Digital twin system for metal corrosion damage monitoring and prediction

Abstract

The invention discloses a digital twinning system for metal corrosion damage monitoring and prediction, and belongs to the technical field of material corrosion monitoring and artificial intelligence crossing. The system comprises a real-time acquisition module, a corrosion image segmentation module, a Q235 finite element model construction module and a digital twinning implementation module; the system collects structured data and unstructured data of the corrosion test; sequentially carrying out corrosion image enhancement and corrosion image segmentation on the collected corrosion image, and carrying out segmentation result analysis to obtain image feature data; performing simulation modeling on different corrosion scenes by using the COMSOL, and outputting simulation results; according to the system, real-time monitoring and high-precision prediction of the corrosion state are achieved, the equipment maintenance cost and the safety risk are remarkably reduced, and an innovative solution is provided for intelligent health management of the service performance of metal materials.

Description

Technical Field

[0001] The present invention belongs to the technical field of material corrosion monitoring and artificial intelligence, and in particular relates to a digital twin system for metal corrosion damage monitoring and prediction. Background Art

[0002] Metal materials are widely used in aerospace, marine engineering, automotive manufacturing, building structures, and other fields due to their excellent mechanical properties and good processability. However, in service environments, metal materials are often affected by a variety of factors such as humidity, temperature, oxygen, and corrosive media, leading to chemical or electrochemical reactions and corrosion. Metal corrosion not only reduces the mechanical properties and service life of the material but can also lead to structural failure and safety accidents, resulting in significant economic losses and casualties. Therefore, monitoring and predicting metal corrosion damage has important engineering significance and social value.

[0003] Currently, the study of metal corrosion damage relies primarily on two approaches: first, obtaining corrosion behavior data through experimental methods such as outdoor exposure tests and indoor accelerated tests; and second, modeling and predicting the corrosion process using numerical simulation techniques such as cellular automata (CA), artificial neural networks (ANN), and the boundary element method (BEM). Although experimental methods offer high accuracy and intuitiveness, they suffer from long test cycles, small sample sizes, and high costs, making them difficult to meet the needs of large-scale structural service performance assessments. While numerical simulation methods can improve efficiency, their prediction accuracy is limited by modeling assumptions and data-driven model constraints, making it difficult to fully and realistically reproduce the corrosion process.

[0004] In recent years, the development of digital twin technology has provided new solutions for modeling, simulation, prediction, and visualization of material service processes. Digital twin is an advanced technology system that enables simultaneous simulation, dynamic updating, and intelligent analysis of the target system's state and behavior through data mapping and real-time interaction between virtual models and physical entities. This technology integrates multi-source heterogeneous sensor data, experimental historical data, and mechanism models to construct high-fidelity, dynamically evolving virtual entity models in a digital space, enabling visualization, predictability, and controllability of material damage processes.

[0005] Furthermore, deep learning, a core technology in the field of artificial intelligence, has been widely applied to tasks such as image recognition, pattern classification, and complex data modeling. It is particularly well-suited for extracting and classifying corrosion image features. Deep learning models (such as U-Net and DeepLabV3+) based on convolutional neural network (CNN) structures can accurately segment and quantitatively analyze metal corrosion images, providing data support for corrosion behavior modeling and prediction.

[0006] However, the following technical bottlenecks still exist in the current research on metal corrosion monitoring and prediction: (1) There is a lack of effective registration and association mechanisms between multi-scale corrosion damage data representations; (2) There is a lack of real-time, stable and reliable data interaction channels between physical entities and virtual models; (3) The semantic extraction and prediction capabilities of corrosion images need to be improved, especially in terms of real-time performance and robustness in dynamic environments.

[0007] In summary, there is an urgent need for a system that integrates digital twin and deep learning technologies for metal corrosion damage monitoring and prediction. It can achieve real-time mapping of physical and virtual spaces, intelligent fusion of multi-source data, and accurate prediction and visualization of corrosion behavior, thereby breaking through the limitations of traditional corrosion monitoring methods and improving the reliability and service safety of key metal components in complex environments. Summary of the Invention

[0008] In view of the above-mentioned problems, the present invention discloses a digital twin system for metal corrosion damage monitoring and prediction.

[0009] The present invention adopts the following technical solutions:

[0010] A digital twin system for metal corrosion damage monitoring and prediction, comprising a real-time acquisition module, a corrosion image segmentation module, a Q235 finite element model construction module, and a digital twin implementation module;

[0011] The real-time acquisition module is used to collect corrosion test structured data and unstructured data; the structured data includes temperature and humidity, and the unstructured data includes corrosion images;

[0012] The corrosion image segmentation module is used to perform corrosion image enhancement and corrosion image segmentation on the collected corrosion image in sequence, and analyze the segmentation results to obtain image feature data; the image feature data includes the area of ​​the corrosion area, corrosion type, color, texture characteristics and edge roughness;

[0013] The Q235 finite element model building module is used to simulate and model different corrosion scenarios using COMSOL, establish a two-dimensional or three-dimensional corrosion simulation model, and output simulation results; the simulation results include corrosion rate, potential distribution, and equivalent stress simulation value;

[0014] The digital twin implementation module is used to align the area, shape, and grayscale distribution characteristics of the corrosion image segmentation with the output data of COMSOL; and train a regression neural network model to realize corrosion depth prediction.

[0015] Furthermore, the digital twin implementation module also includes a system UI display and user interaction sub-module, which is used to design a UI display page to achieve human-computer interaction and system linkage.

[0016] Furthermore, the human-computer interaction and system linkage include historical image query, corrosion evolution playback and prediction data visualization.

[0017] Furthermore, the real-time acquisition module includes: building an experimental platform for metal corrosion physical entities, integrating temperature and humidity sensors, and industrial camera acquisition equipment; and realizing real-time transmission of structured data and unstructured data from physical space to virtual space through serial communication and wireless communication.

[0018] Furthermore, the transmission of the structured data and unstructured data adopts a hybrid wired and wireless communication method; the wired transmission of the structured data includes serial port communication between Unity and Arduino, and the wireless transmission of the structured data includes sending an HTTP request to the WiFi module; the wired transmission of the unstructured data includes camera acquisition and real-time storage, and Unity synchronous reading, and the wireless transmission of the unstructured data includes establishing a TCP connection.

[0019] Furthermore, the eroded image enhancement includes contrast enhancement, flipping and cropping of the eroded image.

[0020] Furthermore, the corrosion image segmentation includes: using a deep learning model to perform semantic segmentation on the collected corrosion image to extract the corrosion area and its morphological features.

[0021] Furthermore, the deep learning model includes U-Net, DeepLab and FCN.

[0022] Furthermore, the digital twin implementation module includes: the neural network model is trained and optimized based on the image enhanced and annotated training data to improve the robustness and real-time performance of the neural network model in complex environments.

[0023] Furthermore, the corrosion depth prediction includes: fusing simulation results with image feature data to establish a depth prediction model to achieve high-precision prediction of corrosion degree and trend.

[0024] Beneficial effects:

[0025] (1) Strong real-time performance: The system realizes real-time perception and virtual-real synchronization of corrosion behavior through high-frequency real-time acquisition of structured and unstructured data and two-way hybrid communication.

[0026] (2) High degree of intelligence: Combining corrosion image segmentation and physical simulation, it can accurately predict the development trend of metal corrosion and effectively support equipment life assessment and risk warning.

[0027] (3) Good visualization effect: By combining two-dimensional or three-dimensional corrosion simulation models with image display, the dynamic simulation of the corrosion process and the intuitive presentation of the evolution process can be achieved.

[0028] (4) Strong versatility: This system is applicable to a variety of metal materials and various corrosion scenarios, such as atmospheric corrosion, marine corrosion, etc., and has good promotion and application prospects.

[0029] (5) High system integration: By integrating sensing technology, AI algorithms and virtual reality platforms, an integrated corrosion monitoring and prediction solution is constructed. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0031] Figure 1 This is a schematic diagram of the module structure of a digital twin system for metal corrosion damage monitoring and prediction according to the present invention;

[0032] Figure 2 This is a schematic diagram of a digital twin system framework for metal corrosion damage monitoring and prediction according to the present invention;

[0033] Figure 3 This is a schematic diagram of the preparation of an image feature data set for a digital twin system for metal corrosion damage monitoring and prediction in the present invention. DETAILED DESCRIPTION

[0034] In order to better understand the technical solution of the present invention, the embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0035] It should be understood that the embodiments described are only a portion of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by persons of ordinary skill in the art without creative work are within the scope of protection of the present invention.

[0036] Example 1

[0037] A digital twin system for metal corrosion damage monitoring and prediction, such as Figure 1As shown, the system includes a real-time acquisition module, a corrosion image segmentation module, a Q235 finite element model construction module and a digital twin implementation module;

[0038] The real-time acquisition module is used to collect corrosion test structured data and unstructured data; the structured data includes temperature and humidity, and the unstructured data includes corrosion images;

[0039] The corrosion image segmentation module is used to perform corrosion image enhancement and corrosion image segmentation on the collected corrosion image in sequence, and analyze the segmentation results to obtain image feature data; the image feature data includes the area of ​​the corrosion area, corrosion type, color, texture characteristics and edge roughness;

[0040] The Q235 finite element model building module is used to simulate and model different corrosion scenarios using COMSOL, establish a two-dimensional or three-dimensional corrosion simulation model, and output simulation results; the simulation results include corrosion rate, potential distribution, and equivalent stress simulation value;

[0041] The digital twin implementation module is used to align the area, shape, and grayscale distribution characteristics of the corrosion image segmentation with the output data of COMSOL; and train a regression neural network model to realize corrosion depth prediction.

[0042] Furthermore, the digital twin implementation module also includes a system UI display and user interaction sub-module, which is used to design a UI display page to achieve human-computer interaction and system linkage.

[0043] Furthermore, the human-computer interaction and system linkage include historical image query, corrosion evolution playback and prediction data visualization.

[0044] Furthermore, the real-time acquisition module includes: building an experimental platform for metal corrosion physical entities, integrating temperature and humidity sensors, and industrial camera acquisition equipment; and realizing real-time transmission of structured data and unstructured data from physical space to virtual space through serial communication and wireless communication.

[0045] Furthermore, the transmission of the structured data and unstructured data adopts a hybrid wired and wireless communication method; the wired transmission of the structured data includes serial port communication between Unity and Arduino, and the wireless transmission of the structured data includes sending an HTTP request to the WiFi module; the wired transmission of the unstructured data includes camera acquisition and real-time storage, and Unity synchronous reading, and the wireless transmission of the unstructured data includes establishing a TCP connection.

[0046] Furthermore, the eroded image enhancement includes contrast enhancement, flipping and cropping of the eroded image.

[0047] Furthermore, the corrosion image segmentation includes: using a deep learning model to perform semantic segmentation on the collected corrosion image to extract the corrosion area and its morphological features.

[0048] Furthermore, the deep learning model includes U-Net, DeepLab and FCN.

[0049] Furthermore, the digital twin implementation module includes: the neural network model is trained and optimized based on the image enhanced and annotated training data to improve the robustness and real-time performance of the neural network model in complex environments.

[0050] Furthermore, the corrosion depth prediction includes: fusing simulation results with image feature data to establish a depth prediction model to achieve high-precision prediction of corrosion degree and trend.

[0051] Example 2

[0052] (1) Construction of digital twin system framework.

[0053] Build an experimental platform for physical entities of metal corrosion, integrating acquisition equipment such as temperature and humidity sensors and industrial cameras; realize real-time transmission of structured and unstructured data from physical space to virtual space through serial and wireless communication; build a three-dimensional digital twin of metal components in virtual space based on the Unity engine to realize data mapping and synchronous display.

[0054] (2) Semantic segmentation and feature extraction of corrosion images.

[0055] Deep learning models such as U-Net and DeepLabV3+ are used to perform semantic segmentation on the collected corrosion images to extract the corrosion areas and their morphological features. Image enhancement and annotation technologies are combined to train and optimize the model to improve its robustness and real-time performance in complex environments.

[0056] (3) Fusion of corrosion scenario simulation modeling and depth prediction.

[0057] Based on finite element simulation platforms such as COMSOL, a metal corrosion damage evolution model is constructed to obtain physical field data such as corrosion depth and stress distribution. The simulation results are integrated with image feature data to establish a depth prediction model to achieve high-precision prediction of corrosion degree and trend.

[0058] (4) System UI display and user interaction.

[0059] Design UI display pages to realize human-computer interaction and system linkage, including historical image query, corrosion evolution playback, prediction data visualization and other functions.

[0060] Example 3

[0061] A digital twin system for metal corrosion damage monitoring and prediction, its overall structure includes:

[0062] (1) Construction of digital twin system framework.

[0063] Digital twin system framework Figure 2 As shown, it mainly includes physical entities, data acquisition systems, and virtual twin space construction.

[0064] 1) Physical entity and data acquisition system.

[0065] Test platform construction: Q235 cold-rolled steel plates were selected as corrosion specimens and subjected to standard degreasing, cleaning and drying treatments; exposure corrosion tests were conducted in natural or artificial climate environments for a period of 30 days.

[0066] Sensor layout and data collection equipment: Temperature and humidity data are collected via a DHT22 sensor, with an Arduino control board serving as the control unit. A Dahua A7500CG20 industrial camera is used to periodically capture corrosion images with a 2-second sampling period. An ESP8266 WiFi module enables wireless data upload, with serial communication serving as a wired backup channel.

[0067] 2) Construction of virtual twin space.

[0068] Twin model construction: Build a digital three-dimensional metal component model based on the Unity engine; analyze the temperature, humidity and image data collected by sensors in real time and map them to the corresponding positions in the virtual twin space; the UI interface displays real-time temperature, humidity, image streams and historical data query functions, and supports looping of images by time period.

[0069] Data interaction mechanism: Serial port reading, HTTP requests and TCP sockets are used to achieve two-way communication between virtual space and physical space; the Unity side uses the coroutine mechanism to achieve non-blocking data reception, improving the efficiency of UI interface updates; sensor data and image information are classified, stored and processed in structured and unstructured formats.

[0070] (2) Semantic segmentation and feature extraction of corrosion images.

[0071] The schematic diagram of image feature dataset preparation is as follows Figure 3 As shown in the figure, the collected images are enhanced (flip, rotation, scaling, color change); Labelme is used to semantically annotate the images, and the segmentation targets include pearlite (corrosion area) and ferrite (background); training set, validation set and test set are constructed, and the ratio is divided into 8:1:1.

[0072] Model selection and training: U-Net and DeepLabV3+ were selected as the core segmentation models, and structural adjustments and parameter tuning were performed. Multi-scale backbone feature extraction networks (such as ResNet and EfficientNet) were introduced for model comparison experiments. The model used the Dice loss function and the Intersection over Union (IoU) metric to evaluate performance, and ultimately derived the optimal model for deployment.

[0073] (3) Fusion of corrosion simulation modeling and depth prediction.

[0074] Finite element simulation model construction: Use COMSOL to establish a two-dimensional or three-dimensional corrosion simulation model; including steps such as material setting, electrochemical corrosion parameter setting, meshing, and corrosion depth output; output simulation value data of physical quantities such as corrosion depth, stress field, corrosion rate, and potential distribution.

[0075] Data fusion and depth prediction model: Align features such as area, shape, and grayscale distribution extracted from image segmentation with COMSOL output data; train regression neural network models (such as LSTM+CNN and Transformer) to predict corrosion depth; use model pruning, quantization, and other technologies for inference optimization, enabling real-time deployment and interactive display in Unity.

[0076] (4) System UI display and user interaction.

[0077] The virtual space UI interface can be divided into three parts: the left side is the real-time monitoring module of the metal corrosion damage service environment, the middle is the real-time image display, and the right side is the historical query; the system supports historical image time period screening and corrosion trend curve generation; it supports interactive operations such as twin rotation, scaling, annotation, and screenshots.

[0078] The embodiments of the present invention are described in detail above. Specific examples are used herein to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core ideas. At the same time, for those skilled in the art, according to the ideas of the present invention, there may be changes in the specific implementation methods and application scopes. In summary, the contents of this specification should not be understood as limiting the present invention.

Claims

1. A digital twin system for metal corrosion damage monitoring and prediction, comprising a real-time acquisition module, a corrosion image segmentation module, a Q235 finite element model construction module, and a digital twin implementation module. The real-time acquisition module is used to collect structured and unstructured data from corrosion tests; the structured data includes temperature and humidity, and the unstructured data includes corrosion images. The corrosion image segmentation module is used to perform corrosion image enhancement and corrosion image segmentation on the collected corrosion image in sequence, and analyze the segmentation results to obtain image feature data; the image feature data includes the area of ​​the corrosion area, corrosion type, color, texture characteristics and edge roughness; The Q235 finite element model building module is used to simulate different corrosion scenarios using COMSOL and output simulation results; the simulation results include corrosion rate, potential distribution, and equivalent stress simulation value; The digital twin implementation module is used to align the area, shape, and grayscale distribution characteristics of the corrosion image segmentation with the output data of COMSOL; And train the regression neural network model to realize corrosion depth prediction.

2. The digital twin system according to claim 1, characterized in that: The digital twin implementation module also includes a system UI display and user interaction sub-module, which is used to design a UI display page to achieve human-computer interaction and system linkage.

3. The digital twin system according to claim 2, characterized in that: The human-computer interaction and system linkage include historical image query, corrosion evolution playback and prediction data visualization.

4. The digital twin system according to claim 1, characterized in that: The real-time acquisition module includes: building an experimental platform for metal corrosion physical entities, integrating temperature and humidity sensors, and industrial camera acquisition equipment; and realizing real-time transmission of structured data and unstructured data from physical space to virtual space through serial communication and wireless communication.

5. The digital twin system according to claim 1, characterized in that: The transmission of structured data and unstructured data adopts a hybrid communication mode of wired and wireless communication; the wired transmission of structured data includes serial port communication between Unity and Arduino, and the wireless transmission of structured data includes sending HTTP requests to the WiFi module; the wired transmission of unstructured data includes camera acquisition and real-time storage, and Unity synchronous reading, and the wireless transmission of unstructured data includes establishing a TCP connection.

6. The digital twin system according to claim 1, characterized in that: The eroded image enhancement includes contrast enhancement, flipping and cropping of the eroded image.

7. The digital twin system according to claim 1, characterized in that: The corrosion image segmentation includes: using a deep learning model to perform semantic segmentation on the collected corrosion image, and extracting the corrosion area and its morphological features.

8. The digital twin system according to claim 7, characterized in that: The deep learning models include U-Net, DeepLab and FCN.

9. The digital twin system according to claim 1, characterized in that: The digital twin implementation module includes: the neural network model is trained and optimized based on image enhancement and labeled training data to improve the robustness and real-time performance of the neural network model in complex environments.

10. The digital twin system according to claim 1, characterized in that: The corrosion depth prediction includes: fusing simulation results with image feature data to establish a depth prediction model to achieve high-precision prediction of corrosion degree and trend.

Images