of the internal defect is calculated using the following formula. : in, and These are pre-calibrated coefficients; S515. Record the position coordinates of the internal defect along the weld direction and the defect depth. Equivalent size The location coordinates are output as internal defect data to the data fusion processor; The specific steps of step S54 are as follows: S541. The data fusion processor iterates through the coordinates of each position on the weld, and determines the equivalent size of the internal defect at the corresponding coordinate. If the area of ​​the surface defect is greater than the preset first threshold, or the area of ​​the surface defect is greater than the preset second threshold, then the coordinates of this location are determined to be a defect point. S542. Based on the number and severity of defects on the entire weld, output the overall weld rating as qualified, repairable, or scrap.

Images