Digital twin manufacturing and testing system for key parts of aerospace equipment

Abstract

The invention discloses a digital twin manufacturing and testing system for key parts of aerospace equipment, which belongs to the field of intelligent manufacturing and comprises an intelligent manufacturing module, a surface detection module, a nondestructive inspection module and a mechanical detection module. The intelligent manufacturing module digitalizes the manufacturing process by means of a digital twinning technology, so that the processing process state and the subsequent detection result can be mapped conveniently; the surface detection module is mainly used for visual detection, two-dimensional or three-dimensional visual information is used, an artificial intelligence algorithm is combined, and surface machining quality detection is achieved; the nondestructive inspection module is used for detecting internal defects of the parts by virtue of internal defect inspection equipment; the mechanical detection module is mainly used for simulating detection of mechanical characteristics and life cycle of parts under dynamic working conditions of stress, temperature and the like. According to the method, digital mapping of the intelligent manufacturing process, the surface characteristics, the internal characteristics and the mechanical properties is realized by constructing a neural network architecture, so that the aim of gradually replacing physical testing with digital testing is fulfilled.

Description

Technical Field

[0001] The present invention belongs to the field of intelligent manufacturing, and specifically relates to a digital twin intelligent manufacturing and testing system for key components of aerospace equipment. Background Art

[0002] As the aerospace industry moves towards the fast lane of development, countries are paying more and more attention to improving the digitalization level of aerospace equipment manufacturing and testing processes. Aerospace equipment is a complex system that integrates multiple subsystems and mechanical components, with a wide variety of supporting parts, complex and changeable manufacturing processes, and many testing links. The processes are highly correlated and mutually restricted. In the manufacturing and testing process of key parts of aerospace equipment, due to the wide variety of parts, unclear test plans, unclear boundary conditions, and the need for a large number of repeated physical tests during the design and manufacturing process, the cost of physical testing is high and the cycle is long, which greatly restricts the improvement of aerospace equipment design efficiency. Therefore, the digital twin intelligent manufacturing and testing system may be one of the effective ways to solve the above problems.

[0003] Existing intelligent manufacturing and testing systems mostly rely on a single testing device to test the characteristics of key components. They cannot cover a full range of testing functions such as surface defects, internal defects, and mechanical properties, and cannot be applied to multi-variety component testing tasks. Therefore, it is difficult to meet the digital twin intelligent manufacturing and testing needs of key components of aerospace equipment.

[0004] For example, Chinese patent application CN116579660A discloses a method and interactive detection system for the processing quality inspection of aircraft engine parts, but the system mainly relies on two-dimensional and three-dimensional drawings to build a management system for digital inspection, and has poor coverage of the internal and mechanical properties of parts; Chinese patent CN115034147B discloses an intelligent manufacturing system based on digital twins, but the system mainly relies on physical sensors to collect data to calculate processing errors and then generate correction signals, and has poor coverage of the types of key parts inspection; Chinese patent application CN117825521A mainly addresses the problem of low coverage of ultrasonic detection equipment probes, and designs an automated production line to rotate the workpiece to achieve inspection.

[0005] In the process of aerospace equipment design and intelligent manufacturing, the high cost, long cycle and limited mode of physical testing of key components are the key issues that restrict the improvement of design efficiency. However, the existing manufacturing and testing systems generally have a single detection type and have not achieved the integration of multi-type and multi-platform detection modes at the equipment and digital information levels. Therefore, it is urgent to develop a digital twin intelligent manufacturing and testing system for key components of aerospace equipment. Summary of the invention

[0006] In order to solve the problems of high testing cost and long cycle caused by numerous physical tests in the design process of key parts of aerospace equipment, the present invention provides a digital twin intelligent manufacturing and testing system for key parts of aerospace equipment. A small sample data set is established through the data collected by the four modules of intelligent manufacturing module, surface detection module, nondestructive flaw detection module and mechanical detection module during the detection process. Combined with the actual data under actual working conditions, a neural network architecture is constructed to realize the fusion of small sample data and actual data and the weight sharing in the mapping process of each module. With the help of reinforcement learning mechanism, continuous training is performed to iterate the weights, and the digital mapping of intelligent manufacturing process-surface characteristics-internal characteristics-mechanical properties is realized to complete the overall digital twin intelligent manufacturing and testing system, so as to achieve the purpose of gradually replacing physical tests with digital tests. With the help of digital twin technology, the present invention achieves the purpose of gradually replacing physical tests with digital tests, so that the test cost and cycle of the design process of key parts of aerospace equipment such as engine blades and blades are reduced.

[0007] In order to achieve the above object, the present invention adopts the following technical scheme:

[0008] A digital twin intelligent manufacturing and testing system for key components of aerospace equipment, including an intelligent manufacturing module, a surface inspection module, a non-destructive flaw detection module, and a mechanical inspection module;

[0009] The intelligent manufacturing module uses digital twin technology to digitize the manufacturing process, which facilitates mapping of processing parameters with the detection results of the surface detection module, non-destructive testing module, and mechanical detection module, converges to the real working conditions, and realizes digital prediction of the processing process;

[0010] The surface inspection module mainly uses visual inspection, based on two-dimensional or three-dimensional visual information, combined with artificial intelligence algorithms, to detect surface processing quality;

[0011] The non-destructive testing module detects internal defects of key components with the help of internal defect detection equipment;

[0012] The mechanical detection module simulates the mechanical properties of key components under stress and temperature dynamic conditions and performs life cycle detection;

[0013] The intelligent manufacturing module is used to collect processing parameters and geometric data of key components, the surface inspection module is used to collect surface feature data of key components, the non-destructive testing module is used to collect internal damage and pore data of key components, and the mechanical testing module is used to collect mechanical property data of key components. A small sample data set is established through geometric data, surface feature data, internal damage, pore data, and mechanical property data. Combined with the actual data under actual working conditions, a neural network is constructed to achieve the fusion of small sample data and actual data and the weight sharing in the mapping process of the intelligent manufacturing module, surface inspection module, non-destructive testing module, and mechanical inspection module. With the help of reinforcement learning mechanism, continuous training and weight iteration are performed to realize the digital mapping of intelligent manufacturing process-surface characteristics-internal characteristics-mechanical properties.

[0014] Furthermore, the key components of the aerospace equipment include blisks and blades.

[0015] Furthermore, the intelligent manufacturing module processes key components of aerospace equipment, collects the processing parameters of the components based on digital twin technology, completes digital modeling of the processing process, serves as the mapping basis for digital twin intelligent manufacturing and testing, and realizes digital prediction of the processing process.

[0016] Furthermore, the surface detection module completes the key point analysis required for modeling of key components through two-dimensional or three-dimensional visual information, digitally enhances the surface features near the key points of key components based on generative AI and deep learning algorithms, and completes the inspection of the processing quality of key components based on the enhanced surface features; the key points include holes and curved surfaces.

[0017] Furthermore, the nondestructive testing module uses eddy current or ultrasonic digital information to detect internal damage and pores of key components according to the intensity of the detection signal and the abnormalities of oscillation in the time domain and frequency domain. At the same time, it scans key areas including holes and curved surfaces according to the processing characteristics of different key components, and improves the detection efficiency by adaptively adjusting the scanning density of preset key areas and non-key areas; the non-key areas refer to regular planes.

[0018] Furthermore, the mechanical detection module simulates the dynamic working conditions of key components under real working conditions by setting dynamic working condition information, combines the small sample data set of mechanical properties collected by the mechanical detection module with the actual data under actual working conditions, and fits the degradation law of mechanical properties of key components under specified stress and temperature conditions, providing a reference for adjusting processing parameters in the processing process and inferring the results of surface and internal defect detection.

[0019] Furthermore, the digital mapping of the intelligent manufacturing process-surface characteristics-internal characteristics-mechanical properties relies on the shared mode of multi-task learning to construct a shared neural network, design the input layer and hidden layer of the shared neural network, and use the data of the intelligent manufacturing module and the surface detection module in the constructed small sample data set as input, and the data of the intelligent manufacturing module and the surface detection module under actual working conditions as comparison input, and complete the shared learning of the input layer and the hidden layer with the help of the shared learning kernel; use the internal damage, pore data and mechanical properties as the output of the neural network, design the loss function, and use the reinforcement learning mechanism to continuously train and iterate the weights, so that the model of the digital mapping of the intelligent manufacturing process-surface characteristics-internal characteristics-mechanical properties converges to the real working condition, and constructs a digital mapping model.

[0020] Furthermore, the nondestructive flaw detection module includes an X-ray flaw detector, an eddy current flaw detector or an ultrasonic flaw detector.

[0021] Furthermore, the processing quality includes the quality of knife marks, surface cracks, roughness, and processing deformation.

[0022] Beneficial effects:

[0023] 1. The present invention digitizes the manufacturing process with the help of digital twin technology by setting up an intelligent manufacturing module, meets the processing requirements of multiple varieties of key components, constructs a digital model, replaces the traditional method of first processing and then testing, and combines with the subsequent digital twin test model to achieve the purpose of predicting the processing quality results during the processing and adjusting the process parameters.

[0024] 2. The present invention sets up a surface detection module, a non-destructive flaw detection module and a mechanical detection module, relies on the real data provided by specific detection equipment, and uses artificial intelligence algorithms to realize the fusion of virtual and real data, thereby solving the problems of incomplete acquisition of physical test data and high cost during the test process.

[0025] 3. The present invention constructs a neural network architecture, relies on the small sample data sets established by the four modules in the process of intelligent manufacturing and testing, combines the actual data, realizes the fusion of small sample data and actual data and the weight sharing in the mapping process of each module, and uses the reinforcement learning mechanism to continuously train and iterate the weights to realize the digital mapping of intelligent manufacturing process-surface characteristics-internal characteristics-mechanical properties. It can realize the digital prediction or test of mechanical properties based on intelligent manufacturing parameters and surface characteristics, and realize the reverse deduction of processing conditions based on the expected mechanical properties, thereby reducing the number of physical experiments to a certain extent and improving the test efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 This is a principle block diagram of the digital twin intelligent manufacturing and testing system for key components of aerospace equipment of the present invention;

[0027] Figure 2 It is a working principle diagram of the surface detection module of the present invention;

[0028] Figure 3 This is a preliminary experimental data diagram of the surface detection module of the present invention. DETAILED DESCRIPTION

[0029] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0030] The present invention proposes a digital twin intelligent manufacturing and testing system for key parts of aerospace equipment, including an intelligent manufacturing module, a surface detection module, a non-destructive flaw detection module and a mechanical detection module. The intelligent manufacturing module, the surface detection module, the non-destructive flaw detection module and the mechanical detection module have communication and data collection functions.

[0031] like Figure 1 As shown, the working process of the digital twin intelligent manufacturing and testing system of key parts of aerospace equipment of the present invention is: first, the processing of key parts of aerospace equipment is digitized by the intelligent manufacturing module with the help of digital twin technology, and then the surface detection module, non-destructive testing module and mechanical detection module are used to complete the testing process of key parts of aerospace equipment through artificial intelligence algorithms. By constructing a neural network, a small sample data set established based on the data collected by the four modules is fused with the actual working condition data, and converged to the real working condition direction to realize the digital mapping of intelligent manufacturing process-surface characteristics-internal characteristics-mechanical properties, and realize digital testing to gradually replace physical testing, so as to solve the problems of high cost and long cycle caused by numerous physical tests in component design. Furthermore, the intelligent manufacturing module digitizes the processing process with the help of digital twin technology, which is convenient for mapping the state of the processing process with subsequent test results.

[0032] Specifically, the intelligent manufacturing module processes key components of aerospace equipment, digitally processes the parameters of the components based on digital twin technology, completes the digital modeling of the overall processing process by collecting intelligent manufacturing processing data, and provides a digital model foundation.

[0033] Furthermore, the surface inspection module mainly uses visual inspection based on two-dimensional or three-dimensional visual information and combines artificial intelligence algorithms to achieve surface processing quality inspection.

[0034] Specifically, the surface detection module uses two-dimensional or three-dimensional visual information, combined with the model features of key components of aerospace equipment, to digitally enhance the surface features near the key points of components based on generative AI and deep learning algorithms, and judges the surface defects through the sparsity information of the point cloud model to complete the inspection of component processing quality.

[0035] Furthermore, the nondestructive flaw detection module detects internal defects of components with the aid of internal defect detection equipment.

[0036] Specifically, the nondestructive testing module uses internal defect detection equipment to detect the internal structure of parts through electromagnetic or acoustic signals such as eddy current or ultrasound, and judges whether there are any abnormalities inside according to different response signals in the time domain or frequency domain, that is, according to the detection data analysis, the internal defects of parts such as pores and cracks inside the parts are digitally detected. At the same time, according to the processing characteristics of different parts, specific areas are scanned in a focused manner to improve the detection efficiency by adaptively adjusting the scanning density of preset key areas and non-key areas.

[0037] Furthermore, the mechanical detection module mainly simulates the detection of the mechanical properties and life cycle of key components under dynamic conditions such as stress and temperature.

[0038] Specifically, the mechanical detection module combines the material properties of existing key components and other conditions, sets stress temperature instruction information according to the mechanical properties to be tested, and applies different loads (i.e. Figure 1 The applied load in the simulation is used to simulate the dynamic mechanics, temperature and other environments of the parts under real working conditions, and the degradation law of the mechanical properties of the parts under specific working conditions is fitted by big data, so as to predict the mechanical properties and life cycle of the parts. Specifically, the neural network structure is constructed based on the shared mode of multi-task learning, and the input layer and hidden layer of the shared neural network are designed. The data of the intelligent manufacturing module and the surface characteristic module in the constructed small sample data set are used as input, and the data of the two modules under the actual working conditions are used as comparison input. With the help of the kernel of shared learning, the shared learning of the input layer and the hidden layer is completed, and the fusion of small sample data and measured data and the weight sharing in the mapping process of each module are realized. With the internal characteristics and mechanical properties as the output of the neural network, the loss function is designed, and the weight iteration is continuously trained with the help of the reinforcement learning mechanism, so that the digital mapping model of the intelligent manufacturing process-surface characteristics-internal characteristics-mechanical properties converges to the real working condition. The digital mapping model constructed in this way can not only realize the digital prediction or test of mechanical properties based on intelligent manufacturing parameters and surface characteristics, but also reversely realize the reverse inference of processing conditions based on the expected mechanical properties, thereby achieving the purpose of replacing physical testing with digital twin testing.

[0039] like Figure 2As shown, the surface detection module first analyzes the key points of key component modeling, and plans the path for scanning components of different types and working conditions using a two-dimensional or three-dimensional scanning robot based on a deep reinforcement learning robot path planning algorithm, thereby realizing the key point analysis for key component modeling, realizing the deep reinforcement learning robot path planning algorithm, completing a comprehensive scan of key components, and building a basic digital model of the surface features of key components; then using the generative AI-based point cloud upsampling algorithm to complete the local sparse area positioning of the digital model, relying on the surface defect detection algorithm for point cloud deep learning, realizing the defect probability inference for key components, determining the type and location information of surface defects, and finally completing the point cloud update to confirm the defect location. The specific algorithm is as follows: first, the high-frequency point cloud is extracted through a graph filter:

[0040] ;

[0041] in, is a high-pass filter, is the adjacency matrix of the point cloud, is the identity matrix.

[0042] ;

[0043] in, is the current point, is the current point The adjacent points of are the elements of the adjacency matrix A, , P is an arbitrary matrix, R represents a rational number, is the dimension of the matrix, , i, j represent the sequence number of the point, For the current point The extraction filtering result. The current point High frequency point.

[0044] Secondly, the Earth Mover’s Distance (EMD) loss function is used to constrain the geometric consistency between the upsampled point cloud and the true point cloud, and the uniformity loss function is used to ensure the uniform distribution of the upsampled point cloud on the surface.

[0045] The EMD loss function is as follows:

[0046] ;

[0047] in, is the upsampled point cloud, is the target point cloud, is the mapping function of the point pair, is the EMD loss function, for The point cloud in Represents the Euclidean distance between the sample point and the mapped point. Represents the Euclidean distance operation.

[0048] The uniformity loss function is as follows:

[0049] ;

[0050] in, is a set of points, is a global uniform function, is a local uniform function, is the uniformity loss function, , for Finally, through the analysis of the EMD loss function and the uniform distribution of the point cloud, the point cloud upsampling algorithm based on generative AI is analyzed based on the Chamfer evaluation index.

[0051] The Chamfer evaluation index is as follows:

[0052] ;

[0053] in, is the evaluation index of point cloud, for The sampling points in for The sampling points in .

[0054] It is further explained that the surface inspection module can complete digital inspection of the processing quality of parts under a variety of different working conditions, including but not limited to quality characteristics such as knife marks, surface cracks, roughness, and processing deformation.

[0055] like Figure 3 The following is the preliminary experimental data of the surface inspection module. First, the deep reinforcement learning path planning algorithm is used to generate a scanning path based on the sample features. The robot executes this path to complete the sample scanning and obtain the original point cloud data of the part surface. Then, the point cloud upsampling algorithm based on generative AI is used to complete the local sparse point cloud upsampling processing, and the surface flatness is calculated on the upsampled point cloud data. The preliminary experiment is conducted on the same sample five times (corresponding to Figure 3The first to fifth groups of the flatness detection experiment were carried out, and the deviation percentages were 0.026%, 0.208%, 0.098%, 0.094%, and 0.042%, respectively. The deviation percentage here refers to the ratio of the deviation between the experimentally tested flatness and the actual flatness to the actual flatness. The average deviation percentage of the five groups of experiments was 0.0936%. This result shows that the surface detection module test data is highly consistent with the actual experimental data, proving that the surface detection module of the present invention can be used for the detection of key components of aerospace equipment.

[0056] Specifically, the above two-dimensional or three-dimensional scanning robot completes a comprehensive scan of the blade parts by planning the scanning path of the blade parts; then uses the generative AI-based point cloud upsampling algorithm to complete the positioning of the local sparse area of ​​the digital model, and completes the detection of the hole features and surface flatness features of the blade parts. By using digital detection, the process of scanning and detecting multiple surfaces is reduced, and hole feature and surface flatness data samples are established. Through the hole and surface feature data of different blade parts, rapid comparison and detection can be carried out to improve test efficiency.

[0057] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. It should be pointed out that for those skilled in the art, any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A digital twin intelligent manufacturing and testing system for key components of aerospace equipment, characterized in that: Including intelligent manufacturing module, surface inspection module, non-destructive testing module and mechanical inspection module; The intelligent manufacturing module uses digital twin technology to digitize the manufacturing process, which facilitates mapping of processing parameters with the detection results of the surface detection module, non-destructive testing module, and mechanical detection module, converges to the real working conditions, and realizes digital prediction of the processing process; The surface inspection module mainly uses visual inspection, based on two-dimensional or three-dimensional visual information, combined with artificial intelligence algorithms, to detect surface processing quality; The non-destructive testing module detects internal defects of key components with the help of internal defect detection equipment; The mechanical detection module simulates the mechanical properties of key components under stress and temperature dynamic conditions and performs life cycle detection; The intelligent manufacturing module is used to collect processing parameters and geometric data of key components, the surface inspection module is used to collect surface feature data of key components, the non-destructive testing module is used to collect internal damage and pore data of key components, and the mechanical testing module is used to collect mechanical property data of key components. A small sample data set is established through geometric data, surface feature data, internal damage, pore data, and mechanical property data. Combined with the actual data under actual working conditions, a neural network is constructed to achieve the fusion of small sample data and actual data and the weight sharing in the mapping process of the intelligent manufacturing module, surface inspection module, non-destructive testing module, and mechanical inspection module. With the help of reinforcement learning mechanism, continuous training and weight iteration are performed to realize the digital mapping of intelligent manufacturing process-surface characteristics-internal characteristics-mechanical properties.

2. According to claim 1, a digital twin intelligent manufacturing and testing system for key components of aerospace equipment is characterized in that: The key components of the aerospace equipment include blisks and blades.

3. According to claim 1, a digital twin intelligent manufacturing and testing system for key components of aerospace equipment is characterized in that: The intelligent manufacturing module processes key components of aerospace equipment, collects the processing parameters of key components based on digital twin technology, completes digital modeling of the processing process, serves as the mapping basis for digital twin intelligent manufacturing and testing, and realizes digital prediction of the processing process.

4. The digital twin intelligent manufacturing and testing system for key components of aerospace equipment according to claim 3 is characterized in that: The surface inspection module completes the key point analysis required for modeling key components through two-dimensional or three-dimensional visual information, digitally enhances the surface features near the key points of key components based on generative AI and deep learning algorithms, and completes the inspection of the processing quality of key components based on the digitally enhanced surface features; the key points include holes and curved surfaces.

5. The digital twin intelligent manufacturing and testing system for key components of aerospace equipment according to claim 4, characterized in that: The nondestructive testing module detects internal damage and pores of key components through eddy current or ultrasonic digital information according to the intensity of the detection signal and the abnormalities of oscillation in the time domain and frequency domain. At the same time, it scans key areas including holes and curved surfaces according to the processing characteristics of different key components, and improves the detection efficiency by presetting the scanning density of key areas and non-key areas and adaptively adjusting them; the non-key areas refer to regular planes.

6. The digital twin intelligent manufacturing and testing system for key components of aerospace equipment according to claim 5, characterized in that: The mechanical detection module simulates the dynamic working conditions of key components under real working conditions by setting dynamic working condition information, combines the small sample data set of mechanical properties collected by the mechanical detection module with the actual data under actual working conditions, and fits the degradation law of mechanical properties of key components under specified stress and temperature conditions, providing a reference for adjusting processing parameters in the processing process and inferring the results of surface and internal defect detection.

7. The digital twin intelligent manufacturing and testing system for key components of aerospace equipment according to claim 6 is characterized in that: The digital mapping of the intelligent manufacturing process-surface characteristics-internal characteristics-mechanical properties is based on a shared mode of multi-task learning to construct a shared neural network, design the input layer and hidden layer of the shared neural network, take the data of the intelligent manufacturing module and the surface detection module in the constructed small sample data set as input, take the data of the intelligent manufacturing module and the surface detection module under actual working conditions as comparison input, and complete the shared learning of the input layer and the hidden layer with the help of the shared learning kernel; take the internal damage, pore data and mechanical properties as the output of the neural network, design the loss function, and use the reinforcement learning mechanism to continuously train and iterate the weights, so that the model of the digital mapping of the intelligent manufacturing process-surface characteristics-internal characteristics-mechanical properties converges to the real working condition, and constructs a digital mapping model.

8. The digital twin intelligent manufacturing and testing system for key components of aerospace equipment according to claim 1, characterized in that: The nondestructive flaw detection module includes an X-ray flaw detector, an eddy current flaw detector or an ultrasonic flaw detector.

9. The digital twin intelligent manufacturing and testing system for key components of aerospace equipment according to claim 4, characterized in that: The processing quality includes the quality of knife marks, surface cracks, roughness, and processing deformation.

Images