A quality control system and method for additive manufacturing

Abstract

This invention discloses an additive manufacturing forming quality control system and method. The additive manufacturing forming quality control system includes a temperature acquisition module, a layer height acquisition module, a current acquisition module, a neural network model training and storage module, a data center storage module, and a central system forming quality parameter adjustment and identification module. This additive manufacturing forming quality control system and method converts one-dimensional data into two-dimensional image data through various data acquisition modules, studies the influence of different process parameters on the deposited cladding layer, and uses a neural network model for welding quality detection. It solves the problems of traditional control systems, such as the inability to update and iterate data, low identification accuracy, poor controllability of forming quality, and difficulty in quickly determining welding process parameters and optimizing welding quality and process parameters during additive manufacturing.

Description

Technical Field

[0001] This invention relates to the field of additive manufacturing technology, specifically to an additive manufacturing forming quality control system and method. Background Technology

[0002] Additive manufacturing is an advanced manufacturing technology that discretizes a digital 3D model into several 2D planar slices, and uses a coded program to drive a motion mechanism to stack material layer by layer until a 3D part is formed. WireArc Additive Manufacturing (WAAM) has advantages such as low cost, high efficiency, and high flexibility, making it particularly suitable for manufacturing large metal parts. However, in the arc welding process, the molten pool is large and the heat input is high. As the number of cladding layers increases, the heat dissipation conditions of the workpiece become increasingly poor, leading to a rapid increase in the solidification time of the molten pool. In actual welding, it is very difficult to control the shape of the weld bead after the molten pool solidifies, resulting in difficulty in controlling the shape of the final welded part. At the same time, during the WAAM forming process, the arc heat source undergoes a long thermal cycle, and the deposition path in actual welding is diverse and complex, with an extremely uneven temperature field distribution. Therefore, the formed part undergoes complex thermal cycles. This inevitably leads to large residual stress inside the formed part, and may even result in defects such as micro-cracks and deformations, making it difficult to control the forming quality. Therefore, studying the evolution of temperature and stress fields during WAAM forming and further optimizing related welding process parameters through algorithms is of great significance for realizing automated additive manufacturing and improving the quality of formed parts.

[0003] Patent CN105880808A provides a method for controlling the forming morphology of GMAW additive manufacturing co-directional forming, which effectively suppresses the height and dimensional differences generated at the ends of parts during the co-directional forming process, reduces defects generated during the forming process, and provides reliable technical support for the quality control of GMAW additive manufacturing forming.

[0004] Patent CN109262109A provides a feedback control device and method for the morphology of TIG arc additive manufacturing, which can realize the rapid forming of complex three-dimensional structures and significantly increase the forming efficiency.

[0005] Patent CN106956060A provides a method for actively controlling the interlayer temperature in arc additive manufacturing using electromagnetic induction heating. This method can achieve induction heating or forced cooling, thereby realizing the purpose of actively controlling the interlayer temperature and providing an effective method for actively controlling the interlayer temperature in arc additive manufacturing.

[0006] Patent CN111037052B provides a system and method for detecting and compensating the morphology of arc additive manufacturing, which solves the problem of poor measurement accuracy in existing systems for detecting and compensating the morphology of arc additive manufacturing.

[0007] Patent CN112975054A provides a device and method for controlling the surface temperature of parts during arc additive manufacturing, which can reduce interlayer waiting time and improve the mechanical properties and forming efficiency of arc additive formed parts.

[0008] Patent CN109514066A provides a device for controlling interlayer temperature based on electron beam fused wire additive manufacturing, which can achieve temperature control without waiting for heat dissipation, saving time and improving production efficiency.

[0009] Patent CN112703099A provides a method and system for additive manufacturing using closed-loop temperature control, which controls the heat dissipation rate of the cooling system by sensing signals generated by the sensing system to achieve efficient temperature control.

[0010] Patent CN115007969A provides a surface forming quality control method for CMT+P arc additive manufacturing. By controlling the process parameters of the contour forming path and the internal scanning path, as well as the welding torch posture control method, the multi-layer stacking forming process combined with interlayer temperature control achieves stable droplet transition and stable solidification of the molten pool, thereby achieving geometric dimensional accuracy and surface forming quality control.

[0011] Patent CN106180986A provides an active control method for the forming quality of arc additive manufacturing. Based on the relationship between the cladding layer process parameters, the external cooling system control parameters, the interlayer temperature, and the forming quality, an analytical model of the forming quality characterization quantity and the forming process parameters is established to ensure good surface quality and mechanical properties of arc additive manufacturing.

[0012] Patent CN110340358A provides a method for gradient control of process parameters in additive manufacturing, which solves the problem of unstable laser forming caused by the heat accumulation effect of the layer-by-layer printing process in existing laser additive manufacturing technology.

[0013] Patent CN110802300A provides a device and method for controlling forming accuracy and quality in the process of electric arc additive manufacturing. It solves the problems of large heat input and poor forming accuracy of formed parts in electric arc additive manufacturing, while ensuring forming efficiency and realizing the control of forming accuracy and quality in the process of electric arc additive manufacturing.

[0014] Patent CN112122741A provides a weld bead forming control system and its parameter optimization method in the process of electric arc additive manufacturing, which can effectively reduce the change of molten pool and improve the stability of the deposition process. Through continuous feedback adjustment, the process parameters in the welding process can be effectively optimized.

[0015] Patent CN114372725A provides a forming monitoring system and method for additive manufacturing systems based on digital twins, which realizes closed-loop feedback control of the manufacturing system, improves the forming accuracy of parts, and saves manpower and material resources for the design and manufacture of new products.

[0016] Patent CN112008198B provides a quality control system and method for aluminum alloy electric arc additive manufacturing, which overcomes the problem of difficult-to-control forming quality in aluminum alloy electric arc additive manufacturing and has the advantages of high manufacturing efficiency, low cost and stable forming quality.

[0017] Patent CN113762366A provides a method and system for predicting and controlling the forming state in additive manufacturing, which can accurately predict various quality problems in additive manufacturing and achieve precise control.

[0018] Patent CNCN110788324A provides a method for parallel control of part deformation and precision during additive manufacturing, realizing high-precision and high-performance additive manufacturing with a one-step ultra-short process. It has high processing accuracy, and the parts can be directly used, which has strong practical application value.

[0019] Patent CN106670623A provides a device for actively controlling the interlayer temperature in arc additive manufacturing, which not only improves the forming efficiency, morphological quality and mechanical properties of parts, but also has the advantages of high temperature control accuracy.

[0020] Among the descriptions of the various patents above, both closed-loop and open-loop control methods have largely solved the problems of temperature management and process parameter optimization for additive manufacturing parts; they have also proposed good solutions for monitoring and controlling the forming morphology of various complex shaped components. However, for specific arc additive manufacturing methods applied to large-scale, multi-layer, and multi-channel applications, especially those involving additive manufacturing forming control systems and methods based on iterative learning, there is currently no iterative learning-based additive manufacturing forming control system and method that can comprehensively address the complex temperature field distribution, part deformation, and optimization of process parameters.

[0021] This application addresses the challenges of achieving high forming accuracy and quality in large arc additive manufacturing parts, as well as the issues of automatically optimizing welding process parameters and realizing highly automated additive manufacturing. It provides an additive manufacturing forming quality control system and method based on iterative learning, solving the problems of traditional processes failing to meet the necessary high-precision forming requirements and guaranteeing the mechanical properties of the formed parts.

[0022] In view of this, in-depth research was conducted on the above issues, which led to the creation of this case. Summary of the Invention

[0023] The purpose of this invention is to provide an additive manufacturing forming quality control system and method to solve the problems mentioned in the background art, such as the difficulty in achieving the necessary high-precision formed parts through traditional processes and the difficulty in guaranteeing the mechanical properties of the formed parts.

[0024] To achieve the above objectives, the present invention provides the following technical solution: an additive manufacturing forming quality control system;

[0025] The additive manufacturing forming quality control system includes a temperature acquisition module, a layer height acquisition module, a current acquisition module, a neural network model training and storage module, a data center storage module, and a central system forming quality parameter control and identification module.

[0026] The temperature acquisition module collects temperature field data of the total cladding layer in the additive manufacturing process using an infrared camera. Since the temperature field data is too large, one-dimensional temperature data in the X and Y directions is extracted and further converted into two-dimensional image temperature data, which enhances the data's processability and retains more original information, thereby improving the reliability of the temperature data.

[0027] The layer height acquisition module collects layer height data of the cladding layer during the additive manufacturing process using a laser contour scanner. To improve the accuracy of the layer height data, it extracts data in the X and Y directions, sequentially numbers and converts it into two-dimensional image layer height data, which serves as the input to the neural network model to improve the recognition accuracy.

[0028] The current acquisition module collects current data during the additive manufacturing process through a Hall sensor, further converts it into two-dimensional image current data, and corresponds it one-to-one with temperature data to enhance the recognition capability of the system model. At the same time, it can optimize process parameters and provide the optimal additive manufacturing process parameters for the next additive manufacturing process.

[0029] The neural network model training and storage module extracts effective features from the data of each module and uses them as input to the neural network model. After iterative training through the input layer, hidden layer, and output layer, a well-preserved recognition model is finally saved for the central system to call for recognition.

[0030] The data center storage module stores all the data acquisition modules of the system, and also has the function of adding or removing modules, which brings convenience to system optimization and upgrading;

[0031] The central system forming quality parameter control and identification module identifies and controls the forming quality of the additive manufacturing process by calling data from various modules of the system. At the same time, it provides different optimized process parameters for each type of additive manufacturing, further improving the controllability of the forming quality of additive manufacturing.

[0032] The temperature acquisition module, layer height acquisition module, current acquisition module, neural network model training and storage module, data center storage module, and central system forming quality parameter control and identification module are all integrated into the additive manufacturing forming quality control system software; the software monitors the working conditions of the welding process in real time, corrects welding parameters, and controls the weld formation quality; while controlling the welding parameters, it records the welding parameters corresponding to the working conditions.

[0033] The data monitored in real time by the system corresponds to the real-time welding conditions. By analyzing and processing the monitored data, the actual welding conditions can be restored to the greatest extent. During the welding process, if a certain type of defect occurs in a certain area of ​​the weld, the temperature, current, and layer height data will be recorded, further forming a set of physical data corresponding to the weld quality and physical domain information. The physical data serves as the training input data for the neural network, and each newly formed physical data can serve as the training input data for the neural network to iteratively update the predictive ability of the prediction model.

[0034] The additive manufacturing forming control system possesses good compatibility and modularity, enabling it to accommodate and operate multiple modules simultaneously while maintaining high interoperability among them. Each module in this system has independent development and maintenance capabilities, with clearly defined interfaces and functions; this allows modules to be easily added, replaced, or upgraded without adversely affecting the overall system. To ensure compatibility between different modules, the system adopts standardized interfaces and communication protocols, allowing modules to run seamlessly in different hardware and operating system environments, and easily integrate with third-party modules or systems, making the system more scalable.

[0035] Preferably, the temperature acquisition module mainly records the temperature field data during the additive manufacturing process using an infrared camera, further selects temperature node data in the X and Y directions, combines them into time series data, and further converts the temperature data into two-dimensional image data to enhance the data's processability and retain more effective information, thereby improving the accuracy of predicting and identifying weld defects.

[0036] The defects generally include weld collapse, weld porosity, weld protrusion, and molten droplet spatter. Data collection is based on additive manufacturing under different welding parameters, recording the weld defect data generated in each layer to reduce the defects generated in the next additive manufacturing process. At the same time, an additive manufacturing database system is established to store temperature data in DATA_Image_Temp, laying the groundwork for the next automated additive manufacturing process.

[0037] Preferably, the layer height acquisition module mainly uses a laser contour scanner to scan the layer height data of each layer. To improve the visualization and processability of the data and increase the accuracy of recognition, the one-dimensional data is converted into two-dimensional image information for storage. Furthermore, layer height data in the horizontal and vertical directions of each cladding layer is selected as the output result. The position coordinates in the X and Y directions are subdivided according to the length and width of each cladding layer to select the corresponding layer height data. The interval t in the X and Y directions is an important basis for dividing the data points. The size of t directly affects the amount of data, thus affecting the accuracy of the data.

[0038] Preferably, the current acquisition module mainly acquires current data during the additive manufacturing process using Hall sensors and numbers the data according to the time series; selects appropriate process parameters, keeps other parameters constant, gradually increases the current value for welding, and uses Hall sensors to record the current data of each layer in real time to study the main influence of current on the additive manufacturing process.

[0039] Current is a crucial parameter affecting the additive manufacturing process and a major factor leading to excessively high temperatures. Therefore, matching the collected current information with the information collected by the temperature module, and matching the corresponding defect temperature with the current information, helps improve the accuracy of neural network model classification and recognition, and further explains the relationship between heat, current, and defects in arc additive manufacturing.

[0040] Preferably, the temperature acquisition module and the current acquisition module establish a temperature-current-defect array to further correlate the relationship between temperature, current and defects, and generate two-dimensional image information accordingly. This allows for further research on the relationship between defect formation and process parameters, while also improving the accuracy of model prediction and further improving the quality control of arc additive manufacturing.

[0041] The correspondence between the temperature data and the current data is specifically explained as follows: the current value refers to the average current change at the location where the cladding layer is deposited over a period of time, and the temperature value is the average temperature at that location after one layer of cladding layer has been deposited. Furthermore, if a corresponding defect occurs at this location, the corresponding location is labeled as loc1, loc2, loc3, etc., in sequence. Here, loc1 indicates that the defect type at this location is 1, loc2 indicates that the defect type at this location is 2, and so on, forming an independent array index to provide data support for further automated intelligent identification.

[0042] Preferably, the neural network model training and storage module mainly extracts effective image information as feature input data for the neural network system by calling the data from the temperature acquisition module and the floor height acquisition module;

[0043] The two-dimensional image is split into R, G, B, H, S, and V channels, and the pixel value of each channel is calculated sequentially. The difference between the pixel value of each channel and the corresponding pixel value of the previous frame is used to obtain the T value. The T value is compared with a preset threshold TH. If the T value is within the threshold range, the next step of image processing is performed to further extract the image's feature information as training data for the neural network model. If the T value is not within the preset threshold range, the TH threshold is adjusted according to the working conditions, and the comparison with the T value continues until it meets the preset threshold. Finally, the effective feature information is used as training data to save the current neural network model for subsequent defect classification.

[0044] Preferably, the data center storage module mainly includes a layer height acquisition module, a current acquisition module, a temperature acquisition module, a current and temperature data module, and a neural network model training and storage module. The data center storage module can store multiple modules, providing more interfaces for subsequent system upgrades and further improving the system's ability to control the quality of additive manufacturing.

[0045] Preferably, the central system forming quality parameter control and identification module mainly coordinates the data from various data acquisition modules for iterative updates, further improving the system's quality forming control capability and providing data and process references for achieving automated high-precision additive manufacturing;

[0046] The additive manufacturing forming quality control system has two modes: an iterative data update identification mode and a database-based identification mode. Each new additive manufacturing system can call each module to collect new data and train the model, updating the new data to each module to continuously reduce the system model identification error and improve the system's forming accuracy control capability.

[0047] A method for quality control in additive manufacturing involves storing physical data collected by various sensors in a system storage module, converting one-dimensional data into two-dimensional data, and labeling different types of defective welds and normal welds. Furthermore, the labeled data is used as input data for training a neural network, and the trained model is saved. In the next additive manufacturing process, the model is invoked to detect weld quality in real time and output the weld quality type and corresponding welding process parameters.

[0048] Preferably, the various sensors include a temperature sensor, a Hall sensor, and a laser profile scanner; they output temperature data, current data, and weld bead height data respectively; the current data, temperature data, and weld bead height data at each time point are treated as an array, corresponding to the current weld quality, which can retain information in the additive manufacturing process to the maximum extent and enhance the neural network model's ability to identify different types of defects.

[0049] Compared with the prior art, the beneficial effects of the present invention are: the additive manufacturing forming quality control system and method convert one-dimensional data into two-dimensional image data through various data acquisition modules, study the influence of different process parameters on the deposited cladding layer, and use a neural network model to detect the welding quality; it solves the problems of traditional control systems that cannot update and iterate data, have low recognition accuracy, poor controllability of forming quality, and difficulty in quickly determining welding process parameters and welding quality and optimizing process parameters in the additive manufacturing process. Attached Figure Description

[0050] Figure 1 This is a schematic diagram of the interface of the additive manufacturing forming control system of the present invention;

[0051] Figure 2 This is a schematic diagram of the modules included in the additive manufacturing forming control system of the present invention;

[0052] Figure 3 This is a schematic diagram of the data acquisition and processing method of the present invention;

[0053] Figure 4 This is a schematic diagram of the temperature acquisition module of the present invention;

[0054] Figure 5 This is a schematic diagram of the current acquisition module of the present invention;

[0055] Figure 6 This is a schematic diagram of the layer height acquisition module of the present invention;

[0056] Figure 7 This is a schematic diagram showing the location for selecting surface data of the cladding layer in this invention;

[0057] Figure 8This is a schematic diagram of the temperature-current-defect data correlation steps of the present invention;

[0058] Figure 9 This is a schematic diagram of the neural network model training and storage module of the present invention;

[0059] Figure 10 This is a schematic diagram of the data center storage module of the present invention;

[0060] Figure 11 This is a schematic diagram of the process of the central system forming quality parameter control and identification module of the present invention;

[0061] Figure 12 This is a schematic diagram illustrating the correlation between one-dimensional data converted to two-dimensional data and weld defects in this invention;

[0062] Figure 13 This is a schematic diagram of the neural network training model of the present invention. Detailed Implementation

[0063] 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.

[0064] Please see Figure 1-13 The present invention provides the following technical solution: an additive manufacturing forming quality control system includes a temperature acquisition module, a layer height acquisition module, a current acquisition module, a neural network model training and storage module, a data center storage module, and a central system forming quality parameter control and identification module;

[0065] The temperature acquisition module collects temperature field data of the total cladding layer in the additive manufacturing process using an infrared camera. Since the temperature field data is too large, one-dimensional temperature data in the X and Y directions is extracted and further converted into two-dimensional image temperature data, which enhances the data's processability and retains more original information, thereby improving the reliability of the temperature data.

[0066] The temperature acquisition module mainly records the temperature field data during the additive manufacturing process using an infrared camera, further selects temperature node data in the X and Y directions, combines them into time series data, and then converts the temperature data into two-dimensional image data to enhance the data's processability and retain more effective information, thereby improving the accuracy of predicting and identifying weld defects.

[0067] Defects generally include weld collapse, weld porosity, weld protrusion, and molten droplet spatter; specific temperature data acquisition modules include... Figure 4As shown, data acquisition is based on additive manufacturing under different welding parameters, recording weld defect data for each layer to reduce defects in subsequent additive manufacturing processes. Simultaneously, an additive manufacturing database system is established, storing temperature data in DATA_Image_Temp to prepare for future automated additive manufacturing. A schematic diagram of the selected locations for cladding layer surface data is shown below. Figure 7 As shown.

[0068] The layer height acquisition module collects layer height data of the cladding layer during the additive manufacturing process using a laser contour scanner. To improve the accuracy of the layer height data, the data in the X and Y directions is extracted, sequentially numbered, and converted into two-dimensional image layer height data, which serves as the input for the neural network model to improve the recognition accuracy.

[0069] The layer height acquisition module primarily uses a laser contour scanner to scan the layer height data of each layer. To improve data visualization, processability, and recognition accuracy, the one-dimensional data is converted into two-dimensional image information for storage. Further, layer height data is selected from the horizontal and vertical directions of each cladding layer as the output. The X and Y coordinates are subdivided according to the length and width of each cladding layer to select the corresponding layer height data. The interval t in the X and Y directions is a crucial basis for dividing the data points; the size of t directly affects the amount of data, thus affecting the data accuracy. Figure 7 As shown. In this invention, t is selected based on one-hundredth of the length in the X and Y directions as the data sampling interval.

[0070] The current acquisition module collects current data during the additive manufacturing process through Hall sensors, further converts it into two-dimensional image current data, and corresponds it one-to-one with temperature data to enhance the recognition ability of the system model. At the same time, it can optimize process parameters and provide the optimal additive manufacturing process parameters for the next additive manufacturing process.

[0071] The current acquisition module mainly collects current data during the additive manufacturing process using Hall sensors and numbers the data in time series. By selecting appropriate process parameters and keeping other parameters constant, the current value is gradually increased during welding. The Hall sensors are used to record the current data of each layer in real time to study the main impact of current on the additive manufacturing process.

[0072] Current is a crucial parameter affecting the additive manufacturing process and a major factor leading to excessively high temperatures. Therefore, matching the collected current information with the information collected by the temperature module, and matching the corresponding defect temperature with the current information, helps improve the accuracy of neural network model classification and recognition, and further explains the relationship between heat, current, and defects in arc additive manufacturing. The specific process for current data acquisition and conversion is as follows: Figure 5 As shown.

[0073] To further correlate the temperature-current-defect relationship, the temperature acquisition module and the current acquisition module established a temperature-current-defect array and generated two-dimensional image information based on it. This allowed for further research into the relationship between defect formation and process parameters, while also improving the accuracy of model predictions and further enhancing the quality control of arc additive manufacturing.

[0074] The correspondence between temperature and current data is explained as follows: the current value refers to the average current change at the location where the cladding layer is deposited over a specific period of time, and the temperature value is the average temperature at that location after one layer of cladding layer has been deposited. Furthermore, if a defect occurs at this location, the corresponding location is labeled as loc1, loc2, loc3, etc., sequentially. Here, loc1 indicates defect type 1, loc2 indicates defect type 2, and so on, forming an independent array index to provide data support for further automated intelligent identification. A detailed diagram illustrating the temperature-current-defect data correlation is shown below. Figure 8 As shown.

[0075] The neural network model training and storage module extracts effective features from the data of each module and uses them as input to the neural network model. After iterative training through the input layer, hidden layer, and output layer, a well-preserved recognition model is finally saved for the central system to call for recognition.

[0076] The neural network model training and storage module mainly extracts effective image information as feature input data for the neural network system by calling data from the temperature acquisition module and the floor height acquisition module.

[0077] The 2D image is split into R, G, B, H, S, and V channels, and the pixel value of each channel is calculated sequentially. The difference between the pixel value of each channel and the corresponding pixel value of the previous frame is used to obtain the T value. The T value is compared with a preset threshold TH. If the T value is within the threshold range, the next step of image processing is performed to further extract the image's feature information as training data for the neural network model. If the T value is not within the preset threshold range, the TH threshold is adjusted according to the working conditions, and the comparison with the T value continues until it meets the preset threshold. Finally, the effective feature information is used as training data, and the current neural network model is saved for subsequent defect classification. The specific neural network model training and storage module is as follows: Figure 9 As shown.

[0078] The data center storage module stores all the data acquisition modules of the system, and also has the function of adding or removing modules, which brings convenience to system optimization and upgrade;

[0079] The data center storage module mainly includes a layer height acquisition module, a current acquisition module, a temperature acquisition module, a current and temperature data module, and a neural network model training and storage module. This module can store multiple modules, providing more interfaces for subsequent system upgrades and further improving the system's ability to control additive manufacturing quality. (The data center storage module is described as follows...) Figure 10 As shown.

[0080] The central system forming quality parameter control and identification module identifies and controls the forming quality of the additive manufacturing process by calling data from various modules of the system. At the same time, it provides different optimized process parameters for each type of additive manufacturing, further improving the controllability of the forming quality of additive manufacturing.

[0081] The central system's forming quality parameter control and identification module primarily coordinates the iterative updates of data from various data acquisition modules, further enhancing the system's quality forming control capabilities and providing data and process references for achieving automated, high-precision additive manufacturing. The specific implementation process is as follows: Figure 11 As shown.

[0082] There are two modes in the additive manufacturing forming quality control system: one is the iterative update data recognition mode, and the other is the database-based recognition mode. Each time a new additive manufacturing system is developed, it can call each module to collect new data and train the model, and update each module with new data to continuously reduce the system model recognition error and improve the system's forming accuracy control capability.

[0083] Software interface design for additive manufacturing forming quality control system, such as Figure 1 As shown, the integrated software can efficiently call the corresponding modules to work, achieving an integrated effect of additive manufacturing control. The temperature acquisition module, layer height acquisition module, current acquisition module, neural network model training and storage module, data center storage module, and central system forming quality parameter control and identification module are all integrated into the additive manufacturing forming quality control system software. The software monitors the welding process in real time, corrects welding parameters, and controls weld formation quality. While controlling welding parameters, it also records the welding parameters corresponding to the working conditions.

[0084] The system monitors data in real time, which corresponds to the real-time welding conditions. By analyzing and processing the monitored data, the actual welding conditions can be reproduced to the greatest extent possible. During the welding process, if a certain type of defect occurs in a certain area of ​​the weld, the temperature, current, and layer height data will be recorded, further forming a set of physical data corresponding to the weld quality and physical domain information. The physical data serves as the training input data for the neural network, and each newly formed physical data can also serve as the training input data for the neural network to iteratively update the predictive ability of the prediction model.

[0085] The additive manufacturing forming control system has good compatibility and modularity, enabling it to accommodate and operate multiple modules simultaneously, such as... Figure 2 As shown, these modules maintain a high degree of interoperability. Each module in this system has the capability for independent development and maintenance, with clearly defined interfaces and functions; this allows modules to be easily added, replaced, or upgraded without adversely affecting the entire system. To ensure compatibility between different modules, the system adopts standardized interfaces and communication protocols, allowing modules to run seamlessly in different hardware and operating system environments, and to easily integrate with third-party modules or systems, making the system more scalable.

[0086] The additive manufacturing forming quality control method stores physical data collected by various sensors in the system storage module, further converts one-dimensional data into two-dimensional data, and annotates different types of defective weld beads and normal weld beads. Further, the annotated data is used as input data for training a neural network, and the trained model is saved. In the next additive manufacturing process, the model is called to detect weld bead quality in real time and outputs the weld bead quality type and corresponding welding process parameters, such as... Figure 3 As shown.

[0087] Various sensors, including temperature sensors, Hall effect sensors, and laser profile scanners, output temperature data, current data, and weld bead height data, respectively. The current data, temperature data, and weld bead height data at each time point are compiled into an array, corresponding to the current weld quality. This maximizes the retention of information during the additive manufacturing process and enhances the neural network model's ability to identify different types of defects.

[0088] The contents not described in detail in this specification are prior art known to those skilled in the art. Although embodiments of the present invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A quality control system for additive manufacturing, characterized in that: The additive manufacturing forming quality control system includes a temperature acquisition module, a layer height acquisition module, a current acquisition module, a neural network model training and storage module, a data center storage module, and a central system forming quality parameter control and identification module. The temperature acquisition module collects temperature field data of the total cladding layer in the additive manufacturing process using an infrared camera. Since the temperature field data is too large, one-dimensional temperature data in the X and Y directions is extracted and further converted into two-dimensional image temperature data. The layer height acquisition module acquires layer height data of the cladding layer during the additive manufacturing process using a laser contour scanner. To improve the accuracy of the layer height data, it extracts data in the X and Y directions, sequentially numbers them, and converts them into two-dimensional image layer height data. The current acquisition module collects current data during the additive manufacturing process through a Hall sensor, further converts it into two-dimensional image current data, and corresponds it one-to-one with temperature data to enhance the recognition capability of the system model. At the same time, it can optimize process parameters and provide the optimal additive manufacturing process parameters for the next additive manufacturing process. The neural network model training and storage module extracts effective features from the data of each module and uses them as input to the neural network model. After iterative training through the input layer, hidden layer, and output layer, a well-preserved recognition model is finally saved for the central system to call for recognition. The data center storage module stores all the data acquisition modules of the system, and also has the function of adding or removing modules, which brings convenience to system optimization and upgrading; The central system forming quality parameter control and identification module identifies and controls the forming quality of the additive manufacturing process by calling data from various modules of the system. At the same time, it provides different optimized process parameters for each type of additive manufacturing, further improving the controllability of the forming quality of additive manufacturing. The temperature acquisition module, layer height acquisition module, current acquisition module, neural network model training and storage module, data center storage module, and central system forming quality parameter control and identification module are all integrated into the additive manufacturing forming quality control system software; the software monitors the working conditions of the welding process in real time, corrects welding parameters, and controls the weld formation quality; while controlling the welding parameters, it records the welding parameters corresponding to the working conditions. The data monitored by the system in real time corresponds to the real-time welding conditions; by analyzing and processing the monitored data, the actual welding conditions can be reconstructed. During the welding process, if a certain type of defect appears in a certain area of ​​the weld, the temperature, current, and layer height data will be recorded, further forming a set of physical data corresponding to the weld quality and physical domain information. The physical data serves as the training input data for the neural network, and each newly formed physical data can also serve as the training input data for the neural network to iteratively update the predictive ability of the prediction model. The additive manufacturing quality control system has good compatibility and modularity, enabling it to accommodate and operate multiple modules simultaneously. To ensure compatibility between different modules, the system adopts standardized interfaces and communication protocols, allowing modules to run seamlessly in different hardware and operating system environments, and can be easily integrated with third-party modules or systems, making the system more scalable. The temperature acquisition module records temperature field data during the additive manufacturing process using an infrared camera, further selects temperature node data in the X and Y directions, combines them into time series data, and further converts the temperature data into two-dimensional image data to enhance the data's processability and retain more effective information, thereby improving the accuracy of predicting and identifying weld defects. The defects include weld collapse, weld porosity, weld protrusion, and molten droplet spatter. Data collection is based on additive manufacturing under different welding parameters, recording the weld defect data generated in each layer to reduce the defects generated in the next additive manufacturing. At the same time, an additive manufacturing database system is established to store temperature data in DATA_Image_Temp, laying the groundwork for the next automated additive manufacturing. In order to further correlate the relationship between temperature, current and defects, the temperature acquisition module and the current acquisition module established a temperature-current-defect array and generated two-dimensional image information based on it to further study the relationship between defect formation and process parameters. The correspondence between the temperature data and the current data is specifically explained as follows: the current value refers to the average current change at the location where the cladding layer is deposited over a period of time, and the temperature value is the average temperature at that location after one layer of cladding layer has been deposited. Furthermore, if a corresponding defect occurs at this location, the corresponding location is labeled as loc1, loc2, loc3, etc., in sequence. Here, loc1 indicates that the defect type at this location is 1, loc2 indicates that the defect type at this location is 2, and so on, forming an independent array index to provide data support for further automated intelligent identification.

2. The additive manufacturing quality control system according to claim 1, characterized in that: The layer height acquisition module scans the layer height data of each layer using a laser contour scanner. To improve the visualization and processability of the data and increase the accuracy of recognition, the one-dimensional data is converted into two-dimensional image information for storage. Furthermore, layer height data in the horizontal and vertical directions of each cladding layer is selected as the output result. The position coordinates in the X and Y directions are subdivided according to the length and width of each cladding layer to select the corresponding layer height data. The interval t in the X and Y directions is an important basis for dividing the data points. The size of t directly affects the amount of data, thus affecting the accuracy of the data.

3. The additive manufacturing quality control system according to claim 1, characterized in that: The current acquisition module collects current data during the additive manufacturing process using Hall sensors and numbers the data along with the time sequence. By selecting appropriate process parameters and keeping other parameters constant, the current value is gradually increased during welding. The Hall sensors are used to record the current data of each layer in real time to study the impact of current on the additive manufacturing process.

4. The additive manufacturing quality control system according to claim 1, characterized in that: The neural network model training and storage module further extracts effective image information as feature input data for the neural network system by calling data from the temperature acquisition module and the floor height acquisition module. The two-dimensional image is split into R, G, B, H, S, and V channels, and the pixel value of each channel is calculated sequentially. The difference between the pixel value of each channel and the corresponding pixel value of the previous frame is used to obtain the T value. The T value is compared with a preset threshold TH. If the T value is within the threshold range, the next step of image processing is performed to further extract the image's feature information as training data for the neural network model. If the T value is not within the preset threshold range, the TH threshold is adjusted according to the working conditions, and the comparison with the T value continues until it meets the preset threshold. Finally, the effective feature information is used as training data to save the current neural network model for subsequent defect classification.

5. The additive manufacturing quality control system according to claim 1, characterized in that: The data center storage module includes a layer height acquisition module, a current acquisition module, a temperature acquisition module, a current and temperature data module, and a neural network model training and storage module. The data center storage module can store multiple modules, providing more interfaces for subsequent system upgrades and further improving the system's ability to control the quality of additive manufacturing.

6. The additive manufacturing quality control system according to claim 1, characterized in that: The central system forming quality parameter control and identification module coordinates the data from various data acquisition modules for iterative updates, further improving the system's quality forming control capability and providing data and process references for achieving automated high-precision additive manufacturing. The additive manufacturing forming quality control system has two modes: an iterative data update identification mode and a database-based identification mode. Each new additive manufacturing system can call each module to collect new data and train the model, updating the new data to each module to continuously reduce the system model identification error and improve the system's forming accuracy control capability.

7. An additive manufacturing quality control system according to any one of claims 1-6, characterized in that: The control method of the control system includes: storing physical data collected by various sensors in the system storage module, further converting one-dimensional data into two-dimensional data, and labeling different types of defective welds and normal welds; further, using the labeled data as input data for neural network training, saving the trained model, and calling the model in the next additive manufacturing process to detect weld quality in real time, and outputting the weld quality type and corresponding welding process parameters.

8. The additive manufacturing quality control system according to claim 7, characterized in that: The various sensors include temperature sensors, Hall effect sensors, and laser profile scanners; they output temperature data, current data, and weld bead height data respectively; the current data, temperature data, and weld bead height data at each time point are compiled into an array, corresponding to the current weld quality, which can retain information in the additive manufacturing process and enhance the neural network model's ability to identify different types of defects.

Images