A defect intelligent perception and backtracking repair system for additive and subtractive hybrid manufacturing process and a control method thereof

Abstract

The application provides an intelligent sensing and backtracking repair system for defects in an additive-subtractive composite manufacturing process and a control method thereof. A monitoring subsystem obtains real-time images of an additive melt pool and transmits the images to a computing subsystem. The computing subsystem processes the real-time images of the additive melt pool, infers the possibility of defects and the positions of the defects in the current additive layer during the additive process according to a defect judgment reference model, and outputs the possibility and the positions to a communication subsystem. The communication subsystem is connected with an additive-subtractive central control system through a field bus, and according to the possibility and the positions of the defects given by the computing subsystem, the communication subsystem instructs a motion system to move a subtractive machining head to the positions where the defects are likely to be generated to remove the defects by subtractive machining. An intelligent learning subsystem is used to optimize the accuracy of the defect judgment reference model. The application realizes closed-loop management of manufacturing defects in the additive process, weakens the problem of excessive coupling, and improves the quality and manufacturing accuracy of products.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of additive and subtractive hybrid manufacturing and computer-aided intelligent material processing, in particular to an intelligent sensing and backtracking repair system for defects in an additive and subtractive hybrid manufacturing process and a control method thereof. BACKGROUND

[0002] Additive and subtractive hybrid manufacturing is a combined processing and manufacturing method combining the high efficiency and high flexibility of additive manufacturing with the high precision of subtractive manufacturing. For large and complex components, traditional additive manufacturing often cannot meet the requirements of geometric size precision and surface quality. After hybrid subtractive manufacturing, subsequent or in-process machining can significantly improve the final precision of the additive manufacturing part, so that it meets the requirements of use performance, and therefore this hybrid manufacturing technology has broad application prospects. However, due to the high coupling of additive manufacturing and subtractive manufacturing in the manufacturing process, although the final precision of the structure is guaranteed by subtractive manufacturing, if the size of the part after additive manufacturing is greatly out of tolerance, the cutting amount and the difficulty of subtractive path planning will be increased, and even the accumulation of defects will occur, which will adversely affect the hybrid manufacturing process. Therefore, in the additive and subtractive hybrid manufacturing process, online identification and sensing of forming defects in the additive process and defect backtracking repair in the manufacturing process are of great significance to guarantee the quality and precision of the hybrid manufacturing product. SUMMARY

[0003] The application proposes an intelligent sensing and backtracking repair system for defects in an additive and subtractive hybrid manufacturing process and a control method thereof, aiming at the problem of sensitive forming and process defects in the additive and subtractive hybrid manufacturing process, especially in the electric arc additive and subtractive hybrid manufacturing process, and the increase of cutting amount and the complexity of path planning in subtractive manufacturing. The application realizes closed-loop management of manufacturing defects in the additive process, reduces the excessive coupling problem caused by defects in the traditional additive process and subtractive process, and further improves the flexibility of the additive and subtractive manufacturing process and guarantees the quality and manufacturing precision of the hybrid manufacturing product.

[0004] The application is realized by the following scheme:

[0005] An intelligent sensing and backtracking repair system for defects in an additive and subtractive hybrid manufacturing process:

[0006] The system specifically comprises a monitoring subsystem, a communication subsystem, a computing subsystem and an intelligent learning subsystem.

[0007] The monitoring subsystem comprises a high-dynamic welding camera, an infrared thermal imager and a structured light sensor. Real-time images of the additive molten pool are obtained by the high-dynamic welding camera, the infrared thermal imager and the structured light sensor, and are transmitted to the computing subsystem after integration.

[0008] The computing subsystem processes data of real-time images of the additive melt pool, infers the possibility of defects and the position of defects in the current additive layer in the additive process according to a defect judgment reference model, and outputs to the communication subsystem;

[0009] The communication subsystem is connected with the additive and subtractive central control system through a field bus, and informs the motion system to move the subtractive machining head to the position where the defects are likely to be generated for subtractive machining to remove the defects according to the possibility of defects and the position of defects given by the computing subsystem;

[0010] The intelligent learning subsystem is used to optimize the accuracy of the defect judgment reference model.

[0011] Further, the high-dynamic welding camera is installed on the additive machining head by a side shaft to collect passive visible light vision images of the additive melt pool in real time; the infrared thermal imager is installed on the additive machining head by a side shaft to collect infrared vision images and temperature field data in front and back of the additive melt pool; and the structured light sensor collects active visible light vision images of the additive melt pool in real time to identify surface profile information of the current deposition layer of the additive melt pool.

[0012] The monitoring subsystem processes the collected passive visible light vision images, infrared vision images and active visible light vision images of the additive melt pool in real time through multi-information fusion image processing, obtains real-time images of the additive melt pool, integrates the real-time images of the additive melt pool, the temperature field data in front and back of the additive melt pool and the surface profile information of the current deposition layer of the additive melt pool according to absolute additive time, and transmits the integrated data to the computing subsystem after packaging.

[0013] Further, the computing subsystem extracts the morphology and size information of the melt pool from the real-time images of the additive melt pool through computer graphics algorithms or neural networks, performs filtering, and accumulates the additive process in time to form a melt pool morphology and size dataset of the whole additive process; extracts the temperature and temperature gradient information in front and back of the melt pool from the temperature field data in front and back of the melt pool, performs filtering, and accumulates the additive process in time to form a temperature field dataset of the whole additive process; reconstructs the surface profile curve from the surface profile point cloud data of the current deposition layer, calculates the deposition layer size, and performs real-time registration with the deposition layer size and deposition layer cross-sectional profile designed in the path planning stage, performs filtering, and accumulates the additive process in time to form a surface profile and deposition size dataset of the whole additive process;

[0014] The computing subsystem infers the possibility of defects and the position of defects in the current additive layer in the additive process according to the melt pool morphology and size dataset of the whole additive process, the temperature field dataset of the whole additive process and the surface profile and deposition size dataset of the whole additive process, and outputs to the communication subsystem according to the defect judgment reference model.

[0015] Further, the defect judgment reference model is based on an artificial neural network, the artificial neural network takes the molten pool morphology and size information of the full additive process, the temperature field data set of the full additive process, the surface profile and deposition size data set of the full additive process as input variables, and takes the defect generation possibility as an output variable; the possibility of defect generation is inferred after each layer of additive process is completed, and then the spatial coordinates of the defect generation position are traced back from the historical data set according to the defect generation time.

[0016] Further, the intelligent learning subsystem generates artificial defects through additive and subtractive process experiments, integrates the virtual set of the molten pool morphology and size information of the full additive process, the temperature field data set of the full additive process, the surface profile and deposition size data set of the full additive process of the artificial defect generation process into a training data set of the defect judgment reference model, and improves the accuracy of the defect judgment reference model through iterative learning.

[0017] Further, the intelligent learning subsystem obtains a large amount of theoretical data of defect generation through numerical simulation of the molten pool flow of the additive and subtractive process, constructs a virtual camera in the simulation software to collect data of the additive process at the same position, obtains the virtual set of the molten pool morphology and size information of the full additive process, the temperature field data set of the full additive process, the surface profile and deposition size data set of the full additive process, takes the virtual set as a training data set of the defect judgment reference model, and improves the accuracy of the defect judgment reference model through iterative learning.

[0018] A control method of an additive and subtractive composite manufacturing process defect intelligent perception and backtracking repair system:

[0019] The monitoring subsystem obtains real-time images of the additive molten pool, and transmits the integrated images to the computing subsystem;

[0020] The computing subsystem processes the real-time images of the additive molten pool, infers the possibility of defect generation and the position of defect generation of the current additive layer in the additive process according to the defect judgment reference model, and outputs to the communication subsystem;

[0021] The communication subsystem is connected with the additive and subtractive central control system through a field bus, and according to the defect generation possibility and the defect generation position given by the computing subsystem, the subsystem informs the motion system to move the subtractive machining head to the position where the defect is likely to be generated to remove the defect by subtractive machining;

[0022] The intelligent learning subsystem is used to optimize the accuracy of the defect judgment reference model.

[0023] Advantages of the present application

[0024] The application realizes closed-loop management of manufacturing defects in the additive process, reduces the excessive coupling problem caused by defects in the traditional additive process and subtractive process, and further improves the flexibility of the additive and subtractive manufacturing process and ensures the quality and manufacturing precision of the composite manufacturing product. BRIEF DESCRIPTION OF DRAWINGS

[0025] Figure 1 is a structural diagram of the application;

[0026] Figure 2 is a workflow of the application. DETAILED DESCRIPTION

[0027] The technical solutions in the embodiments of the application will be clearly and completely described below with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are only part of the embodiments of the application, rather than all the embodiments of the application. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor fall within the protection scope of the application.

[0028] In combination Figures 1-2 ,

[0029] An additive and subtractive composite manufacturing process defect intelligent perception and backtracking repair system:

[0030] As Figure 1 the system of the application specifically comprises: a monitoring subsystem, a communication subsystem, a computing subsystem and an intelligent learning subsystem;

[0031] The monitoring subsystem comprises a high-dynamic welding camera, an infrared thermal imager and a structured light sensor; the high-dynamic welding camera is installed on the additive processing head by means of a side shaft, and is used for collecting passive visible light vision images of an additive molten pool in real time; the infrared thermal imager is installed on the additive processing head by means of a side shaft, and is used for collecting infrared vision images of the additive molten pool and temperature field data in front and back of the additive molten pool; the structured light sensor (a rear active linear laser vision sensor) is used for collecting active visible light vision images of the additive molten pool in real time, and identifying surface profile information of a current deposition layer of the additive molten pool;

[0032] The monitoring subsystem performs multi-information fusion real-time image processing on the collected passive visible light vision images, infrared vision images and active visible light vision images of the additive molten pool, obtains real-time images of the additive molten pool, and integrates the real-time images of the additive molten pool, the temperature field data in front and back of the additive molten pool and the surface profile information of the current deposition layer of the additive molten pool according to absolute additive time, and then transmits the integrated data to the computing subsystem after packaging.

[0033] The computing subsystem extracts the morphology and size information of the molten pool from the real-time image of the additive molten pool by computer graphics algorithm or neural network, filters, and time accumulates the additive process to form the molten pool morphology and size dataset of the whole additive process; extracts the front and rear temperature of the molten pool, temperature gradient and other information from the front and rear temperature field data of the molten pool, filters, and time accumulates the additive process to form the temperature field dataset of the whole additive process; reconstructs the surface profile curve from the surface profile point cloud data of the current deposition layer, calculates the deposition layer size, and real-time registers with the deposition layer size and deposition layer cross section profile designed in the path planning stage, filters, and time accumulates the additive process to form the surface profile and deposition size dataset of the whole additive process;

[0034] The computing subsystem infers the possibility of defect generation and the position of defect generation of the current additive layer in the additive process according to the molten pool morphology and size dataset of the whole additive process, the temperature field dataset of the whole additive process, and the surface profile and deposition size dataset of the whole additive process, and outputs to the communication subsystem according to the defect judgment reference model.

[0035] The communication subsystem is connected with the additive and subtractive central control system through the field bus, and according to the defect generation possibility and the defect generation position given by the computing subsystem, the subsystem informs the motion system to move the subtractive machining head to the position where the defect is likely to be generated to remove the defect by subtractive machining; since the judgment of defects is a semi-quantitative inference process based on experience, it is difficult to model the forward model by mathematical reasoning, and the judgment model adopts machine learning algorithm.

[0036] The intelligent learning subsystem is used to optimize the precision of the defect judgment reference model. The intelligent learning subsystem serves to optimize the precision of the defect judgment reference model, and provides training dataset for the machine learning algorithm model by combining numerical simulation and process test, so that it has the precision to meet the performance requirements.

[0037] The defect judgment reference model is based on artificial neural network, and the artificial neural network takes the molten pool morphology and size information of the whole additive process, the temperature field dataset of the whole additive process, and the surface profile and deposition size dataset of the whole additive process as input variables, and takes the defect generation possibility as output variable; infers the possibility of defect generation after each additive process, and then traces back to the spatial coordinates of the defect generation position from the historical dataset according to the time of defect generation.

[0038] The intelligent learning subsystem integrates the virtual set of the molten pool shape and size information of the whole additive process, the temperature field data set of the whole additive process, the surface profile and deposition size data set of the whole additive process into a training data set of a defect judgment reference model, and improves the accuracy of the defect judgment reference model through iterative learning.

[0039] The intelligent learning subsystem obtains a large number of defect generation theoretical data through additive-subtractive process molten pool flow numerical simulation, constructs a virtual camera in the simulation software to collect data of the additive process at the same position, obtains the molten pool shape and size information of the whole additive process, the temperature field data set of the whole additive process, the surface profile and deposition size data set of the whole additive process, and takes the virtual set as a training data set of a defect judgment reference model, and improves the accuracy of the defect judgment reference model through iterative learning.

[0040] A control method of an additive-subtractive composite manufacturing process defect intelligent perception and backtracking repair system:

[0041] The monitoring subsystem obtains real-time images of the additive molten pool, and transmits the images to the computing subsystem after integration;

[0042] The computing subsystem processes the real-time images of the additive molten pool, infers the possibility of defect generation and the position of defect generation of the current additive layer in the additive process according to the defect judgment reference model, and outputs to the communication subsystem;

[0043] The communication subsystem is connected with the additive-subtractive central control system through a field bus, and removes the defect by moving the subtractive machining head to the position where the defect is likely to be generated according to the possibility of defect generation and the position of defect generation given by the computing subsystem;

[0044] The intelligent learning subsystem is used to optimize the accuracy of the defect judgment reference model.

[0045] The above describes a kind of additive-subtractive composite manufacturing process defect intelligent perception and backtracking repair system and its control method, the principle and implementation mode of the present application are described, the above example is only used to help understand the method and core idea of the present application;At the same time, for the general technical personnel in the art, according to the idea of the present application, in specific implementation mode and application range will have changes, the above description should not be understood as the limitation of the present application.

Claims

1. An intelligent defect perception and backtracking repair system for additive-subtractive hybrid manufacturing process, characterized in that: the system specifically comprises a monitoring subsystem, a communication subsystem, a computing subsystem and an intelligent learning subsystem; the monitoring subsystem comprises a high dynamic welding camera, an infrared thermal imager and a structured light sensor; real-time images of the additive molten pool are obtained by the high dynamic welding camera, the infrared thermal imager and the structured light sensor, and are transmitted to the computing subsystem after integration; the computing subsystem processes the real-time images of the additive molten pool, infers the possibility of defect generation and the location of defect generation in the current additive layer during the additive process according to a defect judgment reference model, and outputs to the communication subsystem; the communication subsystem is connected with the additive-subtractive central control system through the field bus, and according to the defect generation possibility and the defect generation location given by the computing subsystem, the subsystem instructs the motion system to move the subtractive machining head to the location where the defect is likely to be generated for subtractive machining to remove the defect; the computing subsystem extracts the morphology and size information of the molten pool from the real-time images of the additive molten pool by computer graphics algorithms or neural networks, filters, and accumulates the additive process in time to form the molten pool morphology and size dataset of the whole additive process; extracts the temperature and temperature gradient information in front of and behind the molten pool from the temperature field data in front of and behind the molten pool, filters, and accumulates the additive process in time to form the temperature field dataset of the whole additive process; reconstructs the surface profile curve from the surface profile point cloud data of the current deposition layer, calculates the deposition layer size, and performs real-time registration with the deposition layer size and deposition layer cross section profile designed in the path planning stage, filters, and accumulates the additive process in time to form the surface profile and deposition size dataset of the whole additive process; the computing subsystem infers the possibility of defect generation and the location of defect generation in the current additive layer during the additive process according to the molten pool morphology and size dataset of the whole additive process, the temperature field dataset of the whole additive process, and the surface profile and deposition size dataset of the whole additive process, and outputs to the communication subsystem; the defect judgment reference model is based on an artificial neural network, which takes the molten pool morphology and size information of the whole additive process, the temperature field dataset of the whole additive process, and the surface profile and deposition size dataset of the whole additive process as input variables, and takes the defect generation possibility as output variable; infers the possibility of defect generation after each additive process, and then traces back to the spatial coordinates of the defect generation location from the historical dataset according to the defect generation time; the intelligent learning subsystem is used to optimize the accuracy of the defect judgment reference model.

2. The system according to claim 1, characterized in that: The high-dynamic welding camera is installed on the additive manufacturing head by a side shaft to collect passive visible light vision images of the additive melt pool in real time; the infrared thermal imager is installed on the additive manufacturing head by a side shaft to collect infrared vision images of the additive melt pool and temperature field data in front and back of the additive melt pool; the structured light sensor collects active visible light vision images of the additive melt pool in real time to identify surface profile information of the current deposition layer of the additive melt pool. The monitoring subsystem performs multi-information fusion real-time image processing on the collected passive visible light vision images, infrared vision images and active visible light vision images of the additive melt pool to obtain real-time images of the additive melt pool, and integrates the real-time images of the additive melt pool, temperature field data in front and back of the additive melt pool and surface profile information of the current deposition layer of the additive melt pool according to absolute additive time, and transmits the integrated data to the computing subsystem after packaging.

3. The system according to claim 1, characterized in that: The intelligent learning subsystem generates artificial defects through additive manufacturing process test, integrates melt pool morphology and size information of the whole additive process, temperature field data set of the whole additive process, virtual set of surface profile and deposition size data set of the whole additive process generated in the process of generating the artificial defects into a training data set of a defect judgment reference model, and improves the accuracy of the defect judgment reference model through iterative learning.

4. The system according to claim 1, characterized in that: The intelligent learning subsystem obtains a large amount of theoretical data of defect generation through additive process melt pool flow numerical simulation, constructs a virtual camera in the simulation software to collect data of the additive process at the same position, and obtains melt pool morphology and size information of the whole additive process, temperature field data set of the whole additive process, virtual set of surface profile and deposition size data set of the whole additive process, and uses the virtual set as a training data set of a defect judgment reference model to improve the accuracy of the defect judgment reference model through iterative learning.

5. A control method applied to the system according to any one of claims 1-4, characterized in that: The monitoring subsystem obtains real-time images of the additive melt pool and transmits the integrated data to the computing subsystem; The computing subsystem processes the real-time images of the additive melt pool, infers the possibility of defect generation and the position of defect generation of the current additive layer in the additive process according to the defect judgment reference model, and outputs the data to the communication subsystem; The communication subsystem is connected with the additive and subtractive central control system through a field bus, and informs the motion system to move the subtractive manufacturing head to the position where the defect is likely to be generated to remove the defect according to the possibility of defect generation and the position of defect generation given by the computing subsystem; The intelligent learning subsystem is used to optimize the accuracy of the defect judgment reference model.

Images