Online monitoring system and method for liquid level of molten pool surface

Abstract

The invention provides a molten pool level online monitoring system and method, and the system comprises a data collection module, a data preprocessing module, a model training and optimizing module, an angle detection and compensation module, a nonlinear data processing module, a system testing and optimizing module, a display module, a communication module, and a control module. The online real-time monitoring of the molten pool level can be realized, the production automation level is improved, when the data acquisition module reaches a detection position, the control system triggers a signal through the coil, the camera is started to capture the images of the liquid level of the slag surface and the liquid level of the copper solution surface of the current molten pool, the online real-time monitoring function of the molten pool level is realized, and the production efficiency is improved. Compared with a traditional manual measurement or intermittent monitoring method, the method has the advantages that the detection efficiency can be greatly improved, the dependence on manual operation is reduced, and the automation level of the production process is improved.

Description

Technical Field

[0001] The invention belongs to the technical field of molten pool surface liquid level monitoring in a copper smelting electric furnace process, and relates to an online molten pool surface liquid level monitoring system and method. Background Art

[0002] Smelting is generally a pyrometallurgical process in which metal materials and other auxiliary materials are put into a smelting furnace for melting and tempering. In the high-temperature smelting furnace, the materials undergo certain physical and chemical changes to produce crude metal or metal concentrate and slag. In the metal smelting production process, it is often necessary to control the liquid level of the high-temperature metal melt in the smelting furnace.

[0003] At present, the copper smelting electric furnace process production process relies on traditional manual methods to measure the height of the copper molten pool slag surface. The operation process requires two people to collaborate in the use of semi-automatic equipment combined with a tape measure to conduct on-site measurements. One person measures while the other assists in recording data. The inspection and measurement cycle of all copper molten pool slag surface detection points takes about 15 minutes, and the half-hour cycle detection. The on-site ambient temperature is high, the labor intensity of operators is high, and data updates are slow, which cannot provide timely and accurate data support for production process adjustments.

[0004] The traditional copper molten pool slag surface measurement method can no longer meet the essential needs of optimizing production process control indicators driven by data, improving product quality, and reducing production energy consumption. Due to the temperature difference between the molten copper and the copper slag, the slurry hanging height of the detection rod when passing through the molten copper and the copper slag is different. Therefore, the copper molten pool slag surface can be detected according to the above-mentioned characteristic items. Summary of the invention

[0005] The purpose of the present invention is to provide a molten pool surface level online monitoring system and method to address the problems of low accuracy, slow response and reliance on manual operation in the prior art of molten pool surface level monitoring, so as to achieve real-time and accurate online monitoring of the slag surface and copper solution surface levels in the molten pool, and promote data-driven production process optimization and control.

[0006] To this end, the present invention adopts the following technical solutions: A molten pool surface liquid level online monitoring system, comprising: Data acquisition module, data preprocessing module, model training and optimization module, angle detection and compensation module, nonlinear data processing module and system testing and optimization module; The data acquisition module is used to collect image data of the slag surface of the molten pool and the liquid level of the copper solution in the molten pool, and includes an industrial camera and a probe that can operate in a high temperature environment. The probe is used to measure and detect the liquid level position of the slag surface of the molten pool and the liquid level of the copper solution in the molten pool, and the camera is used to capture images of the slag surface of the molten pool and the liquid level of the copper solution in the molten pool; The data preprocessing module is used to normalize and enhance the image data to improve the accuracy of subsequent processing. It adopts an image segmentation processing model to achieve image segmentation of the slag surface of the molten pool and the copper solution surface of the molten pool; The model training and optimization module is used to construct, train and optimize the image segmentation model, which uses the U-Net convolutional neural network based on deep learning to perform high-precision segmentation of the molten pool slag surface and the molten pool copper solution surface on the image data; The angle detection and compensation module is used to detect and compensate the influence of the camera shooting angle change on the measurement result in real time, and includes an angle sensor for monitoring the tilt angle change of the probe to compensate for the measurement error; The nonlinear data processing module is used to perform nonlinear fitting and error correction on the measured data, which includes a polynomial fitting algorithm model, and is used to collect the measured data of the molten pool level, including the probe measurement data and the actual liquid level height, to ensure that the data covers the possible operating range, and to standardize the data, using the probe measurement data as input data and the actual liquid level height output data for normalization processing, in order to improve the accuracy of liquid level height measurement; The system testing and optimization module is used to evaluate the overall performance of the system and perform corresponding optimization. When the measurement error exceeds a preset threshold, the system can send an alarm signal and automatically adjust parameters.

[0007] Also includes a display module, a communication module and a control module; The control module is electrically connected to the data acquisition module, the data preprocessing module, the model training and optimization module, the angle detection and compensation module, the nonlinear data processing module, the system testing and optimization module and the display module respectively through the communication module; The display module includes a field operation screen, which is used to display and monitor measurement data in real time; The communication module is used to control the data transmission between the control module and the data acquisition module, the data preprocessing module, the model training and optimization module, the angle detection and compensation module, the nonlinear data processing module, the system testing and optimization module and the display module; The control module is used for signal triggering, data transmission and automatic control logic.

[0008] The control module includes a trigger signal processing module, a data storage and transmission module, and a control logic module; wherein: The trigger signal processing module is used to detect the trigger signal of the probe and start the image capture process; The data storage and transmission module is used to manage the coil triggering and the data reading and writing of the holding register, ensuring data synchronization with the on-site operation screen; The control logic module is used to trigger the coil signal according to the detection position arrival signal on the probe, and display the slag surface height and copper solution surface height information of the molten pool after the holding register receives the written data.

[0009] A method for online monitoring of the liquid level on the surface of a molten pool, using the above monitoring system, comprises the following steps: Step 1: Data acquisition: When the probe in the data acquisition module moves to the height to be detected, the probe generates a trigger signal through the coil, and the signal is transmitted to the control module through the communication module. After receiving the trigger signal, the control module immediately issues a command to start the industrial camera in the data acquisition module to capture the image. The camera transmits the captured image data of the slag surface and the copper solution surface of the molten pool to the display module through the communication module to realize real-time display of the image; Step 2: Data preprocessing: Use the data preprocessing module to normalize at least 1,000 images containing the slag surface and the copper solution surface of the molten pool to reduce the interference of external factors such as light and temperature and improve the consistency of the data; input the processed image data into the trained and optimized image segmentation processing model, perform image segmentation based on the U-Net model, and output the segmentation results of the slag surface and the copper solution surface of the molten pool; extract the boundary information of the slag surface and the copper solution surface of the molten pool from the segmentation results for subsequent liquid level calculation; Step 3: Model construction and training: Use the model training and optimization module to adjust the model structure and hyperparameters to adapt to the data characteristics and ensure that the model can process image data in complex process environments; use cross-validation technology to prevent the model from overfitting and ensure that it has good performance on the test set; Specifically, the hyperparameters include learning rate, batch size, etc. Response time optimization: Test the measurement data of the probe, correct the error through the nonlinear fitting model, and ensure that the ratio of the system measurement error to the manual measurement error does not exceed 10%; evaluate the delay from the probe trigger to the data transmission, and comprehensively consider the temperature decay rate of the probe; Camera parameter calibration: by adjusting the external and internal parameters of the camera, calibrate the data acquisition module to ensure that the image geometry accuracy is consistent with the actual environment; Specifically, the external parameters include the standard position of the distance probe and the six standard angle values, and the internal parameters include the focal length, distortion coefficient, and scale; Step 4: Angle detection and compensation: Use the angle detection and compensation module to detect the angle change of the probe in real time, and use the compensation calculation method to correct the liquid level measurement error; Specifically, the angle sensor is used to monitor the tilt angle of the probe in real time and compensate for it. In addition, according to the measurement errors caused by the changes in the angle of the probe caused by different on-site production conditions, the corresponding angle compensation logic is developed. The tilt angle of the probe is divided into six reference standards: upper left, middle left, lower left, upper right, middle right, and lower right, so as to overcome the measurement errors caused by the angle changes on the two-dimensional plane of the probe. By calibrating with the standard ruler, the deviation calibration compensation on the three-dimensional surface caused by the forward and backward tilt movement of the probe is ensured, so as to ensure the accuracy of the angle compensation calculation of the measurement data. Step 5: Error processing: Use the nonlinear data processing module to select the nonlinear fitting model parameter threshold, train the model according to the data distribution characteristics, apply the fitting model to real-time data, perform error correction, and improve measurement accuracy; The nonlinear data processing module includes a polynomial fitting algorithm model, which collects the measured data of the molten pool level, including the probe measurement data and the actual liquid level height, ensures that the data covers the possible operating range, standardizes the data and uses the probe measurement data as input and the actual liquid level height as output for normalization: in, To measure data for the probe; Choose an nth-order polynomial for the nonlinear relationship: in, is the actual liquid level height; Secondly, the order of the error function is constructed. Too low an order may lead to underfitting, while too high an order may lead to overfitting and calculation of fitting errors. After testing, the appropriate order is selected. The error function is: in, is the optimization coefficient, To minimize the error, For the probe measurement data, is the actual liquid level height; The algorithm is further fitted and the polynomial fitting is converted into a linear algebra matrix solution method. Finally, the performance of the fitting model is verified through a test data set to improve the accuracy of liquid level measurement. Real-time acquisition of original measurement data, through nonlinear fitting, calculates the corrected data, and outputs the corrected liquid level height. The error between the corrected liquid level height and the actual manual measurement is no more than 10%, which is used for subsequent control and display; Step 6: System Optimization: The system testing and optimization module is used to evaluate system performance and optimize it, using mean square error (MSE) and coefficient of determination. Optimize and modify the model; When the measurement error exceeds the preset threshold, an alarm signal is sent and the system parameters are automatically adjusted; performance evaluation detects the stability, response speed and measurement accuracy of the system under different working conditions by running tests; parameter optimization automatically adjusts the system parameters according to the evaluation results to ensure the optimal performance of the system in the actual production environment; alarm mechanism When the system measurement error or response time exceeds the preset threshold, an alarm signal is triggered to prompt the operator to check and adjust.

[0010] The beneficial effects of the present invention are: The present invention determines the slurry hanging height of the molten copper liquid surface and the slag surface on the probe by the visible light color of the molten copper liquid surface and the slag surface of the molten copper pool, obtains the liquid level height data of the slag surface of the molten copper pool in real time and accurately, and timely adjusts and optimizes the production process parameters of the electric furnace process, and checks and handles the production abnormality alarm, so that the production organization of the electric furnace process is more efficient and the emergency disposal is more timely; specifically: 1. The present invention realizes online real-time monitoring of the liquid level of the molten pool surface by setting a data acquisition module, thereby improving the level of production automation. When the data acquisition module reaches the detection position, the control system triggers a signal through the coil, and the system automatically turns on the camera to capture the image of the current slag surface and the copper solution surface of the molten pool, thereby realizing the online real-time monitoring function of the liquid level of the molten pool surface. Compared with the traditional manual measurement or intermittent monitoring method, the present invention can reduce manual intervention, improve the degree of automation of the production process, and significantly improve the detection efficiency; 2. The present invention segments the collected images by setting a data preprocessing module and a model training and optimization module, accurately measures the height of the slag surface and the copper solution surface of the molten pool, and adjusts the built-in parameters and external parameters of the camera and optimizes the correlation coefficient to adapt to the complex process environment of the copper smelting electric furnace. Even under adverse conditions such as high temperature, dust and light changes, the slag surface and the copper solution surface of the molten pool can still be accurately segmented. Compared with the problem of reduced measurement accuracy of traditional technologies in complex environments, the present invention ensures that the measurement of the slag surface and the copper solution surface of the molten pool can still maintain high-precision image segmentation under complex working conditions, ensuring the measurement accuracy of the molten pool liquid level; 3. The present invention eliminates the measurement error caused by the change of the probe angle by setting an angle detection and compensation module, and ensures the consistency and accuracy of the height measurement by accurately calculating and correcting the angle deviation of the probe and calibrating it with the standard ruler. Compared with the traditional method without the angle correction function, the present invention effectively reduces the measurement error caused by the tilt of the probe, and ensures the consistency and reliability of the measurement data by calibrating it with the standard ruler. 4. The present invention improves the overall accuracy of the system by setting up a nonlinear data processing module. The nonlinear data processing module can process and compare the actual measurement data and perform data correction for nonlinear changes in complex process environments. Compared with traditional linear processing methods, it can effectively cope with nonlinear changes under complex process conditions, significantly improve the overall accuracy of the system, and meet the needs of high-precision measurement. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] Figure 1 It is a schematic diagram of the structure of the system of the present invention.

[0012] In the figure, 1-data acquisition module, 2-data preprocessing module, 3-model training and optimization module, 4-angle detection and compensation module, 5-nonlinear data processing module, 6-system testing and optimization module, 7-display module, 8-communication module, 9-control module. DETAILED DESCRIPTION

[0013] The technical solution of the present invention is described below in conjunction with the accompanying drawings and implementation methods.

[0014] like Figure 1 As shown, a molten pool surface liquid level online monitoring system includes a data acquisition module 1, a data preprocessing module 2, a model training and optimization module 3, an angle detection and compensation module 4, a nonlinear data processing module 5 and a system testing and optimization module 6; wherein: The data acquisition module 1 is used to collect image data of the slag surface and the copper solution surface of the molten pool, and includes a probe and an industrial camera that can work in a high temperature environment and has the function of measuring the liquid level of the copper solution surface and the slag surface of the molten pool.

[0015] The data preprocessing module 2 is used to normalize and enhance the collected images through an image segmentation model to achieve image segmentation of the slag surface of the molten pool and the copper solution surface of the molten pool.

[0016] The model training and optimization module 3 is used to build, train and optimize the image segmentation model. It uses the deep learning U-Net convolutional neural network to train and optimize the image segmentation model to ensure high-precision molten pool liquid level segmentation. When training the model, a large amount of molten pool slag surface and molten pool copper solution surface image data is required so that the model can accurately segment the image data.

[0017] The angle detection and compensation module 4 is used to monitor the tilt angle of the probe in real time through the angle sensor and compensate for it. In addition, according to the change of the probe angle in different on-site production conditions, there is a measurement error. The corresponding angle compensation logic is developed. By monitoring the tilt angle of the probe, it is divided into six reference standards: upper left, middle left, lower left, upper right, middle right, and lower right, so as to overcome the measurement error caused by the angle change on the two-dimensional plane of the probe. By calibrating with the standard ruler, the deviation on the three-dimensional surface caused by the forward and backward tilt movement of the probe is calibrated and compensated, thereby ensuring the accuracy of the measurement data.

[0018] The nonlinear data processing module 5 is used for performing nonlinear fitting and error correction on the measured data, which includes a polynomial fitting algorithm model, and the model collects the measured data of the molten pool level, including the probe measurement data and the actual liquid level height, to ensure that the data covers the possible operating range.

[0019] The system testing and optimization module 6 is used to evaluate system performance and optimize it. This embodiment uses the mean square error MSE and the coefficient of determination , optimize and correct the model, and when the measurement error exceeds the preset threshold, send an alarm signal and automatically adjust the system parameters.

[0020] The present invention further comprises a display module 7 , a communication module 8 and a control module 9 .

[0021] The control module 9 realizes electrical connection and data interaction with the data acquisition module 1, the data preprocessing module 2, the model training and optimization module 3, the angle detection and compensation module 4, the nonlinear data processing module 5, the system testing and optimization module 6 and the display module 7 through the communication module 8.

[0022] The display module 7 includes a field operation screen, which is used to display and monitor the measurement data in real time; the display module 7 displays the visible light images of the current molten pool slag surface and copper solution surface and the captured probe image, displays the identified molten pool slag surface and copper solution surface height data, and records and displays the historical identification results. Through a clear and intuitive interface, it provides operators with real-time data monitoring and trend analysis support, which facilitates timely adjustment of production processes.

[0023] The communication module 8 is used for stable and efficient communication between the control module 9 and the data acquisition module 1, the data preprocessing module 2, the model training and optimization module 3, the angle detection and compensation module 4, the nonlinear data processing module 5, the system testing and optimization module 6 and the display module 7, so as to realize data sharing among different modules and ensure the coordination and real-time performance of the whole system.

[0024] The control module 9 is used for signal triggering, data transmission and automatic control logic; specifically, the control module 9 adopts a PLC controller, and its program design uses ladder diagrams and function block diagrams for programming.

[0025] The control module 9 includes a trigger signal processing module, a data storage and transmission module, and a control logic module; wherein the trigger signal processing module is responsible for detecting the trigger signal of the probe and starting the image capture process to ensure the real-time data acquisition; the data storage and transmission module is used to manage the coil trigger and the data reading and writing of the holding register, to achieve data synchronization with the on-site operation screen, and to ensure the consistency of the overall operation of the system; the control logic module is used to trigger the coil signal according to the signal that the probe reaches the detection position, and after the holding register receives the written data, the molten pool slag surface height and the molten pool copper solution surface height information are fed back to the display module 7 in real time for the operator to view.

[0026] A method for online monitoring of the liquid level on the surface of a molten pool, using the above monitoring system, comprises the following steps: Step 1, data acquisition: When the probe in the data acquisition module 1 moves to the height to be detected, the probe generates a trigger signal through the coil, and the signal is transmitted to the control module 9 through the communication module 8. After receiving the trigger signal, the control module 9 immediately issues an instruction to start the industrial camera in the data acquisition module 1 to capture the image. The camera transmits the captured image data of the slag surface and the copper solution surface of the molten pool to the display module 7 through the communication module 8 to realize real-time display of the image.

[0027] Step 2: Data preprocessing: Use data preprocessing module 2 to normalize at least 1,000 image data containing the slag surface and the copper solution surface of the molten pool to reduce the interference of external factors such as light and temperature and improve data consistency; input the processed image data into the trained and optimized image segmentation processing model, perform image segmentation based on the U-Net model, and output the segmentation results of the slag surface and the copper solution surface of the molten pool; extract the boundary information of the slag surface and the copper solution surface of the molten pool from the segmentation results for subsequent liquid level height calculation.

[0028] Step 3: Model construction and training: Use the model training and optimization module 3 to adjust the model structure and hyperparameters to adapt to the data characteristics and ensure that the model can process image data in complex process environments; use cross-validation technology to prevent the model from overfitting and ensure that it has good performance on the test set; Specifically, the hyperparameters include learning rate, batch size, etc.

[0029] Response time optimization: Test the measurement data of the probe, correct the error through the nonlinear fitting model, and ensure that the ratio of the system measurement error to the manual measurement error does not exceed 10%; evaluate the delay from the probe trigger to the data transmission, and comprehensively consider the temperature decay speed of the probe. The system response time is set to 20ms.

[0030] Camera parameter calibration: by adjusting the external and internal parameters of the camera, calibrate the data acquisition module 1 to ensure that the image geometry accuracy is consistent with the actual environment; Specifically, the external parameters include the standard position of the distance probe, six standard angle values, etc., and the internal parameters include focal length, distortion coefficient, scale, etc.

[0031] In addition, in order to adapt to the changes in illumination in the copper smelting furnace process environment, the correlation coefficient of the image processing model is adjusted to improve the accuracy of the segmentation results.

[0032] Step 4: Angle detection and compensation: The angle detection and compensation module 4 is used to detect the angle change of the probe in real time, and the compensation calculation method is used to correct the liquid level measurement error.

[0033] Specifically, the inclination angle of the probe is monitored in real time and compensated by an angle sensor, and corresponding angle compensation logic is developed for the measurement errors caused by the changes in the probe angle due to different on-site production conditions. The inclination angle of the probe is divided into six reference standards of upper left, middle left, lower left, upper right, middle right and lower right to overcome the measurement error caused by the angle change on the two-dimensional plane of the probe, and the deviation on the three-dimensional surface caused by the forward and backward tilting movement of the probe is calibrated and compensated by calibration with a standard ruler to ensure the accuracy of the angle compensation calculation of the measurement data; the angle deviation of the probe is accurately calculated and corrected, and calibrated with a standard ruler to ensure the consistency and accuracy of the height measurement. Compared with the traditional method that does not have the angle correction function, the present invention effectively reduces the measurement error caused by the inclination of the probe, and the consistency and reliability of the measurement data are ensured by calibration with a standard ruler. According to the angle deviation of the probe, the correction value of the actual liquid level height is calculated to ensure the accuracy of the measurement result.

[0034] Step 5: Error processing: Use the nonlinear data processing module 5 to select an appropriate nonlinear fitting model parameter threshold, train the model according to the data distribution characteristics, apply the fitting model to real-time data, perform error correction, and improve measurement accuracy.

[0035] The nonlinear data processing module 5 includes a polynomial fitting algorithm model, which collects the measured data of the molten pool level, including the probe measurement data and the actual liquid level height, ensures that the data covers the possible operating range, standardizes the data and uses the probe measurement data as input and the actual liquid level height as output for normalization processing: in, To measure data for the probe; Choose an nth-order polynomial for the nonlinear relationship: in, is the actual liquid level height; Secondly, the order of the error function is constructed. Too low an order may lead to underfitting, while too high an order may lead to overfitting and calculation of fitting errors. After testing, the appropriate order is selected. The error function is: in, is the optimization coefficient, To minimize the error, For the probe measurement data, is the actual liquid level height; The algorithm is further fitted and the polynomial fitting is converted into a linear algebra matrix solution method. Finally, the fitting model performance is verified through a test data set to improve the accuracy of liquid level measurement.

[0036] The original measurement data is collected in real time and the corrected data is calculated through nonlinear fitting. The error between the output corrected liquid level height and the actual manual measurement is no more than 10%, which is used for subsequent control and display.

[0037] Step 6: System Optimization: System Testing and Optimization Module 6 is used to evaluate system performance and optimize it, using mean square error (MSE) and coefficient of determination. , optimize and correct the model.

[0038] When the measurement error exceeds the preset threshold, an alarm signal is sent and the system parameters are automatically adjusted; performance evaluation detects the stability, response speed and measurement accuracy of the system under different working conditions by running tests; parameter optimization automatically adjusts the system parameters according to the evaluation results to ensure the optimal performance of the system in the actual production environment; alarm mechanism When the system measurement error or response time exceeds the preset threshold, an alarm signal is triggered to prompt the operator to check and adjust.

Claims

1. An online monitoring system for the liquid level of a molten pool, characterized in that: include: Data acquisition module (1), data preprocessing module (2), model training and optimization module (3), angle detection and compensation module (4), nonlinear data processing module (5) and system testing and optimization module (6); The data acquisition module (1) is used to collect image data of the slag surface of the molten pool and the liquid level of the copper solution in the molten pool, and comprises an industrial camera and a probe that can operate in a high temperature environment, the probe being used to measure and detect the liquid level position of the slag surface of the molten pool and the liquid level of the copper solution in the molten pool, and the camera being used to capture images of the slag surface of the molten pool and the liquid level of the copper solution in the molten pool; The data preprocessing module (2) is used to perform normalization and enhancement processing on the image data to improve the accuracy of subsequent processing. It adopts an image segmentation processing model to achieve image segmentation of the slag surface of the molten pool and the copper solution surface of the molten pool; The model training and optimization module (3) is used to construct, train and optimize the image segmentation model, which uses a U-Net convolutional neural network based on deep learning to perform high-precision segmentation of the molten pool slag surface and the molten pool copper solution surface on the image data; The angle detection and compensation module (4) is used to detect and compensate in real time the influence of the camera shooting angle change on the measurement result, and comprises an angle sensor for monitoring the tilt angle change of the probe, so as to compensate for the measurement error; The nonlinear data processing module (5) is used to perform nonlinear fitting and error correction on the measured data, and includes a polynomial fitting algorithm model, which is used to collect the measured data of the molten pool liquid level, including the probe measurement data and the actual liquid level height, to ensure that the data covers the possible operating range, to standardize the data, to use the probe measurement data as input data and to normalize the actual liquid level height output data, in order to improve the accuracy of liquid level height measurement; The system testing and optimization module (6) is used to evaluate the overall performance of the system and perform corresponding optimization. When the measurement error exceeds a preset threshold, the system can send an alarm signal and automatically adjust parameters.

2. The on-line monitoring system for the liquid level of the molten pool surface according to claim 1 is characterized in that: It also includes a display module (7), a communication module (8) and a control module (9); The control module (9) is electrically connected to the data acquisition module (1), the data preprocessing module (2), the model training and optimization module (3), the angle detection and compensation module (4), the nonlinear data processing module (5), the system testing and optimization module (6), and the display module (7) respectively through the communication module (8); The display module (7) comprises an on-site operation screen, which is used to display and monitor the measurement data in real time; The communication module (8) is used to control the control module (9) to respectively transmit data with the data acquisition module (1), the data preprocessing module (2), the model training and optimization module (3), the angle detection and compensation module (4), the nonlinear data processing module (5), the system testing and optimization module (6) and the display module (7); The control module (9) is used for signal triggering, data transmission and automatic control logic.

3. The on-line monitoring system for the molten pool surface level according to claim 2 is characterized in that: The control module (9) comprises a trigger signal processing module, a data storage and transmission module, and a control logic module; wherein: The trigger signal processing module is used to detect the trigger signal of the probe and start the image capture process; The data storage and transmission module is used to manage the coil triggering and the data reading and writing of the holding register, ensuring data synchronization with the on-site operation screen; The control logic module is used to trigger the coil signal according to the detection position arrival signal on the probe, and display the slag surface height and copper solution surface height information of the molten pool after the holding register receives the written data.

4. A method for online monitoring of the liquid level on the molten pool surface, characterized in that: The monitoring system according to claim 3 is applied, comprising the following steps: The first step is data acquisition: when the probe in the data acquisition module (1) moves to the height position to be detected, the probe generates a trigger signal through the coil, and the signal is transmitted to the control module (9) through the communication module (8). After receiving the trigger signal, the control module (9) immediately issues an instruction to start the industrial camera in the data acquisition module (1) to capture images. The camera transmits the captured image data of the slag surface and the copper solution surface of the molten pool to the display module (7) through the communication module, so as to realize real-time display of the image; Step 2: Data preprocessing: using the data preprocessing module (2), normalizing at least 1,000 images of the slag surface and the copper solution surface of the molten pool to reduce interference from external factors such as light and temperature and improve data consistency; inputting the processed image data into a trained and optimized image segmentation processing model, performing image segmentation based on a U-Net model, and outputting the segmentation results of the slag surface and the copper solution surface of the molten pool; extracting the boundary information of the slag surface and the copper solution surface of the molten pool from the segmentation results for subsequent liquid level calculation; Step 3: Model construction and training: Use the model training and optimization module (3) to adjust the model structure and hyperparameters to adapt to the data characteristics and ensure that the model can process image data in complex process environments. Use cross-validation technology to prevent the model from overfitting and ensure that it has good performance on the test set. Specifically, the hyperparameters include learning rate, batch size, etc. Response time optimization: Test the measurement data of the probe, correct the error through the nonlinear fitting model, and ensure that the ratio of the system measurement error to the manual measurement error does not exceed 10%; evaluate the delay from the probe trigger to the data transmission, and comprehensively consider the temperature decay rate of the probe; Camera parameter calibration: by adjusting the external and internal parameters of the camera, calibrate the data acquisition module (1) to ensure that the image geometry accuracy is consistent with the actual environment; Specifically, the external parameters include the standard position of the distance probe and the six standard angle values, and the internal parameters include the focal length, distortion coefficient, and scale; Step 4: Angle detection and compensation: Use the angle detection and compensation module (4) to detect the angle change of the probe in real time, and use the compensation calculation method to correct the liquid level measurement error; Specifically, the angle sensor is used to monitor the tilt angle of the probe in real time and compensate for it. In addition, according to the measurement errors caused by the changes in the angle of the probe caused by different on-site production conditions, the corresponding angle compensation logic is developed. The tilt angle of the probe is divided into six reference standards: upper left, middle left, lower left, upper right, middle right, and lower right, so as to overcome the measurement errors caused by the angle changes on the two-dimensional plane of the probe. By calibrating with the standard ruler, the deviation calibration compensation on the three-dimensional surface caused by the forward and backward tilt movement of the probe is ensured, so as to ensure the accuracy of the angle compensation calculation of the measurement data. Step 5: Error processing: Use the nonlinear data processing module (5) to select the threshold of the nonlinear fitting model parameters, train the model according to the data distribution characteristics, apply the fitting model to the real-time data, perform error correction, and improve the measurement accuracy; The nonlinear data processing module (5) includes a polynomial fitting algorithm model, which collects the measured data of the molten pool level, including the probe measurement data and the actual liquid level height, ensures that the data covers the possible operating range, standardizes the data and uses the probe measurement data as input and the actual liquid level height as output for normalization processing: in, To measure data for the probe; Choose an nth-order polynomial for the nonlinear relationship: in, is the actual liquid level height; Secondly, the order of the error function is constructed. Too low an order may lead to underfitting, while too high an order may lead to overfitting and calculation of fitting errors. After testing, the appropriate order is selected. The error function is: in, is the optimization coefficient, To minimize the error, For the probe measurement data, is the actual liquid level height; The algorithm is further fitted and the polynomial fitting is converted into a linear algebra matrix solution method. Finally, the performance of the fitting model is verified through a test data set to improve the accuracy of liquid level measurement. Real-time acquisition of original measurement data, through nonlinear fitting, calculates the corrected data, and outputs the corrected liquid level height. The error between the corrected liquid level height and the actual manual measurement is no more than 10%, which is used for subsequent control and display; Step 6: System Optimization: The system testing and optimization module (6) is used to evaluate system performance and optimize it, using mean square error (MSE) and coefficient of determination (CDR). Optimize and modify the model; When the measurement error exceeds a preset threshold, the control module (9) sends an alarm signal and automatically adjusts the system parameters; the performance evaluation detects the stability, response speed and measurement accuracy of the system under different working conditions by running tests; the parameter optimization automatically adjusts the system parameters according to the evaluation results to ensure the optimal performance of the system in the actual production environment; the alarm mechanism triggers an alarm signal when the system measurement error or response time exceeds a preset threshold, prompting the operator to check and adjust.

Images