A temperature-variable adaptive electrofusion welding control method and system

Abstract

The present invention discloses a temperature-variable adaptive electrofusion welding control method and system, which relates to the field of intelligent welding technology. The method includes the following steps: S1: Install a high-precision temperature sensor, and conduct inspections and pre-treatments on each structure; S2: Fit the resistance value of the welding equipment based on the adaptive least squares method, analyze the relationship between temperature and resistance value changes, and adjust the welding temperature of the welding equipment based on the analysis results; S3: Select and inspect the materials to be welded, and perform preheating treatment on the materials to be welded before electrofusion welding; S4: Conduct real-time analysis and processing of the data based on the improved two-stream graph convolutional network, predict the optimal welding temperature of the welded joint, and perform adaptive adjustment; S5: Conduct post-heat treatment on the welded materials after electrofusion welding is completed. The present invention can accurately control the temperature change during the welding process, is applicable to different welding materials, improves the welding quality while reducing human errors.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent welding, and particularly to a temperature-variable adaptive electrofusion welding control method and system. Background Art

[0002] In the traditional electrofusion welding process, the welding parameters are usually set fixedly and cannot be automatically adjusted according to the change of ambient temperature. This results in large fluctuations in welding quality under different temperature conditions.

[0003] For example, in a low-temperature environment, the melting speed of the welding material becomes slower, and higher welding energy is required to achieve good welding results. However, if the welding parameters at normal temperature are still used, it may lead to insufficient welding and a decrease in the strength of the weld. On the contrary, in a high-temperature environment, the melting speed of the welding material accelerates. If the welding parameters are not adjusted in time, overheating may occur, resulting in coarse-grained weld structure and a decline in welding quality.

[0004] The influence of ambient temperature on electrofusion welding is multi-faceted. First of all, temperature affects the physical properties of the welding material, such as thermal conductivity, specific heat capacity, etc. The changes in these properties directly affect heat transfer and melting state during the welding process.

[0005] Secondly, ambient temperature also affects the performance of the welding equipment. For example, in a low-temperature environment, the output power of the welding power supply may be affected, resulting in unstable welding current. At the same time, temperature changes also affect the service life and performance of the welding electrodes.

[0006] In addition, different welding materials have different sensitivities to temperature. Some materials are prone to welding defects such as cracks and pores under large temperature changes. Therefore, a control method that can automatically adjust welding parameters according to ambient temperature changes is needed to ensure the stability of welding quality.

[0007] With the development of modern industry, the requirements for welding quality are getting higher and higher. Especially in some key fields, such as aerospace, automotive manufacturing, petrochemical industry, etc., the quality of welding directly affects the safety and reliability of products.

[0008] In these fields, the welded joints often need to withstand the tests of harsh environments such as high temperature, high pressure, and corrosion. Therefore, the stability and reliability of welding quality must be ensured. The traditional electrofusion welding method can no longer meet these high requirements, and a more advanced welding control method is needed.

[0009] To solve the above problems, the present invention proposes a temperature-variable adaptive electrofusion welding control method and system. Summary of the Invention

[0010] The main object of the present invention is to provide a temperature-change adaptive electrofusion welding control method and system, which can effectively solve the problems in the background art.

[0011] To achieve the above object, the present invention adopts the following technical solutions:

[0012] A temperature-change adaptive electrofusion welding control method includes the following steps:

[0013] S1: Install a high-precision temperature sensor and connect the temperature sensor to the welding control system; then perform parameter setting, inspection and maintenance of the welding equipment, and signal transmission debugging of the welding equipment.

[0014] S2: Detect the temperature change of the welding equipment under constant voltage conditions, respectively collect the voltage, current and temperature change data of the welding equipment, fit the resistance value of the welding equipment based on the adaptive least squares method, analyze the relationship between the temperature of the welding equipment and the resistance value, and adjust the welding temperature of the welding equipment based on the analysis results.

[0015] S3: Select and inspect the materials to be welded, and pre-treat the materials to be welded before electrofusion welding.

[0016] S4: During the electrofusion welding process, the temperature sensor continuously monitors the temperature data of the welding joint and the surrounding welding environment and transmits it to the welding control system in real time. The welding control system performs real-time analysis and processing on the collected data based on the improved two-stream graph convolutional network, predicts the optimal welding temperature of the welding joint, and performs adaptive adjustment of the welding equipment temperature.

[0017] S5: Perform post-heat treatment on the welded materials after electrofusion welding.

[0018] Preferably, the adaptive least squares method in S2 is specifically as follows:

[0019] According to the internal circuit design rules of the welding equipment, combined with Kirchhoff's voltage law and current law, establish an equivalent circuit model, discretize the model, pre-treat the voltage and current data of the collected welding equipment, and perform resistance value fitting based on the following least squares method:

[0020]

[0021] where s is the time; β is the identification parameter matrix; K s is the gain value; y s is the system output; δ s is the input-output data vector; Q s is the covariance matrix; α s is the forgetting factor; T is the transpose of the vector;

[0022]

[0023]

[0024] Among them, γ is a random number between (0.9, 1); ξ is an adjustable parameter; ε s is the residual at time s;

[0025] Based on the adaptive genetic algorithm, an adaptive dynamic adjustment of the forgetting factor is performed. The specific adaptive genetic algorithm is as follows:

[0026] According to the value range interval of the forgetting factor, an initial population is randomly generated. The individual represents the value of the forgetting factor, and the reciprocal of the fitting error is used as the fitness function.

[0027] Based on the method of combining roulette wheel selection and tournament selection, individuals are selected. Specifically, roulette wheel selection is performed 2 times to select a pair of individuals, and the fitness of these 2 individuals is compared. The individual with the higher fitness is selected. This process is repeated until the selection is full.

[0028] Through the crossover probability p a and the mutation probability p b to control the occurrence frequency of chromosome crossover and mutation. Among them, the crossover probability p a and the mutation probability p b are specifically as follows:

[0029]

[0030] Among them, f max is the maximum fitness value in the population of the adaptive genetic algorithm; f bmax is the larger one of the fitness values of two chromosomes; is the average fitness value of the population; c 1 、c 2 、c 3 and c 4 are random numbers between (0, 1];

[0031] Set the maximum number of generations of evolution and the maximum continuous retention generation of the optimal individual in the population. Based on the above selection method, crossover probability and mutation probability, a new generation of population is generated until the termination condition is met.

[0032] Preferably, the preprocessing in S3 is to perform preheating treatment on the materials to be welded. The preheating treatment includes cleaning the surface of the welding materials, preheating the surface and the whole of the materials to be welded, and heat preservation of the welding materials.

[0033] Preferably, in step S3, corresponding welding auxiliary materials are also prepared according to welding process requirements, and the welding auxiliary materials are inspected and pretreated. The inspection of the welding auxiliary materials includes the model selection of the welding auxiliary materials, the cleanliness of the welding auxiliary materials, the shelf life of the welding auxiliary materials, and the state of the welding auxiliary materials. The pretreatment of the welding auxiliary materials includes the cleaning of the welding auxiliary materials, the change of the state of the welding auxiliary materials, and the preheating treatment of the welding auxiliary materials. The state of the welding auxiliary materials is the liquid, solid, and gaseous states of the welding materials.

[0034] Preferably, the welding auxiliary materials include welding shielding gas, welding solvent, welding filler material, anti-spatter agent, and slag cleaning agent.

[0035] Preferably, in step S4, the welding control system collects the welding state data of the welding materials and performs corresponding analysis based on the improved dual-stream graph convolutional network:

[0036] Construct a graph structure according to the entities and relationships in the welding process. Different types of welding materials, various equipment in the welding process, and environmental factors are used as nodes. The different states of the same type of welding materials and different components of complex welding equipment are also subdivided into nodes, and each node contains corresponding physical information. Then, establish an edge between the welding material node and the welding equipment node to represent the interaction relationship between the material and the equipment; establish an edge between the wire feeder node and the welding material node to represent the wire feeder node providing welding wire to the welding torch; establish an edge between different welding equipment nodes to represent the cooperation relationship between different welding equipment; establish an edge between the automatic control equipment in the welding system and other equipment nodes to represent the transmission relationship of control signals; establish an edge between the welding material node and the environment node to reflect the influence of the environment on the material; establish an edge between the environmental humidity node and the welding material node to represent the corrosion effect of humidity on the welding material;

[0037] Based on the relationship between the nodes and edges established above, construct a topological structure described as G=(V,F), where V is the set of nodes; F is the set of edges connecting the nodes; use the adjacency matrix R to represent the relationship between the nodes. If there is an edge between node a and node b, then R ab =1, otherwise R ab =0; use the degree matrix D to represent the number of edges directly connected to the nodes, and its calculation method is expressed by the following equation:

[0038]

[0039] Combine the adjacency matrix R, the degree matrix D, and the node feature matrix X to obtain an enhanced node feature matrix Specifically as follows:

[0040]

[0041] Aggregate the neighbor features of computing node a

[0042]

[0043] wherein, represents the feature representation vector of the (i - 1)-th layer of aggregated neighbor features of node a; N(a) represents the set of neighbor nodes of the node; RELU(·) represents the activation function;

[0044] Capture the temporal dependence between data based on the gated recurrent unit model, and output the welding state features and welding data features. The gated recurrent unit model is specifically as follows:

[0045] cz t = σ(W cz X t + W czh h t-1 + b cz );

[0046] gz t = σ(W gz X t + W gzh h t-1 + b gz );

[0047]

[0048] wherein, cz t is the reset gate at time t; gz t is the update gate at time t; is the memory cell at time t; h t is the hidden state at time t; σ(·) and tanh(·) are activation functions; X t is the input at time t; W cz , W gz and W h are the weight matrices of the input X t in the reset gate, update gate and memory cell respectively; W czh , W gzh and W hh are the weight matrices of the hidden state h t-1 in the reset gate, update gate and memory cell respectively; b cz , b gz and b h are the offsets of the reset gate, update gate and memory cell respectively.

[0049] Preferably, a two-stream graph convolutional network is used to input the welding state features and welding data features output by the gated recurrent unit model into two identical two-stream convolutional networks respectively. First, the data is input into the convolutional module for convolution operation:

[0050]

[0051] where F is the output of the convolutional module; x in is the input of the convolutional module; J is the number of partitioning strategies; ω j is the weight matrix; ω 1j and ω 2j are both weight matrices of the normalized embedded Gaussian function;

[0052] Then, the output of the convolutional module is input into the modeling module:

[0053]

[0054] where F' is the output of the modeling module; F AMP is the attention map;

[0055] F AMP = G × f SE (f avg (G));

[0056] where f SE is the attention score learned through the SE module; f avg is the average pooling;

[0057] Finally, the output features of the two-stream convolutional network are concatenated and input into the prediction module.

[0058] The prediction module is set with 16 hidden layers, and each hidden layer has 128 neurons. For the l-th (l = 1, …, 16) hidden layer, there is;

[0059]

[0060] where Y l is the output of the l-th layer; σ' is the activation function of the l-th layer; W l is the weight of the l-th layer; Y l is the output of the (l - 1)-th layer; B l is the bias of the l-th layer;

[0061] Then, adversarial training is carried out. During the training process, by maximizing the prediction accuracy and minimizing the prediction error, the parameters of the target task prediction module are updated based on the adaptive genetic algorithm, and the mean square error is used as the loss function of the target.

[0062] Preferably, in S4, the voltage of the welding equipment is adjusted to reach the optimal welding temperature of the predicted welding joint, and multi-layer welding treatment is performed on the welding joint.

[0063] A temperature-variable adaptive electrofusion welding control system includes:

[0064] Equipment detection module: used to install and connect a high-precision temperature sensor; then perform parameter setting, inspection and maintenance of the welding equipment, and signal transmission debugging of the welding equipment;

[0065] Equipment analysis module: used to detect the temperature change of the welding equipment under constant voltage conditions, respectively collect the voltage, current and temperature change data of the welding equipment, fit the resistance value of the welding equipment based on the adaptive least squares method, and analyze the relationship between the temperature of the welding equipment and the change of the resistance value, and adjust the welding temperature of the welding equipment based on the analysis results;

[0066] Pretreatment module: used to select and inspect the materials to be welded, and also perform pretreatment on the materials to be welded before electrofusion welding;

[0067] Adaptive adjustment module: used to receive the temperature data of the welding joint and the surrounding welding environment continuously monitored by the temperature sensor in real time during the electrofusion welding process, perform real-time analysis and processing on the collected data based on the improved two-stream graph convolutional network, predict the optimal welding temperature of the welding joint, and perform adaptive adjustment of the welding equipment temperature;

[0068] Post-treatment module: used to perform post-heat treatment on the welding materials after the electrofusion welding is completed.

[0069] Preferably, the equipment detection module includes a sensor detection unit and an equipment detection and processing unit;

[0070] The sensor detection unit is used to install and connect a high-precision temperature sensor;

[0071] The equipment detection and processing unit is used to perform parameter setting, inspection and maintenance of the welding equipment, and signal transmission debugging of the welding equipment;

[0072] The equipment analysis module includes a resistance fitting unit and a temperature analysis unit;

[0073] The resistance fitting unit is used to detect the temperature change of the welding equipment under constant voltage conditions, respectively collect the voltage, current and temperature change data of the welding equipment, and fit the resistance value of the welding equipment based on the adaptive least squares method;

[0074] The temperature analysis unit is used to analyze the relationship between the temperature of the welding equipment and the change of the resistance value based on the fitting result, and adjust the welding temperature of the welding equipment based on the analysis results;

[0075] The adaptive adjustment module includes a temperature prediction unit and a temperature adjustment unit;

[0076] The temperature prediction unit is used to continuously monitor the temperature data of the welding joint and the surrounding welding environment by a temperature sensor during the electrofusion welding process, and receive in real time the temperature data of the welding joint and the surrounding welding environment continuously monitored by the temperature sensor. Based on the improved two-stream graph convolutional network, the collected data is analyzed and processed in real time, and the optimal welding temperature of the welding joint is predicted;

[0077] The temperature adjustment unit is used to adaptively adjust the temperature of the welding equipment according to the optimal welding temperature predicted by the temperature prediction unit.

[0078] Compared with the prior art, the present invention provides a temperature-variable adaptive electrofusion welding control method and system, which have the following beneficial effects:

[0079] The present invention fits the resistance value of the welding equipment based on the adaptive least squares method, analyzes the relationship between the temperature of the welding equipment and the change of the resistance value, and realizes the precise control of the temperature of the welding equipment; constructs a graph structure according to the entities and relationships in the welding process, predicts the optimal welding temperature of the welding joint during the welding process based on the improved two-stream graph convolutional network, and controls the welding equipment to perform multi-layer precise welding; also improves the quality of the welding joint based on preprocessing and postprocessing operations; the present invention can accurately control the temperature change during the welding process, is applicable to different welding materials, and can reduce human errors while improving the welding quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0080] Figure 1 It is the flowchart of the method mentioned in Embodiment 1 of the present invention;

[0081] Figure 2 It is the flowchart of the adaptive genetic algorithm mentioned in Embodiment 1 of the present invention;

[0082] Figure 3 It is the schematic diagram of the architecture of the improved two-stream graph convolutional network mentioned in Embodiment 1 of the present invention;

[0083] Figure 4 It is the system block diagram mentioned in Embodiment 2 of the present invention;

[0084] Figure 5 It is the flowchart of the welding temperature regulation of the present invention.

[0085] In the figure:

[0086] 100. Equipment detection module; 110. Sensor detection unit; 120. Equipment detection and processing unit; 200. Equipment analysis module; 210. Resistance fitting unit; 220. Temperature analysis unit; 300. Preprocessing module; 400. Adaptive adjustment module; 410. Temperature prediction unit; 420. Temperature adjustment unit; 500. Postprocessing module. Detailed implementation manner

[0087] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments.

[0088] Embodiment 1:

[0089] Please refer to Figures 1 - 3 , a temperature change adaptive electrofusion welding control method of the present invention includes the following steps:

[0090] S1: Install a high-precision temperature sensor and connect the temperature sensor to the welding control system; then perform parameter setting, inspection and maintenance of the welding equipment, and signal transmission debugging of the welding equipment; specifically as follows:

[0091] Install a high-precision temperature sensor to ensure that it can accurately monitor the temperature of the welding area and the surrounding environment. Connect the sensor to the welding control system and perform debugging to ensure the accurate transmission and processing of temperature data.

[0092] Check the parameter settings of the welding equipment to ensure that it is in a normal working state. Check and maintain the welding power supply, electrodes, fixtures, etc. to ensure the stability of the welding process.

[0093] S2: Detect the temperature change of the welding equipment under constant voltage conditions, respectively collect the voltage, current and temperature change data of the welding equipment, fit the resistance value of the welding equipment based on the adaptive least squares method, analyze the relationship between the temperature of the welding equipment and the resistance value change, and adjust the welding temperature of the welding equipment based on the analysis results; specifically as follows:

[0094] During the electrofusion welding process, under different temperature conditions, the welding quality of the welding material may fluctuate greatly. Therefore, it is necessary to precisely control the temperature of the welding equipment.

[0095] According to the internal circuit design rules of the welding equipment, establish an equivalent circuit model in combination with Kirchhoff's voltage law and current law, perform discretization processing on the model, preprocess the collected voltage and current data of the welding equipment, and perform resistance value fitting based on the following least squares method:

[0096]

[0097] Among them, s is the moment; β is the identification parameter matrix; K s is the gain value; y s is the system output; δ s is the input-output data vector; Q s is the covariance matrix; α is the forgetting factor; T is the transpose of the vector;

[0098]

[0099]

[0100] Among them, γ is a random number between (0.9, 1); ξ is an adjustable parameter; ε s is the residual at the moment s;

[0101] Based on the adaptive genetic algorithm, the adaptive dynamic adjustment of the forgetting factor is carried out. The specific steps of the adaptive genetic algorithm are as follows:

[0102] Referring to Figure 2 , the initial population is randomly generated according to the value range of the forgetting factor. The individual represents the value of the forgetting factor, and the reciprocal of the fitting error is used as the fitness function.

[0103] Based on the method of combining roulette wheel selection and tournament selection, individuals are selected. Specifically, roulette wheel selection is performed 2 times to select a pair of individuals. Compare the fitness of these 2 individuals, and the individual with higher fitness is selected. Repeat this process until the selection is full.

[0104] The occurrence frequencies of chromosome crossover and mutation are controlled by the crossover probability p a and the mutation probability p b . Among them, the crossover probability p a and the mutation probability p b are specifically as follows:

[0105]

[0106] Among them, f max is the maximum fitness value in the population of the adaptive genetic algorithm; f bmax is the larger one of the fitness values of the two chromosomes; f is the average fitness value of the population; c 1 , c 2 , c 3 and c 4 are random numbers between (0, 1];

[0107] Set the maximum number of generations of evolution and the maximum number of consecutive generations that the optimal individual in the population can be maintained. Generate a new generation of population based on the above selection method, crossover probability, and mutation probability until the termination condition is met.

[0108] In the adaptive least squares method, the value of the forgetting factor has a crucial impact on the performance of the model. The forgetting factor is used to weigh the contributions of old data and new data to the estimation of model parameters. A suitable forgetting factor can enable the model to better track the dynamic changes of the system. The adaptive ant colony algorithm has a powerful global search ability. It can search for the optimal forgetting factor in a relatively complex parameter space. Through continuous iterative search, it is possible to find the value of the forgetting factor that enables the least squares method model based on the forgetting factor to more accurately fit the actual data. The adaptive adjustment of the forgetting factor can make the model more sensitive to new data. By optimizing the forgetting factor, the adaptive ant colony algorithm can make the least squares method model based on the forgetting factor update its parameters faster to adapt to new data changes. The ant colony algorithm simulates the process of ants foraging and guides the search direction through the transmission and update of pheromones. During the optimization process of the forgetting factor, it can gradually narrow the search range based on the feedback information during the search process to find the optimal forgetting factor. This can greatly reduce the time and workload of parameter selection compared to traditional methods such as manual adjustment or simple enumeration.

[0109] A temperature sensor is also used to continuously monitor the temperature of the welding equipment. By combining Joule's law and Ohm's law, the temperature-resistance relationship curve is analyzed. Based on the results of temperature monitoring and analysis, a temperature model of the welding area is established. This model can describe the relationship between temperature and factors such as time and welding parameters, and can control the temperature adjustment of the welding equipment according to the temperature model.

[0110] S3: Select and inspect the materials to be welded. Before electrofusion welding, the materials to be welded are also pretreated as follows:

[0111] Before welding, the welding materials are preheated. Preheating can reduce the cold brittleness of the welding materials and improve the quality and reliability of the welded joints.

[0112] Select welding materials suitable for temperature-variable adaptive electrofusion welding to ensure good welding performance at different temperatures. Inspect and pre-treat the welding materials, such as cleaning and drying, to improve the welding quality.

[0113] According to the welding process requirements, prepare the corresponding auxiliary materials, including welding shielding gas, welding solvent, welding filler material, anti-spatter agent, and slag cleaning agent; among them, for example, the welding shielding gas can select inert gases such as argon and helium, which can prevent the material from being oxidized, and at the same time can improve the welding speed and achieve better penetration control; for example, the welding solvent generally includes melting flux and sintered flux, which can be selected according to actual needs; carbon dioxide gas protection can also be used to prevent spatter.

[0114] S4: During the electrofusion welding process, the temperature sensor continuously monitors the temperature data of the welded joint and the surrounding welding environment and transmits it to the welding control system in real time. The welding control system performs real-time analysis and processing on the collected data based on the improved dual-stream graph convolutional network, predicts the optimal welding temperature of the welded joint, and performs adaptive adjustment of the welding equipment temperature; specifically as follows:

[0115] The improved dual-stream graph convolutional network constructs a graph structure according to the entities and relationships in the welding process, and takes different types of welding materials as independent nodes respectively. For example, if there are different welding materials such as steel and aluminum, a node can be created for each material. These nodes can contain physical property information of the materials, such as melting point, thermal conductivity, specific heat capacity, etc. For the same type of welding material, the nodes can also be further divided according to their different states. For instance, the unwelded material, the material being welded, and the welded material are taken as different nodes respectively, and each node can record the temperature, hardness, microstructure and other characteristics of the material in that state.

[0116] Take various devices used in the welding process, such as welding machines, welding torches, wire feeders, etc. as nodes. These nodes can contain parameter information of the devices, such as power, current, voltage, etc. For a complex welding system, different components of the device can also be taken as sub-nodes. For example, the electrode, nozzle, handle, etc. of the welding torch can be taken as sub-nodes of the welding torch node respectively, and each sub-node can record the specific attributes of that component.

[0117] Considering the welding environmental factors, take the environmental temperature, humidity, air pressure, etc. as environmental nodes. These nodes can record the environmental conditions at the welding site in real time, which have an important impact on the prediction of temperature changes. The geographical location, wind direction and other factors of the welding site can also be taken as extended environmental nodes to further improve the prediction accuracy.

[0118] Then establish edges between the welding material nodes and the welding equipment nodes to represent the interaction relationship between the material and the equipment. For example, an edge can be established from the welding material node to the welding torch node, indicating that the material is being welded by the welding torch. This edge can be assigned weights, representing parameters such as the welding strength and speed. For the relationship between the wire feeder and the welding material, an edge can be established from the wire feeder node to the welding material node, indicating that the wire feeder provides welding wire for the welding material. The weight of the edge can be determined according to factors such as wire feeding speed and wire diameter.

[0119] Connect different welding equipment nodes to represent their collaborative relationship. For example, an edge can be established between the welding machine node and the welding torch node, indicating that the welding machine provides power for the welding torch. The weight of the edge can be determined according to parameters such as current magnitude and voltage stability. If there is an automated control device in the welding system, such as a robot controller, it can be connected to other device nodes to represent the transmission relationship of control signals.

[0120] Establish an edge between the welding material node and the environment node to reflect the influence of the environment on the material. For example, an edge can be established from the welding material node to the environmental temperature node to indicate that the temperature of the material is affected by the environmental temperature. The weight of the edge can be determined according to factors such as the heat transfer coefficient and the range of environmental temperature change. The edge between the environmental humidity node and the welding material node can represent the corrosion effect of humidity on the welding material, and the weight of the edge can be determined according to factors such as the humidity level and the corrosion resistance of the material.

[0121] Then, assign state characteristics and data characteristics to both the nodes and the edges; for the welding material node, the data characteristics can include the chemical composition, physical properties, current temperature, hardness, etc. of the material. These information can be obtained through experimental measurements or material specification sheets. The data characteristics of the welding equipment node can include the model, parameter settings, operating status, etc. of the equipment. For example, the characteristics of the electric welding machine node can include output current, voltage, power factor, etc.; the characteristics of the welding torch node can include nozzle diameter, gas flow rate, electrode wear degree, etc. The characteristics of the environment node can be the real-time measured environmental parameters such as temperature, humidity, air pressure, etc. The state characteristics are generally qualitative states, such as indoor environment or outdoor environment, presence or absence of ventilation equipment, etc.

[0122] The data characteristics of the edge can be determined according to the relationship represented by the edge. For example, the characteristics of the material-equipment edge can include welding speed, welding current, welding pressure, etc.; the characteristics of the equipment-equipment edge can include signal transmission delay, priority of control instructions, etc.; the characteristics of the material-environment edge can include heat transfer coefficient, degree of influence of humidity on the material, etc. The state characteristics are generally whether the edge exists.

[0123] Refer to Figure 3 , construct a graph structure according to the entities and relationships in the welding process. Based on the established relationships between the nodes and edges above, construct a topological structure described as G=(V,F), where V is the set of nodes; F is the set of edges connecting the nodes; use the adjacency matrix R to represent the relationship between the nodes. If there is an edge between node a and node b, then R ab =1, otherwise R ab =0; use the degree matrix D to represent the number of edges directly connected to the node, and its calculation method is expressed by the equation as follows:

[0124]

[0125] Combine the adjacency matrix R, the degree matrix D and the node feature matrix X to obtain the enhanced node feature matrix Specifically as follows:

[0126]

[0127] Calculate the aggregated neighbor features of node a

[0128]

[0129] where represents the feature representation vector of the (i - 1)-th layer of aggregated neighbor features of node a; N(a) represents the set of neighbor nodes of the node; RELU(·) represents the activation function;

[0130] Based on the gated recurrent unit model to capture the temporal dependence between data, output the welding state features and welding data features. The gated recurrent unit model is specifically as follows:

[0131] cz t = σ(W cz X t + W czh h t-1 + b cz );

[0132] gz t = σ(W gz X t + W gzh h t-1 + b gz );

[0133]

[0134] where cz t is the reset gate at time t; gz t is the update gate at time t; h~ t is the memory cell at time t; h t is the hidden state at time t; σ(·) and tanh(·) are activation functions; X t is the input at time t; W cz , W gz and W h are the weight matrices of the input X t in the reset gate, update gate, and memory cell respectively; W czh , W gzh and W hh are the weight matrices of the hidden state h t-1 in the reset gate, update gate, and memory cell respectively; b cz , b gz and b h are the offsets of the reset gate, update gate, and memory cell respectively.

[0135] Input the welding state and welding data output by the gated recurrent unit model into two identical two-stream convolutional networks respectively. First, input the data into the convolutional module for convolutional operations:

[0136]

[0137] Among them, F is the output of the convolution module; x in is the input of the convolution module; J is the number of partitioning strategies; ω j is the weight matrix; ω 1j and ω 2j are both weight matrices of the normalized embedded Gaussian function;

[0138] Then, the output of the convolution module is input into the modeling module:

[0139]

[0140] Among them, F' is the output of the modeling module; F AMP is the attention map;

[0141] F AMP =G×f SE (f avg (G));

[0142] Among them, f SE is the attention score learned through the SE module; f avg is average pooling;

[0143] Finally, the output features of the dual-stream convolutional network are concatenated and input into the prediction module,

[0144] The prediction module is set with 16 hidden layers, and each hidden layer has 128 neurons. For the l-th (l = 1, …, 16) hidden layer, there is;

[0145]

[0146] Among them, Y l is the output of the l-th layer; σ' is the activation function of the l-th layer; W l is the weight of the l-th layer; Y l is the output of the (l - 1)-th layer; B l is the bias of the l-th layer;

[0147] The graph convolutional network model is trained using the collected welding data, and then adversarial training is carried out. During the training process, by maximizing the prediction accuracy and minimizing the prediction error, the parameters of the target task prediction module are updated based on the adaptive genetic algorithm, and the mean square error is used as the loss function of the target.

[0148] Based on the trained model, new welding state data and welding data are input, and the improved dual - stream graph convolutional network is used to predict the temperature change situation. The predicted temperature change situation is applied to the control and optimization of the welding process, such as adjusting welding parameters, taking cooling measures, etc., to ensure welding quality and safety.

[0149] Taking the graph convolutional network as Model 1, the dual - stream graph convolutional network as Model 2, and the improved dual - stream graph convolutional network of this embodiment as Model 3, temperature predictions are made for the same welding material based on the three models, and the prediction results are statistically analyzed. The specific prediction results can be referred to Table 1:

[0150] Table 1 Results of temperature predictions for the same welding material by three models

[0151]

[0152] As can be seen from the table, the RMSE of the model of this embodiment for predicting the temperature at 15 min, 30 min, and 60 min in the future are 0.0421, 0.0428, and 0.0461 respectively; the MAE are 0.0316, 0.0321, and 0.0334 respectively; the MAPE are 5.9029, 6.0391, and 7.1054 respectively. Compared with Model 1 and Model 2, the deviation between the predicted value and the true value of the model is smaller, and the prediction result is more accurate.

[0153] S5: After the electro - fusion welding is completed, post - heat treatment is carried out on the welding material. Specifically as follows:

[0154] After welding, post - heat treatment is carried out on the welding material. Post - heat treatment can eliminate welding residual stress and improve the service life of the welded joint.

[0155] Embodiment Two:

[0156] Please refer to Figure 4 , a temperature - change adaptive electro - fusion welding control system of the present invention includes:

[0157] The device detection module 100 includes a sensor detection unit 110 and a device detection and processing unit 120; the sensor detection unit 110 is used to install and connect a high - precision temperature sensor; the device detection and processing unit 120 is used for parameter setting, inspection and maintenance of the welding device, signal transmission debugging of the welding device, selecting the welding material to be welded, and inspecting and pre - processing the welding material to be welded;

[0158] The device analysis module 200 includes a resistance fitting unit 210 and a temperature analysis unit 220; the resistance fitting unit 210 is used to detect the temperature change of the welding device under constant voltage conditions, collect the voltage, current and temperature change data of the welding device respectively, and fit the resistance value of the welding device based on the adaptive least squares method; the temperature analysis unit 220 is used to analyze the relationship between the temperature of the welding device and the change of the resistance value, and adjust the welding temperature of the welding device based on the analysis result.

[0159] The pretreatment module 300: is used to pretreat the materials to be welded before electrofusion welding.

[0160] The adaptive adjustment module 400 includes a temperature prediction unit 410 and a temperature adjustment unit 420; the temperature prediction unit 410 is used to receive the temperature data continuously monitored by the temperature sensor for the welding joint and the surrounding welding environment in real time during the electrofusion welding process, perform real-time analysis and processing on the collected data based on the improved dual-stream graph convolutional network, and predict the optimal welding temperature of the welding joint; the temperature adjustment unit 420 is used to adaptively adjust the temperature of the welding device according to the optimal welding temperature predicted by the temperature prediction unit 410.

[0161] The post-treatment module 500: is used to post-treat the welding materials after the electrofusion welding is completed.

[0162] Embodiment 3:

[0163] Please refer to Figure 5 , Figure 5 , which is the electrical signal control flow chart of the temperature sensor and the welding device in the present invention. This kind of control can generally be carried out by a PLC controller. The temperature sensor monitors the surrounding environment temperature in real time, and instantaneously feeds back the temperature control information to the temperature control unit based on its own structure design. When the surrounding environment temperature changes, the temperature value of the temperature sensor will change. When it exceeds its set temperature value, it will send a temperature control signal to the controller. After receiving the temperature change signal, the controller performs temperature control on the temperature control system, such as heating control and cooling control. Generally speaking, the temperature control here is mainly carried out by increasing or decreasing the current. Because the surrounding environment temperature does not change suddenly, there is no need to stop the welding action during the temperature control here. When the temperature adjustment reaches the standard, continue welding at the existing welding temperature. If it does not reach the standard, the temperature adjustment system continues to adjust until it reaches the standard and then continues welding. The time during the temperature adjustment is not too long, so there is no need to stop the welding action, which ensures the welding stability and welding efficiency.

[0164] The above are only the preferred specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention, according to the technical solution and inventive concept of the present invention, making equivalent substitutions or changes, shall be covered by the protection scope of the present invention.

Claims

1. A temperature-dependent adaptive electric fusion welding control method, characterized in that: The steps include: S1: Install high-precision temperature sensors and connect them to the welding control system; then perform parameter setting, inspection and maintenance of the welding equipment, as well as signal transmission debugging of the welding equipment; S2: Detect the temperature change of the welding equipment under constant voltage conditions, collect the voltage, current and temperature change data of the welding equipment respectively, and fit the resistance value of the welding equipment based on the adaptive least squares method. The adaptive least squares method is as follows: According to the internal circuit design rules of welding equipment, combined with Kirchhoff's voltage law and current law, a circuit equivalent model is established, and the model is discretized. The collected voltage and current data of the welding equipment are preprocessed, and the resistance value is fitted based on the following least squares method: ;in, for the moment; is the identification parameter matrix; is the gain value; Output for the system; is the input and output data vector; is the covariance matrix; For the forgetting factor; is the transpose of a vector; ; ;in, for A random number between is an adjustable parameter; for The residual of the moment; And analyze the relationship between the temperature of the welding equipment and the change of the resistance value, and adjust the welding temperature of the welding equipment based on the analysis result; S3: Select and inspect the materials to be welded, and pre-treat the materials to be welded before electrofusion welding; S4: During the electrofusion welding process, the temperature sensor continuously monitors the temperature data of the welding joint and the surrounding welding environment, and transmits it to the welding control system in real time. The welding control system analyzes and processes the collected data in real time based on the improved dual-flow graph convolutional network, predicts the optimal welding temperature of the welding joint, and performs adaptive adjustment of the temperature of the welding equipment; S5: After the electric fusion welding is completed, the welding materials are subjected to post-heat treatment.

2. A temperature-dependent adaptive electric fusion welding control method according to claim 1, characterized in that: In S2: The adaptive dynamic adjustment of the forgetting factor is performed based on an adaptive genetic algorithm, and the adaptive genetic algorithm is specifically as follows: The initial population is randomly generated according to the range of the forgetting factor. The individual represents the value of the forgetting factor, and the inverse of the fitting error is used as the fitness function. The method of selecting individuals is based on the mixed selection of roulette and tournament. Specifically, the roulette selection is performed twice, a pair of individuals is selected, and the fitness of the two individuals is compared. The individual with higher fitness is selected, and this process is repeated until the number of individuals is full. By crossover probability and mutation probability Control the frequency of chromosome crossover and mutation, where the crossover probability and mutation probability The details are as follows: ; ;in, is the maximum fitness value in the adaptive genetic algorithm population; is the one with the larger fitness value among the two chromosomes; is the average fitness value of the population; , , and for A random number between The maximum number of evolutionary generations and the maximum number of consecutive generations of the optimal individuals in the population are set, and a new generation of population is generated based on the above selection method, crossover probability and mutation probability until the termination condition is met.

3. A temperature-dependent adaptive electric fusion welding control method according to claim 1, characterized in that: The pretreatment in S3 is preheating the material to be welded, and the preheating includes cleaning the surface of the welding material, preheating the surface and the whole of the material to be welded, and keeping the welding material warm.

4. A temperature-dependent adaptive electric fusion welding control method according to claim 3, characterized in that: In S3, corresponding welding auxiliary materials are prepared according to the welding process requirements, and the welding auxiliary materials are inspected and pretreated. The inspection of the welding auxiliary materials includes the model selection of the welding auxiliary materials, the cleanliness of the welding auxiliary materials, the shelf life of the welding auxiliary materials and the state of the welding auxiliary materials. The pretreatment of the welding auxiliary materials includes the cleaning of the welding auxiliary materials, the state change of the welding auxiliary materials and the preheating treatment of the welding auxiliary materials. The state of the welding auxiliary materials includes the liquid, solid and gaseous states of the welding materials.

5. A temperature-dependent adaptive electric fusion welding control method according to claim 4, characterized in that: The welding auxiliary materials include welding shielding gas, welding solvent, welding filling material, anti-spatter agent and slag cleaning agent.

6. A temperature-dependent adaptive electric fusion welding control method according to claim 1, characterized in that: In S4, the welding control system collects welding state data of the welding material and performs corresponding analysis based on the improved dual flow graph convolutional network: The graph structure is constructed according to the entities and relationships in the welding process. Different types of welding materials, various types of equipment in the welding process, and environmental factors are used as nodes. The nodes of different states of the same type of welding materials and different components of complex welding equipment are also subdivided. Each node contains corresponding physical information; then the edge between the welding material node and the welding equipment node is established to represent the interactive relationship between the material and the equipment; the edge from the wire feeder node to the welding material node is established to represent the wire feeder node to the welding gun to provide welding wire; the edge between different welding equipment nodes is established to represent the collaborative relationship between different welding equipment; the edge between the automation control equipment and other equipment nodes in the welding system is established to represent the transmission relationship of the control signal; the edge between the welding material node and the environment node is established to reflect the impact of the environment on the material; the edge between the environmental humidity node and the welding material node is established to represent the corrosive effect of humidity on the welding material; Based on the relationship between nodes and edges established above, construct The topological structure of is a node set; is the set of edges connecting nodes; using the adjacency matrix Represents the relationship between nodes. If the node and nodes If there is an edge between ,otherwise ; Cost matrix It represents the number of edges directly connected to the node, and its calculation method is expressed by the following equation: ; The adjacency matrix , degree matrix and the node feature matrix Combined to obtain the enhanced node feature matrix , as follows: ; Compute nodes Aggregate neighbor features : ;in, Representation Node No. The layer aggregates the feature representation vector of neighbor features; Represents the set of neighbor nodes of a node; represents an activation function; based on the gated recurrent unit model, the temporal dependency between data is captured, and welding state features and welding data features are output. The gated recurrent unit model is specifically as follows: ; ; ; ;in, for The reset gate of the moment; for The door of renewal at all times; for Memory cells of the moment; for The hidden state of the moment; and is the activation function; for Input of time; , and They are the reset gate, update gate and input in the memory cell respectively. The weight matrix of , and They are reset gate, update gate and hidden state in memory cell respectively. The weight matrix of , and They are the reset gate, update gate and offset of memory cells respectively.

7. A temperature-dependent adaptive electric fusion welding control method according to claim 6, characterized in that: The welding state features and welding data features output by the gated recurrent unit model are respectively input into two identical dual-stream convolutional networks using a dual-stream graph convolutional network. First, the data is input into the convolution module for convolution operation: ;in, is the output of the convolution module; Is the input of the convolution module; is the number of partition strategies; is the weight matrix; and They are all weight matrices of normalized embedded Gaussian functions; Then input the output of the convolution module into the modeling module: ;in, is the output of the modeling module; is the attention map; ;in, is the attention score learned by the SE module; is average pooling; Finally, the output features of the two-stream convolutional network are concatenated and input into the prediction module. The prediction module sets 16 hidden layers, each with 128 neurons. Layer hidden layer, there are; ;in, For the The output of the layer; For the The activation function of the layer; For the The weight of the layer; For the The output of the layer; For the During the training process, the parameters of the target task prediction module are updated based on the adaptive genetic algorithm by maximizing the prediction accuracy and minimizing the prediction error, and the mean square error is used as the target loss function.

8. A temperature-dependent adaptive electric fusion welding control method according to claim 7, characterized in that: In S4, the voltage of the welding equipment is adjusted to reach the predicted optimal welding temperature of the welding joint, and the welding joint is subjected to multi-layer welding.

9. A temperature-dependent adaptive electric fusion welding control system, applicable to a temperature-dependent adaptive electric fusion welding control method according to any one of claims 1 to 8, characterized in that: include: Equipment detection module (100): used to install and connect high-precision temperature sensors; and then perform parameter setting, inspection and maintenance of welding equipment, as well as signal transmission debugging of welding equipment; Equipment analysis module (200): used to detect temperature changes of welding equipment under constant voltage conditions, respectively collect voltage, current and temperature change data of the welding equipment, fit the resistance value of the welding equipment based on the adaptive least squares method, and analyze the relationship between the temperature of the welding equipment and the change of the resistance value, and adjust the welding temperature of the welding equipment based on the analysis result; Pre-processing module (300): used to select and check the materials to be welded, and to pre-process the materials to be welded before electrofusion welding; An adaptive adjustment module (400): used to receive temperature data of the welding joint and the surrounding welding environment continuously monitored by the temperature sensor in real time during the electric fusion welding process, analyze and process the collected data in real time based on the improved dual-flow graph convolution network, predict the optimal welding temperature of the welding joint, and perform adaptive adjustment of the temperature of the welding equipment; Post-processing module (500): used for performing post-heat treatment on welding materials after the electrofusion welding is completed.

10. A temperature-dependent adaptive electric fusion welding control system, applicable to a temperature-dependent adaptive electric fusion welding control method according to any one of claims 1 to 8, characterized in that: The device detection module (100) comprises a sensor detection unit (110) and a device detection processing unit (120); The sensor detection unit (110) is used to install and connect a high-precision temperature sensor; The equipment detection processing unit (120) is used for parameter setting, inspection and maintenance of welding equipment, as well as signal transmission debugging of welding equipment; The device analysis module (200) comprises a resistance fitting unit (210) and a temperature analysis unit (220); The resistance fitting unit (210) is used to detect temperature changes of the welding equipment under constant voltage conditions, collect voltage, current and temperature change data of the welding equipment respectively, and fit the resistance value of the welding equipment based on an adaptive least squares method; The temperature analysis unit (220) is used to analyze the relationship between the temperature of the welding equipment and the change in the resistance value based on the fitting result, and to adjust the welding temperature of the welding equipment based on the analysis result; The adaptive adjustment module (400) comprises a temperature prediction unit (410) and a temperature adjustment unit (420); The temperature prediction unit (410) is used for, during the electric fusion welding process, continuously monitoring the temperature data of the welding joint and the surrounding welding environment by the temperature sensor, and receiving the temperature data of the welding joint and the surrounding welding environment continuously monitored by the temperature sensor in real time, analyzing and processing the collected data in real time based on the improved dual-flow graph convolution network, and predicting the optimal welding temperature of the welding joint; The temperature adjustment unit (420) is used to adaptively adjust the temperature of the welding equipment according to the optimal welding temperature predicted by the temperature prediction unit (410).

Images