Online monitoring system and method based on machine vision and artificial intelligence algorithm

Abstract

The invention relates to the field of steel structure cutting monitoring analysis, in particular to an online monitoring system and method based on machine vision and an artificial intelligence algorithm. The method comprises the following steps: collecting real-time cutting monitoring videos and multi-dimensional monitoring parameters of a steel structure; carrying out adaptive filtering noise reduction and initial frame space position calculation on the steel structure real-time cutting monitoring video to generate an initial cutter position coordinate; dynamic optical flow tracking is carried out based on the initial cutter position coordinates, time sequence cutting trajectory fitting is carried out, and a multi-time-point cutting trajectory is constructed; performing steel structure thermal coupling effect evolution on the multi-dimensional monitoring parameters, and performing dynamic cutting track rendering according to a multi-time-point cutting track to construct a real-time cutting rendering model; and tool wear state prediction and steel structure thermal damage trend analysis are carried out on the real-time cutting rendering model, and a cutting abnormal fault adjustment strategy is constructed. The stability, efficiency and precision of the steel structure cutting process are improved.

Description

Technical Field

[0001] The present invention relates to the field of steel structure cutting monitoring and analysis, and in particular to an online monitoring system and method based on machine vision and artificial intelligence algorithms. Background Art

[0002] As a vital component of modern manufacturing, steel structure cutting is widely used in a variety of fields, including construction, bridges, and ships, fulfilling the critical tasks of structural processing and precision cutting. During the steel structure cutting process, cutting quality directly affects the accuracy and stability of the final product. With the continuous advancement of industrial technology, especially the development of artificial intelligence and automation, the accuracy and production efficiency of steel structure cutting have been significantly improved. However, abnormal conditions during the cutting process, such as unstable cutting speeds, temperature fluctuations, tool wear, and operational errors, can lead to fluctuations in cutting quality and even process failures, affecting overall production progress and quality.

[0003] Traditional methods for monitoring steel structure cutting rely heavily on manual inspections, periodic checks, and local testing. These methods often suffer from slow response, misjudgment, and low detection efficiency in practical applications. Especially when faced with complex cutting environments and large amounts of production data, traditional methods struggle to meet the demands of real-time monitoring and precise analysis. Therefore, to meet the demands of modern steel structure cutting for efficiency and intelligence, a new, intelligent online monitoring method for steel structure cutting is urgently needed. Summary of the Invention

[0004] In order to solve the above technical problems, the present invention proposes an online monitoring system and method based on machine vision and artificial intelligence algorithms to solve at least one of the above technical problems.

[0005] To achieve the above object, the present invention provides an online monitoring method based on machine vision and artificial intelligence algorithm, comprising the following steps: Step S1: collecting real-time steel structure cutting monitoring video and multi-dimensional monitoring parameters; performing adaptive filtering and noise reduction and initial frame spatial position calculation on the real-time steel structure cutting monitoring video to generate initial tool position coordinates; Step S2: Perform dynamic optical flow tracking based on the initial tool position coordinates, perform temporal cutting trajectory fitting, and construct a multi-time point cutting trajectory; Step S3: Evolution of the thermal coupling effect of the steel structure is performed on the multi-dimensional monitoring parameters, and dynamic cutting trajectory rendering is performed according to the multi-time point cutting trajectory to construct a real-time cutting rendering model; Step S4: Predict tool wear status and analyze steel structure thermal damage trends based on the real-time cutting rendering model, and build a cutting abnormality fault adjustment strategy; Step S5: Calculating the deviation of the cutting trajectory at each time point and making defect attribution inferences to obtain the cutting defect factors; Step S6: Fine-tune the intelligent cutting parameters according to the cutting defect factors, and perform iterative machine learning optimization based on the cutting abnormality fault adjustment strategy to build an iterative cutting optimization engine.

[0006] By collecting real-time steel structure cutting surveillance video and multi-dimensional monitoring parameters (such as temperature, vibration, and pressure), this system comprehensively covers data from all aspects of the cutting process, ensuring that critical changes at every moment are captured without omission. Real-time video data is often affected by factors such as noise and interference. Adaptive filtering and noise reduction algorithms can effectively reduce these unwanted noises, improve image quality, and enable more accurate subsequent tool positioning. By calculating the spatial position of the initial frame, the cutting tool's starting position can be accurately determined, providing a precise starting point for subsequent tool motion tracking and cutting trajectory fitting. This allows the system to accurately track the tool's entire cutting path. The optical flow algorithm accurately tracks the tool's dynamic motion based on image features between consecutive frames. Even in complex cutting environments, optical flow tracking provides stable tool position change information, avoiding errors caused by illumination variations and noise in traditional methods. By performing time-series fitting on the cutting trajectory, the cutting trajectory can be accurately reconstructed. This not only provides accurate cutting path information but also provides strong data support for subsequent anomaly detection and fault prediction. Multi-point trajectory fitting can reflect subtle changes in the cutting process, helping to identify potential anomalies. During the cutting process, heat conduction, thermal stress, and temperature fluctuations affect the physical properties of steel structures. By simulating and analyzing thermal coupling effects, material deformation, hardening, or weakening during the cutting process can be more accurately predicted, enabling a comprehensive assessment of steel cutting quality. By combining the time-series cutting trajectory with the thermal coupling effects and using dynamic rendering technology to generate a real-time virtual cutting model, this not only allows operators to observe the cutting process in real time but also effectively simulates various cutting scenarios, providing multi-faceted support for subsequent fault analysis and optimization. By analyzing the real-time cutting rendering model, the system can predict tool wear. Real-time monitoring and prediction of cutting tool wear can promptly identify tool wear issues, preventing inaccurate cutting or equipment damage caused by excessive wear. High temperatures during the cutting process can cause thermal damage to steel structures, impacting cutting results and steel quality. Thermal damage trend analysis can proactively identify thermal damage risks and enable effective intervention to ensure material integrity and quality during the cutting process. By predicting tool wear and thermal damage, the system can automatically generate troubleshooting strategies. These strategies can help operators adjust cutting parameters or replace tools in real-time to ensure production efficiency and cutting quality. By calculating cutting path deviations at multiple points in time, any deviations that occur during the cutting process can be accurately identified. This provides a scientific basis for detecting changes in cutting accuracy and can promptly identify deviations caused by equipment failure, improper operation, or material issues. By calculating cutting path deviations and combining them with multi-dimensional monitoring data, the system can automatically infer the cause of defects. Whether due to tool problems, equipment failure, or inherent material defects, the system can accurately attribute the defect and help operators quickly resolve the problem.Based on cutting defect factors, the system can automatically fine-tune cutting parameters. Through real-time analysis of deviations, it automatically adjusts parameters such as cutting speed, temperature, and pressure to ensure optimal cutting performance, reduce defect rates, and improve cutting accuracy and production efficiency. By continuously collecting and analyzing new data, the system continuously learns and optimizes its cutting control strategies. Each iterative optimization makes the cutting process more intelligent and enables more accurate identification and resolution of problems in actual operations. Machine learning optimization also enables the system to adapt to new cutting requirements and environmental changes, further improving automation and stability.

[0007] In this specification, an online monitoring system based on machine vision and artificial intelligence algorithms is provided, which is used to perform the online monitoring method based on machine vision and artificial intelligence algorithms as described above, including: A video optimization module is used to collect real-time steel structure cutting monitoring videos and multi-dimensional monitoring parameters; perform adaptive filtering and noise reduction on the real-time steel structure cutting monitoring videos and calculate the initial frame spatial position to generate initial tool position coordinates; Cutting trajectory module, which is used to perform dynamic optical flow tracking based on the initial tool position coordinates, perform time-series cutting trajectory fitting, and construct multi-time point cutting trajectories; The cutting trajectory rendering module is used to perform the evolution of the thermal coupling effect of the steel structure on the multi-dimensional monitoring parameters and perform dynamic cutting trajectory rendering based on the multi-time point cutting trajectory to build a real-time cutting rendering model; The fault adjustment module is used to predict tool wear status and analyze steel structure thermal damage trends based on the real-time cutting rendering model, and to build a cutting abnormality fault adjustment strategy; The trajectory deviation module is used to calculate the cutting trajectory deviation of multiple cutting points one by one, and make defect attribution inferences to obtain the cutting defect factors; The iterative learning module is used to fine-tune intelligent cutting parameters according to cutting defect factors, and to perform iterative machine learning optimization based on cutting abnormal fault adjustment strategies to build an iterative cutting optimization engine.

[0008] The present invention uses a video optimization module to collect real-time steel structure cutting monitoring video and multi-dimensional monitoring parameters (such as temperature and vibration). This multi-dimensional data provides comprehensive cutting information, providing rich input for subsequent analysis. Videos of the cutting process are often subject to noise, especially in strong light or vibrating environments. Using an adaptive filtering noise reduction algorithm, image noise can be effectively reduced, improving image clarity. This makes subsequent tool position identification and cutting path tracking more accurate, reducing errors caused by poor image quality. Calculating the spatial position of the initial frame in the video determines the initial tool position, providing a precise starting point for subsequent tool tracking and cutting trajectory analysis. An accurate initial position is crucial for tracking the entire cutting process and ensures efficient and stable dynamic optical flow tracking. Optical flow tracking technology tracks the tool's trajectory in real time by analyzing pixel movement between different frames in the image. Even during the cutting process due to noise or changes in ambient light, the optical flow algorithm maintains high-precision tracking, providing an accurate dynamic trajectory for subsequent cutting path analysis. Based on the initial tool position coordinates, time-series cutting trajectory fitting accurately captures the tool's motion path and gradually builds cutting trajectory models at multiple moments. This not only helps accurately reproduce the tool's motion process but also reveals potential cutting issues, such as unstable tool paths and cutting accuracy issues. This module provides cutting trajectory data at every moment, enabling the system to observe tool position changes in real time over time, thereby analyzing the accuracy and efficiency of the cutting process. During the cutting process, effects such as heat conduction and thermal stress between the tool and the steel structure significantly impact cutting quality. By analyzing multi-dimensional monitoring parameters, the evolution of thermal coupling effects can be modeled, enabling real-time prediction of issues such as temperature unevenness and thermal damage during the cutting process. Dynamic rendering of the cutting trajectory allows the system to simulate the cutting process in a virtual environment. Based on cutting trajectory data at multiple points in time, the dynamic rendering module constructs a real-time cutting rendering model and displays changes in temperature, stress, and other parameters during the cutting process. This helps operators fully understand the cutting process, identify potential issues, and adjust operations promptly. Combining thermal coupling effects with cutting trajectory rendering, the real-time cutting rendering model provides operators with a visual representation of the cutting process. The virtual model accurately displays the cutting path and its impact, facilitating immediate decision-making and fault analysis. Tool wear is an inevitable problem during steel cutting, but this module can predict tool wear status by analyzing real-time cutting rendering models. By monitoring parameters such as cutting trajectory, cutting speed, and pressure, wear trends can be identified in advance, avoiding reduced cutting quality or equipment damage caused by tool wear. Thermal damage to steel structures during cutting affects material properties. By analyzing thermal damage trends in this module, the system can identify problems such as uneven heat conduction or localized overheating. Early identification of these issues can help adjust the cutting process, thereby reducing the negative impact of thermal damage.Based on wear prediction and thermal damage analysis results, the system generates real-time adjustment strategies for cutting anomalies. These strategies address tool replacement, cutting parameter adjustments, and cooling system optimization, providing operators with scientific and timely adjustment recommendations. By calculating deviations from the cutting trajectory at each point in time, the system can identify tool deviations from the planned path in real time. These deviations are caused by equipment failure, material issues, or operator error. Prompt identification of deviations can prevent further problems. Combining these deviations with sensor data, the system can infer the cause of the defect. For example, if the deviation is excessive, tool wear or uneven material may be the cause. The accuracy of defect attribution provides key clues for subsequent troubleshooting. This module helps identify potential defect factors during the cutting process and reduce errors caused by environmental, operational, or equipment issues. It also helps improve cutting accuracy and stability. Based on the factors causing the cutting defect, the system automatically adjusts cutting parameters (such as cutting speed, pressure, and temperature) to optimize the cutting process. Fine-tuning allows for targeted adjustments to ensure optimal cutting performance. Through an iterative learning module, the system continuously optimizes the cutting process based on historical data and real-time feedback. Machine learning optimization enables the system to adapt to different cutting environments and working conditions. Through continuous learning and adjustment, it improves the stability, efficiency, and precision of the cutting process. Through continuous feedback and optimization, the system ultimately establishes a highly efficient iterative optimization engine. This engine continuously adjusts and optimizes cutting parameters during the production process to ensure cutting quality and production efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] Figure 1 This is a flow chart of the steps of an online monitoring method based on machine vision and artificial intelligence algorithms of the present invention; Figure 2 Detailed implementation flow chart of step S1; Figure 3 Detailed implementation flow chart of step S2; Figure 4 Schematic diagram of the detailed implementation steps of step S3. DETAILED DESCRIPTION

[0010] It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0011] This application provides an online monitoring system and method based on machine vision and artificial intelligence algorithms. The execution entities of the online monitoring system and method based on machine vision and artificial intelligence algorithms include, but are not limited to, the following: mechanical equipment, data processing platforms, cloud server nodes, network upload devices, etc. that are equipped with the system, which can be regarded as general computing nodes of this application. The data processing platform includes, but is not limited to, at least one of an audio and image management system, an information management system, and a cloud data management system.

[0012] See also Figures 1 to 4 The present invention provides an online monitoring method based on machine vision and artificial intelligence algorithm, which includes the following steps: Step S1: collecting real-time steel structure cutting monitoring video and multi-dimensional monitoring parameters; performing adaptive filtering and noise reduction and initial frame spatial position calculation on the real-time steel structure cutting monitoring video to generate initial tool position coordinates; Step S2: Perform dynamic optical flow tracking based on the initial tool position coordinates, perform temporal cutting trajectory fitting, and construct a multi-time point cutting trajectory; Step S3: Evolution of the thermal coupling effect of the steel structure is performed on the multi-dimensional monitoring parameters, and dynamic cutting trajectory rendering is performed according to the multi-time point cutting trajectory to construct a real-time cutting rendering model; Step S4: Predict tool wear status and analyze steel structure thermal damage trends based on the real-time cutting rendering model, and build a cutting abnormality fault adjustment strategy; Step S5: Calculating the deviation of the cutting trajectory at each time point and making defect attribution inferences to obtain the cutting defect factors; Step S6: Fine-tune the intelligent cutting parameters according to the cutting defect factors, and perform iterative machine learning optimization based on the cutting abnormality fault adjustment strategy to build an iterative cutting optimization engine.

[0013] By collecting real-time steel structure cutting surveillance video and multi-dimensional monitoring parameters (such as temperature, vibration, and pressure), this system comprehensively covers data from all aspects of the cutting process, ensuring that critical changes at every moment are captured without omission. Real-time video data is often affected by factors such as noise and interference. Adaptive filtering and noise reduction algorithms can effectively reduce these unwanted noises, improve image quality, and enable more accurate subsequent tool positioning. By calculating the spatial position of the initial frame, the cutting tool's starting position can be accurately determined, providing a precise starting point for subsequent tool motion tracking and cutting trajectory fitting. This allows the system to accurately track the tool's entire cutting path. The optical flow algorithm accurately tracks the tool's dynamic motion based on image features between consecutive frames. Even in complex cutting environments, optical flow tracking provides stable tool position change information, avoiding errors caused by illumination variations and noise in traditional methods. By performing time-series fitting on the cutting trajectory, the cutting trajectory can be accurately reconstructed. This not only provides accurate cutting path information but also provides strong data support for subsequent anomaly detection and fault prediction. Multi-point trajectory fitting can reflect subtle changes in the cutting process, helping to identify potential anomalies. During the cutting process, heat conduction, thermal stress, and temperature fluctuations affect the physical properties of steel structures. By simulating and analyzing thermal coupling effects, material deformation, hardening, or weakening during the cutting process can be more accurately predicted, enabling a comprehensive assessment of steel cutting quality. By combining the time-series cutting trajectory with the thermal coupling effects and using dynamic rendering technology to generate a real-time virtual cutting model, this not only allows operators to observe the cutting process in real time but also effectively simulates various cutting scenarios, providing multi-faceted support for subsequent fault analysis and optimization. By analyzing the real-time cutting rendering model, the system can predict tool wear. Real-time monitoring and prediction of cutting tool wear can promptly identify tool wear issues, preventing inaccurate cutting or equipment damage caused by excessive wear. High temperatures during the cutting process can cause thermal damage to steel structures, impacting cutting results and steel quality. Thermal damage trend analysis can proactively identify thermal damage risks and enable effective intervention to ensure material integrity and quality during the cutting process. By predicting tool wear and thermal damage, the system can automatically generate troubleshooting strategies. These strategies can help operators adjust cutting parameters or replace tools in real-time to ensure production efficiency and cutting quality. By calculating cutting path deviations at multiple points in time, any deviations that occur during the cutting process can be accurately identified. This provides a scientific basis for detecting changes in cutting accuracy and can promptly identify deviations caused by equipment failure, improper operation, or material issues. By calculating cutting path deviations and combining them with multi-dimensional monitoring data, the system can automatically infer the cause of defects. Whether due to tool problems, equipment failure, or inherent material defects, the system can accurately attribute the defect and help operators quickly resolve the problem.Based on cutting defect factors, the system can automatically fine-tune cutting parameters. Through real-time analysis of deviations, it automatically adjusts parameters such as cutting speed, temperature, and pressure to ensure optimal cutting performance, reduce defect rates, and improve cutting accuracy and production efficiency. By continuously collecting and analyzing new data, the system continuously learns and optimizes its cutting control strategies. Each iterative optimization makes the cutting process more intelligent and enables more accurate identification and resolution of problems in actual operations. Machine learning optimization also enables the system to adapt to new cutting requirements and environmental changes, further improving automation and stability.

[0014] In the embodiment of the present invention, see Figure 1 , is a schematic flow chart of the steps of an online monitoring method based on machine vision and artificial intelligence algorithm of the present invention. In this example, the steps of the online monitoring method based on machine vision and artificial intelligence algorithm include: Step S1: collecting real-time steel structure cutting monitoring video and multi-dimensional monitoring parameters; performing adaptive filtering and noise reduction and initial frame spatial position calculation on the real-time steel structure cutting monitoring video to generate initial tool position coordinates; In this embodiment, a high-definition camera system is configured to ensure sufficient camera resolution (e.g., 1080p or higher) to clearly capture details during the cutting process. Furthermore, the camera should have a good dynamic range to accommodate varying lighting conditions. The camera's mounting position should generally be selected at an angle that provides full coverage of the cutting area, ensuring that the contact area between the tool and the steel remains within the camera's field of view. Multiple cameras can be used for multi-view monitoring. The frame rate for video capture should be set; 30 frames per second or higher is recommended to capture subtle changes during the rapid cutting process. Ensure that the video capture software is capable of recording and saving data in real time. During the cutting process, video data is recorded in real time to ensure the integrity of each cutting operation. Data files are generated and timestamped for subsequent analysis. Appropriate sensors are selected to monitor multiple parameters during the cutting process, including a temperature sensor (for monitoring cutting temperature), a vibration sensor (for monitoring tool vibration), and a force sensor (for monitoring cutting force). Ensure that the accuracy and sensitivity of each sensor meet the experimental requirements. For example, a temperature sensor should have a measurement accuracy of ±1°C, and a vibration sensor should be able to detect acceleration changes of 1g. During the cutting process, data from each sensor is collected in real time and time-synchronized with the video data. The acquisition frequency is set, typically consistent with the video frame rate (e.g., collecting data once per second) to ensure data consistency. The collected multi-dimensional monitoring parameters are stored in a database, recording the sensor readings and their time information at each time point to form a multi-dimensional monitoring data table. The collected real-time cutting surveillance video undergoes preprocessing, primarily involving denoising and image quality enhancement. Adaptive filtering algorithms (such as mean filtering or Gaussian filtering) are used to reduce noise caused by low light or rapid motion. Filter parameters, such as the filter window size, are set to ensure optimal image processing results. Typically, the window size is set to 3x3 or 5x5, but this can be adjusted based on actual conditions. After denoising, the signal-to-noise ratio (SNR) is calculated to evaluate the processing effectiveness. A threshold for the SNR (e.g., SNR ≥ 20 dB) is set to determine whether the noise reduction has met the target. If the noise reduction is unsatisfactory, the filtering parameters are adjusted and reprocessing is performed until satisfactory results are achieved. The first frame of the denoised cutting surveillance video is selected as the initial frame. Ensure that this image frame is clear and contains the complete outline of the tool for subsequent spatial position calculations. Use computer vision techniques (such as edge detection algorithms) to extract the tool's outline information to ensure accurate identification of the tool's position. Calculate the tool's spatial position coordinates using image processing algorithms (such as the Hough transform). Set a coordinate system to facilitate subsequent data analysis, typically using a Cartesian coordinate system. Record the tool's coordinate position (e.g., X, Y) in the initial frame and associate this position with real-time monitoring parameters and store it in a database to form a table of the tool's initial position data.

[0015] Step S2: Perform dynamic optical flow tracking based on the initial tool position coordinates, perform temporal cutting trajectory fitting, and construct a multi-time point cutting trajectory; In this embodiment, an appropriate optical flow tracking algorithm, such as the Lucas-Kanade optical flow method or the Farneback optical flow method, is selected. These algorithms are suitable for motion estimation between video frames and can effectively capture the dynamic changes of the tool during the cutting process. Algorithm parameters, such as the window size and the number of optical flow calculation iterations, are set to ensure stability and accuracy under different cutting conditions. Consecutive frames are extracted from the previously denoised cutting surveillance video, ensuring a consistent time interval between frames (e.g., 30 frames per second) to facilitate subsequent optical flow tracking. The tool's initial position coordinates are marked in the initial frame as the starting point for optical flow tracking. This position is clearly marked so that the subsequent algorithm can accurately identify it. Optical flow is calculated for the extracted consecutive frames, and the selected optical flow algorithm is used to estimate the tool's motion in each frame. The algorithm calculates the motion vector of each pixel between consecutive frames, focusing on changes in the tool area. The motion vectors for each frame are recorded to form a motion dataset for subsequent analysis. A motion vector threshold is set to filter out invalid motion below the set value (e.g., motion vectors less than 1 pixel). The calculated motion vectors are verified to ensure that they accurately reflect the dynamic changes of the tool. The tool's trajectory during the cutting process can be verified by comparing it with actual video observations. Key verification parameters, such as the accuracy and stability of the optical flow calculation, should be recorded to ensure that the tracking results meet expectations. Tool motion data obtained from optical flow tracking should be collected and organized, including the tool position coordinates (X, Y) and corresponding timestamps at each time point. Data timeliness and integrity should be ensured to prevent data loss from impacting subsequent analysis. Data windows can be set (e.g., every five frames as a window) to facilitate segmented processing and analysis of the cutting trajectory. Appropriate fitting methods, such as polynomial fitting or spline interpolation, should be selected to fit the tool's cutting trajectory. These methods can generate a smooth cutting trajectory based on the collected tool position data. The fitting order should be set (e.g., quadratic or cubic polynomial) to ensure that the fitting results accurately reflect the tool's trajectory. Trajectory fitting should be performed on the collected tool position data to generate the corresponding cutting trajectory model. Key fitting parameters, such as the fitting error (e.g., root mean square error (RMSE)) and the R² value, should be recorded to evaluate the fitting performance. Generate a visualization of the fitted cutting trajectory to help visualize the tool's motion trajectory and compare it with the actual trajectory to ensure the rationality of the fitting results. Integrate the fitted cutting trajectories from each time period to form a complete multi-time point cutting trajectory dataset. Ensure that the dataset contains tool position and motion information at each time point. Sorting the trajectories according to timestamps ensures data continuity and time sequence, facilitating subsequent analysis and processing. Store the constructed multi-time point cutting trajectory in a database to form a cutting trajectory data table, recording the tool position, motion status, and related monitoring parameters at each time point.Use visualization tools to generate multi-time point cutting trajectory diagrams to help analyze the tool's motion path and dynamic changes during the cutting process.

[0016] Step S3: Evolution of the thermal coupling effect of the steel structure is performed on the multi-dimensional monitoring parameters, and dynamic cutting trajectory rendering is performed according to the multi-time point cutting trajectory to construct a real-time cutting rendering model; In this embodiment, data is integrated from the multi-dimensional monitoring parameters (such as temperature, vibration, cutting force, etc.) obtained in the previous steps to ensure that all parameters are aligned on the same timeline to form a unified data set. Each data point should include a timestamp and its corresponding monitoring value. The temperature data is preprocessed, and noise is reduced using methods such as moving average or weighted average filtering to ensure data continuity and stability. Filter parameters are set, such as a window size of 5 data points, to ensure smoothness. Based on the collected monitoring parameters, a thermal coupling effect model is constructed, taking into account factors such as thermal conduction, thermal convection, and thermal radiation of the material during the cutting process. The model's physical parameters, such as thermal conductivity, specific heat capacity, and density, are set to ensure model accuracy. The thermal coupling effect is simulated using finite element analysis (FEA) methods, and the analysis grid size (such as a grid resolution of 1 mm) is set to improve calculation accuracy. The temperature distribution and heat flux density are recorded at each time point. The thermal coupling model is run to simulate the temperature evolution of the steel structure during the cutting process. The temperature distribution diagram at key moments is recorded to analyze the thermal response under different cutting conditions. Evaluate the results of thermal coupling effects, such as the maximum temperature, temperature gradient, and degree of thermal damage in the cutting area, to ensure accurate reflection of thermal behavior during the cutting process. Using the multi-point cutting trajectory data generated in the previous step, ensure that the tool position coordinates and corresponding monitoring parameters (such as temperature) are compiled at each time point. Set an update interval (e.g., once per second) to facilitate real-time rendering. Discretize the cutting trajectory, dividing it into multiple segments to achieve smooth dynamic rendering during rendering. Use computer graphics techniques to construct a dynamic cutting rendering model and select a suitable rendering engine (such as OpenGL or Unity3D) for real-time display. Ensure that the model reflects the tool motion and thermal coupling effects during the cutting process. In the rendered model, integrate the tool motion trajectory, the thermal distribution of the cutting area, and the material properties. Set visual effects parameters such as lighting, material, and transparency to enhance the model's realism. During the rendering process, update the tool position and status in real time, and use interpolation algorithms (such as linear interpolation or Bezier curve interpolation) to smooth the tool trajectory transitions and ensure a smooth and natural rendering effect. Based on the evolution of the thermal coupling effect, the color or material of the cutting area is dynamically changed to reflect temperature changes and the degree of thermal damage. For example, when the temperature exceeds the set threshold, the color of the cutting area gradually changes from blue to red. The dynamic cutting rendering model is integrated with the results of the thermal coupling effect analysis to ensure that the model can simultaneously display the dynamic effects of tool movement and material thermal changes. The model is tested under different cutting conditions, and the frame rate of the program operation (such as maintaining it above 30 frames per second) and the rendering delay time (such as controlling it within 50 milliseconds) are recorded to ensure real-time and smoothness. The rendering results are verified and compared with observations of the actual cutting process to ensure the accuracy and authenticity of the model. Any inconsistencies that arise during the verification process are recorded and necessary adjustments are made.Optimize rendering parameters and algorithms to improve rendering performance and quality, such as reducing unnecessary calculations and improving graphics processing efficiency.

[0017] Step S4: Predict tool wear status and analyze steel structure thermal damage trends based on the real-time cutting rendering model, and build a cutting abnormality fault adjustment strategy; In this embodiment, characteristic parameters related to tool wear, including cutting force, cutting speed, tool temperature, and cutting time, are extracted from previous monitoring data. Data integrity and continuity are ensured, and relevant parameters are recorded at each time point. A feature extraction method is established, such as using statistical features (mean, standard deviation, peak value, etc.) and frequency domain features (Fourier transform) to extract tool wear characteristics. These features should effectively reflect changes in tool performance during the cutting process. An appropriate machine learning algorithm (such as a random forest, support vector machine, or neural network) is selected to construct a tool wear prediction model. The model's input parameters are set as the extracted features, and its output parameter is the tool wear state (e.g., degree of wear). The model is trained and validated using historical datasets, with training parameters (e.g., learning rate, number of iterations, etc.) set to ensure model accuracy and stability. Cross-validation can be used to evaluate the model's performance on different datasets. The real-time monitoring parameters are input into the trained wear prediction model to perform real-time predictions of the tool wear state. The prediction results, including degree of wear and prediction confidence, are recorded. Analyze the wear prediction results and set warning thresholds (e.g., triggering an alert when the wear exceeds 20%) to facilitate timely maintenance measures. Collect temperature and heat flux data from the steel structure during the cutting process from the output of the thermal coupling model, ensuring data timeliness and integrity. Record temperature changes and heat conduction at each time point. Set an analysis window (e.g., analyze data every 10 seconds), summarize temperature changes within each window, and calculate key statistical indicators (such as average temperature, maximum temperature, and temperature fluctuation). Based on the collected thermal damage data, select an appropriate trend analysis model (such as time series analysis or regression analysis) to analyze the impact of temperature changes on thermal damage to the steel structure. Set model parameters and train and validate using historical data to ensure the model accurately reflects thermal damage trends. Record evaluation metrics (such as root mean square error (RMSE) and coefficient of determination (R²)) to assess its reliability. Use the constructed thermal damage analysis model to perform trend predictions on real-time monitored temperature data. Record the prediction results and analyze the thermal damage trends of the steel structure during the cutting process. Compare the prediction results with actual monitoring data to verify the accuracy of the model and adjust and optimize the model as necessary. Design a framework for adjusting cutting failures based on tool wear prediction results and thermal damage trend analysis. Set goals for the adjustment strategy, such as reducing tool wear rate and thermal damage. Determine the adjustment range for cutting parameters. For example, when tool wear exceeds a set threshold, reduce the cutting speed or adjust the cutting depth to reduce the wear rate. Implement the adjustment strategy during the actual cutting process and monitor the cutting results. Record changes in key parameters during implementation, such as cutting force, temperature, and tool status, to verify the effectiveness of the strategy. Evaluate the effectiveness of the adjustment strategy by comparing the cutting results before and after implementation (such as changes in tool wear rate and reduction in thermal damage).Statistical methods such as t-tests can be used to analyze the results.

[0018] Step S5: Calculating the deviation of the cutting trajectory at each time point and making defect attribution inferences to obtain the cutting defect factors; In this embodiment, deviation values ​​are calculated for each time point along the cutting trajectory. This process requires traversing all time points and using a predefined deviation calculation formula to generate deviation data for each time point. The deviation values ​​and their corresponding timestamps are recorded to form a cutting trajectory deviation data table. A deviation threshold (e.g., ±5 mm) is set to facilitate subsequent defect identification and analysis. For each time point along the cutting trajectory, deviation values ​​are calculated for each time point. This process requires traversing all time points and using a predefined deviation calculation formula to generate deviation data for each time point. The deviation values ​​and their corresponding timestamps are recorded to form a cutting trajectory deviation data table. A deviation threshold (e.g., ±5 mm) is set to facilitate subsequent defect identification and analysis. Before inferring defect attribution, characteristic data related to the cutting process is collected, including cutting parameters (e.g., cutting speed, cutting depth), environmental conditions (e.g., temperature, humidity), and tool status (e.g., wear and tool material). This characteristic data is correlated with the deviation data to provide necessary information support for subsequent defect analysis. A defect attribution analysis model is constructed based on the collected deviation values ​​and characteristic data. Select appropriate statistical analysis methods (such as logistic regression, decision trees, or random forests) to identify the primary factors influencing cutting defects. The model's input parameters are set as deviation values ​​and relevant features, and the output parameters are the defect category or attribution result. During model training, use historical datasets for training and validation to ensure model accuracy. Input real-time monitoring data into the trained attribution analysis model to infer defect attribution. Record the defect inference results at each time point, including the factors contributing to the defect and their corresponding weights. Analyze the inference results to identify the primary factors contributing to cutting defects. For example, if the inference results indicate that tool wear is the primary factor causing cutting deviation, further analysis and treatment can be performed based on the wear condition. Verify the defect attribution results by combining observations and records from the actual cutting process to ensure their rationality and accuracy. This verification can be performed through expert review or field observations. Record key data and findings during the verification process to ensure transparency and traceability of the attribution analysis.

[0019] Step S6: Fine-tune the intelligent cutting parameters according to the cutting defect factors, and perform iterative machine learning optimization based on the cutting abnormality fault adjustment strategy to build an iterative cutting optimization engine.

[0020] In this embodiment, based on the cutting defect factors identified in the previous step, the specific impact of each factor on cutting quality is analyzed. For example, if tool wear is identified as the primary defect factor, the specific impact of the wear degree on cutting performance (such as cutting accuracy and surface quality) needs to be analyzed. Historical cutting data is collected to correlate the defect factors with cutting parameters (such as cutting speed, cutting depth, and feed rate). A data model is established to determine which cutting parameters need to be adjusted based on the defect factors. Based on the defect factor analysis results, a fine-tuning strategy for the cutting parameters is set. If it is found that excessive cutting speed causes cutting deviation, the cutting speed can be reduced (for example, by 10%) when the wear level exceeds a certain threshold. The fine-tuning range and step size are set, such as a cutting depth adjustment range of ±2 mm and a cutting speed adjustment range of ±10%. Ensure that the fine-tuning strategy can minimize tool wear and thermal damage while ensuring cutting quality. The fine-tuning strategy is implemented during the actual cutting process, and the cutting performance is monitored in real time. Key parameters (such as cutting force, temperature, and tool status) and cutting quality indicators (such as cutting accuracy and surface finish) are recorded for each cut. Evaluate the results, comparing the cutting performance before and after fine-tuning. If the cutting quality improves significantly after fine-tuning (e.g., reduced deviation and wear), document the successful fine-tuning strategy and incorporate it into the standard operating procedure for subsequent cutting processes. Collect new data obtained during the fine-tuning process, including changes in cutting parameters, feedback on cutting quality, and real-time monitoring data of defect factors. Ensure the timeliness and integrity of the data to facilitate subsequent machine learning training. Clean and organize the collected data to remove outliers and noise to ensure dataset quality. Statistical analysis methods (such as Z-score normalization) can be used to identify and address anomalous data. Select an appropriate machine learning algorithm (such as random forest, gradient boosting tree, or neural network) and build an iterative learning model. Set the model's input parameters to include cutting parameters, monitoring data, and defect factors, and the output parameters to include cutting quality indicators (such as cutting accuracy and surface quality). Train and validate the model using collected historical data. Set training parameters (such as the learning rate and number of iterations) to ensure model accuracy and stability. After each cutting task, update and optimize the trained model using new data. Record the model's performance on different datasets and evaluate its generalization capabilities through cross-validation. Evaluate the model's predictive performance, using metrics such as root mean square error (RMSE) and coefficient of determination (R²) to measure model accuracy. If the model's predictions are poor, adjust model parameters or reselect features. Integrate intelligent cutting parameter fine-tuning strategies with iterative learning models to build an iterative cutting optimization engine. The engine should be able to receive real-time monitoring data from the cutting process and automatically analyze and adjust cutting parameters. Define the engine's workflow, including data input, parameter adjustment, effect evaluation, and feedback mechanisms, to ensure the automation and intelligence of the entire cutting process.During the cutting process, the optimization engine dynamically adjusts cutting parameters. Based on real-time monitoring data and prediction results, the engine automatically fine-tunes parameters and monitors cutting performance in real time. Each parameter change and cutting result are recorded for subsequent analysis and optimization. Optimization reports are regularly generated, summarizing the engine's operating status, parameter adjustment records, and cutting performance analysis. A feedback mechanism is established to continuously feed monitoring data and optimization results during the cutting process back to the iterative learning model to promote continuous learning and optimization. The optimization engine's performance is regularly evaluated, and necessary upgrades and improvements are implemented to ensure its long-term stability and effectiveness. This completes a comprehensive cutting optimization system to improve cutting quality and efficiency.

[0021] In this embodiment, refer to Figure 2 , is a flowchart of the detailed implementation steps of step S1. In this embodiment, the detailed implementation steps of step S1 include: Collect real-time cutting monitoring videos and multi-dimensional monitoring parameters of steel structures; Adaptively filtering and denoising the real-time cutting monitoring video of the steel structure, thereby constructing a filtered and denoised cutting monitoring video; Perform visual recognition of cutting tools on filtered and noise-reduced cutting surveillance videos and mark cutting tools; Perform initial frame space position calculation on the cutting tool to generate initial tool position coordinates.

[0022] In this embodiment, a high-resolution camera (such as a 1080p or 4K camera) is configured to ensure clear capture of every detail during the cutting process. The camera should be installed at an optimal viewing angle, providing a comprehensive view of the cutting area and ensuring no blind spots. In addition to video monitoring, multi-dimensional monitoring sensors, including temperature sensors, vibration sensors, and laser rangefinders, are integrated to record temperature changes, vibration intensity, and cutting depth in real time during the cutting process. The sampling rate of these sensors should be set to 10 times per second to capture rapidly changing dynamic parameters. The monitoring system is activated to begin real-time acquisition of cutting video and sensor data. The video stream should use a compressed encoding scheme (such as H.264 or H.265) to reduce storage space, maintain video quality, and ensure smooth data transmission. The captured monitoring video and multi-dimensional monitoring parameters are stored on a local server or in a cloud database to ensure data security and accessibility. The data storage format should be configured to ensure that the video and monitoring parameters are stored synchronously for easy subsequent analysis. The collected video and sensor data should be regularly checked for integrity and accuracy. Quality inspection standards should be established, such as a video frame loss rate of less than 1% and sensor readings within a reasonable range. Before processing captured surveillance video, first decompose the video into consecutive frames. Use video processing software (such as OpenCV) to extract each frame for subsequent processing. The frame rate should be maintained at the same rate as the original video to ensure smooth video playback. Perform a preliminary inspection of each frame to determine its clarity and stability. Mark frames with severe blur or jitter to provide a basis for subsequent noise reduction. Select an appropriate adaptive filtering algorithm (such as median filtering or adaptive Gaussian filtering) for video noise reduction. Median filtering is particularly suitable for removing salt-and-pepper noise, while adaptive Gaussian filtering is more effective at handling complex noise. Filter parameters, such as the filter window size (e.g., 3x3 or 5x5), should be set to accommodate varying levels of noise interference. Parameter selection should be based on a preliminary analysis of the noise characteristics of the cut video. Apply the adaptive filtering algorithm to each frame for noise reduction, removing background noise and unnecessary interference. The processed video frames should maintain clarity while removing most noise. After processing, verify the noise reduction results to ensure that the filtered frames have significantly improved quality. Record the processing time and effectiveness of each frame for subsequent evaluation. Before performing visual recognition of cutting tools, a labeled training dataset is required. Collect images of different types of cutting tools, annotated with their location and category. This dataset should contain at least 1,000 images to ensure effective model training. Select an appropriate visual recognition algorithm (such as YOLO, SSD, or Faster R-CNN) to train the tool recognition model, using appropriate hyperparameter settings (such as learning rate and batch size) to optimize model performance. Filtered and denoised surveillance video frames are then fed into the trained tool recognition model for real-time tool recognition.The model automatically analyzes each frame, identifies the location of the cutting tool, and marks its bounding box. A confidence threshold (e.g., 0.5) is set to ensure that only recognition results with a confidence level above this threshold are recorded to reduce false positives. The tool recognition results for each frame are recorded, including tool type, location, and confidence level. This provides basic data for subsequent analysis and optimization. The accuracy of the tool recognition model is regularly evaluated, and cases of false positives and missed positives are collected for iterative model training and optimization. Based on the tool recognition results, the bounding box coordinates of the tool in each frame are extracted. The 2D coordinates (x, y) need to be converted to 3D coordinates (x, y, z) for more accurate position calculation. The camera's intrinsic and extrinsic parameters are set to facilitate coordinate conversion and depth calculation. Intrinsic parameters include focal length and principal point coordinates, while extrinsic parameters include camera position and orientation. Depth information is acquired in each frame using a depth sensor or stereo vision system. Combined with the identified tool bounding box coordinates, the tool's 3D position is calculated. The accuracy of the depth information should be within ±1 cm to ensure accurate position calculation. Record the initial tool position coordinates for each calculation, including timestamp, coordinate values, and related frame information, for subsequent monitoring and analysis. Verify the calculated tool position coordinates to ensure they are within a reasonable range. Set valid position ranges (e.g., tool height must be within a specific range) to filter out unreasonable data. Store valid tool position coordinates in a database, synchronized with surveillance video and cutting parameters, to ensure data integrity and traceability.

[0023] In this embodiment, the specific steps of performing adaptive filtering and noise reduction on the real-time steel structure cutting monitoring video to construct the filtered and noise-reduced cutting monitoring video are as follows: Performing frame-by-frame image delay calculation on the real-time steel structure cutting monitoring video and extracting all inter-frame delay values; Calculating an average of the inter-frame delay values; Performing delay mutation abnormal video frame detection on the real-time steel structure cutting monitoring video according to the average value, and extracting the delay abnormal video frame; Perform inter-frame delay interpolation optimization on the abnormally delayed video frames to obtain inter-frame delay reconstructed surveillance video; Perform overbrightness analysis on the steel cutting area of ​​the inter-frame delayed reconstructed surveillance video and extract the overbrightness area of ​​the steel cutting; Calculate the pixel value of the surrounding area based on the bright area of ​​steel cutting; Performing brightness neutralization processing based on the pixel values ​​of the surrounding area to obtain an overbrightness neutralization monitoring video; Adaptive filtering and denoising is performed on overly bright and neutral surveillance videos, thereby constructing filtered and denoised cut surveillance videos.

[0024] In this embodiment, real-time segmented surveillance video is decomposed into individual image frames. A frame extraction frequency is set (for example, 30 frames per second) to ensure that dynamic changes in the segmentation can be captured during processing. Each frame should be saved in the same format (such as JPEG or PNG) for subsequent analysis. The timestamp of each frame is recorded to ensure inter-frame delay can be calculated. Timestamp accuracy should be in the millisecond range to accurately measure delay. Time differences are calculated between adjacent image frames to obtain the delay value for each pair of adjacent frames. The delay value calculation formula is: Delay value = current frame timestamp - previous frame timestamp. All calculated delay values ​​are recorded in a table. To ensure accuracy, a valid range for delay values ​​is set (for example, delay values ​​should be between 0 and 1000 milliseconds), and outliers (such as negative or extreme values) are filtered out. Statistical analysis is performed on the stored inter-frame delay values ​​to calculate their average. The statistical method is set to the arithmetic mean, i.e., average delay = ∑(delay value) / N, where N is the total number of delay values. Key parameters of the statistical process, such as the maximum, minimum, and standard deviation of the delay values, are recorded to facilitate subsequent anomaly detection analysis. To better understand the distribution of latency values, you can plot a latency histogram or box plot. This visualization helps identify central trends and outliers in latency values, facilitating subsequent anomaly detection. Mark the calculated average value on the chart for subsequent comparison with other latency values. Set anomaly detection criteria based on the average latency value. For example, you can set a threshold (such as 1.5 times the average latency value) as a reference for abnormal latency. Any latency value exceeding this threshold is considered an anomaly. Record the calculation process of the anomaly threshold for subsequent analysis and adjustment. Iterate through all latency values ​​and identify and extract frames with latency values ​​exceeding the set threshold. These frames are considered to have abnormal latency. Record the timestamps and latency values ​​of abnormal frames for subsequent analysis and optimization. Detected abnormal latency frames are stored in a separate folder and tagged with information including the anomaly type, timestamp, and latency value for subsequent review and analysis. Select an appropriate interpolation algorithm (such as linear, bilinear, or cubic interpolation) to process abnormal latency frames. Set the interpolation method based on the actual situation to ensure smoothness and continuity of the video. Record the basis for selecting the interpolation algorithm, such as the dynamic characteristics of the video content and the expected interpolation effect. For the identified delayed abnormal video frames, apply the selected interpolation algorithm to generate intermediate frames to fill the time gap. The interpolation process between each pair of abnormal frames should ensure that the generated frames are visually consistent with the style of the original video. Record each step of the interpolation process, including the original abnormal frame, the new frame after interpolation and its timestamp, to ensure the integrity of subsequent data. Merge the new frame after interpolation with the original video frame to generate an inter-frame delayed reconstructed surveillance video. Ensure that the frame rate of the new video is consistent with the original video (such as 30 frames / second) and maintain the smoothness of the video.Perform a quality check on the reconstructed video to ensure there are no noticeable artifacts or artifacts. Record the reconstructed video parameters, such as the total number of frames and duration. Select an appropriate image processing algorithm (such as threshold segmentation or adaptive histogram equalization) to perform brightness analysis on the reconstructed surveillance video. Set a brightness threshold (e.g., a brightness value greater than 200) to identify overly bright areas. Record the brightness analysis parameter settings for subsequent analysis and adjustment. Iterate through each frame of the reconstructed surveillance video and apply the selected brightness analysis algorithm to extract overly bright areas within the steel cutting process. These areas indicate abnormalities during the cutting process. Record the location and brightness value of each overly bright area for subsequent analysis. Store the extracted overly bright areas in a database and tag them with the frame number, location coordinates, and brightness value for subsequent review and analysis. Determine the surrounding area of ​​the overly bright area. This can be set to pixels within a certain radius (e.g., 5 pixels) around the overly bright area. This step facilitates the calculation of the average brightness value of the surrounding area. Record the parameters defining the surrounding area for subsequent processing. For each extracted overly bright area, calculate the pixel value (e.g., RGB average) of its surrounding area. Calculate the average brightness and color values ​​by traversing all pixels in the surrounding area. Record the values ​​of the surrounding pixels for each overbright area for subsequent brightness neutralization processing. Select an appropriate brightness neutralization algorithm (such as gain control or gamma correction) to address the overbright areas. Set neutralization parameters (such as target brightness) to ensure that the brightness of overbright areas returns to a reasonable range. Record the basis for selecting the neutralization algorithm and the parameter settings for subsequent analysis and adjustment. Apply the selected brightness neutralization algorithm to each overbright area, adjusting the brightness of the overbright pixels to within the specified range. When synthesizing new image frames, ensure that the brightness of the surrounding areas is not affected. Record the results of each processing, including the change in brightness before and after processing, to evaluate the neutralization effect. Combine the brightness neutralization frame with the reconstructed surveillance video to generate an overbright neutralized surveillance video. Ensure that the synthesized video is smooth and free of noticeable discontinuities. Perform a quality check on the generated neutralized surveillance video to ensure that the visual quality meets the expected standards. Select an appropriate adaptive filtering algorithm (such as adaptive median filtering or bilateral filtering) for noise reduction. Set filtering parameters (such as the filter window size) to suit the characteristics of the video content. Record the selected noise reduction algorithm and parameter settings for subsequent evaluation and adjustment. Apply an adaptive filtering algorithm to the generated overly bright and neutral surveillance video to perform noise reduction processing to remove background noise and unnecessary interference in the video. The processed video frames should maintain clarity while eliminating most noise. Record the processing time and effect of each frame for subsequent evaluation. Combine the noise-reduced video frames into the final filtered, noise-reduced, and cut surveillance video, and record the parameters of the final video, such as the total number of frames and duration. Perform a quality check on the final video to ensure that it meets the expected standards in terms of visual effects and surveillance effects.

[0025] In this embodiment, refer to Figure 3 , is a flowchart of the detailed implementation steps of step S2. In this embodiment, the detailed implementation steps of step S2 include: Perform dynamic cutting process behavior analysis on the wave noise reduction cutting monitoring video to generate dynamic cutting process behavior characteristics; Mining the cutting process logic based on dynamic cutting process behavior characteristics to generate cutting process logic relationships; Based on the initial tool position coordinates, dynamic optical flow tracking is performed on the wave denoising cutting monitoring video to extract the tool dynamic optical flow tracking path; Mark tool landing points at multiple time points based on the tool's dynamic optical flow tracking path to generate tool landing point coordinates at multiple time points; According to the logical relationship of the cutting process, the time-series cutting trajectory is fitted to the tool landing point coordinates at multiple time points to construct a multi-time point cutting trajectory.

[0026] In this example, noise-reduced cutting surveillance video is used as the data source. Video clarity and stability are ensured to facilitate subsequent behavioral analysis. The video frame rate is set to 30 frames per second to capture subtle changes during the cutting process. The video is decomposed frame by frame, and each frame is extracted for subsequent feature extraction and analysis. Computer vision techniques (such as edge detection and contour extraction) are applied to analyze the process behavior characteristics during the cutting process. Feature extraction parameters, such as the thresholds of the edge detection algorithm (such as the high and low thresholds of the Canny algorithm), are set to improve feature extraction accuracy. Dynamic cutting process behavior features are generated by analyzing key parameters such as the contact point between the tool and the material, cutting speed, and cutting depth. The timestamp and corresponding frame information of each feature are recorded. The extracted dynamic cutting process behavior features are stored in a database and organized and categorized. A feature data table can be created to record each feature's type, timestamp, value, and other information for subsequent analysis and mining. A feature change curve graph is generated using visualization tools to assist in analyzing behavioral change trends during the cutting process and identify key cutting event points. Based on the extracted dynamic cutting process behavior features, a logical relationship model of the cutting process is constructed. Association rule mining algorithms (such as the Apriori algorithm or the FP-Growth algorithm) can be used to analyze the relationships between process features. Set minimum support and confidence thresholds (e.g., support 0.2, confidence 0.7) to identify important logical relationships. Run the association rule mining algorithm to generate logical relationships between cutting process features. For example, if a certain combination of cutting depth and cutting speed occurs frequently, a "logical relationship between cutting depth and cutting speed" can be established. Record the mined logical relationships, including information such as feature combinations, support, and confidence, for subsequent analysis and application. Analyze the mined cutting process logical relationships to identify potential process optimization points. For example, identify the optimal parameter combination under certain cutting conditions to provide a basis for subsequent cutting optimization. Select a suitable optical flow algorithm (such as the Lucas-Kanade optical flow method or the Horn-Schunck optical flow method) for dynamic tool optical flow tracking. Set algorithm parameters, such as the window size and number of iterations, to ensure the accuracy and stability of the optical flow calculation. Prepare the initial position coordinates of the tool to ensure association with subsequent optical flow tracking results. Dynamic optical flow tracking is performed on noise-reduced cutting surveillance video to calculate the tool's trajectory in each frame. By analyzing brightness changes between adjacent frames, the tool's direction and speed are identified. The tool's dynamic optical flow path is recorded at each moment, including timestamps and coordinates, for subsequent extraction of the tool's impact point. The tool's motion path, derived from optical flow tracking, is stored in a database, recording each key point along the path and its corresponding time. A path data table can be created to facilitate subsequent analysis. A visualization of the tool's motion trajectory is generated to facilitate analysis of the tool's trajectory and behavioral characteristics during the cutting process.Extract the tool's landing point coordinates based on the dynamic optical flow tracing path. Set the extraction interval (e.g., extract a landing point once per second) to ensure that tool landing point information is captured at multiple time points. Record the timestamp and coordinates of each landing point for subsequent analysis and fitting. Store the extracted tool landing point coordinates at multiple time points in a database, recording relevant information for each landing point, including time, coordinates, and optical flow path. Create a landing point data table to ensure data traceability and integrity for subsequent analysis and optimization. Select an appropriate time-series trajectory fitting algorithm (such as polynomial fitting, spline fitting, or linear regression) to fit the tool's landing point coordinates. Set fitting parameters, such as the polynomial order or spline smoothness, to ensure fitting accuracy. Prepare fitting data, including the tool's landing point coordinates at multiple time points and their corresponding time information. Apply the selected fitting algorithm to perform time-series fitting on the tool's landing point coordinates at multiple time points to generate the tool's cutting trajectory. Ensure that the fitting results accurately reflect the tool's motion during the cutting process. Record key parameters and results from the fitting process, including the fitting function form, fitting error, and goodness of fit. Visualize the fitted cutting trajectory to help analyze tool motion trends and cutting efficiency. Generate a 3D trajectory graph to more intuitively display the tool's cutting path. Analyze the fitting results, compare the actual cutting trajectory with the fitted trajectory, and identify potential optimization points.

[0027] In this embodiment, refer to Figure 4 , is a flowchart of the detailed implementation steps of step S3. In this embodiment, the detailed implementation steps of step S3 include: Perform time-series cutting temperature calculation on multi-dimensional monitoring parameters to generate time-series cutting temperature parameters; Perform temperature fluctuation analysis on the sequential cutting temperature parameters to generate cutting temperature fluctuation characteristics; Perform cutting vibration frequency statistics on multi-dimensional monitoring parameters to extract cutting vibration frequency; According to the cutting temperature fluctuation characteristics, the steel structure thermal coupling effect evolution is carried out on the wave noise reduction cutting monitoring video to generate a steel structure thermal coupling effect model; The steel structure thermal coupling effect model is dynamically rendered based on the multi-time point cutting trajectory to build a real-time cutting rendering model.

[0028] In this embodiment, multi-dimensional monitoring parameter data is collected, including cutting temperature, cutting speed, and cutting depth, as captured by temperature sensors. This data should be recorded using a unified time base to ensure data timeliness and consistency. A data collection frequency is set (e.g., once per second) to ensure sufficient temporal resolution for subsequent analysis. The collected temperature data is organized and arranged chronologically to generate time-series cutting temperature parameters. The temperature value at each time point is calculated and its corresponding timestamp is recorded. For temperature data, a threshold is set to filter out abnormal values ​​below a reasonable range (e.g., temperatures below room temperature) to ensure data reliability. The generated time-series cutting temperature parameters are stored in a database, and a temperature data table is created, recording timestamps, temperature values, and related monitoring parameters. This ensures data traceability and integrity. Fluctuation analysis is performed on the time-series cutting temperature parameters to calculate the mean, standard deviation, and fluctuation amplitude of the temperature. A method for calculating the fluctuation amplitude is set, such as the difference between the maximum and minimum values, and the temperature fluctuation characteristics for each time period are recorded. A time window is set for the temperature fluctuation analysis (e.g., every 10 seconds) to facilitate statistical analysis within each window. The extracted temperature fluctuation characteristics are stored in a database, recording the fluctuation characteristics for each time window, including the mean, standard deviation, and fluctuation amplitude, for subsequent analysis and mining. A visualization of the fluctuation characteristics is generated to help analyze trends and patterns in temperature fluctuations and identify potential abnormal cutting events. Abnormal temperature fluctuations during the cutting process are detected based on a set fluctuation characteristic threshold (for example, a fluctuation amplitude exceeding a certain value is considered abnormal). The timestamp and fluctuation characteristics of each detection are recorded for subsequent analysis. Data related to cutting vibration is collected from multi-dimensional monitoring parameters, such as sampling data from vibration sensors. The sampling frequency of the vibration data is set (for example, 100 times per second) to ensure that rapidly changing vibration signals are captured. The timestamp and corresponding vibration amplitude of the vibration data are recorded for subsequent frequency analysis. Time-domain analysis is performed on the collected vibration data to calculate the frequency spectrum of the vibration signal. The fast Fourier transform (FFT) algorithm can be used to convert the time-domain signal into a frequency-domain signal to extract the main vibration frequency components. The main frequency components and their amplitudes are recorded to identify vibration frequency characteristics that affect the cutting process. The extracted cutting vibration frequencies are stored in a database, and a vibration frequency data table is created to record information such as timestamps, main frequencies, and their amplitudes for subsequent analysis. A spectrum diagram is generated to help analyze the distribution of vibration frequencies and identify vibration modes that affect cutting quality. Based on the extracted cutting temperature fluctuation characteristics and vibration frequencies, a thermal coupling effect model for steel structures is established. This model should consider the mutual influence between temperature and vibration during the cutting process. Model parameters such as material thermal conductivity and specific heat capacity are set to ensure the physical authenticity of the model. Finite element analysis (FEA) methods are used to simulate the thermal coupling effect and analyze the thermal impact on the steel structure during the cutting process.Simulate the evolution of thermal coupling effects under different cutting conditions based on the time-series cutting temperature parameters and vibration frequency. Record the results of each simulation, including temperature distribution, stress distribution, etc., for subsequent analysis and optimization. Collect cutting trajectory data at multiple time points, including the dynamic landing coordinates and time information of the tool. Ensure the integrity of the data for trajectory rendering. Set the time interval for trajectory rendering (such as rendering once per second) to ensure the smoothness and real-time nature of the rendering process. Dynamically render the cutting trajectory of the tool based on multi-point cutting trajectory data and the thermal coupling effect model. Use 3D rendering software (such as Unity or Blender) to build a real-time cutting rendering model. During the rendering process, visualize the impact of the thermal coupling effect to show the thermal distribution and stress changes of the tool during the cutting process.

[0029] In this embodiment, the specific steps of performing steel structure thermal coupling effect evolution on the wave noise reduction cutting monitoring video according to the cutting temperature fluctuation characteristics to generate the steel structure thermal coupling effect model are as follows: Positioning the steel to be cut based on the wave noise reduction cutting monitoring video; Performing an initial three-dimensional morphological structure analysis on the steel to be cut to generate initial morphological structure characteristics of the steel to be cut; Conduct 3D point cloud modeling of the initial morphological and structural features of the steel to be cut, and construct a 3D model of the steel structure; Identify the surface cutting area based on the morphological and structural characteristics of the steel to be cut; Based on the cutting temperature fluctuation characteristics, the heat conduction characteristics of the surface cutting area are analyzed to extract the heat conduction characteristics of the steel structure surface cutting; Conduct thermal deformation analysis on the surface cutting area according to the initial morphological and structural characteristics of the steel to be cut to generate thermal deformation data of the cutting area; According to the heat conduction characteristics of steel structure surface cutting and the thermal deformation data of the cutting area, the thermal coupling effect evolution of the steel structure three-dimensional model is carried out to generate a steel structure thermal coupling effect model.

[0030] In this embodiment, the cutting monitoring video after wave noise reduction processing is used as the data source. Ensure that the video meets the requirements in terms of lighting conditions and clarity to facilitate subsequent steel positioning. Set the video frame rate to 30 frames per second to ensure that the dynamic changes in the cutting process can be captured. Process the video frame by frame, extract each frame image, and perform necessary image enhancement operations (such as contrast adjustment and edge enhancement) to improve the accuracy of subsequent positioning. Use a deep learning target detection algorithm (such as YOLO or FasterR-CNN) to locate the steel to be cut. According to the trained model, target detection is performed on each frame image to identify and mark the steel area to be cut. Set the detection confidence threshold (such as 0.5) to ensure that only detection results with a confidence level higher than this value are recorded. Record the coordinates and timestamp of each positioning result for subsequent analysis. By performing three-dimensional reconstruction of the located steel area to be cut, it is first necessary to obtain multi-view image data of the area. Images from different angles can be collected using multiple cameras or motion shooting. Set image capture parameters, such as the interval between shots at each angle (e.g., one image every 15 degrees), to ensure the stability of the reconstruction. Use computer vision techniques (such as structured light or stereo vision) to perform an initial 3D morphological analysis. By matching multi-view images, extract feature points and build a 3D model. Record key parameters extracted during the morphological analysis, such as the number of feature points and reconstruction accuracy (e.g., ±0.5 mm), for subsequent analysis. Based on the results of the initial 3D morphological analysis, extract 3D point cloud data. Record the coordinates and color information of each point to form a point cloud model. Set the point cloud density (e.g., no less than 1000 points per cubic meter) to ensure model detail and accuracy. Process the initial point cloud data using point cloud processing algorithms (e.g., filtering and resampling) to remove noise and unnecessary points. Downsampling can be performed using a voxel grid method to reduce computational complexity. Record key parameters during the processing, such as the effectiveness of denoising and the change in the number of point clouds (e.g., from an original 500,000 points to 300,000 points). A 3D model of the steel structure is constructed using the generated point cloud data. Surface reconstruction can be performed using 3D modeling tools to ensure the smoothness and integrity of the model. Based on the initial morphological and structural characteristics of the steel to be cut, the surface is analyzed to identify the cutting area. Image segmentation techniques (such as threshold segmentation or region growing) can be used to extract the cutting area. Characteristic parameters of the cutting area, such as surface defects and cutting depth, are set to improve recognition accuracy. The identified surface cutting area is marked, and its position, area, and characteristic information are recorded. Ensure that the characteristic data of each cutting area is traceable. The cutting area data is stored in a database to form a cutting area characteristic data table. Based on the cutting temperature fluctuation characteristics, a heat conduction model for the steel structure surface cutting is constructed. Model parameters, such as thermal conductivity and specific heat capacity, are set to ensure the physical authenticity of the model.Record the material's thermal properties. For example, the thermal conductivity of steel is typically in the range of 50-60 W / (m·K). Use finite element analysis (FEA) to simulate the heat conduction characteristics of the cutting area and analyze the impact of temperature changes on heat conduction during the cutting process. Record the results of the heat conduction simulation, including temperature distribution and heat flux, to facilitate subsequent analysis and optimization. Construct a thermal deformation analysis model based on the initial morphological and structural characteristics of the steel to be cut. Set parameters for the thermal deformation analysis, such as the material's coefficient of thermal expansion (CTE) (for example, the CTE of steel is approximately 11×10^-6 / K). Record key parameters used in the model to ensure analysis accuracy. Use thermodynamic analysis to perform thermal deformation analysis on the surface cutting area, simulating the deformation caused by temperature changes during the cutting process. Record the results of each analysis step, including deformation and stress distribution in the cutting area, for subsequent analysis. Integrate a thermal coupling effect model based on the heat conduction characteristics of the steel structure surface and the thermal deformation data in the cutting area. Set the parameters of the coupling model to ensure model accuracy and reliability. Record key parameters and results from the integration process for subsequent verification and optimization. Analyze the integrated thermal coupling effect model using finite element analysis to simulate the interaction between heat conduction and deformation during the cutting process. Record key results from each simulation step for subsequent evaluation and optimization. Verify the generated thermal coupling effect model, comparing the simulation results with the measured data to ensure model accuracy. Record key parameters and results from the verification process. Store the thermal coupling effect model and its analysis results in a database for subsequent use and analysis.

[0031] In this embodiment, the specific steps of step S4 are: Perform cutting evolution simulation in future periods on the real-time cutting rendering model to generate cutting evolution simulation data; Perform cutting tool spatiotemporal variation analysis on cutting evolution simulation data to generate cutting tool spatiotemporal variation law; Predict tool wear status based on the spatiotemporal variation of cutting tools to generate tool wear prediction trends; Conduct steel structure thermal damage trend analysis on cutting evolution simulation data to generate steel structure thermal damage trend; According to the tool wear prediction trend and steel structure thermal damage trend, abnormal cutting fault adjustments are made and a cutting fault adjustment strategy is constructed.

[0032] In this embodiment, based on the real-time cutting rendering model, cutting conditions and parameters for future time periods are defined. These conditions include cutting speed, cutting depth, tool type, and material properties. All parameter settings are based on actual cutting data and physical properties. A simulation time span is set (e.g., cutting evolution over the next 5 minutes) and a time step is determined (e.g., once per second) to ensure simulation precision and accuracy. Computational fluid dynamics (CFD) or finite element analysis (FEA) methods are used to dynamically simulate the cutting evolution process. During the simulation, factors such as heat conduction, stress distribution, and material deformation are considered to generate dynamic cutting evolution data. The cutting state at each time step, including tool position, cutting depth, and temperature distribution, is recorded for subsequent analysis and evaluation. The generated cutting evolution simulation data is stored in a database, forming a cutting evolution data table that records cutting state information at each time point, including cutting position, tool state, and heat distribution. Temporal and spatial variation information of the tool, including tool speed, position, angle, and temperature, is extracted from the cutting evolution simulation data. The extraction interval is set (e.g., once per second) to ensure data integrity. Record tool status information at each time point to form a tool status dataset containing the tool's dynamic change characteristics. Statistical analysis methods (such as time series analysis or regression analysis) are used to analyze the tool's spatiotemporal variation data. Develop an analysis model, such as the ARIMA model, to identify patterns and trends in tool variation. Calculate metrics such as the average wear rate and temperature change rate during the cutting process to identify potential wear patterns and variation patterns. Build a tool wear state prediction model based on the tool's spatiotemporal variation patterns. Set model input parameters, including cutting speed, temperature, and material properties, to ensure model accuracy. Select an appropriate prediction algorithm (such as a support vector machine (SVM) or neural network) for wear state prediction, and set the model's training parameters and validation criteria. Train and validate the wear model using historical cutting data and current spatiotemporal variation data. Record accuracy metrics (such as root mean square error (RMSE)) for each prediction to ensure model validity. Predict tool wear for future time periods and generate a wear prediction trend, including estimated wear extent and tool life. Extract thermal damage information from the cutting evolution simulation data, including temperature distribution, heat flux density, and material strength changes. Set the data extraction interval (e.g., synchronized with the spatiotemporal changes of the tool). Record the thermal damage data for the steel structure within each time period to form a thermal damage status dataset. Analyze the thermal damage data using data analysis methods (such as trend analysis or extreme value analysis) to identify thermal damage trends occurring during the cutting process. Evaluate the impact of thermal damage on material properties, calculate indicators such as strength reduction and fatigue life, and record the analysis results. Store the thermal damage trend analysis results in a database, creating a thermal damage trend data table that records key damage indicators and trends.Generate a thermal damage trend graph to help intuitively understand the thermal damage of steel structures during the cutting process, facilitating subsequent safety assessments. Build a framework for adjusting strategies for abnormal cutting faults based on the tool wear prediction trend and the thermal damage trend of steel structures. Set goals for the adjustment strategy, such as extending tool life and reducing thermal damage. Analyze adjustment measures under different fault conditions, including adjusting cutting parameters (such as reducing cutting speed, optimizing cutting depth) and regularly maintaining tools. Implement the adjustment strategy during the actual cutting process and monitor the cutting results and fault conditions. Record key parameters and results during the implementation process to verify the effectiveness of the strategy. Analyze the effects after implementation, including reduced tool wear rate and reduced thermal damage, to ensure the feasibility and effectiveness of the strategy.

[0033] In this embodiment, the specific steps of step S5 are: Obtaining a preset standard cutting log; identifying a standard cutting trajectory according to the standard cutting log; According to the standard cutting trajectory, the cutting offset of the multi-time point cutting trajectory is identified and the cutting offset trajectory is marked; Calculate the cutting trajectory deviation point by point on the cutting offset trajectory and extract the cutting trajectory deviation value at each time point; Perform deep cutting defect visual recognition on the wave noise reduction cutting monitoring video to generate cutting defect features; Classify the defect types according to the cutting defect characteristics and generate cutting defect types; Defect attribution is speculated based on the cutting defect type to obtain the cutting defect factors.

[0034] In this embodiment, a standard cutting log is collected and defined, ensuring that it contains all key parameters of the cutting process, such as cutting speed, cutting depth, tool type, and material properties. The standard cutting log should be based on historical cutting data and best practices to ensure its representativeness and reliability. A format for recording the standard cutting log is established to facilitate subsequent data analysis and extraction. The log should include timestamps, cutting parameters, and actual cutting trajectory information. Standard cutting log data is extracted from the actual cutting process, recording detailed information for each cutting process. Data collection is performed at a sufficiently high frequency (e.g., once per second) to ensure data timeliness and integrity. The collected standard cutting log data is stored in a database, forming a standard cutting log data table for subsequent analysis and application. Standard cutting trajectory information is extracted from the standard cutting log, recording tool position changes during the cutting process. Tool position changes are described using a coordinate system (e.g., a Cartesian coordinate system). Key parameters of the standard cutting trajectory, such as trajectory smoothness and continuity, are defined to ensure that the extracted trajectory accurately reflects the tool motion during the cutting process. The identified standard cutting trajectory is stored in a database, recording the tool position and status information at each point in time. This ensures traceability of the trajectory data. Use visualization tools (such as MATLAB or Python visualization libraries) to visualize the standard cutting trajectory to help you intuitively understand the tool motion during the cutting process. Extract multi-point cutting trajectory data from the real-time monitoring video to record the actual tool position changes during the cutting process. Ensure that the data acquisition frequency is consistent with the standard cutting log (e.g., once per second). Set the extraction time range and adjust it according to the actual cutting process to ensure that critical cutting moments are captured. Compare the extracted multi-point cutting trajectory with the standard cutting trajectory to identify any deviations. Set a deviation threshold (e.g., within a range of ±5 mm is considered normal); any trajectory outside this range is considered a deviation. Record the time point and deviation of each deviation trajectory for subsequent analysis. Set a method for calculating cutting trajectory deviation, typically using Euclidean distance. Record the parameters used in the calculation, such as the time range and threshold settings, to ensure the accuracy of the results. Calculate the deviation of the cutting trajectory at each time point and extract the deviation value at each time point. Ensure that the calculated results reflect the deviation changes during the cutting process. Record the deviation value at each time point for subsequent analysis. The calculated cutting trajectory deviation values ​​are stored in a database, forming a trajectory deviation data table that records the deviation value and corresponding time information at each time point. A deviation change curve is generated to help analyze deviation trends during the cutting process and facilitate subsequent anomaly identification. Based on the noise reduction cutting surveillance video, a deep learning model (such as a convolutional neural network (CNN)) is designed and trained for visual recognition of cutting defects. Ensure that the model's training data includes a variety of cutting defect samples to improve recognition accuracy.Set model training parameters, such as the learning rate, batch size, and number of iterations, to optimize model performance. Input the preprocessed cutting surveillance video into the trained defect recognition model to detect and identify deep cutting defects. Record the timestamp and defect location of each detection result. Set a recognition confidence threshold (e.g., 0.6) to ensure that only defect recognition results with a confidence level above this threshold are recorded. Classify defects based on identified cutting defect characteristics. Set defect classification criteria, such as excessive cutting depth, uneven cutting, and surface roughness. Use clustering analysis methods (e.g., K-means or hierarchical clustering) to cluster defect characteristics to identify defect types with similar characteristics. Store the classified defect types in a database, creating a defect type data table that records the frequency of each defect type and its corresponding defect characteristics. Generate a defect type report summarizing the analysis results to assist in subsequent defect management and resolution. Build a defect attribution inference model based on the cutting defect types. Determine the model input and output parameters, including cutting parameters (e.g., speed, depth, tool type) and defect type. Select an appropriate attribution analysis method (such as a decision tree or random forest) to infer defect attribution and set the model's training parameters to improve accuracy. Use the collected cutting data to train the defect attribution model and infer the factors that cause cutting defects. Record the parameters and results used in the inference process to ensure model validity. Generate a defect factor inference report that summarizes the causes of each defect type to facilitate subsequent cutting process improvements. Store the defect attribution inference results in a database, creating a defect factor data table that records each defect type and its corresponding inferred factors. Regularly evaluate and optimize the inference results to ensure the accuracy and adaptability of the attribution analysis.

[0035] In this embodiment, the specific steps of step S6 are: Intelligent cutting parameter fine-tuning is performed based on the cutting trajectory deviation value and cutting defect factors at each time point to generate a cutting error fine-tuning strategy; Iterative machine learning optimization of cutting abnormality fault adjustment strategies and cutting abnormality fault adjustment strategies to build an iterative cutting optimization engine; Instant cutting control is performed based on an iterative cutting optimization engine, along with full-cycle monitoring, analysis, and processing.

[0036] In this embodiment, cutting trajectory deviation values ​​and cutting defect factor data are collected at each time point. This data should be derived from previous deviation calculations and defect attribution analysis to ensure accuracy and relevance. A data structure is established to associate the deviation values ​​with the corresponding cutting defect factors, forming a data table for subsequent analysis. Based on the cutting trajectory deviation values ​​and defect factors, an intelligent cutting parameter fine-tuning model is constructed. Key parameters affecting cutting quality, such as cutting speed, cutting depth, and feed rate, are set, and their adjustment ranges are defined (for example, the cutting speed can be adjusted within a ±10% range). A machine learning algorithm, such as multivariate linear regression or support vector machine (SVM), is used to build a model to predict the optimal cutting parameters based on the deviations and defect factors. The results of the model analysis are used to generate a cutting error fine-tuning strategy. For example, when the deviation value exceeds a set threshold (e.g., ±5 mm) and a cutting defect is present, the cutting speed and feed rate are adjusted to reduce the error. The generated fine-tuning strategy is recorded in a database, forming a cutting error fine-tuning strategy data table to ensure strategy traceability. Collect implementation data for the abnormal cutting fault adjustment strategy, including post-execution cutting results, tool wear, and thermal damage trends. This data should be recorded in real time during the cutting process to ensure its integrity. Establish an iterative learning cycle (e.g., one iteration after each completed cutting task) to continuously update and optimize the strategy. Build an iterative machine learning optimization model based on the collected historical data. Select an appropriate algorithm (e.g., reinforcement learning or genetic algorithm), set the model parameters, and implement a reward mechanism to promote strategy optimization. Record the model training process to ensure that each iteration improves cutting quality and efficiency. After each cutting task, use the iterative optimization model to evaluate and update the abnormal cutting fault adjustment strategy. Analyze the impact of cutting results on the strategy and optimize adjustment measures. Develop a cutting optimization engine and record each optimization strategy change and cutting results for subsequent analysis and improvement. Establish a real-time cutting control system based on the iterative cutting optimization engine. The system should be able to receive real-time monitoring data (e.g., temperature, vibration, cutting depth, etc.) and adjust the cutting parameters based on the optimization engine's output. Set the control system's response time (e.g., less than 100 milliseconds) to ensure rapid response to changes in the cutting process. During the cutting process, real-time data is monitored and analyzed throughout the entire process. Data analysis tools (such as a real-time data stream analysis platform) are used to dynamically monitor the cutting process and record key performance indicators (KPIs), such as cut quality, tool wear rate, and thermal damage level. Monitoring and analysis reports are regularly generated, summarizing various indicators during the cutting process to help identify potential problems and areas for improvement. Monitoring and analysis results are fed back to the iterative cutting optimization engine to promote continuous learning and optimization of the model. Key data and results from the feedback process are recorded to facilitate subsequent strategy adjustments. The performance and effectiveness of the cutting control system are regularly evaluated, and necessary optimizations and upgrades are performed to ensure the reliability and stability of the system during long-term operation.

[0037] In this embodiment, an online monitoring system based on machine vision and artificial intelligence algorithms is provided, which is used to execute the online monitoring method based on machine vision and artificial intelligence algorithms as described above, including: A video optimization module is used to collect real-time steel structure cutting monitoring videos and multi-dimensional monitoring parameters; perform adaptive filtering and noise reduction on the real-time steel structure cutting monitoring videos and calculate the initial frame spatial position to generate initial tool position coordinates; Cutting trajectory module, which is used to perform dynamic optical flow tracking based on the initial tool position coordinates, perform time-series cutting trajectory fitting, and construct multi-time point cutting trajectories; The cutting trajectory rendering module is used to perform the evolution of the thermal coupling effect of the steel structure on the multi-dimensional monitoring parameters and perform dynamic cutting trajectory rendering based on the multi-time point cutting trajectory to build a real-time cutting rendering model; The fault adjustment module is used to predict tool wear status and analyze steel structure thermal damage trends based on the real-time cutting rendering model, and to build a cutting abnormality fault adjustment strategy; The trajectory deviation module is used to calculate the cutting trajectory deviation of multiple cutting points one by one, and make defect attribution inferences to obtain the cutting defect factors; The iterative learning module is used to fine-tune intelligent cutting parameters according to cutting defect factors, and to perform iterative machine learning optimization based on cutting abnormal fault adjustment strategies to build an iterative cutting optimization engine.

[0038] The present invention uses a video optimization module to collect real-time steel structure cutting monitoring video and multi-dimensional monitoring parameters (such as temperature and vibration). This multi-dimensional data provides comprehensive cutting information, providing rich input for subsequent analysis. Videos of the cutting process are often subject to noise, especially in strong light or vibrating environments. Using an adaptive filtering noise reduction algorithm, image noise can be effectively reduced, improving image clarity. This makes subsequent tool position identification and cutting path tracking more accurate, reducing errors caused by poor image quality. Calculating the spatial position of the initial frame in the video determines the initial tool position, providing a precise starting point for subsequent tool tracking and cutting trajectory analysis. An accurate initial position is crucial for tracking the entire cutting process and ensures efficient and stable dynamic optical flow tracking. Optical flow tracking technology tracks the tool's trajectory in real time by analyzing pixel movement between different frames in the image. Even during the cutting process due to noise or changes in ambient light, the optical flow algorithm maintains high-precision tracking, providing an accurate dynamic trajectory for subsequent cutting path analysis. Based on the initial tool position coordinates, time-series cutting trajectory fitting accurately captures the tool's motion path and gradually builds cutting trajectory models at multiple moments. This not only helps accurately reproduce the tool's motion process but also reveals potential cutting issues, such as unstable tool paths and cutting accuracy issues. This module provides cutting trajectory data at every moment, enabling the system to observe tool position changes in real time over time, thereby analyzing the accuracy and efficiency of the cutting process. During the cutting process, effects such as heat conduction and thermal stress between the tool and the steel structure significantly impact cutting quality. By analyzing multi-dimensional monitoring parameters, the evolution of thermal coupling effects can be modeled, enabling real-time prediction of issues such as temperature unevenness and thermal damage during the cutting process. Dynamic rendering of the cutting trajectory allows the system to simulate the cutting process in a virtual environment. Based on cutting trajectory data at multiple points in time, the dynamic rendering module constructs a real-time cutting rendering model and displays changes in temperature, stress, and other parameters during the cutting process. This helps operators fully understand the cutting process, identify potential issues, and adjust operations promptly. Combining thermal coupling effects with cutting trajectory rendering, the real-time cutting rendering model provides operators with a visual representation of the cutting process. The virtual model accurately displays the cutting path and its impact, facilitating immediate decision-making and fault analysis. Tool wear is an inevitable problem during steel cutting, but this module can predict tool wear status by analyzing real-time cutting rendering models. By monitoring parameters such as cutting trajectory, cutting speed, and pressure, wear trends can be identified in advance, avoiding reduced cutting quality or equipment damage caused by tool wear. Thermal damage to steel structures during cutting affects material properties. By analyzing thermal damage trends in this module, the system can identify problems such as uneven heat conduction or localized overheating. Early identification of these issues can help adjust the cutting process, thereby reducing the negative impact of thermal damage.Based on wear prediction and thermal damage analysis results, the system generates real-time adjustment strategies for cutting anomalies. These strategies address tool replacement, cutting parameter adjustments, and cooling system optimization, providing operators with scientific and timely adjustment recommendations. By calculating deviations from the cutting trajectory at each point in time, the system can identify tool deviations from the planned path in real time. These deviations are caused by equipment failure, material issues, or operator error. Prompt identification of deviations can prevent further problems. Combining these deviations with sensor data, the system can infer the cause of the defect. For example, if the deviation is excessive, tool wear or uneven material may be the cause. The accuracy of defect attribution provides key clues for subsequent troubleshooting. This module helps identify potential defect factors during the cutting process and reduce errors caused by environmental, operational, or equipment issues. It also helps improve cutting accuracy and stability. Based on the factors causing the cutting defect, the system automatically adjusts cutting parameters (such as cutting speed, pressure, and temperature) to optimize the cutting process. Fine-tuning allows for targeted adjustments to ensure optimal cutting performance. Through an iterative learning module, the system continuously optimizes the cutting process based on historical data and real-time feedback. Machine learning optimization enables the system to adapt to different cutting environments and working conditions. Through continuous learning and adjustment, it improves the stability, efficiency, and precision of the cutting process. Through continuous feedback and optimization, the system ultimately establishes a highly efficient iterative optimization engine. This engine continuously adjusts and optimizes cutting parameters during the production process to ensure cutting quality and production efficiency.

[0039] The present invention is therefore intended to be illustrative and non-restrictive in all respects, with the scope of the invention being defined by the appended claims rather than the foregoing description, and all changes that come within the meaning and range of equivalents of the application documents are intended to be embraced therein.

[0040] The foregoing description is intended only to provide specific embodiments of the present invention, which will enable those skilled in the art to understand and implement the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention is not intended to be limited to the embodiments shown herein, but is to be construed in the widest possible manner consistent with the principles and novel features disclosed herein.

Claims

1. An online monitoring method based on machine vision and artificial intelligence algorithm, characterized in that: The following steps are involved: Step S1: collecting real-time steel structure cutting monitoring video and multi-dimensional monitoring parameters; performing adaptive filtering and noise reduction and initial frame spatial position calculation on the real-time steel structure cutting monitoring video to generate initial tool position coordinates; Step S2: Perform dynamic optical flow tracking based on the initial tool position coordinates, perform temporal cutting trajectory fitting, and construct a multi-time point cutting trajectory; Step S3: Evolution of the thermal coupling effect of the steel structure is performed on the multi-dimensional monitoring parameters, and dynamic cutting trajectory rendering is performed according to the multi-time point cutting trajectory to construct a real-time cutting rendering model; Step S4: Predict tool wear status and analyze steel structure thermal damage trends based on the real-time cutting rendering model, and build a cutting abnormality fault adjustment strategy; Step S5: Calculating the deviation of the cutting trajectory at each time point and making defect attribution inferences to obtain the cutting defect factors; Step S6: Fine-tune the intelligent cutting parameters according to the cutting defect factors, and perform iterative machine learning optimization based on the cutting abnormality fault adjustment strategy to build an iterative cutting optimization engine.

2. The online monitoring method based on machine vision and artificial intelligence algorithm according to claim 1 is characterized in that: The specific steps of step S1 are: Collect real-time cutting monitoring videos and multi-dimensional monitoring parameters of steel structures; Adaptively filtering and denoising the real-time cutting monitoring video of the steel structure, thereby constructing a filtered and denoised cutting monitoring video; Perform visual recognition of cutting tools on filtered and noise-reduced cutting surveillance videos and mark cutting tools; Perform initial frame space position calculation on the cutting tool to generate initial tool position coordinates.

3. The online monitoring method based on machine vision and artificial intelligence algorithm according to claim 2 is characterized in that: The specific steps of performing adaptive filtering and noise reduction on the real-time steel structure cutting monitoring video to construct the filtered noise reduction cutting monitoring video are as follows: Performing frame-by-frame image delay calculation on the real-time steel structure cutting monitoring video and extracting all inter-frame delay values; Calculating an average of the inter-frame delay values; Performing delay mutation abnormal video frame detection on the real-time steel structure cutting monitoring video according to the average value, and extracting the delay abnormal video frame; Perform inter-frame delay interpolation optimization on the abnormally delayed video frames to obtain inter-frame delay reconstructed surveillance video; Perform overbrightness analysis on the steel cutting area of ​​the inter-frame delayed reconstructed surveillance video and extract the overbrightness area of ​​the steel cutting; Calculate the pixel value of the surrounding area based on the bright area of ​​steel cutting; Performing brightness neutralization processing based on the pixel values ​​of the surrounding area to obtain an overbrightness neutralization monitoring video; Adaptive filtering and denoising is performed on overly bright and neutral surveillance videos, thereby constructing filtered and denoised cut surveillance videos.

4. The online monitoring method based on machine vision and artificial intelligence algorithm according to claim 1 is characterized in that: The specific steps of step S2 are: Perform dynamic cutting process behavior analysis on the wave noise reduction cutting monitoring video to generate dynamic cutting process behavior characteristics; Mining the cutting process logic based on dynamic cutting process behavior characteristics to generate cutting process logic relationships; Based on the initial tool position coordinates, dynamic optical flow tracking is performed on the wave denoising cutting monitoring video to extract the tool dynamic optical flow tracking path; Mark tool landing points at multiple time points based on the tool's dynamic optical flow tracking path to generate tool landing point coordinates at multiple time points; According to the logical relationship of the cutting process, the time-series cutting trajectory is fitted to the tool landing point coordinates at multiple time points to construct a multi-time point cutting trajectory.

5. The online monitoring method based on machine vision and artificial intelligence algorithm according to claim 1 is characterized in that: The specific steps of step S3 are: Perform time-series cutting temperature calculation on multi-dimensional monitoring parameters to generate time-series cutting temperature parameters; Perform temperature fluctuation analysis on the sequential cutting temperature parameters to generate cutting temperature fluctuation characteristics; Perform cutting vibration frequency statistics on multi-dimensional monitoring parameters to extract cutting vibration frequency; According to the cutting temperature fluctuation characteristics, the steel structure thermal coupling effect evolution is carried out on the wave noise reduction cutting monitoring video to generate a steel structure thermal coupling effect model; The steel structure thermal coupling effect model is dynamically rendered based on the multi-time point cutting trajectory to build a real-time cutting rendering model.

6. The online monitoring method based on machine vision and artificial intelligence algorithm according to claim 1 is characterized in that: The specific steps of performing steel structure thermal coupling effect evolution on the wave noise reduction cutting monitoring video according to the cutting temperature fluctuation characteristics to generate the steel structure thermal coupling effect model are as follows: Positioning the steel to be cut based on the wave noise reduction cutting monitoring video; Performing an initial three-dimensional morphological structure analysis on the steel to be cut to generate initial morphological structure characteristics of the steel to be cut; Conduct 3D point cloud modeling of the initial morphological and structural features of the steel to be cut, and construct a 3D model of the steel structure; Identify the surface cutting area based on the morphological and structural characteristics of the steel to be cut; Based on the cutting temperature fluctuation characteristics, the heat conduction characteristics of the surface cutting area are analyzed to extract the heat conduction characteristics of the steel structure surface cutting; Conduct thermal deformation analysis on the surface cutting area according to the initial morphological and structural characteristics of the steel to be cut to generate thermal deformation data of the cutting area; According to the heat conduction characteristics of steel structure surface cutting and the thermal deformation data of the cutting area, the thermal coupling effect evolution of the steel structure three-dimensional model is carried out to generate a steel structure thermal coupling effect model.

7. The online monitoring method based on machine vision and artificial intelligence algorithm according to claim 1 is characterized in that: The specific steps of step S4 are: Perform cutting evolution simulation in future periods on the real-time cutting rendering model to generate cutting evolution simulation data; Perform cutting tool spatiotemporal variation analysis on cutting evolution simulation data to generate cutting tool spatiotemporal variation law; Predict tool wear status based on the spatiotemporal variation of cutting tools to generate tool wear prediction trends; Conduct steel structure thermal damage trend analysis on cutting evolution simulation data to generate steel structure thermal damage trend; According to the tool wear prediction trend and steel structure thermal damage trend, abnormal cutting fault adjustments are made and a cutting fault adjustment strategy is constructed.

8. The online monitoring method based on machine vision and artificial intelligence algorithm according to claim 1 is characterized in that: The specific steps of step S5 are: Obtaining a preset standard cutting log; identifying a standard cutting trajectory according to the standard cutting log; According to the standard cutting trajectory, the cutting offset of the multi-time point cutting trajectory is identified and the cutting offset trajectory is marked; Calculate the cutting trajectory deviation point by point on the cutting offset trajectory and extract the cutting trajectory deviation value at each time point; Perform deep cutting defect visual recognition on the wave noise reduction cutting monitoring video to generate cutting defect features; Classify the defect types according to the cutting defect characteristics and generate cutting defect types; Defect attribution is speculated based on the cutting defect type to obtain the cutting defect factors.

9. The online monitoring method based on machine vision and artificial intelligence algorithm according to claim 1, characterized in that: The specific steps of step S6 are: Intelligent cutting parameter fine-tuning is performed based on the cutting trajectory deviation value and cutting defect factors at each time point to generate a cutting error fine-tuning strategy; Iterative machine learning optimization of cutting abnormality fault adjustment strategies and cutting abnormality fault adjustment strategies to build an iterative cutting optimization engine; Instant cutting control is performed based on an iterative cutting optimization engine, along with full-cycle monitoring, analysis, and processing.

10. An online monitoring system based on machine vision and artificial intelligence algorithm, characterized in that: The method for performing online monitoring based on machine vision and artificial intelligence algorithm according to claim 1 comprises: A video optimization module is used to collect real-time steel structure cutting monitoring videos and multi-dimensional monitoring parameters; perform adaptive filtering and noise reduction on the real-time steel structure cutting monitoring videos and calculate the initial frame spatial position to generate initial tool position coordinates; Cutting trajectory module, which is used to perform dynamic optical flow tracking based on the initial tool position coordinates, perform time-series cutting trajectory fitting, and construct multi-time point cutting trajectories; The cutting trajectory rendering module is used to perform the evolution of the thermal coupling effect of the steel structure on the multi-dimensional monitoring parameters and perform dynamic cutting trajectory rendering based on the multi-time point cutting trajectory to build a real-time cutting rendering model; The fault adjustment module is used to predict tool wear status and analyze steel structure thermal damage trends based on the real-time cutting rendering model, and to build a cutting abnormality fault adjustment strategy; The trajectory deviation module is used to calculate the cutting trajectory deviation of multiple cutting points one by one, and make defect attribution inferences to obtain the cutting defect factors; The iterative learning module is used to fine-tune intelligent cutting parameters according to cutting defect factors, and to perform iterative machine learning optimization based on cutting abnormal fault adjustment strategies to build an iterative cutting optimization engine.

Images