Machine vision-based molten steel temperature control system and control method
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- MCC CAPITAL ENGINEERING & RESEARCH INC LTD
- Filing Date
- 2025-01-20
- Publication Date
- 2026-07-21
AI Technical Summary
Traditional contact temperature sensors are easily damaged, inaccurate, and affect the uniformity of molten steel temperature detection. Existing machine vision systems are complex, costly, and difficult to deploy in industrial settings.
A machine vision-based steel temperature control system is adopted, which uses a CCD camera to acquire flame images, edge computing devices to perform image preprocessing and deep learning model inference, and a host computer to perform temperature control. The temperature control is optimized through a lightweight convolutional neural network and a PID controller.
It achieves high-accuracy detection of molten steel temperature, reduces human error, ensures precise temperature control, improves production efficiency and product quality, and reduces system complexity and cost.
Smart Images
Figure CN122431440A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of metallurgical automation control technology, specifically to a steel temperature control system and control method based on machine vision. Background Technology
[0002] The iron and steel metallurgical industry is the foundation of modern industry, and accurate detection of molten steel temperature is crucial for ensuring the quality of steel products and production efficiency. Traditional methods for detecting molten steel temperature typically rely on contact temperature sensors, such as thermocouples or infrared thermometers. While these methods can provide temperature information to some extent, they have some inherent limitations.
[0003] First, contact sensors can be affected by the high temperature and corrosive environment of molten steel, leading to sensor damage or inaccurate measurement data. Second, contact measurements may interfere with the flow of molten steel, affecting its uniformity and quality. Furthermore, these methods typically require frequent maintenance and calibration, increasing operating costs.
[0004] To overcome these challenges, non-contact temperature detection technologies have emerged. Among them, machine vision-based methods have attracted attention due to their ability to provide real-time, non-destructive detection. Machine vision technology infers temperature by analyzing image or spectral information of the molten steel surface, avoiding direct contact with the molten steel. Although machine vision technology has many theoretical advantages, it still faces some challenges in practical applications. For example, the brightness and color of the molten steel surface can be affected by various factors, such as furnace lighting conditions, steel composition, and the formation of oxide scale. These factors can increase the complexity and uncertainty of image analysis. Furthermore, existing machine vision systems often require complex image processing algorithms and high-performance computing resources, which not only increases the complexity and cost of the system but also limits its deployment and application in industrial settings. Summary of the Invention
[0005] To address the problems in the prior art, this application provides a machine vision-based steel temperature control system and method, which can at least partially solve the problems existing in the prior art.
[0006] On the one hand, this application proposes a machine vision-based steel temperature control system, comprising:
[0007] CCD camera to capture real-time images of flames at the furnace mouth of molten steel smelting furnace;
[0008] An edge computing device is connected to the CCD camera and is equipped with a flame recognition model. The edge computing device acquires the furnace opening flame image and uses the flame recognition model to identify the current molten steel temperature based on the furnace opening flame image. The flame recognition model is obtained by training a preset convolutional neural network based on the furnace opening flame image.
[0009] The host computer is connected to the edge computing device to obtain the current molten steel temperature and adjust the heating parameters of the molten steel furnace according to the current molten steel temperature in order to control the molten steel temperature in the molten steel furnace.
[0010] In some embodiments, the edge computing device is communicatively connected to the CCD camera via a high-speed industrial bus communication module.
[0011] In some embodiments, the CCD camera communicates with the edge computing device via GMSL, and the edge computing device communicates with the host computer via USB.
[0012] On the other hand, this application proposes a machine vision-based method for controlling molten steel temperature, based on the machine vision-based molten steel temperature control system described in any of the above embodiments, the method comprising:
[0013] CCD cameras capture real-time images of flames at the furnace opening of steel smelting furnaces;
[0014] Edge computing devices use a flame recognition model to identify the current temperature of molten steel based on the furnace flame image;
[0015] The host computer adjusts the heating parameters of the molten steel furnace according to the current molten steel temperature in order to control the temperature of the molten steel in the molten steel furnace.
[0016] In some embodiments, the edge computing device uses a flame recognition model to identify the current molten steel temperature based on the furnace opening flame image, including:
[0017] The furnace flame image is preprocessed to obtain the target image;
[0018] The target image is input into the flame recognition model, which then identifies the current molten steel temperature based on the furnace flame image.
[0019] In some embodiments, the edge computing device preprocesses the furnace flame image to obtain a target image, including:
[0020] The region of interest is selected from the furnace flame image to obtain an intermediate image;
[0021] The intermediate image is then subjected to noise reduction, brightness adjustment, and contrast enhancement to obtain the target image.
[0022] In some embodiments, the training process of the flame recognition model is as follows:
[0023] Obtain N images of the flame at the furnace opening of the molten steel furnace at N time points and the temperature of the molten steel in the molten steel furnace measured by a temperature measuring gun, where N is a positive integer greater than 2;
[0024] The furnace flame image at each moment is preprocessed to obtain the target image;
[0025] The target image corresponding to the furnace flame image at a certain moment is input into a preset convolutional neural network to obtain the steel temperature at that moment predicted by the convolutional neural network.
[0026] Based on the mapping table between flame characteristics and molten steel temperature, determine the molten steel temperature corresponding to the flame characteristics of the furnace mouth flame image at that moment.
[0027] Based on the flame characteristics of the furnace flame image at that moment, corresponding to the molten steel temperature, and the molten steel temperature in the smelting furnace measured by the temperature measuring gun at that moment, the parameters of the convolutional neural network are adjusted to obtain the first updated convolutional neural network.
[0028] The target image corresponding to the furnace flame image at another time is then input into the updated convolutional neural network to obtain the flame features extracted by the first updated convolutional neural network.
[0029] Based on the mapping table between flame characteristics and molten steel temperature, determine the molten steel temperature corresponding to the flame characteristics of the furnace mouth flame image at another moment.
[0030] Based on the flame characteristics of the furnace flame image at another moment corresponding to the molten steel temperature and the molten steel temperature in the smelting furnace measured by the temperature measuring gun at another moment, the parameters of the first updated convolutional neural network are adjusted to obtain the second updated convolutional neural network.
[0031] This process continues until the Nth updated convolutional neural network is obtained, which is then used as the flame recognition model.
[0032] In some embodiments, the host computer performs the training process of the flame recognition model.
[0033] In some embodiments, the host computer adjusts the heating parameters of the molten steel furnace according to the current molten steel temperature to control the molten steel temperature in the molten steel furnace, including:
[0034] Determine whether the current temperature of the molten steel has reached the preset target temperature;
[0035] If the current molten steel temperature does not reach the preset target temperature, the PID controller dynamically adjusts the oxygen blowing volume of the oxygen lance in real time according to the current molten steel temperature and the target temperature in order to control the molten steel temperature in the smelting furnace.
[0036] In some embodiments, after determining whether the current molten steel temperature has reached a preset target temperature, the method further includes:
[0037] If the current temperature of the molten steel reaches the preset target temperature, the oxygen blowing lance will be controlled to stop blowing oxygen.
[0038] The machine vision-based steel temperature control system and method provided in this application uses an edge computing device as its core. A CCD camera is responsible for acquiring images of the furnace flame, while the edge computing device handles image preprocessing and deep learning model inference calculations. A host computer monitors and controls the entire system. The edge computing device is equipped with a lightweight embedded AI framework, directly running a complex neural network model for health monitoring of industrial equipment. This design allows the model inference calculation process to be completed on the edge computing device, occupying almost no computing resources of the host computer, thus improving the system's response speed and data processing efficiency. By employing advanced image recognition technology and deep learning algorithms, precise image analysis of the molten steel surface can be performed, achieving high-accuracy detection of molten steel temperature, reducing human error, ensuring the precision of temperature control during steel production, and ultimately improving product quality and production efficiency. Attached Figure Description
[0039] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. In the drawings:
[0040] Figure 1 This is a schematic diagram of a machine vision-based steel temperature control system provided in one embodiment of this application.
[0041] Figure 2 This is a schematic flowchart of a machine vision-based steel temperature control method provided in an embodiment of this application.
[0042] Figure 3 This is a partial flowchart of a machine vision-based steel temperature control method provided in one embodiment of this application.
[0043] Figure 4 This is a partial flowchart of a machine vision-based steel temperature control method provided in one embodiment of this application.
[0044] Figure 5 This is a partial flowchart of a machine vision-based steel temperature control method provided in one embodiment of this application.
[0045] Figure 6 This is a partial flowchart of a machine vision-based steel temperature control method provided in one embodiment of this application.
[0046] Figure 7 This is a control logic block diagram of a machine vision-based molten steel temperature control system provided in one embodiment of this application.
[0047] Figure 8 This is a schematic diagram of the physical structure of an electronic device provided in an embodiment of this application. Detailed Implementation
[0048] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the embodiments of this application will be further described in detail below with reference to the accompanying drawings. Here, the illustrative embodiments and their descriptions are used to explain this application, but are not intended to limit this application. It should be noted that, unless otherwise specified, the embodiments and features in the embodiments of this application can be arbitrarily arranged.
[0049] The terms “first,” “second,” etc., used in this document are not intended to specifically refer to order or sequence, nor are they used to limit this application; they are merely used to distinguish elements or operations described using the same technical terms.
[0050] The terms “include,” “including,” “have,” “contain,” etc., used in this article are all open-ended terms, meaning that they include but are not limited to.
[0051] The term "and / or" as used in this document includes any or all of the items mentioned.
[0052] Figure 1 This is a schematic diagram of a machine vision-based steel temperature control system according to an embodiment of this application, as shown below. Figure 1 As shown in the embodiment of this application, the machine vision-based steel temperature control system includes:
[0053] CCD (Charge-Coupled Device) camera 10, which acquires real-time images of flames at the furnace mouth of a steel smelting furnace;
[0054] Edge computing device 20 is connected to CCD camera 10 and is equipped with a flame recognition model. Edge computing device 20 acquires the furnace mouth flame image and uses the flame recognition model to identify the current molten steel temperature based on the furnace mouth flame image. The flame recognition model is obtained by training a preset convolutional neural network based on the furnace mouth flame image.
[0055] The host computer 30 is connected to the edge computing device 20 to obtain the current molten steel temperature and adjust the heating parameters of the molten steel furnace according to the current molten steel temperature in order to control the molten steel temperature in the molten steel furnace.
[0056] Specifically, the CCD camera 10 can be a high-resolution industrial CCD camera, which acquires images of the furnace flame 1 in real time. These images are a key data source for analyzing the temperature of molten steel. The image acquisition frequency of the camera can be adjusted according to the specific needs of the steelmaking process to ensure accurate capture of flame images in rapidly changing temperature environments. A deep learning model (flame recognition model) can be deployed on an edge computing device 20 such as an NVIDIA Jetson to analyze the images using deep learning technology. By recognizing the characteristics of the flame at different temperatures, the image data is converted into temperature readings. This temperature data is sent to a host computer 30 in real time. Based on the received temperature information, the host computer 30 automatically adjusts the heating parameters of the steelmaking furnace to precisely control the temperature of the molten steel in the furnace. The preset convolutional neural network can be a lightweight CNN neural network for easy deployment in mobile or embedded devices.
[0057] In other words, the machine vision-based molten steel temperature control system provided in this application embodiment uses an edge computing device 20 as its core. A CCD camera 10 is responsible for acquiring images of the furnace flame, while the edge computing device 20 handles image preprocessing and deep learning model inference calculations. The host computer 30 monitors and controls the entire system. The edge computing device 20 is equipped with a lightweight embedded AI framework, directly running a complex neural network model used for health monitoring of industrial equipment. This design allows the model inference calculation process to be completed on the edge computing device 20, consuming almost no computing resources from the host computer 30, thus improving the system's response speed and data processing efficiency.
[0058] Therefore, the machine vision-based molten steel temperature control system provided in this application, by employing advanced image recognition technology and deep learning algorithms, can perform precise image analysis on the surface of molten steel, thereby achieving high-accuracy detection of molten steel temperature, reducing human error, ensuring the accuracy of temperature control during steel production, and thus improving product quality and production efficiency.
[0059] As part of the intelligent steelmaking process, this system can be seamlessly integrated with existing metallurgical automation control systems, providing closed-loop feedback for temperature control.
[0060] In some embodiments, the edge computing device 20 can communicate with the CCD camera 10 via a high-speed industrial bus communication module.
[0061] Specifically, the edge computing device 20 communicates with the CCD camera 10 through a high-speed industrial bus communication module to acquire multi-source heterogeneous data, and uses the deployed deep learning model to perform real-time calculations. The calculation results are directly fed back to the host computer 30, realizing intelligent control with low latency and high security.
[0062] In some embodiments, the CCD camera 10 communicates with the edge computing device 20 via GMSL, and the edge computing device 20 communicates with the host computer 30 via USB.
[0063] Specifically, this embodiment optimizes the communication methods between the CCD camera 10 and the edge computing device 20, as well as between the edge computing device 20 and the host computer 30. The selection of these communication methods optimizes data transmission efficiency and system performance.
[0064] Based on the same inventive concept, this application also provides a machine vision-based method for controlling molten steel temperature, which is based on the machine vision-based molten steel temperature control system described in any of the above embodiments. Figure 2 This is a schematic flowchart of a machine vision-based steel temperature control method according to an embodiment of this application, as shown below. Figure 2 As shown in the embodiments of this application, the steel temperature control method based on machine vision includes:
[0065] S1: CCD camera acquires real-time images of flames at the furnace mouth of molten steel smelting furnace;
[0066] S2. The edge computing device uses a flame recognition model to identify the current molten steel temperature based on the furnace flame image;
[0067] S3. The host computer adjusts the heating parameters of the molten steel furnace according to the current molten steel temperature in order to control the molten steel temperature in the molten steel furnace.
[0068] The machine vision-based steel temperature control method provided in this application uses an edge computing device as its core. A CCD camera is responsible for acquiring images of the furnace flame, while the edge computing device handles image preprocessing and deep learning model inference calculations. A host computer monitors and controls the entire system. The edge computing device is equipped with a lightweight embedded AI framework, directly running a complex neural network model used for health monitoring of industrial equipment. This design allows the model inference calculation process to be completed on the edge computing device, consuming almost no computing resources from the host computer, thus improving the system's response speed and data processing efficiency.
[0069] Therefore, the machine vision-based molten steel temperature control system provided in this application, by employing advanced image recognition technology and deep learning algorithms, can perform precise image analysis on the surface of molten steel, thereby achieving high-accuracy detection of molten steel temperature, reducing human error, ensuring the accuracy of temperature control during steel production, and thus improving product quality and production efficiency.
[0070] like Figure 3 As shown, in some embodiments, step S2 above may include:
[0071] S21. Preprocess the furnace flame image to obtain the target image;
[0072] S22. Input the target image into the flame recognition model, so that the flame recognition model can identify the current molten steel temperature based on the furnace flame image.
[0073] Specifically, the acquired raw images are preprocessed to enhance image quality and adapt to the input requirements of subsequent deep learning models.
[0074] like Figure 4 As shown, in some embodiments, step S21 above may include:
[0075] S211. Select the furnace flame portion as the region of interest on the furnace flame image to obtain an intermediate image;
[0076] S212. The intermediate image is denoised, its brightness is adjusted, and its contrast is enhanced to obtain the target image.
[0077] Specifically, preprocessing steps may include selecting the area of the furnace flame as a region of interest (ROI), filtering the image to remove noise, adjusting brightness, and enhancing contrast to improve image quality. Providing accurate image data for feature extraction and temperature prediction is crucial for improving image recognition accuracy because it reduces noise in the image, allowing the model to more accurately identify the characteristics of the flame in its current state. Furthermore, the target image size can be adjusted, for example, to 224×224, to fit the input requirements of subsequent deep learning models.
[0078] like Figure 5 As shown, in some embodiments, the training process of the flame recognition model is as follows:
[0079] S4. Obtain N images of the furnace mouth flame at the molten steel smelting furnace at different times and the temperature of the molten steel in the molten steel smelting furnace measured by a temperature measuring gun, where N is a positive integer greater than 2.
[0080] S5. Preprocess the furnace flame image at each moment to obtain the target image;
[0081] S6. Input the target image corresponding to the furnace flame image at a certain moment into a preset convolutional neural network to obtain the steel temperature at that moment predicted by the convolutional neural network.
[0082] S7. Based on the mapping table between flame characteristics and molten steel temperature, determine the molten steel temperature corresponding to the flame characteristics of the furnace mouth flame image at that moment.
[0083] S8. Based on the flame characteristics of the furnace flame image at that moment, corresponding to the molten steel temperature, and the molten steel temperature in the smelting furnace measured by the temperature measuring gun at that moment, adjust the parameters of the convolutional neural network to obtain the first updated convolutional neural network.
[0084] S9. Continue to input the target image corresponding to the furnace flame image at another moment into the updated convolutional neural network to obtain the flame features extracted by the first updated convolutional neural network;
[0085] S10. Based on the mapping table between flame characteristics and molten steel temperature, determine the molten steel temperature corresponding to the flame characteristics of the furnace mouth flame image at another moment.
[0086] S11. Based on the flame characteristics of the furnace flame image at another time corresponding to the molten steel temperature and the molten steel temperature in the smelting furnace measured by the temperature measuring gun at another time, adjust the parameters of the first updated convolutional neural network to obtain the second updated convolutional neural network.
[0087] S12. This process continues until the Nth updated convolutional neural network is obtained, and the Nth updated convolutional neural network is used as the flame recognition model.
[0088] Specifically, the preprocessed image data is used to train a lightweight convolutional neural network (CNN) model. For example, all flame data can be divided into three states for image classification, with a training, validation, and test set ratio of 8:1:1. Upsampling can be used to balance the data for categories with smaller sample sizes. The CNN image classification model is constructed, including its structure and key parameters such as the number of hidden layers, the connections within hidden layers, the number of convolutional kernels, the number of output feature maps per layer, and the number of training iterations. Furthermore, data augmentation techniques such as random flipping and random color adjustments are used to improve the model's generalization ability and robustness. The preprocessed images are then input into the CNN model, and the network is trained by iterating according to the pre-defined number of iterations.
[0089] In some embodiments, the host computer performs the training process of the flame recognition model. That is, the training process of the flame recognition model is performed on the host computer. After the model is trained, it is deployed to the edge computing device, enabling the edge computing device to preprocess and perform online judgment on the images of the furnace flames captured by the CCD camera.
[0090] like Figure 6 As shown, in some embodiments, the host computer adjusts the heating parameters of the molten steel furnace according to the current molten steel temperature to control the molten steel temperature in the molten steel furnace, including:
[0091] S31. Determine whether the current temperature of the molten steel has reached the preset target temperature;
[0092] S32. If the current molten steel temperature does not reach the preset target temperature, the PID controller is used to dynamically adjust the oxygen blowing volume of the oxygen blowing lance in real time according to the current molten steel temperature and the target temperature in order to control the molten steel temperature in the smelting furnace.
[0093] S33. If the current molten steel temperature reaches the preset target temperature, control the oxygen blowing lance to stop blowing oxygen.
[0094] Specifically, a PID controller is used to dynamically adjust the oxygen flow rate of the oxygen lance. The PID controller calculates the necessary adjustment based on the difference between the set target temperature and the real-time monitored temperature to precisely control the molten steel temperature. By adjusting the proportional (P), integral (I), and derivative (D) parameters, the PID controller optimizes the temperature control process, reduces temperature fluctuations, and improves steelmaking efficiency and product quality. When the system detects that the molten steel temperature has not reached the preset target temperature, it controls the oxygen lance to continue blowing oxygen to raise the temperature. Once the molten steel temperature reaches or exceeds the endpoint temperature, the PID controller sends a signal to stop oxygen blowing and pours the molten steel into the ladle, achieving precise control of the steelmaking process. This feedback-based control mechanism allows the system to automatically adjust heating parameters to ensure that the molten steel is smelted under optimal conditions, thereby improving the quality and consistency of the final product. In this way, the system not only improves the automation level of the steelmaking process but also reduces the need for manual intervention, enhancing the safety and reliability of the entire system.
[0095] To better understand this application, the following detailed description of the machine vision-based molten steel temperature control system and control method provided in this application is provided through a specific embodiment.
[0096] like Figure 1 As shown in the figure, this embodiment provides a machine vision-based steel temperature control system, including a CCD camera 10, an edge computing device 20, and a host computer 30. The dashed arrows in the figure represent the preparatory work before system operation; processes I, II, and III represent image acquisition, model training, and algorithm deployment, respectively. The solid arrows represent the information transmission path during system operation; processes IV, V, VI, and VII represent real-time flame image acquisition, inference calculation, result visualization, and signal transmission, respectively.
[0097] CCD camera 10 acquires a large number of multi-state images of the furnace mouth flame, which are stored, preprocessed, and trained in the host computer 30 to obtain a flame recognition model. The flame recognition model is deployed to the edge computing device 20. When the system is running, the furnace mouth flame images acquired by CCD camera 10 are input into the edge computing device 20. The edge computing device 20 infers and identifies the type of flame and calculates the current temperature of the molten steel. The edge computing device 20 transmits the calculated molten steel temperature result to the host computer 30. The host computer 30 determines whether the molten steel has reached the endpoint temperature. If it has not, the host computer 30 controls the oxygen lance 2 to continue blowing oxygen; if the endpoint temperature has been reached, the host computer 30 controls the mechanical mechanism 3 to perform steel tapping or other operations.
[0098] Specifically, before the system can run, a flame recognition model needs to be built and deployed to prepare for inference calculations during system operation. Processes I, II, and III each include the following five steps:
[0099] Step 1: Data Collection
[0100] Images of the furnace flame 1 are acquired in real time using a high-resolution industrial CCD camera 10. These images are a key data source for analyzing the temperature of molten steel. The camera's image acquisition frequency can be adjusted according to the specific needs of the steelmaking process to ensure accurate capture of flame images in rapidly changing temperature environments. The image pixel size is 2560×1440.
[0101] Step 2: Image Preprocessing
[0102] The acquired raw images are stored and preprocessed on the host computer 30. The preprocessing steps include selecting the area of the furnace flame as the region of interest (ROI), filtering the image to remove noise, and adjusting brightness and contrast to enhance image quality. The image size is adjusted to 224×224 to fit the input requirements of the subsequent deep learning model.
[0103] Step 3: Model Training
[0104] Preprocessed image data was used to train a lightweight convolutional neural network (CNN) model. All flame data was divided into three states for image classification, with a training, validation, and test set ratio of 8:1:1. Upsampling was used to balance the data for classes with smaller sample sizes. A CNN image classification model was constructed, including the CNN's structure and key parameters, such as the number of hidden layers, the connections within hidden layers, the number of convolutional kernels, the number of output feature maps per layer, and the number of training iterations. Furthermore, data augmentation techniques such as random flipping and random color adjustments were used to improve the model's generalization ability and robustness. The preprocessed images were then input into the CNN model, and the network was trained iteratively according to the pre-defined number of iterations.
[0105] During model training, learning rate scheduling and early stopping strategies can be used to monitor the validation set loss and accuracy, respectively. Using both strategies simultaneously optimizes the training process. Learning rate scheduling helps the model converge more effectively during training, while early stopping stops training promptly when the model begins to overfit, avoiding resource waste and maintaining optimal performance on the validation set.
[0106] Step 4: Algorithm Deployment
[0107] The flame recognition model is deployed to the edge computing device 20, enabling it to preprocess and make online judgments on the images of the furnace flame 1 captured by the industrial CCD camera 10.
[0108] Step 5: Temperature Calculation
[0109] A mapping is established between the flame type and the current temperature of the molten steel measured by the temperature gun, yielding the corresponding molten steel temperature for the current flame type. The edge computing device 20 outputs the calculation results to the host computer 30 in real time. A user-friendly interface is installed on the host computer 30, allowing operators to intuitively monitor real-time temperature data and historical temperature changes. The interface provides real-time data monitoring and historical data analysis functions, facilitating fault diagnosis and process optimization by operators, thereby achieving more precise temperature control.
[0110] When the system is running, its control logic is as follows: Figure 7 As shown:
[0111] During system operation, the industrial CCD camera 10 first acquires images of the furnace flame in real time and transmits these images to the edge computing device 20 for image preprocessing to enhance image quality and provide accurate data support for subsequent temperature inference.
[0112] Then, the preprocessed image is input into the flame recognition model deployed on the edge computing device 20. The model uses deep learning algorithms to reason about the flame state and identify the characteristics of the flame.
[0113] Secondly, the edge computing device 20 combines the flame recognition results with the temperature data measured by the temperature measuring gun to calculate the current temperature of the molten steel. At this point, a PID controller is introduced. This controller dynamically adjusts the oxygen blowing rate of the oxygen lance 5 based on the set target temperature and the real-time monitored temperature difference to precisely control the molten steel temperature. The parameters (Kp, Ki, Kd) of the PID controller can be adjusted according to the system response characteristics to achieve optimal temperature control.
[0114] Next, the host computer 30 receives real-time temperature data transmitted by the edge computing device 20 and determines whether the current temperature has reached the preset steelmaking endpoint temperature. Based on the output of the PID controller, the host computer 30 sends control signals to the oxygen lance 5 or other mechanical mechanisms 3 to adjust the oxygen blowing volume or execute the steel tapping action.
[0115] Finally, the system continuously monitors temperature changes and control effectiveness in real time, providing feedback to the PID controller for parameter adjustments and optimization of the control strategy, ensuring the accuracy and stability of temperature control. By introducing the PID controller, the system can more precisely control the molten steel temperature, reduce temperature fluctuations, and improve steelmaking efficiency and product quality. The introduction of PID control makes temperature control more automated and intelligent, reducing the need for manual intervention and enhancing the safety and reliability of the entire system.
[0116] In addition, the system can also have a fault self-diagnosis function, which can monitor the working status of each component of the system in real time, and automatically alarm and provide troubleshooting suggestions when a fault occurs, thereby enhancing the safety and stability of the system.
[0117] Therefore, the machine vision-based steel temperature control system and method provided in this embodiment not only improves the automation level of the steelmaking process but also reduces the need for manual intervention, thereby enhancing the safety and reliability of the entire system. The application of edge computing technology gives this system significant advantages in data processing speed, resource utilization efficiency, and system stability.
[0118] By deploying edge computing devices in industrial settings, real-time data processing and decision-making are achieved, significantly improving system response speed and reliability. This application approach offers at least the following benefits: Firstly, processing data near the data source reduces the time it takes for data to reach a central server, thus lowering system response latency—crucial for metallurgical processes requiring real-time control. Secondly, edge computing reduces the amount of data transmitted over the network; specifically, only the processed results need to be sent to the metallurgical central control system (e.g., a host computer), saving bandwidth and reducing data transmission delays caused by network congestion. Thirdly, the distributed nature of edge computing devices enhances system fault tolerance; even if the central control system fails, edge computing devices can still operate independently, ensuring the continuity and stability of the steelmaking process. Fourthly, edge computing devices can process sensitive data locally without transmitting data to a remote server, reducing the risk of data leakage. Fifthly, edge computing allows the system to dynamically allocate computing resources based on actual needs, improving resource utilization efficiency and reducing energy consumption.
[0119] A deep learning-based temperature prediction model can accurately identify flame characteristics to determine the current smelting state and accurately predict the temperature of molten steel. The model analyzes preprocessed image data, extracts flame characteristics under different smelting states, and uses these characteristics for state prediction. It maps the molten steel temperatures measured by traditional temperature guns to the corresponding flame states. Through training with a large amount of molten steel temperature data, the model achieves high-precision prediction, thus significantly improving detection accuracy.
[0120] Edge computing technology optimizes the image recognition inference process, significantly improving the overall efficiency and performance of the system. The computationally intensive inference process of image recognition is executed entirely on edge computing devices. This not only fully leverages the low latency and high response speed advantages of edge computing but also places almost no burden on the host computer for intelligent control of the entire steelmaking process, greatly reducing the computational resource burden on the host computer. Since data processing and analysis are completed near the data source, data transmission requirements are effectively reduced, saving valuable bandwidth resources and improving the real-time performance of data processing. Furthermore, this design enhances the system's reliability and robustness, as critical image recognition tasks can be completed independently on local edge devices even in the event of unstable or interrupted network connections. The host computer only needs to receive the computation results transmitted from the edge device and perform subsequent processing. This collaborative model not only improves computational efficiency but also reduces dependence on the central server, making the entire system more flexible and scalable.
[0121] Because the amount of image data requiring real-time computation is large, the industrial CCD camera 10 communicates with the edge computing device 20 via GMSL; while since the edge computing device 20 only transmits temperature data to the host computer 30, the communication method between the edge computing device 20 and the host computer 30 is USB. Thus, by optimizing the communication protocol, efficient data transmission between the industrial CCD camera 10 and the edge computing device 20, and between the edge computing device 20 and the host computer 30, is achieved.
[0122] The machine vision-based steel temperature control system and method provided in this embodiment not only avoids the damage and maintenance problems of traditional contact sensors, but also significantly improves the automation level and product quality of the steelmaking process through real-time monitoring and intelligent control. By introducing deep learning technology, this system can more intelligently identify and judge the state of molten steel, providing an innovative temperature monitoring solution for the steelmaking industry.
[0123] Figure 8 This is a schematic diagram of the physical structure of an electronic device provided in an embodiment of this application, as shown below. Figure 8 As shown, the electronic device may include a processor 301, a communications interface 302, a memory 303, and a communication bus 304, wherein the processor 301, the communications interface 302, and the memory 303 communicate with each other via the communication bus 304. The processor 301 may call logical instructions in the memory 303 to execute the methods described in any of the above embodiments.
[0124] Furthermore, the logical instructions in the aforementioned memory 303 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.
[0125] This embodiment discloses a computer program product, which includes a computer program stored on a non-transitory computer-readable storage medium. The computer program includes program instructions, and when the program instructions are executed by a computer, the computer can perform the methods provided in the above-described method embodiments.
[0126] This embodiment provides a computer-readable storage medium storing a computer program that causes the computer to perform the methods provided in the above-described method embodiments.
[0127] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.
[0128] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.
[0129] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.
[0130] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.
[0131] In the description of this specification, the references to terms such as "an embodiment," "a specific embodiment," "some embodiments," "for example," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.
[0132] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of this application. It should be understood that the above descriptions are merely specific embodiments of this application and are not intended to limit the scope of protection of this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.
Claims
1. A machine vision-based steel temperature control system, characterized in that, include: CCD camera to capture real-time images of flames at the furnace mouth of molten steel smelting furnace; An edge computing device is connected to the CCD camera and is equipped with a flame recognition model. The edge computing device acquires the furnace opening flame image and uses the flame recognition model to identify the current molten steel temperature based on the furnace opening flame image. The flame recognition model is obtained by training a preset convolutional neural network based on the furnace opening flame image. The host computer is connected to the edge computing device to obtain the current molten steel temperature and adjust the heating parameters of the molten steel furnace according to the current molten steel temperature in order to control the molten steel temperature in the molten steel furnace.
2. The steel temperature control system according to claim 1, characterized in that, The edge computing device is connected to the CCD camera via a high-speed industrial bus communication module.
3. The steel temperature control system according to claim 1 or 2, characterized in that, The CCD camera communicates with the edge computing device via GMSL, and the edge computing device communicates with the host computer via USB.
4. A method for controlling molten steel temperature based on machine vision, characterized in that, The method of the machine vision-based molten steel temperature control system according to any one of claims 1 to 3 includes: CCD cameras capture real-time images of flames at the furnace opening of steel smelting furnaces; Edge computing devices use a flame recognition model to identify the current temperature of molten steel based on the furnace flame image; The host computer adjusts the heating parameters of the molten steel furnace according to the current molten steel temperature in order to control the temperature of the molten steel in the molten steel furnace.
5. The method according to claim 4, characterized in that, The edge computing device uses a flame recognition model to identify the current molten steel temperature based on the furnace flame image, including: The furnace flame image is preprocessed to obtain the target image; The target image is input into the flame recognition model, which then identifies the current molten steel temperature based on the furnace flame image.
6. The method according to claim 4, characterized in that, The edge computing device preprocesses the furnace flame image to obtain a target image, including: The region of interest is selected from the furnace flame image to obtain an intermediate image; The intermediate image is then subjected to noise reduction, brightness adjustment, and contrast enhancement to obtain the target image.
7. The method according to claim 4 or 5, characterized in that, The training process of the flame recognition model is as follows: Obtain N images of the flame at the furnace opening of the molten steel furnace at N time points and the temperature of the molten steel in the molten steel furnace measured by a temperature measuring gun, where N is a positive integer greater than 2; The furnace flame image at each moment is preprocessed to obtain the target image; The target image corresponding to the furnace flame image at a certain moment is input into a preset convolutional neural network to obtain the steel temperature at that moment predicted by the convolutional neural network. Based on the mapping table between flame characteristics and molten steel temperature, determine the molten steel temperature corresponding to the flame characteristics of the furnace mouth flame image at that moment. Based on the flame characteristics of the furnace flame image at that moment, corresponding to the molten steel temperature, and the molten steel temperature in the smelting furnace measured by the temperature measuring gun at that moment, the parameters of the convolutional neural network are adjusted to obtain the first updated convolutional neural network. The target image corresponding to the furnace flame image at another time is then input into the updated convolutional neural network to obtain the flame features extracted by the first updated convolutional neural network. Based on the mapping table between flame characteristics and molten steel temperature, determine the molten steel temperature corresponding to the flame characteristics of the furnace mouth flame image at another moment. Based on the flame characteristics of the furnace flame image at another moment corresponding to the molten steel temperature and the molten steel temperature in the smelting furnace measured by the temperature measuring gun at another moment, the parameters of the first updated convolutional neural network are adjusted to obtain the second updated convolutional neural network. This process continues until the Nth updated convolutional neural network is obtained, which is then used as the flame recognition model.
8. The method according to claim 7, characterized in that, The host computer executes the training process of the flame recognition model.
9. The method according to claim 4, characterized in that, The host computer adjusts the heating parameters of the molten steel furnace according to the current molten steel temperature to control the temperature of the molten steel in the furnace, including: Determine whether the current temperature of the molten steel has reached the preset target temperature; If the current molten steel temperature does not reach the preset target temperature, the PID controller dynamically adjusts the oxygen blowing volume of the oxygen lance in real time according to the current molten steel temperature and the target temperature in order to control the molten steel temperature in the smelting furnace.
10. The method according to claim 9, characterized in that, After determining whether the current molten steel temperature has reached the preset target temperature, the method further includes: If the current temperature of the molten steel reaches the preset target temperature, the oxygen blowing lance will be controlled to stop blowing oxygen.