A multi-arc parallel additive manufacturing arc length adaptive detection and intelligent feedback control system and method

Abstract

The application discloses a kind of multi-arc parallel additive manufacturing arc length adaptive detection and intelligent feedback control system and method, system includes: image acquisition unit: for collecting molten pool image of arc additive manufacturing process;Adaptive parameter adjustment unit: for dynamically adjusting the parameter of image processing unit according to molten pool image statistics characteristics;Image processing unit: for processing the molten pool image collected, detects arc length;Duty cycle calculation and control unit: for calculating duty cycle adjustment amount by proportional control algorithm according to the deviation of detected arc length and preset target value, and generating control instruction;Communication unit: for sending control instruction to arc additive manufacturing heat source equipment using MQTT protocol, and receiving external instruction;Control unit: for coordinating the operation of each unit, and supporting the data synchronization and state management of multi-channel collaborative work subsystem.The application improves the quality and stability of large component additive manufacturing.

Description

Technical Field

[0001] This invention relates to the field of additive manufacturing technology, and in particular to a multi-arc parallel additive manufacturing arc length adaptive detection and intelligent feedback control system and method. Background Technology

[0002] Wafer Additive Manufacturing (WAAM) is an important metal additive manufacturing technology with broad application prospects in the manufacture of large-size metal components due to its high forming efficiency and relatively low cost. The process stability of WAAM is crucial to the forming quality, and the arc length is one of the key parameters affecting the stability of the WAAM process and the behavior of the molten pool. Unstable arc lengths can lead to fluctuations in melt depth, melt width, and build-up height, thereby affecting the geometric accuracy, microstructure, and mechanical properties of the final formed part.

[0003] Traditional WAAM process control often relies on fixed process parameters or requires manual adjustments by operators based on experience. This approach struggles to handle the complex dynamic changes during arc additive manufacturing (such as fluctuations in wire melting speed, changes in molten pool morphology, and environmental interference), leading to unstable arc lengths and making it difficult to guarantee the stability and consistency of mass production.

[0004] Existing automated control systems may use electrical signals (such as voltage and current) for feedback control, but electrical signals are affected by various factors and cannot directly and accurately reflect the actual arc length and molten pool state. Some vision-based detection methods may be sensitive to the intense light and spatter in arc additive manufacturing, have poor robustness, and be difficult to adapt to different arc additive manufacturing conditions, thus limiting their application in actual WAAM processes.

[0005] Therefore, how to achieve accurate, real-time, and adaptive detection of the arc length in the WAAM process, and how to perform intelligent feedback control based on this, is a problem that WAAM technology urgently needs to solve. Summary of the Invention

[0006] To address the technical problem of accurate detection and stable control of arc length in existing electric arc additive manufacturing technologies, this invention provides an adaptive detection and intelligent feedback control system and method for arc length in multi-arc parallel additive manufacturing. This system achieves real-time high-precision detection of arc length through computer vision technology, and adaptively adjusts the molten pool image processing parameters and duty cycle based on the detection results to maintain a stable arc length, thereby improving the quality and stability of additive manufacturing of large components.

[0007] On the one hand, to achieve the above objectives, the present invention provides a multi-arc parallel additive manufacturing arc length adaptive detection and intelligent feedback control system, comprising:

[0008] Image acquisition unit: used to acquire images of the molten pool during the arc additive manufacturing process;

[0009] Adaptive parameter adjustment unit: used to dynamically adjust the parameters of the image processing unit according to the statistical characteristics of the molten pool image;

[0010] Image processing unit: used to process the acquired molten pool image and detect the arc length;

[0011] Duty cycle calculation and control unit: used to calculate the duty cycle adjustment amount based on the detected deviation between the arc length and the preset target value, and generate control commands;

[0012] Communication unit: used to send the control commands to the arc additive manufacturing heat source equipment using the MQTT protocol, and to receive external commands;

[0013] Control unit: Used to coordinate the operation of each unit and support data synchronization and status management of multi-channel collaborative subsystems.

[0014] Preferably, the adaptive parameter adjustment unit includes:

[0015] Image statistical analysis module: used to calculate the average brightness and standard deviation of the molten pool image;

[0016] Binarization threshold dynamic adjustment module: used to update the binarization threshold based on the average brightness and standard deviation, and limit it within a first preset range;

[0017] Morphological kernel adaptive module: used to adjust the size of the denoising kernel according to the noise level, wherein the size of the denoising kernel is positively correlated with the noise intensity of the melt pool image;

[0018] Parameter History Recording Module: Used to record historical data of adjustments and track the trend of parameter changes.

[0019] Preferably, the acquired molten pool image is processed, including:

[0020] The original acquired color molten pool image is converted into a grayscale image, and a contrast-limited adaptive histogram equalization method is applied to enhance the contrast.

[0021] Based on peak state detection, the peak period and the base period of the electric arc additive manufacturing process are distinguished, and the arc length is detected during the base period;

[0022] Splash interference is removed by opening morphological operations, and non-detection areas are shielded by masking.

[0023] The algorithm uses contour detection, image segmentation, or connected component analysis to identify the largest area contour as the suspected molten pool region and automatically detects the tungsten electrode position. If the tungsten electrode is incorrectly positioned, the algorithm manually selects the tungsten electrode to cover the automatically detected tungsten electrode position.

[0024] The image binarization algorithm converts the processed grayscale image into a black and white image. The pixels in the suspected molten pool area are white, and the square area enclosed by the coordinates of the four outermost vertices in the suspected molten pool area is taken as the molten pool scanning area.

[0025] A line-by-line scanning analysis is performed within the molten pool scanning area to calculate the proportion of the suspected molten pool area in each row of pixels. Based on the morphological differences between the filament and the molten pool, a line-by-line analysis is performed from top to bottom. The critical position where the proportion of white pixels in the first row exceeds the threshold is determined by the dynamic threshold method. The critical position is used as the boundary baseline between the filament and the molten pool. The morphological characteristics of the filament are elongated, and the morphological characteristics of the molten pool are diffused.

[0026] The vertical distance from the tip of the tungsten electrode to the baseline of the wire-molten pool boundary is calculated to obtain the arc length, and the arc length is smoothed.

[0027] Preferably, the duty cycle calculation and control unit includes:

[0028] Reference control module: used to set the reference duty cycle, reference arc length, and adjustment coefficient;

[0029] Adaptive adjustment module: used to adjust the duty cycle value based on the deviation between the measured arc length and the reference arc length and the adjustment coefficient, and control the adjustment range of the duty cycle within a second preset range.

[0030] Preferably, the communication unit includes:

[0031] Multi-channel data aggregation module: used to encapsulate the duty cycle parameter adjustment data of each printhead channel into a JSON format message and publish it to the waam / expected channel of MQTT;

[0032] Heartbeat mechanism module: Used to periodically send status information to the waam / heartbeat topic of MQTT;

[0033] Will message configuration module: used to automatically publish offline notifications when an abnormal disconnection occurs;

[0034] Parameter modification module: Used to subscribe to the MQTT waam / expected channel of the arc additive manufacturing heat source equipment control computer, obtain the duty cycle calculated in real time, and send it to the arc additive manufacturing heat source equipment via PLC using the EtherCAT communication protocol to modify the parameters of the arc additive manufacturing heat source equipment.

[0035] Preferably, the multi-channel collaborative working subsystem includes:

[0036] Independent channel processor: used to configure a dedicated image processing unit and adaptive parameter adjustment unit for each additive manufacturing printhead;

[0037] Data synchronization and status management module: used to achieve synchronous transmission of multiple images through edge computing nodes;

[0038] Dynamic configuration loading module: used to update channel parameters and control strategies in real time based on configuration files;

[0039] Multi-channel collaborative module: used to simultaneously process and extract arc length, calculate duty cycle parameter adjustment data, and send duty cycle parameter adjustment data using several channels, while changing the duty cycle parameters of several arc additive manufacturing heat source devices. Each channel corresponds to one printhead.

[0040] Preferably, the system further includes a user interface unit, which is used to display the original molten pool image, processing results, arc length measurement value and duty cycle parameter adjustment value of each channel in real time.

[0041] On the other hand, to achieve the above objectives, the present invention also provides a method for adaptive detection and intelligent feedback control of arc length in multi-arc parallel additive manufacturing, comprising:

[0042] Real-time acquisition of molten pool images during the arc additive manufacturing process, followed by grayscale conversion, contrast enhancement, and binarization.

[0043] Dynamically adjust the binarization threshold and morphological kernel parameters based on the statistical characteristics of the molten pool image;

[0044] The tungsten electrode tip and the top of the molten pool are located by contour detection, and the arc length is calculated.

[0045] Based on the arc length deviation, the duty cycle is adjusted according to a preset ratio;

[0046] The adjusted duty cycle is sent to the heat source control computer via the MQTT protocol, and then sent to the arc additive manufacturing heat source equipment via the EtherCAT communication protocol using the PLC, forming a closed-loop control.

[0047] Preferably, adjusting the morphological kernel parameters includes:

[0048] The noise level of the molten pool image is evaluated using the Laplacian operator, and the morphological kernel size is dynamically adjusted based on the noise level. The parameter adjustment history is recorded for offline optimization and fault diagnosis.

[0049] Preferably, calculating the arc length includes:

[0050] In the base period image, splash interference is removed by opening operation, and non-detection areas are masked by masking.

[0051] The algorithm uses contour detection, image segmentation, or connected component analysis to identify the largest area contour as the suspected molten pool region and automatically detects the tungsten electrode position. If the tungsten electrode is incorrectly positioned, the algorithm manually selects the tungsten electrode to cover the automatically detected tungsten electrode position.

[0052] The image binarization algorithm converts the processed grayscale image into a black and white image. The pixels in the suspected molten pool area are white, and the square area enclosed by the coordinates of the four outermost vertices in the suspected molten pool area is taken as the molten pool scanning area.

[0053] A line-by-line scanning analysis is performed within the molten pool scanning area to calculate the proportion of the suspected molten pool area in each row of pixels. Based on the morphological differences between the filament and the molten pool, a line-by-line analysis is performed from top to bottom. The critical position where the proportion of white pixels in the first row exceeds the threshold is determined by the dynamic threshold method. The critical position is used as the boundary baseline between the filament and the molten pool. The morphological characteristics of the filament are elongated, and the morphological characteristics of the molten pool are diffused.

[0054] The vertical distance from the tip of the tungsten electrode to the baseline of the wire-molten pool boundary is calculated to obtain the arc length, and the arc length is smoothed.

[0055] Compared with the prior art, the present invention has the following advantages and technical effects:

[0056] (1) This invention achieves high-precision real-time detection of arc length, providing a basis for precise control; introduces an adaptive parameter adjustment mechanism to improve the robustness of the system in different additive manufacturing environments of large components; and effectively stabilizes the arc and improves the molten pool morphology and forming quality by intelligently adjusting the duty cycle through arc length feedback.

[0057] (2) The system of the present invention has strong real-time performance and fast response speed, and can respond to process changes in a timely manner. Through modular design, it is easy to expand and maintain, and supports multi-channel collaborative work. It adopts the standard communication protocol MQTT, which is easy to integrate with existing large component additive manufacturing equipment and host computer, and is easy to realize remote monitoring and control.

[0058] (3) This invention reduces the reliance on operator skills and improves the level of automation and production efficiency. Attached Figure Description

[0059] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments and descriptions of this application are used to explain this application and do not constitute an undue limitation of this application. In the drawings:

[0060] Figure 1 This is a schematic diagram of a multi-arc parallel additive manufacturing arc length adaptive detection and intelligent feedback control system according to an embodiment of the present invention;

[0061] Figure 2 This is a flowchart of a method for adaptive detection and intelligent feedback control of arc length in multi-arc parallel additive manufacturing according to an embodiment of the present invention;

[0062] Figure 3 This is a reference diagram of the wire-molten pool boundary baseline in an embodiment of the present invention. Detailed Implementation

[0063] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0064] It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0065] This section describes the simultaneous control of multiple electric arcs:

[0066] In this embodiment, each printhead is equipped with a camera, and the four cameras are connected to an edge computing board or microcomputer. The edge computing board or microcomputer is responsible for camera control and sending video data (via a network communication interface implemented through RTP socket - Real-time Transport Protocol) to the central computing server. The central computing server extracts the arc length of the molten pool of each printhead from the video image (one-to-one correspondence between camera and printhead / molten pool), makes control decisions based on the arc length, calculates the optimal duty cycle, and sends the duty cycle data of the printhead obtained from the decision calculation to the arc additive manufacturing heat source equipment control computer via the MQTT communication protocol. The arc additive manufacturing heat source equipment control computer sends the data to the arc additive manufacturing heat source equipment via PLC using the EtherCAT communication protocol to modify its duty cycle parameter.

[0067] This embodiment proposes a multi-arc parallel additive manufacturing arc length adaptive detection and intelligent feedback control system, such as... Figure 1 ,include:

[0068] Image acquisition unit: used to acquire images of the molten pool during the arc additive manufacturing process;

[0069] Adaptive parameter adjustment unit: used to dynamically adjust the parameters of the image processing unit according to the statistical characteristics of the molten pool image;

[0070] Image processing unit: used to process the acquired molten pool image and detect the arc length;

[0071] Duty cycle calculation and control unit: used to calculate the duty cycle adjustment amount based on the detected deviation between the arc length and the preset target value, and generate control commands;

[0072] Communication unit: used to send the control commands to the arc additive manufacturing heat source equipment using the MQTT protocol, and to receive external commands;

[0073] Control unit: Used to coordinate the operation of each unit and support data synchronization and status management of multi-channel collaborative subsystems.

[0074] Specifically, the system adopts a modular design, with each functional unit cooperating with each other to achieve real-time detection of the arc length and intelligent control of the duty cycle.

[0075] Furthermore, the adaptive parameter adjustment unit includes:

[0076] Image statistical analysis module: used to calculate the average brightness and standard deviation of the molten pool image;

[0077] Binarization threshold dynamic adjustment module: used to update the binarization threshold based on the average brightness and standard deviation, and limit it within a first preset range;

[0078] Morphological kernel adaptive module: used to adjust the size of the denoising kernel according to the noise level, wherein the size of the denoising kernel is positively correlated with the noise intensity of the melt pool image;

[0079] Parameter History Recording Module: Used to record historical data of adjustments and track the trend of parameter changes.

[0080] Specifically, to adapt to different environments and conditions in electric arc additive manufacturing, this embodiment adopts an adaptive parameter adjustment mechanism, implemented through the AdaptiveParameters class:

[0081] Image statistical analysis module: The first step of the algorithm is the statistical analysis of the molten pool image. It obtains key information by calculating the average brightness value and standard deviation of the input molten pool image. The average brightness value reflects the overall illumination level of the molten pool image, while the standard deviation characterizes the contrast and noise level of the molten pool image. These basic statistical quantities provide an important basis for subsequent parameter adjustments, enabling the system to intelligently adjust according to the actual characteristics of the molten pool image.

[0082] The binarization threshold dynamic adjustment module is a core component of the algorithm. It dynamically adjusts the binarization threshold based on the statistical features of the molten pool image. The system adjusts the base threshold according to the average brightness value, while also considering the standard deviation to adapt to contrast variations in the molten pool image. To ensure processing effectiveness, the threshold is limited to a preset minimum and maximum value range and fine-tuned based on the molten pool image distribution characteristics (left / right skew) to make the binarization result more accurate.

[0083] The morphological kernel adaptive module first evaluates the noise level of the melt pool image using the Laplacian operator, then calculates the base kernel size based on the melt pool image dimensions. The kernel size is dynamically adjusted as the noise level changes, ensuring effective removal of spatter while avoiding over-smoothing. The system ensures that the kernel size is odd and limits it to a reasonable range of 3 to 21 to maintain consistent processing results.

[0084] The kernel size for tungsten electrode detection is updated using a feature analysis approach. The system employs Canny edge detection and Hough transform to analyze the feature sizes in the molten pool image, determining the appropriate kernel size by calculating the median size of typical features. The kernel size is dynamically adjusted based on feature sizes and the molten pool image size to ensure accurate capture of tungsten electrode features. The system ensures the kernel size is odd and limits it to a reasonable range of 9 to 51 to maintain detection accuracy.

[0085] Parameter History Record Module: The parameter history record is a crucial component of the algorithm, documenting all historical data related to parameter adjustments. The system records the current frame number, statistical features of the molten pool image (average brightness, standard deviation), and adjusted parameter values ​​(binarization threshold, spatter kernel size, tungsten electrode kernel size). This historical data not only helps track parameter trends but also provides vital reference for subsequent algorithm optimization, enabling the system to continuously improve and optimize processing results.

[0086] Adaptive parameter adjustment mainly includes the following mechanisms:

[0087] Image statistical analysis: Analysis of the mean and standard deviation of molten pool image brightness;

[0088] Binarization threshold adjustment: The binarization threshold is dynamically adjusted based on the characteristics of the molten pool image;

[0089] Morphological kernel size adjustment: Adjust the size of the denoising kernel according to the noise level of the melt pool image;

[0090] Tungsten electrode detection parameter adjustment: Adjust the tungsten electrode detection parameters according to the detection results;

[0091] Parameter history record: Records the history of parameter changes for subsequent analysis and optimization.

[0092] Further processing of the acquired molten pool images includes:

[0093] The original acquired color molten pool image is converted into a grayscale image, and a contrast-limited adaptive histogram equalization method is applied to enhance the contrast.

[0094] Based on peak state detection, the peak period and the base period of the electric arc additive manufacturing process are distinguished, and the arc length is detected during the base period;

[0095] Splash interference is removed by opening morphological operations, and non-detection areas are shielded by masking.

[0096] The maximum area contour is identified as a suspected molten pool region using contour detection, image segmentation, or connected component analysis algorithms. The tungsten electrode position is automatically detected using algorithms such as Hough transform. If the tungsten electrode is incorrectly located, the tungsten electrode is manually selected to cover the tungsten electrode and the tungsten electrode position is automatically detected.

[0097] The image binarization algorithm converts the processed grayscale image into a black and white image. The pixels in the suspected molten pool area are white, and the square area enclosed by the coordinates of the four outermost vertices in the suspected molten pool area is taken as the molten pool scanning area.

[0098] A line-by-line scanning analysis is performed within the molten pool scanning area to calculate the proportion of the suspected molten pool area in each row of pixels. Based on the morphological differences between the filament and the molten pool, and since the filament is generally above the molten pool, the analysis proceeds line by line from top to bottom. A dynamic thresholding method is used to determine the critical position where the proportion of white pixels in the first row exceeds a threshold. This critical position is used as the filament-molten pool boundary baseline (e.g., ...). Figure 3 Among them, the morphological characteristics of the filament are slender, and the morphological characteristics of the molten pool are diffused.

[0099] The vertical distance from the tip of the tungsten electrode to the baseline of the wire-molten pool boundary is calculated to obtain the arc length, and the arc length is smoothed.

[0100] Specifically, the acquired molten pool images are processed, including grayscale conversion, contrast enhancement, binarization, morphological processing, contour detection, and tungsten electrode and molten pool top positioning, in order to detect the arc length.

[0101] In this embodiment, the arc length detection adopts a computer vision-based method, mainly implemented through the ArcLengthDetector class, and its core algorithm is as follows:

[0102] Grayscale image: Create a copy of the molten pool image for processing. If the molten pool image is a color image (such as an RGB three-channel color image), it is converted to a grayscale image. Image contrast enhancement methods (such as the CLAHE algorithm) are used to enhance the contrast of the molten pool image.

[0103] Peak Recognition: A peak recognition algorithm is proposed to detect whether the image is a peak image. If it is a peak image, no processing is performed; if it is a base image, processing is performed to obtain the arc length.

[0104] Binarization: For peak images, the grayscale image is binarized. The algorithm iterates through each pixel in the grayscale image and compares each pixel value with the set threshold self.bi_min. If the pixel value is greater than the threshold, the pixel is set to 255 (white) in the output image. If the pixel value is less than or equal to the threshold, the pixel is set to 0 (black) in the output image.

[0105] The morphological algorithm for removing spatter includes kernel creation and opening operations. Kernel creation: Creates a matrix of 1s as its structuring element, defining the size of the spatter kernel, such as (3,3) or (5,5), and specifying the data type as an 8-bit unsigned integer. Opening operation: A combination of two basic morphological operations: erosion followed by dilation.

[0106] Algorithm principle:

[0107] Erosion: The kernel slides across the molten pool image, and the center pixel remains 1 only when all pixels covered by the kernel are 1. This shrinks the white area (value 255) and removes small white spots (splatter).

[0108] Dilation: The kernel slides across the molten pool image. As long as one pixel in the area covered by the kernel is 1, the center pixel is set to 1. This expands the white area and restores the size of the main features.

[0109] Effect of spatter removal: Opening operations can remove white spots (spatters) smaller than the core while maintaining the integrity of key features (such as welds). For larger spatter, the core size needs to be increased accordingly.

[0110] The impact of parameter selection: The size of the spatter core determines the size of the spatter that can be removed. The larger the core, the larger the spatter that can be removed, but it may also affect the main features. The shape of the core (here, a rectangle is used) will also affect the processing effect.

[0111] The mask processing algorithm works as follows: The mask is a binary image of the same size as the original melt pool image. Pixel values ​​in the mask are either 0 or 255 (black or white). White areas (255) represent areas to be retained, and black areas (0) represent areas to be masked. As needed, the areas to be masked are set; for example, if there is a watermark in the melt pool image, the mask for that area is set to 0. A bitwise AND operation is performed between the mask and the original melt pool image. In the resulting image, areas where the mask is black will be completely masked (become black), while areas where the mask is white will retain their original pixel values.

[0112] Contour Detection: Using OpenCV's contour detection functionality, all external contours are found in the binarized image. It only detects the outermost contour (RETR_EXTERNAL) and uses a simple approximation method (CHAIN_APPROX_SIMPLE) to reduce the number of contour points, retaining only keypoints. This improves processing efficiency and reduces memory usage.

[0113] Maximum contour identification: Among all detected contours, the one with the largest area is identified as the weld pool. This is based on the assumption that weld pools are generally larger than wire and other features. The area of ​​each contour is calculated using the `cv2.contourArea` function, and then the contour with the largest area is found using the `max` function in conjunction with the `key` parameter.

[0114] Within the defined molten pool area, a line-by-line scan analysis is performed to calculate the proportion of the molten pool in each row of pixels. Based on the morphological differences between the filament and the molten pool—the filament is elongated and the molten pool is diffused—the system uses a dynamic threshold method to determine the critical position where the duty cycle of the first row exceeds the threshold, which serves as the baseline for the filament-molten pool boundary.

[0115] Arc length calculation: The distance from the tip of the tungsten electrode to the wire-molten pool boundary baseline is calculated, i.e., the arc length. Finally, the pixel distance is converted into the actual physical distance using a pixel-to-millimeter conversion ratio. This measurement is crucial for controlling the quality of arc additive manufacturing.

[0116] Furthermore, the duty cycle calculation and control unit includes:

[0117] Reference control module: used to set the reference duty cycle, reference arc length, and adjustment coefficient;

[0118] Adaptive adjustment module: used to adjust the duty cycle value based on the deviation between the measured arc length (or actual arc length) and the reference arc length and the adjustment coefficient, and control the adjustment range of the duty cycle within a second preset range; in this embodiment, this step uses empirical formulas, or PID fuzzy logic control algorithms, deep learning algorithms, or reinforcement learning algorithms to calculate the optimal duty cycle process parameter.

[0119] Specifically, in this embodiment, the duty cycle is automatically calculated based on the detected arc length, as follows:

[0120] Input validation is the first step in the algorithm. It first checks whether the input arc length value is valid. If the arc length value is None, the function will directly return None. This is a defensive programming practice that avoids errors caused by null values ​​in subsequent calculations. This validation ensures the robustness of the algorithm, enabling it to gracefully handle exceptional situations.

[0121] The baseline parameter settings are the foundational configuration of the algorithm, defining three key parameters: a baseline duty cycle of 8, representing the standard duty cycle value under ideal conditions; a baseline distance threshold of 7.0 mm, an empirical value used to determine if the arc length is within the ideal range; and an adjustment factor of 1.0, used to control the sensitivity of the duty cycle to changes in arc length. These parameters are selected based on experience with actual arc additive manufacturing processes and can be adjusted according to specific application scenarios.

[0122] The duty cycle calculation logic operates in two cases: when the arc length exceeds a threshold, the system reduces the duty cycle to compensate for the excessive arc length. Specifically, the calculation method involves subtracting the adjustment value (the difference between the arc length and the threshold multiplied by an adjustment coefficient) from the baseline duty cycle, while ensuring the duty cycle does not fall below 6—a safety lower limit. This design prevents quality issues in arc additive manufacturing caused by an excessively low duty cycle when the arc length is too long.

[0123] When the arc length is less than a threshold, the system appropriately increases the duty cycle to compensate for the excessively short arc length. The calculation method is to add the difference between the threshold and the arc length multiplied by an adjustment factor to the baseline duty cycle, while ensuring that the duty cycle does not exceed 10, which is a safety upper limit. This design prevents instability in arc additive manufacturing caused by an excessively high duty cycle when the arc length is too short.

[0124] Finally, the function converts the calculated duty cycle value to an integer and returns it. This rounding operation ensures the usability of the output value, because in practical control systems, the duty cycle usually needs to be used in integer form. The entire algorithm, through this dynamic adjustment, allows the duty cycle to automatically adjust according to the actual arc length, thereby maintaining the stability of the arc additive manufacturing process.

[0125] Key features of duty cycle calculation and control:

[0126] Reference control: Set the reference duty cycle and reference distance threshold;

[0127] Adaptive adjustment: The duty cycle is adjusted based on the deviation between the arc length and the reference threshold;

[0128] Range restrictions: Ensure the duty cycle is within a reasonable range (6%–10%);

[0129] Smoothing: The calculated duty cycle is smoothed to avoid frequent fluctuations.

[0130] Furthermore, the communication unit includes:

[0131] Multi-channel data aggregation module: used to encapsulate the duty cycle and other parameter adjustment data of each printhead channel into a JSON format message and publish it to the waam / expected channel of MQTT;

[0132] Heartbeat mechanism module: Used to periodically send status information to the waam / heartbeat topic of MQTT;

[0133] Will message configuration module: used to automatically publish offline notifications when an abnormal disconnection occurs.

[0134] Specifically, this embodiment uses the MQTT protocol for data communication, implemented through the MQTTClient class:

[0135] Client initialization and connection management form the foundation of the system's architecture. The system creates three independent MQTT client instances: a subscriber, a publisher, and a heartbeat, each with a unique client ID. The system uses thread locks and event mechanisms to manage connection states, ensuring that all clients are correctly connected before commencing operations. Additionally, the system implements a will message to automatically publish an offline status when a client unexpectedly disconnects, improving system reliability.

[0136] The message subscription and processing mechanism is one of the core functions of the system. Subscriber clients listen for all command messages under the "waam / command / #" topic. When a message is received, the system parses the JSON-formatted payload data and executes the corresponding operation based on the command type (e.g., setDuty). The system uses thread locks to protect shared resources (such as active channel information) to ensure data consistency and thread safety. For each command, the system sends a feedback message containing the operation status and timestamp, implementing a complete command-feedback loop.

[0137] The heartbeat mechanism module is a crucial component of the system. It sends a heartbeat message every 2 seconds via a dedicated thread. The heartbeat message contains a counter, a timestamp, and the current status information of all channels, which is published to the "waam / heartbeat" topic. The heartbeat mechanism not only monitors the system status but also synchronizes channel information, ensuring all clients are promptly informed of the system's latest status. The system also implements error handling and retry mechanisms, retries after a short period if a heartbeat transmission fails.

[0138] The status management and feedback system is responsible for maintaining and updating the system's operational status. The system uses a dictionary structure to store information about active channels, including duty cycles and last update times. When a new command is received, the system updates the corresponding channel information and notifies other clients via feedback messages. The system also implements a manual control mode, allowing users to switch control methods by setting flags. All status changes are logged for easy problem tracking and system monitoring.

[0139] The graceful shutdown mechanism in the will message configuration module ensures that the system can safely stop all services. When the system needs to shut down, it first publishes an offline status message, then waits for the message to finish sending, and finally disconnects all client connections in sequence. The system uses flags to control thread termination, ensuring that all threads exit correctly. This design ensures that the system does not lose data during shutdown and can properly clean up resources.

[0140] Parameter modification module: This module is used to subscribe to the MQTT waam / expected channel of the arc additive manufacturing heat source equipment control computer, obtain real-time calculated information such as duty cycle, and send it to the arc additive manufacturing heat source equipment via the PLC using the EtherCAT communication protocol to modify the equipment's parameters. Its specific processing includes:

[0141] The control computer of the arc additive manufacturing heat source equipment sends task status information to a designated MQTT channel within the system, including information such as print start, preset process end, and print end. The feedback control system subscribes to this channel. Upon receiving print start information, it activates the feedback control system. When it receives print start or preset process end information (if there is no preset process, it checks print start; otherwise, it checks preset process end; the preset process refers to a series of thermal energy parameter changes preset based on experience during the arc initiation stage), it begins to detect parameters such as arc length and calculate duty cycle, and sends them to the control computer of the arc additive manufacturing heat source equipment through another channel. Upon receiving print end information, it stops the feedback control system, but the program does not shut down. Instead, it continues to listen to information on the corresponding channel. After the next print start, the feedback control system is restarted, thereby achieving full automation of the entire manufacturing process.

[0142] The method for calculating the duty cycle is as follows:

[0143] Based on the arc length obtained from visual perception (calculated by locating the tungsten electrode tip and the top of the molten pool using contour detection), the duty cycle is obtained using the following formula:

[0144] D(t)=D base +K p ·e(t),

[0145] e(t) = L ref -L(t),

[0146] Because there are maximum and minimum value restrictions, the unified expression is:

[0147]

[0148] In the formula, D(t) is the duty cycle control rate, D base K is the base duty cycle. p To adjust the parameters, e(t) is the arc length error, and L... ref Let L(t) be the reference arc length, L be the actual measured arc length, and D be the expected arc length. min For the minimum duty cycle, D max This represents the maximum duty cycle.

[0149] Assume the measured arc length is 7.2 mm;

[0150] Calculated deviation: 7.2 - 6.5 = 0.7 mm;

[0151] Calculate the adjustment amount: 0.7 × 2.0 = 1.4%;

[0152] Application control: Since the arc length is greater than the reference value, the duty cycle is reduced (reducing heat energy, resulting in lower heat input to the molten pool, faster cooling and solidification, and faster deposition of the deposited layer, which means it is easier to grow taller, thus reducing the arc length).

[0153] Key features of MQTT communication:

[0154] Publish / Subscribe Pattern: Data transmission is implemented using the publish / subscribe mechanism of MQTT;

[0155] Multi-channel data: Supports parallel transmission of data from multiple additive manufacturing printheads;

[0156] Heartbeat mechanism: Ensures the reliability of communication connections through heartbeat messages;

[0157] Will message: Send a will message when the client disconnects abnormally to ensure that the system status is known;

[0158] QoS guarantee: Use QoS=1 to ensure that the message is received at least once.

[0159] Furthermore, the multi-channel collaborative working subsystem includes:

[0160] Independent channel processor: used to configure a dedicated image processing unit and adaptive parameter adjustment unit for each additive manufacturing printhead;

[0161] Data synchronization and status management module: used to realize multi-channel image synchronous transmission through edge computing nodes; in this embodiment, this module has multiple Jetson edge computing nodes, each node controls 4 cameras and pushes images to the industrial computer, for a total of 20 cameras. Parameter control, image acquisition, and image pushing are performed by 5 Jetson edge computing nodes. Image pushing uses a socket interface via TCP / IP protocol. After receiving the images, the industrial computer performs the aforementioned image processing, arc length measurement, duty cycle calculation, MQTT message sending, etc.

[0162] Dynamic configuration loading module: used to update channel parameters and control strategies in real time based on configuration files;

[0163] Multi-channel collaborative module: used to simultaneously process and extract arc length, calculate duty cycle parameter adjustment data, and send duty cycle parameter adjustment data using several channels, while changing the duty cycle parameters of several arc additive manufacturing heat source devices. Each channel corresponds to one printhead.

[0164] Specifically, the system disclosed in this embodiment supports multi-channel collaborative operation, as follows:

[0165] Channel processor management is a fundamental part of the system's architecture. The system creates independent processor instances for each channel, and these processors are responsible for processing the image data for their respective channels. Each processor has its own configuration parameters, including binarization thresholds and morphological operation kernel size. The system uses a dictionary structure to store all channel processors, with the channel ID as the key, for easy access and management. This design ensures that each channel can operate independently while maintaining the overall system's coordination. Image processing and data acquisition are the core functions of the system. Each channel processor independently processes its corresponding video frame, calculating the arc length using image processing algorithms (such as edge detection and binarization), and calculating an appropriate duty cycle based on the arc length. The processing includes steps such as molten pool image preprocessing, feature extraction, and parameter calculation. The system updates the processing results for each channel in real time, including arc length and duty cycle; this data is used for subsequent control decisions.

[0166] Data synchronization and state management are crucial for ensuring multi-channel collaborative operation. The system uses thread locks to protect shared resources and ensure data consistency. The status information of each channel (including duty cycle, arc length, etc.) is updated in real time to the local data structure and MQTT client.

[0167] The system uses a heartbeat mechanism to periodically synchronize the status of all channels, ensuring that information between channels remains synchronized. This design allows the system to monitor and adjust the working status of each channel in real time.

[0168] The control strategy is the decision-making part of the system. The system dynamically calculates the duty cycle based on the arc length of each channel and uses a threshold-based control strategy: when the arc length is greater than the threshold, the duty cycle is decreased; when it is less than the threshold, the duty cycle is increased. This control strategy ensures the stability of the arc additive manufacturing process.

[0169] The system also supports a manual control mode, allowing operators to directly set the channel duty cycle, providing a more flexible control method. The user interface and monitoring system offer an intuitive operation and monitoring interface.

[0170] The system displays the real-time status of each channel through a graphical interface, including duty cycle, arc length, and online status. The interface updates periodically to reflect the latest processing results. The system also provides detailed logging capabilities, recording all important status changes and operational events to facilitate problem diagnosis and system optimization. This design allows operators to easily monitor and control the entire system.

[0171] Key features of multi-channel collaborative operation:

[0172] Channel isolation: Each additive manufacturing printhead has its own independent processor and parameters;

[0173] Flexible configuration: Multiple channel parameters can be defined through configuration files, supporting dynamic configuration;

[0174] Parallel processing: Parallel processing of molten pool images from multiple channels;

[0175] Data aggregation: Aggregating data from multiple channels and publishing it in a unified manner;

[0176] Sequence execution: Supports multi-channel operations performed according to predefined sequences.

[0177] Furthermore, the system also includes a user interface unit, which is used to display in real time the original molten pool image, processing results, arc length measurement value, and duty cycle and other parameter adjustment values ​​of each channel.

[0178] Specifically, to facilitate operator monitoring and troubleshooting, the system provides one or more user interfaces (typically on the main control computer or remote monitoring terminal). This interface displays in real-time raw camera images from each printhead, as well as intermediate processed images (such as binarized images and contour detection images). Simultaneously, key quantitative data, such as the real-time arc length detected by each printhead, the calculated target duty cycle and filament feed speed, and the actual feedback on equipment status, are clearly displayed in numerical, graph, or dashboard formats. Furthermore, the system displays overall operational status information, alarm information, and log records, providing operators with comprehensive process insights and fault diagnosis support.

[0179] Furthermore, in this embodiment, each printhead is equipped with a near-infrared camera using a 940±20nm filter, connected to an edge computing node. The entire system includes multiple edge computing nodes, with several cameras connected to each edge computing node via USB. The edge computing node is responsible for controlling the parameters of the connected cameras and, after preliminary processing of the acquired molten pool images, sending them to the main control computer via the RSTP communication protocol. The main control computer receives molten pool image data from all edge computing nodes, performs subsequent molten pool image processing and arc length extraction, and calculates the required duty cycle and filament feed speed for each printhead according to the process parameter calculation method.

[0180] The system software adopts a modular architecture, mainly including:

[0181] Image acquisition unit: responsible for acquiring images of the molten pool during the arc additive manufacturing process using camera equipment;

[0182] RTPCamera class: Acquires remote video streams via the RTP protocol;

[0183] The CameraManager class manages multiple camera devices.

[0184] Image processing unit: responsible for analyzing the molten pool image and detecting the arc length;

[0185] The ArcLengthDetector class implements an arc length detection algorithm.

[0186] The ArcLengthProcessorAdapter class provides an interface for system integration.

[0187] Adaptive parameter adjustment unit: responsible for adjusting processing parameters according to the characteristics of the molten pool image;

[0188] The AdaptiveParameters class implements an algorithm for adaptive parameter adjustment.

[0189] Communication unit: responsible for internal and external communication within the system;

[0190] The MQTTClient class implements data communication based on the MQTT protocol.

[0191] Control unit: Responsible for overall system control and coordination;

[0192] WAAMSystem class: The main control class of the system, which coordinates the work of various modules;

[0193] The SequenceProcessor class: implements the processing of sequence instructions;

[0194] User interface unit: responsible for visualizing the processing results;

[0195] The CameraDisplay class: Implements multi-channel video display;

[0196] The DisplayPanel class: implements parameter and status display;

[0197] The system configures parameters through a configuration file (config.json), which mainly includes:

[0198] Cameras: Camera device configuration, including ID, name, IP address, port, etc.;

[0199] channels: Process channel configuration, including ID, pixel to millimeter ratio, distance threshold, etc.;

[0200] mqtt: MQTT communication configuration, including server address, port, topic prefix, etc.

[0201] The system supports executing predefined operation sequences via a sequence file (sequence.json);

[0202] The main characteristics of sequence control include:

[0203] Timing control: Controlling the execution timing of instructions through delay time;

[0204] Multi-channel synchronization: Supports simultaneous control of multiple channels;

[0205] Dynamic parameters: Supports dynamic adjustment of parameters such as duty cycle;

[0206] Manual / Automatic Switching: Supports switching between automatic detection and manual control.

[0207] This system achieves high-precision detection of arc length through computer vision technology; it has low processing latency and can respond in real time to changes in the arc additive manufacturing process; it adapts to different arc additive manufacturing environments through adaptive parameter adjustment; its modular design makes the system easy to expand and upgrade; and it supports the parallel operation of multiple additive manufacturing printheads, enabling remote monitoring and control via the MQTT protocol.

[0208] Furthermore, the system also includes a recording unit, which is used to save the original raw image and the processed image as a video, and to save the measured arc length and the calculated duty cycle as a table file.

[0209] The application scenarios for this system include:

[0210] WAAM Additive Manufacturing: Parameter control applied to the arc additive manufacturing process;

[0211] Precision arc additive manufacturing: Applied to precision arc additive manufacturing processes with high precision requirements;

[0212] Automated Arc Additive Manufacturing: Intelligent control applied to arc additive manufacturing robots;

[0213] Quality monitoring in arc additive manufacturing: applied to quality monitoring and early warning in the arc additive manufacturing process;

[0214] Arc additive manufacturing parameter optimization: applied to the automatic optimization of process parameters in arc additive manufacturing.

[0215] Furthermore, the system provided in this embodiment can be implemented on different hardware platforms, including:

[0216] PC platform: The system runs on a high-performance computer.

[0217] Embedded platform: The system runs using embedded devices such as Raspberry Pi, Jetson Nano, etc.

[0218] Cloud-edge collaboration: Molten pool image acquisition and preprocessing are performed on edge devices, while complex calculations are completed in the cloud.

[0219] Furthermore, in addition to the MQTT protocol, the system can also use other communication protocols, such as:

[0220] WebSocket: Used for web-based application scenarios;

[0221] OPC UA: For industrial automation environments;

[0222] DDS: Used for scenarios with higher real-time requirements;

[0223] Furthermore, the arc length detection algorithm in this embodiment can have several variations:

[0224] Deep learning methods: using deep learning models such as CNN and YOLO for object detection;

[0225] 3D reconstruction method: Measurement after 3D reconstruction using binocular or structured light technology;

[0226] Infrared-assisted method: Combining infrared imaging to improve detection accuracy in special environments.

[0227] Furthermore, the duty cycle control strategy of this embodiment can be modified in various ways:

[0228] PID control: Using a PID controller to achieve more precise duty cycle control;

[0229] Neural network control: using neural networks to learn the optimal control strategy;

[0230] Fuzzy control: Using fuzzy logic to handle uncertainties.

[0231] This embodiment also provides a method for adaptive detection and intelligent feedback control of arc length in multi-arc parallel additive manufacturing, including:

[0232] Real-time acquisition of molten pool images in the arc additive manufacturing area, followed by grayscale conversion, contrast enhancement, and binarization processing;

[0233] Dynamically adjust the binarization threshold and morphological kernel parameters based on the statistical characteristics of the molten pool image;

[0234] The tungsten electrode tip and the top of the molten pool are located by contour detection, and the arc length is calculated.

[0235] Based on the arc length deviation, the duty cycle is adjusted according to a preset ratio;

[0236] The adjusted duty cycle is sent to the heat source control computer via the MQTT protocol, and then sent to the arc additive manufacturing heat source equipment via the EtherCAT communication protocol using the PLC, forming a closed-loop control.

[0237] Further, the morphological kernel parameters are adjusted, including:

[0238] The noise level of the molten pool image is evaluated using the Laplacian operator, and the morphological kernel size is dynamically adjusted based on the noise level. The parameter adjustment history is recorded for offline optimization and fault diagnosis.

[0239] Further, calculating the arc length includes:

[0240] In the base period image, splash interference is removed by opening operation, and non-detection areas are masked by masking.

[0241] The maximum area contour is identified as a suspected molten pool region using contour detection, molten pool image segmentation, or connected component analysis algorithms. The tungsten electrode position is automatically detected using algorithms such as Hough transform. If the tungsten electrode is incorrectly located, the tungsten electrode is manually selected to cover the area and the tungsten electrode position is automatically detected.

[0242] The image binarization algorithm converts the processed grayscale image into a black and white image. The pixels in the suspected molten pool area are white, and the square area enclosed by the coordinates of the four outermost vertices in the suspected molten pool area is taken as the molten pool scanning area.

[0243] A line-by-line scanning analysis is performed within the molten pool scanning area to calculate the proportion of the suspected molten pool area in each row of pixels. Based on the morphological differences between the filament and the molten pool, a line-by-line analysis is performed from top to bottom. The critical position where the proportion of white pixels in the first row exceeds the threshold is determined by the dynamic threshold method. The critical position is used as the boundary baseline between the filament and the molten pool. The morphological characteristics of the filament are elongated, while the morphological characteristics of the molten pool are diffused.

[0244] The vertical distance from the tip of the tungsten electrode to the wire-molten pool boundary baseline is calculated to obtain the arc length, and the arc length is smoothed.

[0245] Specifically, such as Figure 2 First, upon system startup, a pre-defined configuration file (e.g., config.json) is loaded. This file contains detailed information on all hardware devices (such as camera parameters for each printhead, edge computing node information, master computer address, heat source control computer configuration, MQTT server parameters, etc.) and process parameters (such as target arc length for each channel, reference duty cycle, reference wire feed speed, adjustment coefficients Kd and Kf, and upper and lower limits for each parameter, etc.). Then, the system instantiates various functional modules based on the configuration file, including the image acquisition and forwarding module on the edge computing node, and the image processing module, arc length detection module, process parameter adaptive adjustment module, process parameter calculation module, and communication module on the master computer. Finally, the internal processing parameters and status variables required by each module are initialized to ensure the system enters a standby state, preparing for subsequent real-time control.

[0246] Multi-channel parallel molten pool image acquisition and transmission stage: For each active printhead, its equipped near-infrared camera captures real-time images of the molten pool in the arc additive manufacturing area. These raw molten pool image data are first sent to the connected edge computing nodes. The edge computing nodes perform preliminary processing on the molten pool images (such as noise reduction and format conversion), and then transmit the video stream to the main control computer in real-time with low latency via the RTP (Real-Time Streaming Protocol). This process is performed independently and in parallel for each printhead, ensuring the timeliness and integrity of the molten pool image data. The image processing module of the main control computer is responsible for receiving and buffering image frames from all edge computing nodes.

[0247] Centralized Arc Length Detection Stage by the Main Control Computer: The main control computer independently executes a high-precision arc length detection algorithm for each printhead image sequence received from each edge computing node. This process typically includes: first, preprocessing the molten pool image, such as converting the color molten pool image to a grayscale image, applying algorithms like CLAHE to enhance image contrast, and performing binarization to separate the arc and molten pool areas from the background. Next, the system performs peak state detection to distinguish between the peak and base periods in the arc additive manufacturing process, and processes the molten pool image accordingly. Subsequently, morphological operations such as opening operations are used to remove noise interference such as arc additive manufacturing spatter, and masking may be combined to shield specific areas. Afterward, the contour of the arc additive manufacturing molten pool is identified and located through contour detection and filament processing algorithms. Simultaneously, the tip position of the tungsten filament electrode and the filament-molten pool boundary baseline are precisely located. Finally, by calculating the vertical distance from the tungsten electrode tip to the filament-molten pool boundary baseline and according to a preset pixel-to-millimeter conversion ratio, the current precise physical arc length value of the printhead is obtained.

[0248] Adaptive Adjustment Stage of Process Parameters: To improve the system's robustness to changes in different materials, arc additive manufacturing conditions, and environment, the adaptive adjustment module within the main control computer operates continuously. This module dynamically adjusts key parameters in the molten pool image processing flow by analyzing the statistical characteristics of the processed molten pool image (such as average brightness, standard deviation, and noise level). For example, it adaptively updates the threshold for binarization based on the overall brightness and contrast of the molten pool image; and adjusts the kernel size and shape of morphological operations (such as opening and closing operations) based on identified noise or spatter. The adjustment history of all parameters is recorded, not only for current real-time optimization but also to provide data support for subsequent algorithm improvements and offline analysis.

[0249] Core process parameter calculation stage (coordinated adjustment of duty cycle and wire feed speed): After obtaining the precise real-time arc length of each printhead, the process parameter calculation module of the main control computer calculates the deviation (Error) based on the preset target arc length (ArcLengthtarget) and the currently measured arc length (ArcLengthmeasured). Based on this deviation, the system uses a specific control strategy (such as the proportional control formula detailed above) to simultaneously calculate the duty cycle (DutyCyclenew) and wire feed speed (WireFeedSpeednew) that need to be adjusted. During the calculation process, the system applies a smoothing algorithm (such as moving average) to avoid frequent and drastic parameter adjustments due to small fluctuations, thereby ensuring the stability of control. At the same time, the calculated duty cycle and wire feed speed are strictly limited within the upper and lower limits of the preset reasonable process window to prevent extreme parameters from damaging the equipment or causing process failure.

[0250] Multi-level data communication and command issuance stage: The main control computer encapsulates the new duty cycle and wire feed speed calculated for each printhead into structured messages (e.g., JSON format, containing printhead ID, target parameter values, etc.). These messages are published to designated MQTT topics via the MQTT communication protocol, and are subscribed to and received by the heat source control computer corresponding to each printhead. After receiving the new process parameter command, the heat source control computer, through its internal PLC (Programmable Logic Controller), uses the high-speed, deterministic EtherCAT industrial Ethernet protocol to precisely send the final control signal to the corresponding heat source controller (adjusting the arc additive manufacturing power output, thereby changing the duty cycle) and wire feed mechanism controller (adjusting the wire feed motor speed, thereby changing the wire feed speed). In addition, the system also has the ability to receive external control commands through MQTT or other predetermined interfaces, such as manual intervention commands from the upper-level monitoring system or operators.

[0251] Real-time results display and system status monitoring phase: To facilitate operator monitoring and debugging, the system provides one or more user interfaces (usually on the main control computer or remote monitoring terminal). This interface displays raw camera images from each printhead in real time, as well as intermediate image results (such as binarized images, contour detection images, etc.). Simultaneously, key quantitative data, such as the real-time arc length detected by each printhead, the calculated target duty cycle and filament feed speed, and the actual feedback equipment status, are clearly displayed in numerical, graph, or dashboard formats. Furthermore, the system displays overall operating status information, alarm information, and log records, providing operators with comprehensive process insights and fault diagnosis basis.

[0252] This embodiment achieves high-precision real-time detection of arc length, providing a foundation for precise control; it introduces an adaptive parameter adjustment mechanism, improving the robustness of the system in different additive manufacturing environments for large components; it effectively stabilizes the arc and improves the molten pool morphology and forming quality by intelligently adjusting the duty cycle through arc length feedback; it reduces reliance on operator skills and improves automation level and production efficiency.

[0253] The above are merely preferred embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A multi-arc parallel additive manufacturing arc length adaptive detection and intelligent feedback control system, characterized in that, include: Image acquisition unit: used to acquire images of the molten pool during the arc additive manufacturing process; Adaptive parameter adjustment unit: used to dynamically adjust the parameters of the image processing unit according to the statistical characteristics of the molten pool image; Image processing unit: used to process the acquired molten pool image and detect the arc length; Duty cycle calculation and control unit: used to calculate the duty cycle adjustment amount based on the detected deviation between the arc length and the preset target value, and generate control commands; Communication unit: used to send the control commands to the arc additive manufacturing heat source equipment using the MQTT protocol, and to receive external commands; Control unit: Used to coordinate the operation of each unit and support data synchronization and status management of multi-channel collaborative subsystems; The adaptive parameter adjustment unit includes: Image statistical analysis module: used to calculate the average brightness and standard deviation of the molten pool image; Binarization threshold dynamic adjustment module: used to update the binarization threshold based on the average brightness and standard deviation, and limit it within a first preset range; Morphological kernel adaptive module: used to adjust the size of the denoising kernel according to the noise level, wherein the size of the denoising kernel is positively correlated with the noise intensity of the melt pool image; Parameter history record module: used to record historical adjustment data and track parameter change trends; The acquired molten pool images are processed, including: The original acquired color molten pool image is converted into a grayscale image, and a contrast-limited adaptive histogram equalization method is applied to enhance the contrast. Based on peak state detection, the peak period and the base period of the electric arc additive manufacturing process are distinguished, and the arc length is detected during the base period; Splash interference is removed by opening morphological operations, and non-detection areas are shielded by masking. The algorithm uses contour detection, image segmentation, or connected component analysis to identify the largest area contour as the suspected molten pool region and automatically detects the tungsten electrode position. If the tungsten electrode is incorrectly positioned, the algorithm manually selects the tungsten electrode to cover the automatically detected tungsten electrode position. The image binarization algorithm converts the processed grayscale image into a black and white image. The pixels in the suspected molten pool area are white, and the square area enclosed by the coordinates of the four outermost vertices in the suspected molten pool area is taken as the molten pool scanning area. A line-by-line scanning analysis is performed within the molten pool scanning area to calculate the proportion of the suspected molten pool area in each row of pixels. Based on the morphological differences between the filament and the molten pool, a line-by-line analysis is performed from top to bottom. The critical position where the proportion of white pixels in the first row exceeds the threshold is determined by the dynamic threshold method. The critical position is used as the boundary baseline between the filament and the molten pool. The morphological characteristics of the filament are elongated, and the morphological characteristics of the molten pool are diffused. The vertical distance from the tip of the tungsten electrode to the baseline of the wire-molten pool boundary is calculated to obtain the arc length, and the arc length is smoothed.

2. The multi-arc parallel additive manufacturing arc length adaptive detection and intelligent feedback control system according to claim 1, characterized in that, The duty cycle calculation and control unit includes: Reference control module: used to set the reference duty cycle, reference arc length, and adjustment coefficient; Adaptive adjustment module: used to adjust the duty cycle value based on the deviation between the measured arc length and the reference arc length and the adjustment coefficient, and control the adjustment range of the duty cycle within a second preset range.

3. The multi-arc parallel additive manufacturing arc length adaptive detection and intelligent feedback control system according to claim 1, characterized in that, The communication unit includes: Multi-channel data aggregation module: used to encapsulate the duty cycle parameter adjustment data of each printhead channel into a JSON format message and publish it to the waam / expected channel of MQTT; Heartbeat mechanism module: Used to periodically send status information to the waam / heartbeat topic of MQTT; Will message configuration module: used to automatically publish offline notifications when an abnormal disconnection occurs; Parameter modification module: Used to subscribe to the MQTT waam / expected channel of the arc additive manufacturing heat source equipment control computer, obtain the duty cycle calculated in real time, and send it to the arc additive manufacturing heat source equipment via PLC using the EtherCAT communication protocol to modify the parameters of the arc additive manufacturing heat source equipment.

4. The multi-arc parallel additive manufacturing arc length adaptive detection and intelligent feedback control system according to claim 1, characterized in that, The multi-channel collaborative working subsystem includes: Independent channel processor: used to configure a dedicated image processing unit and adaptive parameter adjustment unit for each additive manufacturing printhead; Data synchronization and status management module: used to achieve synchronous transmission of multiple images through edge computing nodes; Dynamic configuration loading module: used to update channel parameters and control strategies in real time based on configuration files; Multi-channel collaborative module: used to simultaneously process and extract arc length, calculate duty cycle parameter adjustment data, and send duty cycle parameter adjustment data using several channels, while changing the duty cycle parameters of several arc additive manufacturing heat source devices. Each channel corresponds to one printhead.

5. The multi-arc parallel additive manufacturing arc length adaptive detection and intelligent feedback control system according to claim 1, characterized in that, It also includes a user interface unit, which is used to display the original molten pool image, processing results, arc length measurement value and duty cycle parameter adjustment value of each channel in real time.

Images