A welding-based intelligent monitoring and early warning method and system

Abstract

The application discloses a kind of based on welding intelligent monitoring early warning method and system, it is related to intelligent early warning technical field, specific steps are as follows: obtaining the real-time physical data of welding workshop;Real-time physical data is mapped to digital twin simulation model and is simulated in real time, utilizes digital twin simulation model and neural network to identify dangerous factor;And dangerous factor is analyzed, it is judged whether there is safety hazard to personnel causes harm, and based on the judgment result, preset early warning scheme is called.Safety hazard identification step is as follows: digital twin simulation model is simulated according to real-time physical data, obtains welding simulation data;Welding simulation data is input into hidden danger identification model, identifies and analyzes dangerous factor, judges whether there is safety hazard to personnel causes harm.The application is simulated by real-time acquisition of the physical data of welding workshop and mapping to digital twin simulation model, not only improves the efficiency of monitoring, and enhances the accuracy of result.

Description

Technical Field

[0001] This invention relates to the field of intelligent early warning technology, and more specifically to an intelligent monitoring and early warning method and system based on welding. Background Technology

[0002] Welding technology plays an indispensable role in modern industrial manufacturing, with applications ranging from building structures and automobile manufacturing to aerospace. However, traditional monitoring methods during welding often rely on manual inspection and offline testing, which are not only inefficient but also struggle to ensure real-time performance and accuracy. Especially in complex or critical welding tasks, such as the welding of large steel structures, pressure vessels, and piping systems, undetected defects can lead to significant safety hazards and even catastrophic accidents.

[0003] However, with the advancement of technology, although some online monitoring methods based on sensor technology and computer vision have been proposed and applied to welding process monitoring, these solutions still have limitations. On the one hand, existing intelligent monitoring systems mainly focus on the assessment of welding quality, with relatively weak safety early warning functions for the welding environment; on the other hand, different welding processes and materials have vastly different requirements for monitoring parameters, and existing systems lack sufficient flexibility to adapt to the safety needs under various working conditions, and there is still room for improvement in data processing speed and early warning response time.

[0004] Therefore, developing an intelligent system capable of real-time monitoring of the welding process and possessing efficient early warning capabilities is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0005] In view of this, the present invention provides a welding-based intelligent monitoring and early warning method and system, which overcomes the above-mentioned defects.

[0006] To achieve the above objectives, the present invention adopts the following technical solution:

[0007] A welding-based intelligent monitoring and early warning method, comprising the following steps:

[0008] Obtain real-time physical data of the welding workshop;

[0009] The real-time physical data is mapped to a digital twin simulation model for real-time simulation. The digital twin simulation model and neural network are used to identify risk factors. The risk factors are analyzed to determine whether there are any safety hazards that could cause harm to personnel. Based on the determination results, a preset early warning scheme is invoked.

[0010] Optionally, the real-time physical data includes welding equipment operation data, material data, environmental data, and operator data.

[0011] Optionally, the steps for constructing the digital twin simulation model are as follows:

[0012] Obtain the physical data of the welding workshop, and construct a welding process dataset based on the physical data;

[0013] Based on the physical data of the workshop and the welding process dataset, data modeling is performed for each physical entity to obtain an entity matrix containing multi-level structure and dynamic attribute changes;

[0014] Construct a spatial layout matrix based on the spatial location of each physical entity;

[0015] An environmental state matrix is ​​constructed based on the welding process dataset.

[0016] Obtain information on welding process constraints and manufacturing resource constraints, and automatically identify constraints using natural language processing technology to generate a production constraint matrix;

[0017] Obtain twin modeling feature values ​​based on the entity matrix and the spatial layout matrix;

[0018] Optimization feature values ​​are obtained based on the environmental state matrix and the production constraint matrix;

[0019] The digital twin simulation model is constructed based on the twin modeling feature values ​​and the optimized feature values.

[0020] Optionally, the steps for constructing the welding process dataset are as follows:

[0021] The physical data is preprocessed to generate preprocessed data;

[0022] A causal relationship model is constructed based on the preprocessed data, and safety influencing factors are screened based on the causal relationship model;

[0023] By utilizing a deep learning fusion model and a multimodal attention mechanism, the preprocessed data is fused based on the aforementioned safety factors to generate time-series data reflecting changes in the welding process state.

[0024] The time series data is classified and stored according to the welding process to construct the welding process dataset.

[0025] Optionally, the preprocessing includes cleaning, noise reduction, normalization, and time alignment.

[0026] Optionally, the steps for identifying the safety hazard are as follows:

[0027] The digital twin simulation model performs simulation based on the real-time physical data to obtain welding simulation data;

[0028] The welding simulation data is input into the hazard identification model to identify and analyze the risk factors and determine whether there are any safety hazards that could cause harm to personnel.

[0029] Optionally, the hazard identification model is built on a recurrent neural network and incorporates a multi-attention graph convolutional network module to extract the position and behavioral characteristics of personnel during the welding process, thereby assisting in the judgment of safety hazards.

[0030] A welding-based intelligent monitoring and early warning system includes:

[0031] The real-time data acquisition module is used to acquire real-time physical data of the welding workshop;

[0032] The safety hazard identification module is used to map the real-time physical data to a digital twin simulation model for real-time simulation, use the digital twin simulation model and neural network to identify risk factors, analyze the risk factors to determine whether there are safety hazards that could cause harm to personnel, and invoke a preset early warning scheme based on the judgment result.

[0033] As can be seen from the above technical solution, the present invention provides an intelligent monitoring and early warning method and system based on welding, which has the following beneficial effects compared with the prior art:

[0034] Real-time performance and accuracy: By acquiring physical data from the welding workshop in real time and mapping it to a digital twin simulation model for simulation, near-instantaneous data analysis and risk assessment can be provided, which not only improves the efficiency of monitoring but also enhances the accuracy of the results.

[0035] Proactive prevention of safety hazards: The hazard identification model combines recurrent neural networks (RNN) and multi-attention graph convolutional network modules, which can effectively extract the position and behavioral characteristics of personnel during the welding process, assist in the judgment of safety hazards, and enable the system to issue timely warnings before accidents occur, thereby achieving proactive prevention of potential safety risks and ensuring the safety of workers.

[0036] Improved production efficiency: The introduction of intelligent monitoring and early warning systems reduces downtime caused by safety accidents, promotes the continuity and stability of production, and thus improves overall production efficiency. Attached Figure Description

[0037] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0038] Figure 1 This is a schematic diagram of the overall method flow structure provided by the present invention;

[0039] Figure 2 This is a schematic diagram illustrating the construction process of the digital twin simulation model provided by the present invention. Detailed Implementation

[0040] 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.

[0041] This invention discloses an intelligent monitoring and early warning method based on welding, such as... Figure 1 As shown, the specific steps are as follows:

[0042] Step 1: Obtain real-time physical data of the welding workshop;

[0043] Step 2: Map the real-time physical data to the digital twin simulation model for real-time simulation, use the digital twin simulation model and neural network to identify risk factors; analyze the risk factors to determine whether there are any safety hazards that could cause harm to personnel, and call the preset early warning plan based on the judgment results.

[0044] In one embodiment, real-time physical data includes welding equipment operation data, material data, environmental data, and operator data.

[0045] In one embodiment, the steps for constructing a digital twin simulation model are as follows: Figure 2 As shown, specifically:

[0046] Step 211: Obtain physical data of the welding workshop and construct a welding process dataset based on the physical data;

[0047] Step 212: Based on the physical data of the workshop and the welding process dataset, perform data modeling for each physical entity to obtain an entity matrix containing multi-level structure and dynamic attribute changes;

[0048] Step 213: Construct a spatial layout matrix based on the spatial location of each physical entity;

[0049] Step 214: Construct an environmental state matrix based on the welding process dataset;

[0050] Step 215: Obtain welding process constraints and manufacturing resource constraints, automatically identify constraints using natural language processing technology, and generate a production constraint matrix;

[0051] Step 216: Obtain the twin modeling feature values ​​based on the entity matrix and spatial layout matrix;

[0052] Step 217: Obtain the optimization eigenvalues ​​based on the environmental state matrix and the production constraint matrix;

[0053] Step 218: Construct a digital twin simulation model based on the twin modeling feature values ​​and optimized feature values.

[0054] In one embodiment, the steps for constructing the welding process dataset are as follows:

[0055] Step 2111: Preprocess the physical data to generate preprocessed data;

[0056] Step 2112: Construct a causal relationship model based on the preprocessed data, and screen safety influencing factors based on the causal relationship model;

[0057] Step 2113: Using a deep learning fusion model and a multimodal attention mechanism, the preprocessed data is fused based on safety influencing factors to generate time series data reflecting the changes in the welding process status.

[0058] Step 2114: Classify and store the time series data according to the welding process to construct a welding process dataset.

[0059] Furthermore, the specific steps for constructing the welding process dataset are as follows:

[0060] The physical data is cleaned, denoised, standardized, and time-aligned to generate physical data with a uniform format.

[0061] Based on the physical principles of the welding process and preprocessed physical data, a causal relationship model among variables in the welding process is constructed using a directed acyclic graph (DAG). The impact of different parameters on welding safety is analyzed through Do-Calculus or the potential outcome framework. This identifies which variables have a significant impact on welding safety and thus screens out safety-influencing factors.

[0062] Based on safety influencing factors, a deep learning fusion model is selected. The deep learning fusion model is used to extract features from the preprocessed data based on the selected safety influencing factors. A multimodal attention mechanism is introduced to dynamically generate weights of different modal features, thereby realizing weighted fusion of cross-modal information and generating high-precision time series data that reflects the overall state of the welding process.

[0063] The fused time series data are categorized and stored in a database according to welding processes, and an effective data management mechanism is established for subsequent analysis and decision-making.

[0064] Furthermore, physical data includes production factor data, production behavior data, and scheduling rule data. Among them, production factors include the geometry of the welding workshop, the size of the workshop, the quantity of equipment and materials, and the number of operators; production behavior data includes workshop state transition data; and scheduling rules are the foundation for ensuring the orderly operation of the workshop, which stipulate the interaction and constraints between various elements within the workshop.

[0065] The physical data was acquired using a variety of sampling devices, including position sensors, welding parameter sensors (current, voltage, force, torque) and environmental sensors (temperature, smoke concentration, harmful gas concentration, wind speed, light intensity, sound) deployed in the welding workshop, as well as cameras.

[0066] The collected physical data includes:

[0067] Welding equipment data: This includes the main body data of the welding equipment, as well as operating parameters such as current, voltage, and power, and process parameters such as welding speed and welding time, which reflect the basic state of the welding process and the working condition of the equipment; Welding environment data: Welding parameters such as temperature, air quality (e.g., oxygen content, concentration of harmful gases), wind speed, light intensity, and sound during the welding process.

[0068] Material and workpiece data: including information such as the type, specifications, and quality of welding materials, as well as the size, shape, and material properties of the workpiece;

[0069] Personnel operation data: The identity information and behavioral data of the operators during the welding process, such as operation time, operation method, operation frequency, etc.

[0070] In one embodiment, during storage, the fused data is classified according to different stages of the welding process (such as preheating, welding, cooling, etc.) based on the characteristics of the welding process, and a welding process dataset containing multiple features (such as temperature, humidity, smoke concentration, sound characteristics, current / voltage fluctuations, etc.) is constructed and stored in the database.

[0071] In one embodiment, in step 212, based on the physical data of the workshop, data modeling is performed for each physical entity to obtain an entity matrix containing a multi-level structure and dynamic attribute changes. The multi-level structure includes a basic level of physical entities and a more detailed component level. The dynamic attribute changes include attributes that change over time, such as equipment wear and workpiece deformation. Specifically:

[0072] First, based on the physical data of the welding workshop, the basic attributes of each physical entity are determined, such as name, type, location, and size. A basic data model is created for each physical entity, and these models will serve as the basic elements of the entity matrix. If the physical entity is more complex, its internal structure is further refined, and its key components are identified and modeled. The attributes of the components include component type, function, and offset relative to the entity. Then, sub-elements are created for these components in the entity matrix, and a hierarchical relationship is established between them and the corresponding physical entities.

[0073] Secondly, identify the dynamic attributes of physical entities, which may change over time or due to other factors, such as the wear and tear of equipment or the deformation of workpieces; capture the changing patterns of each dynamic attribute, and assign corresponding data fields to each dynamic attribute for storing real-time or historical data.

[0074] Finally, the data models of the basic level, component level, and dynamic attributes are integrated into a unified entity matrix.

[0075] In one embodiment, a real-time data update mechanism is provided to obtain the latest physical data from IoT devices or other data sources in the workshop. When new data arrives, the corresponding fields in the entity matrix are updated to reflect the current state of the physical entities.

[0076] In one embodiment, in step 215, welding process constraints and manufacturing resource constraints are obtained, and constraints are automatically identified through natural language processing technology. A production constraint matrix containing intelligent constraint identification is constructed. The production constraint matrix can adjust the constraints according to the actual situation and unexpected events in the production process.

[0077] Furthermore, comprehensively collect information on welding process constraints and manufacturing resource constraints. Process constraints include, but are not limited to, welding parameters, process sequence, and quality standards. Manufacturing resource constraints include, but are not limited to, equipment availability, personnel scheduling, and material supply.

[0078] Natural language processing techniques are applied to preprocess and parse the collected constraint information, including text cleaning, word segmentation, part-of-speech tagging, named entity recognition, relation extraction, and semantic understanding, in order to automatically identify and understand the meaning and logic of the constraints.

[0079] Based on the processing results, a production constraint matrix is ​​constructed. The matrix includes rows representing different constraint conditions or resource items, and columns representing the specific content or attributes of the constraints. The identified constraint information is then filled into the matrix to form the production constraint matrix.

[0080] The constraints in the production constraint matrix are analyzed and classified to clearly identify and store different types of constraints, such as process parameter constraints, equipment resource constraints, and personnel scheduling constraints.

[0081] In one embodiment, in step 216, the entity matrix and spatial layout matrix are input to the feature extraction module, and the twin modeling feature values ​​are automatically extracted using a deep learning algorithm;

[0082] In one embodiment, in step 217, the environmental state matrix and the production constraint matrix are input to the optimization analysis module for intelligent analysis and optimization to obtain optimized feature values.

[0083] In one embodiment, the steps for identifying safety hazards are as follows:

[0084] The digital twin simulation model performs simulations based on real-time physical data to obtain welding simulation data;

[0085] Welding simulation data is input into the hazard identification model to identify and analyze risk factors and determine whether there are any safety hazards that could cause harm to personnel.

[0086] In one embodiment, the hazard identification model is built on a recurrent neural network and incorporates a multi-attention graph convolutional network module to extract the position and behavioral characteristics of personnel during the welding process, thereby assisting in the identification of safety hazards.

[0087] Furthermore, the welding simulation data is input into a pre-trained hazard identification model, which is based on a recurrent neural network architecture and constructed by combining historical welding accident cases, expert domain knowledge, and multimodal dataset analysis. The multimodal dataset is constructed in the same way as the welding process dataset and is based on welding accident-related data.

[0088] By using a hazard identification model to perform in-depth analysis of the input welding simulation data, potential hazardous factors in the welding process can be identified and assessed to determine whether there are any safety hazards that may cause harm to personnel.

[0089] When the hazard identification model detects a safety hazard, it automatically generates a safety warning and immediately feeds this warning back to the staff so that they can take timely and appropriate preventive measures.

[0090] All detected and identified problems and their corresponding solutions are recorded in detail and saved in an industrial database. The data in the database serves as a new training set for subsequent optimization and improvement of the hazard identification model.

[0091] Furthermore, in a large welding workshop, an important welding operation was detected. Based on the welding process requirements, the system first determined the safe welding area and set corresponding monitoring thresholds. At the start of the welding operation, video data detected an operator inadvertently approaching the safe welding area. The operator's identity was identified, and their behavioral data was input into a digital twin simulation model for simulation. The simulation data was then input into a hazard identification model. Based on real-time physical data and simulation results, the hazard identification model determined that the operator's behavior might constitute a safety hazard. An alarm was immediately issued, and the alarm location and the operator's specific location were displayed through a visual interface. Simultaneously, the system automatically stopped the welding equipment to prevent potential accidents.

[0092] Upon receiving the alarm, the operator immediately realized the danger of their actions and quickly evacuated to a safe area. The system then recorded the incident and saved the relevant data to the industrial database as a new training set for subsequent optimization and improvement of the hazard identification model.

[0093] This embodiment also discloses a welding-based intelligent monitoring and early warning system, including:

[0094] The real-time data acquisition module is used to acquire real-time physical data of the welding workshop;

[0095] The safety hazard identification module is used to map real-time physical data to a digital twin simulation model for real-time simulation. It uses the digital twin simulation model and neural network to identify risk factors, analyzes the risk factors, determines whether there are safety hazards that could cause harm to personnel, and calls preset early warning schemes based on the judgment results.

[0096] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to the method section.

[0097] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A welding-based intelligent monitoring and early warning method, characterized in that, The specific steps are as follows: Obtain real-time physical data of the welding workshop; The real-time physical data is mapped to a digital twin simulation model for real-time simulation. The digital twin simulation model and neural network are used to identify risk factors. The risk factors are analyzed to determine whether there are any safety hazards that could cause harm to personnel. Based on the determination results, a preset early warning scheme is invoked. The steps for constructing the digital twin simulation model are as follows: Obtain the physical data of the welding workshop, and construct a welding process dataset based on the physical data; Based on the physical data of the workshop and the welding process dataset, data modeling is performed for each physical entity to obtain an entity matrix containing multi-level structure and dynamic attribute changes; Based on the physical data of the welding workshop, determine the basic attributes of each physical entity; create a basic data model for each physical entity, which will serve as the basic elements of the entity matrix; if the physical entity is more complex, further refine its internal structure and identify and model its key components. In the entity matrix, sub-elements are created for these components, and hierarchical relationships are established between them and the corresponding physical entities; Identify the dynamic attributes of physical entities, capture the changing patterns of each dynamic attribute, and assign corresponding data fields to each dynamic attribute for storing real-time or historical data. Integrate the data models of the basic level, component level, and dynamic attributes into a unified entity matrix; The steps for constructing the welding process dataset are as follows: The physical data is preprocessed to generate preprocessed data; A causal relationship model is constructed based on the preprocessed data, and safety influencing factors are screened based on the causal relationship model; By using a deep learning fusion model to extract features from preprocessed data based on screened safety influencing factors, and introducing a multimodal attention mechanism to dynamically generate weights for different modal features, weighted fusion of cross-modal information is achieved, generating time series data reflecting changes in the welding process state. The time series data is classified and stored according to the welding process to construct the welding process dataset.

2. The intelligent monitoring and early warning method based on welding according to claim 1, characterized in that, The real-time physical data includes welding equipment operation data, material data, environmental data, and operator data.

3. The intelligent monitoring and early warning method based on welding according to claim 1, characterized in that, The construction steps of the digital twin simulation model also include: Construct a spatial layout matrix based on the spatial location of each physical entity; An environmental state matrix is ​​constructed based on the welding process dataset. Obtain information on welding process constraints and manufacturing resource constraints, and automatically identify constraints using natural language processing technology to generate a production constraint matrix; Obtain twin modeling feature values ​​based on the entity matrix and the spatial layout matrix; Optimization feature values ​​are obtained based on the environmental state matrix and the production constraint matrix; The digital twin simulation model is constructed based on the twin modeling feature values ​​and the optimized feature values.

4. The intelligent monitoring and early warning method based on welding according to claim 1, characterized in that, The preprocessing includes cleaning, noise reduction, standardization, and time alignment.

5. The intelligent monitoring and early warning method based on welding according to claim 1, characterized in that, The steps for identifying the aforementioned safety hazards are as follows: The digital twin simulation model performs simulation based on the real-time physical data to obtain welding simulation data; The welding simulation data is input into the hazard identification model to identify and analyze the risk factors and determine whether there are any safety hazards that could cause harm to personnel.

6. The intelligent monitoring and early warning method based on welding according to claim 5, characterized in that, The hazard identification model is built on a recurrent neural network and incorporates a multi-attention graph convolutional network module to extract the position and behavioral characteristics of personnel during the welding process, thereby assisting in the identification of safety hazards.

7. A welding-based intelligent monitoring and early warning system, used to execute a welding-based intelligent monitoring and early warning method as described in any one of claims 1-6, characterized in that, include: The real-time data acquisition module is used to acquire real-time physical data of the welding workshop; The safety hazard identification module is used to map the real-time physical data to a digital twin simulation model for real-time simulation, use the digital twin simulation model and neural network to identify risk factors, analyze the risk factors to determine whether there are safety hazards that could cause harm to personnel, and invoke a preset early warning scheme based on the judgment result.

Images