Bridge anti-collision guardrail welding path adaptive planning method and system

Abstract

The present invention discloses a method and system for adaptive planning of the welding path of a bridge anti-collision guardrail, which relates to the technical field of bridge construction. The method is used to solve the problem that the welding quality is affected by factors such as environmental changes and operator experience due to reliance on manual experience and fixed operating parameters, which may cause welding defects and even affect structural safety. Traditional welding path planning is generally based on fixed algorithms or manual adjustments, and cannot be dynamically adjusted according to the on-site environment, welding equipment and operator behavior. There is no real-time feedback mechanism, which leads to the inability to adjust and optimize the path planning in the welding process in a timely manner, thereby affecting work efficiency and quality. The method collects welding site data through sensors, uses an improved Gaussian process regression algorithm to evaluate weld quality, and combines welding behavior prediction to generate an optimization strategy to achieve automatic adjustment and continuous optimization of welding parameters.

Description

Technical Field

[0001] The present invention relates to the technical field of bridge construction, and more particularly to a method and system for adaptively planning a welding path for a bridge anti-collision guardrail. Background Art

[0002] During bridge construction and maintenance, the welding of crash barriers is a critical process, and its quality directly impacts the safety and durability of bridges. Traditional welding processes typically rely on manual operation or automated equipment with pre-set paths, which can meet basic requirements to a certain extent. With the increasing complexity of bridge structural design and the diversification of construction environments, traditional welding path planning often lacks flexibility when faced with guardrail components of varying shapes, sizes, and materials, making it difficult to adapt to complex geometries and changing welding requirements.

[0003] The existing technology has the following deficiencies:

[0004] Relying on manual experience and fixed operating parameters, welding quality is affected by factors such as environmental changes and operator experience, potentially leading to welding defects and even compromising structural safety. Traditional welding path planning is typically based on fixed algorithms or manual adjustments, and cannot dynamically adjust to the site environment, welding equipment, and operator behavior. The lack of a real-time feedback mechanism prevents timely adjustment and optimization of path planning during the welding process, thus impacting work efficiency and quality.

[0005] In view of the above problems, the present invention proposes a solution. Summary of the Invention

[0006] In order to overcome the above-mentioned defects of the prior art, an embodiment of the present invention provides a method and system for adaptively planning a welding path for a bridge anti-collision guardrail, so as to solve the problems raised in the above-mentioned background technology.

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] A bridge anti-collision guardrail welding path adaptive planning method includes the following steps:

[0009] Step S1: Using a variety of pre-set sensors, the temperature, humidity, surface condition of the welding material, and operating parameters of the welding equipment at the welding site are collected in real time. An improved Gaussian process regression algorithm is used to evaluate the weld quality and output the evaluation results of weld strength, weld uniformity, and weld defect rate.

[0010] Step S2: monitoring the welding operator's behavior pattern, including welding speed, welding angle, and welding gun trajectory, and predicting the optimal adjustment direction of the future welding path to obtain welding behavior prediction, and generating a specific welding path optimization strategy based on the weld quality assessment results and welding behavior prediction;

[0011] Step S3: Combined with the welding path optimization strategy, the motion parameters of the welding equipment are automatically adjusted to perform specific adjustment operations, including welding gun angle correction, welding speed adjustment, and weld tracking compensation;

[0012] Step S4: Obtain the adjusted welding parameters and weld quality feedback from the path automatic adjustment module, use deep reinforcement learning technology to optimize the weld quality assessment module and optimization decision engine, and continuously improve the welding path planning and optimization strategy.

[0013] In a preferred embodiment, the method comprises the following steps:

[0014] Step A1: Using a thermocouple sensor and a capacitive humidity sensor, respectively, the temperature and humidity data of the welding area are collected in real time;

[0015] Step A2: Detecting the surface roughness, geometric features, and internal defects of the welding material using a surface scanning device;

[0016] Step A3: Monitor the current fluctuation and vibration frequency of the welding equipment using a current sensor and a vibration sensor to assess the equipment load and stability;

[0017] Step A4: De-noise, normalize, and extract features from the data collected in steps A1 to A3 to generate a multi-dimensional feature vector V, which is expressed as: ; where n is the feature dimension, is the i-th eigenvalue;

[0018] Step A5: Input the eigenvector V into the improved Gaussian process regression model and calculate the comprehensive weld quality score Q using the following formula: ; Where Q represents the weld quality assessment result, represents the weight, Represents the kernel function mapping of the feature vector, b represents the bias, and m represents the dimension of the feature vector;

[0019] Q is split into three sub-indicators: welding strength S, weld uniformity U, and welding defect rate D. Independent neural network models are used to model each of them: ;in, 、 、 They are respectively neural network models based on multi-layer perceptron.

[0020] In a preferred embodiment, the method comprises the following steps:

[0021] Establish a historical welding behavior database to store historical data of welding speed, angle and trajectory;

[0022] Compare the current feature vector V with the feature vector in the historical database Perform cosine similarity matching: ;in, Represents the eigenvector V and the historical eigenvector The similarity between and Denote the eigenvectors V and The i-th eigenvalue of , n represents the dimension of the eigenvector;

[0023] A prediction model is built based on the support vector machine algorithm, which inputs the current feature vector V and outputs the future welding path adjustment direction; ;in, represents the path adjustment resources allocated to the predetermined time period t, Indicates the risk assessment value for this time period, It represents the welding demand forecast value for this time period, and T represents the total time period.

[0024] In a preferred embodiment, the method comprises the following steps:

[0025] According to the angle correction instructions in the optimization strategy, the welding gun angle deviation Δθ is detected in real time and adjusted to the target angle ; where α is the correction coefficient, is the current angle, is the angle compensation;

[0026] Calculate the current speed in real time according to the speed adjustment instruction Difference from target speed , and perform speed compensation: ;in, Indicates the new position after adjustment. Indicates the current location, Indicates the position compensation amount;

[0027] Dynamically compensate for position deviation according to weld tracking instructions , generate a new welding trajectory: ;in, Indicates the new position after adjustment. Indicates the current location, Indicates the position compensation amount.

[0028] In a preferred embodiment, the method comprises the following steps:

[0029] Obtain adjusted welding parameters and weld quality feedback, including adjusted welding gun angle, welding speed, weld position and their corresponding weld quality feedback;

[0030] The model is trained using deep reinforcement learning technology on the processed data. The input parameters of the model include the adjusted welding gun angle, welding speed, weld position and the corresponding weld quality feedback. The output is the continuously improved welding parameters or actions, and the impact of welding parameter adjustments on weld quality is learned.

[0031] Reward function design: Design a multi-dimensional reward function, including positive incentives, negative penalties, and a comprehensive scoring mechanism:

[0032] Positive incentive items: reduction in weld defect rate and improvement in appearance uniformity;

[0033] Negative penalty term: the time loss caused by the welding speed deviating from the optimal value;

[0034] Comprehensive scoring mechanism: Dynamically adjust reward weights based on actual quality feedback;

[0035] The welding parameters are iteratively updated through the policy network, and the welding gun angle and speed are adjusted online to maximize the cumulative reward, forming a closed-loop control.

[0036] The bridge anti-collision guardrail welding path adaptive planning system includes: environmental perception module, weld quality assessment module, welding behavior analysis module, optimization decision engine, path automatic adjustment module and self-learning module, and the signals between each module are connected.

[0037] Environmental perception module: Through a variety of preset sensors, it collects the temperature, humidity, surface condition of welding materials and operating parameters of welding equipment at the welding site in real time, and transmits the collected data to the weld quality assessment module;

[0038] Weld quality assessment module: Receives data from the environmental perception module, uses an improved Gaussian process regression algorithm to assess weld quality, outputs assessment results of weld strength, weld uniformity, and weld defect rate, and sends the assessment results to the welding behavior analysis module and optimization decision engine;

[0039] Welding behavior analysis module: used to monitor the welding operator's behavior pattern, including welding speed, welding angle and welding gun trajectory, and predict the optimal adjustment direction of the future welding path, and transmit the prediction results to the optimization decision engine;

[0040] Optimization decision engine: used to receive data provided by the weld quality assessment module and the welding behavior analysis module, and generate a specific welding path optimization strategy based on the weld quality assessment results and welding behavior prediction, and transmit the optimization strategy to the path automatic adjustment module;

[0041] Automatic path adjustment module: Receives the welding path optimization strategy generated by the optimization decision engine, automatically adjusts the motion parameters of the welding equipment, and performs specific adjustment operations, including welding gun angle correction, welding speed adjustment, and weld tracking compensation;

[0042] Self-learning module: Obtains adjusted welding parameters and weld quality feedback from the automatic path adjustment module, uses deep reinforcement learning technology to optimize the weld quality assessment module and optimization decision engine, and continuously improves welding path planning and optimization strategies.

[0043] The technical effects and advantages of the bridge anti-collision guardrail welding path adaptive planning method and system of the present invention are as follows:

[0044] An improved Gaussian process regression algorithm assesses welding quality in real time, accurately determining weld strength, uniformity, and defect rate. Combined with machine learning technology, it optimizes the welding path, reducing the need for manual intervention and making the welding process more intelligent and automated, thereby improving production efficiency and welding accuracy. The ability to adjust welding parameters based on real-time data enables the welding process to adapt to environmental changes and the varying behavior of operators. By automatically adjusting the motion parameters of the welding equipment, such as the welding gun angle and welding speed, it not only reduces the possibility of human error but also effectively reduces equipment wear and extends its service life. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 This is a structural diagram of the bridge anti-collision guardrail welding path adaptive planning method of the present invention.

[0046] Figure 2 Flowchart of the welding behavior prediction and optimization strategy of the present invention.

[0047] Figure 3 This is a closed-loop flowchart from data acquisition to model deployment in the present invention.

[0048] Figure 4 This is a schematic diagram of the structure of the bridge anti-collision guardrail welding path adaptive planning system of the present invention. DETAILED DESCRIPTION

[0049] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part 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 ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0050] Example 1

[0051] See also Figure 1As shown, the present invention discloses a method for adaptively planning a welding path for a bridge anti-collision guardrail, comprising the steps of:

[0052] Step S1: Using a variety of pre-set sensors, the temperature, humidity, surface condition of the welding material, and operating parameters of the welding equipment at the welding site are collected in real time. An improved Gaussian process regression algorithm is used to evaluate the weld quality and output the evaluation results of weld strength, weld uniformity, and weld defect rate.

[0053] like Figure 2 As shown, step S2 is used to monitor the behavior pattern of the welding operator, including welding speed, welding angle and welding gun trajectory, and predict the optimal adjustment direction of the future welding path to obtain welding behavior prediction, and generate a specific welding path optimization strategy based on the weld quality evaluation result and the welding behavior prediction;

[0054] Step S3: Combined with the welding path optimization strategy, the motion parameters of the welding equipment are automatically adjusted to perform specific adjustment operations, including welding gun angle correction, welding speed adjustment, and weld tracking compensation;

[0055] like Figure 3 As shown, step S4: obtain the adjusted welding parameters and weld quality feedback from the path automatic adjustment module, use deep reinforcement learning technology to optimize the weld quality assessment module and optimization decision engine, and continuously improve the welding path planning and optimization strategy.

[0056] In step S1, the temperature, humidity, surface condition of the welding material, and operating parameters of the welding equipment at the welding site are collected in real time through a variety of preset sensors. The weld quality is evaluated using an improved Gaussian process regression algorithm, and the evaluation results of weld strength, weld uniformity, and weld defect rate are output. The specific contents include:

[0057] Step A1: Real-time acquisition of temperature and humidity data at the welding site using thermocouple sensors and humidity sensors. The thermocouple sensor outputs temperature values ​​by measuring the thermoelectric potential of the welding area, and the humidity sensor outputs humidity values ​​using the capacitive sensing principle.

[0058] Step A2: Real-time detection of the surface condition of the welding material, including scanning the roughness and geometric features of the welding material surface and detecting whether there are cracks or pores inside the material;

[0059] Step A3: Monitor the operating status of the welding equipment using a current sensor and a vibration sensor. The changes in welding current are measured to determine the equipment load. The vibration sensor measures the vibration frequency of the equipment to assess the equipment stability.

[0060] Step A4: Preprocess the temperature, humidity, welding material surface condition and equipment operating parameters at the welding site, including denoising, normalization and feature extraction, to generate a feature vector V for evaluation, expressed as ; where n is the feature dimension, is the i-th eigenvalue, and the improved Gaussian process regression algorithm is used to evaluate the weld quality of the preprocessed eigenvector V. The weld quality evaluation result is calculated by the following formula: ; Where Q represents the weld quality assessment result, represents the weight, Represents the kernel function mapping of the feature vector, b represents the bias, and m represents the dimension of the feature vector;

[0061] Step A5: Split the weld quality assessment results into three parts: weld strength S, weld uniformity U, and weld defect rate D. Use a machine learning model for quantitative analysis:

[0062] Evaluation dimension splitting principle: Welding strength S represents the strength and stability of the weld, affecting the load-bearing capacity of the welded structure. Welding uniformity U reflects the uniformity of the weld. Poor weld uniformity may cause local stress concentration and affect the durability of the structure. Welding defect rate D reflects the defect rate in the weld, such as pores and cracks, which directly affects the quality and safety of welding.

[0063] Through neural networks, independent networks are used to model each indicator separately, retaining the specificity of each indicator. The weld quality assessment results are divided into three parts: welding strength S, weld uniformity U, and welding defect rate D. The mathematical model is constructed as follows: ;in, 、 、 They are respectively neural network models based on multi-layer perceptron.

[0064] In step S2, the welding operator's behavior pattern is monitored, including welding speed, welding angle, and welding gun trajectory, and the optimal adjustment direction of the future welding path is predicted to obtain welding behavior prediction. Based on the weld quality assessment results and welding behavior prediction, a specific welding path optimization strategy is generated, including:

[0065] Monitor the welding operator's welding gun angle changes, generate welding angle data by recording the angle deviation of each welding, monitor the welding operator's welding gun trajectory, generate welding gun trajectory data by recording the trajectory changes of each welding, and establish a database to store welding behavior data;

[0066] Processing the stored welding behavior data, including data cleaning, normalization, and feature extraction, generates feature vectors for path prediction, uses historical behavior data and machine learning algorithms to predict the optimal adjustment direction of the future welding path, and transmits the prediction results to the optimization decision engine;

[0067] Normalizing the preprocessed feature vectors, collecting and storing historical welding behavior data of welding operators to form a historical behavior database, wherein the historical welding behavior data includes past welding speeds, welding angles, and welding gun trajectories;

[0068] Match the current feature vector with the feature vector in the historical behavior database. Specifically, use the cosine similarity algorithm to calculate the similarity between the current feature vector and the historical feature vector. The formula is: ;in, Represents the eigenvector V and the historical eigenvector The similarity between and Denote the eigenvectors V and The i-th eigenvalue of , n represents the dimension of the eigenvector;

[0069] Use the support vector machine algorithm to train historical behavior data and build a prediction model. By maximizing the classification interval, a decision rule for predicting the welding path adjustment direction is established to predict the welding path adjustment direction: the current feature vector is input into the trained support vector machine model to predict the optimal adjustment direction of the future welding path;

[0070] Establish an optimization decision engine: Generate specific welding path optimization strategies based on weld quality assessment results and welding behavior predictions;

[0071] Analyze the welding behavior prediction results to determine the potential deviation trend and adjustment needs of the welding operator's future welding path. Based on the outputs of the evaluation analysis subunit and the behavior analysis subunit, customize the optimization strategy that adapts to the current welding status and predicted welding needs.

[0072] Based on the weld quality assessment results and welding behavior prediction results, a path adjustment plan is formulated, giving priority to path optimization in high-risk areas and complex welding sections. The formula is expressed as: ;in, represents the path adjustment resources allocated to the predetermined time period t, Indicates the risk assessment value for this time period, represents the welding demand forecast value for the time period, and T represents the total time period;

[0073] Based on the identified welding performance bottlenecks, a specific parameter optimization plan is formulated, including welding speed adjustment, welding gun angle correction and weld tracking compensation operations. The specific adjustment formula is: ,in, Indicates the adjusted welding parameters, Indicates the current welding parameters, Indicates the adjustment range;

[0074] Integrate the path adjustment strategy and parameter optimization strategy to generate a comprehensive optimization strategy.

[0075] In step S3, the motion parameters of the welding equipment are automatically adjusted in combination with the welding path optimization strategy, and specific adjustment operations are performed, including welding gun angle correction, welding speed adjustment, and weld tracking compensation. The specific contents include:

[0076] Automatically adjust the welding gun angle according to the angle correction instruction in the welding path optimization strategy: detect the current angle of the welding gun in real time, evaluate the angle deviation, perform angle correction operations, and adjust to the angle recommended by the welding path optimization strategy. The formula is expressed as: ; where α is the correction coefficient, is the current angle, is the angle compensation;

[0077] According to the speed adjustment instruction in the welding path optimization strategy, the moving speed of the welding equipment is adjusted: the current speed of the welding equipment is detected in real time, and the speed adjustment operation is performed according to the result of the speed detection subunit, and adjusted to the speed recommended by the optimization strategy. The formula is expressed as: ;in, Indicates the new position after adjustment. Indicates the current location, Indicates the position compensation amount;

[0078] According to the weld tracking instructions in the welding path optimization strategy, the weld position deviation is dynamically compensated: the weld position deviation is detected in real time, and the motion trajectory of the welding equipment is dynamically adjusted to ensure that the equipment always moves along the predetermined welding path. The formula is expressed as: ;in, Indicates the new position after adjustment. Indicates the current location, Indicates the position compensation amount.

[0079] In step S4, the adjusted welding parameters and weld quality feedback are obtained from the automatic path adjustment module. The weld quality assessment module and the optimization decision engine are optimized using deep reinforcement learning technology to continuously improve the welding path planning and optimization strategy. The specific contents include:

[0080] Obtain adjusted welding parameters and weld quality feedback, including adjusted welding gun angle, welding speed, weld position and their corresponding weld quality feedback;

[0081] The model is trained using deep reinforcement learning technology on the processed data. The input parameters of the model include the adjusted welding gun angle, welding speed, weld position and the corresponding weld quality feedback. The output is the continuously improved welding parameters or actions, and the impact of welding parameter adjustments on weld quality is learned.

[0082] Reward function design: Design a reward function to reflect the quality of welding. If the welding quality improves, the reward increases; if the quality decreases, the penalty increases. The reward function is defined based on the actual weld quality feedback. It includes: the reduction of cracks and pores on the weld surface, the uniformity of the weld appearance, and the time efficiency of the welding speed.

[0083] Model training: Using the collected data set for training, the model continuously adjusts its parameters to maximize the reward. In each training session, the model selects an action based on the current welding parameter state, that is, adjusting the welding parameters and observing the reward feedback. Through multiple iterations, the model continuously optimizes its strategy.

[0084] Through the trained model, the impact of different welding parameters on weld quality is predicted, the welding quality is evaluated, and an improved welding parameter adjustment strategy is generated. Based on the actual feedback during the welding process, the model parameters are adjusted to further improve the welding quality. Based on the feedback mechanism of the deep reinforcement learning model, the welding process is optimized in real time to ensure that each welding task can achieve the optimal quality standards.

[0085] Example 2

[0086] Reference Figure 4 The present invention discloses a bridge anti-collision guardrail welding path adaptive planning system, which includes an environmental perception module, a weld quality assessment module, a welding behavior analysis module, an optimization decision engine, a path automatic adjustment module and a self-learning module, and the modules are signal-connected.

[0087] Environmental perception module: Through a variety of preset sensors, it collects the temperature, humidity, surface condition of welding materials and operating parameters of welding equipment at the welding site in real time, and transmits the collected data to the weld quality assessment module;

[0088] Weld quality assessment module: Receives data from the environmental perception module, uses an improved Gaussian process regression algorithm to assess weld quality, outputs assessment results of weld strength, weld uniformity, and weld defect rate, and sends the assessment results to the welding behavior analysis module and optimization decision engine;

[0089] Welding behavior analysis module: used to monitor the welding operator's behavior pattern, including welding speed, welding angle and welding gun trajectory, and predict the optimal adjustment direction of the future welding path, and transmit the prediction results to the optimization decision engine;

[0090] Optimization decision engine: used to receive data provided by the weld quality assessment module and the welding behavior analysis module, and generate a specific welding path optimization strategy based on the weld quality assessment results and welding behavior prediction, and transmit the optimization strategy to the path automatic adjustment module;

[0091] Automatic path adjustment module: Receives the welding path optimization strategy generated by the optimization decision engine, automatically adjusts the motion parameters of the welding equipment, and performs specific adjustment operations, including welding gun angle correction, welding speed adjustment, and weld tracking compensation;

[0092] Self-learning module: Obtains adjusted welding parameters and weld quality feedback from the automatic path adjustment module, uses deep reinforcement learning technology to optimize the weld quality assessment module and optimization decision engine, and continuously improves welding path planning and optimization strategies.

[0093] The above formulas are all dimensionless and numerical calculations. The formulas are obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The preset parameters in the formulas are set by technicians in this field according to actual conditions.

[0094] The above embodiments may be implemented in whole or in part through software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments may be implemented in whole or in part in the form of a computer program product.

[0095] Those skilled in the art will appreciate that the modules and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application of the technical solution and the invention constraints. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0096] In addition, each functional module in each embodiment of the present application may be integrated into one processing module, or each module may exist physically separately, or two or more modules may be integrated into one module.

[0097] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

[0098] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. The adaptive planning method for welding path of bridge anti-collision guardrail is characterized by: Including steps: Step S1: Using a variety of pre-set sensors, the temperature, humidity, surface condition of the welding material, and operating parameters of the welding equipment at the welding site are collected in real time. An improved Gaussian process regression algorithm is used to evaluate the weld quality and output the evaluation results of weld strength, weld uniformity, and weld defect rate. These include: Step A1: Using a thermocouple sensor and a capacitive humidity sensor, respectively, the temperature and humidity data of the welding area are collected in real time; Step A2: Detecting the surface roughness, geometric features, and internal defects of the welding material using a surface scanning device; Step A3: Monitor the current fluctuation and vibration frequency of the welding equipment using a current sensor and a vibration sensor to assess the equipment load and stability; Step A4: De-noise, normalize, and extract features from the data collected in steps A1 to A3 to generate a multi-dimensional feature vector V, which is expressed as: ; where n is the feature dimension, is the i-th eigenvalue; Step A5: Input the eigenvector V into the improved Gaussian process regression model and calculate the comprehensive weld quality score Q using the following formula: ; Where Q represents the weld quality assessment result, represents the weight, Represents the kernel function mapping of the feature vector, b represents the bias, and m represents the dimension of the feature vector; Q is split into three sub-indicators: welding strength S, weld uniformity U, and welding defect rate D. Independent neural network models are used to model each of them: ;in, 、 、 They are neural network models based on multi-layer perceptron; Step S2: monitoring the welding operator's behavior pattern, including welding speed, welding angle, and welding gun trajectory, and predicting the optimal adjustment direction of the future welding path to obtain welding behavior prediction, and generating a specific welding path optimization strategy based on the weld quality assessment results and welding behavior prediction; Step S3: Combined with the welding path optimization strategy, the motion parameters of the welding equipment are automatically adjusted to perform specific adjustment operations, including welding gun angle correction, welding speed adjustment, and weld tracking compensation; Step S4: Obtain the adjusted welding parameters and weld quality feedback from the path automatic adjustment module, use deep reinforcement learning technology to optimize the weld quality assessment module and optimization decision engine, and continuously improve the welding path planning and optimization strategy.

2. The bridge anti-collision guardrail welding path adaptive planning method according to claim 1 is characterized in that ； Establish a historical welding behavior database to store historical data of welding speed, angle and trajectory; Compare the current feature vector V with the feature vector in the historical database Perform cosine similarity matching: ;in, Represents the eigenvector V and the historical eigenvector The similarity between and Denote the eigenvectors V and The i-th eigenvalue of , n represents the dimension of the eigenvector; A prediction model is built based on the support vector machine algorithm, which inputs the current feature vector V and outputs the future welding path adjustment direction; ;in, represents the path adjustment resources allocated to the predetermined time period t, Indicates the risk assessment value for this time period, It represents the welding demand forecast value for this time period, and T represents the total time period.

3. The bridge anti-collision guardrail welding path adaptive planning method according to claim 2 is characterized by: According to the angle correction instructions in the optimization strategy, the welding gun angle deviation Δθ is detected in real time and adjusted to the target angle ; where α is the correction coefficient, is the current angle, is the angle compensation; Calculate the current speed in real time according to the speed adjustment instruction Difference from target speed , and perform speed compensation: ;in, Indicates the new position after adjustment. Indicates the current location, Indicates the position compensation amount; Dynamically compensate for position deviation according to weld tracking instructions , generate a new welding trajectory: ;in, Indicates the new position after adjustment. Indicates the current location, Indicates the position compensation amount.

4. The bridge anti-collision guardrail welding path adaptive planning method according to claim 1 is characterized by: Obtain adjusted welding parameters and weld quality feedback, including adjusted welding gun angle, welding speed, weld position and their corresponding weld quality feedback; Deep reinforcement learning technology is used to train the model on the processed data. The input parameters of the model include the adjusted welding gun angle, welding speed, weld position and its corresponding weld quality feedback. The output is the continuously improved welding parameters or actions, and the impact of welding parameter adjustment on weld quality is learned.

5. The bridge anti-collision guardrail welding path adaptive planning method according to claim 4 is characterized by: Reward function design: Design a multi-dimensional reward function, including positive incentives, negative penalties, and a comprehensive scoring mechanism: Positive incentive items: reduction in weld defect rate and improvement in appearance uniformity; Negative penalty term: the time loss caused by the welding speed deviating from the optimal value; Comprehensive scoring mechanism: Dynamically adjust reward weights based on actual quality feedback; The welding parameters are iteratively updated through the policy network, and the welding gun angle and speed are adjusted online to maximize the cumulative reward, forming a closed-loop control.

6. A bridge guardrail welding path adaptive planning system, used to implement the bridge guardrail welding path adaptive planning method according to any one of claims 1 to 5, characterized in that: Environmental perception module: Through a variety of preset sensors, it collects the temperature, humidity, surface condition of welding materials and operating parameters of welding equipment at the welding site in real time, and transmits the collected data to the weld quality assessment module; Weld quality assessment module: Receives data from the environmental perception module, uses an improved Gaussian process regression algorithm to assess weld quality, outputs assessment results of weld strength, weld uniformity, and weld defect rate, and sends the assessment results to the welding behavior analysis module and optimization decision engine; Welding behavior analysis module: used to monitor the welding operator's behavior pattern, including welding speed, welding angle and welding gun trajectory, and predict the optimal adjustment direction of the future welding path, and transmit the prediction results to the optimization decision engine; Optimization decision engine: used to receive data provided by the weld quality assessment module and the welding behavior analysis module, and generate a specific welding path optimization strategy based on the weld quality assessment results and welding behavior prediction, and transmit the optimization strategy to the path automatic adjustment module; Automatic path adjustment module: Receives the welding path optimization strategy generated by the optimization decision engine, automatically adjusts the motion parameters of the welding equipment, and performs specific adjustment operations, including welding gun angle correction, welding speed adjustment, and weld tracking compensation; Self-learning module: Obtains adjusted welding parameters and weld quality feedback from the automatic path adjustment module, uses deep reinforcement learning technology to optimize the weld quality assessment module and optimization decision engine, and continuously improves welding path planning and optimization strategies.

Images