Method and device for detecting abnormality of real-time temperature field information on surface of casting blank

Abstract

The invention discloses a casting blank surface real-time temperature field information anomaly detection method and device. The method comprises the steps that S1, casting blank surface infrared image data are collected; s2, infrared image data are sent into a target detection model to be processed, and a casting blank surface temperature abnormal area is obtained through analysis; and S3, extracting temperature data from a temperature abnormal area on the surface of the casting blank, and then performing secondary judgment on the abnormal area by using the trained K-means clustering algorithm to position a temperature abnormal state result. Based on the infrared thermal image technology and the machine learning technology, thermal image data are collected in the casting blank moving process and applied to casting blank surface real-time temperature field information abnormal state detection, and the accuracy and stability of casting blank surface temperature detection are improved.

Description

Technical Field

[0001] The present invention relates to the field of steelmaking continuous casting process and quality control, and in particular to a method, device and electronic equipment for detecting abnormality of real-time temperature field information on the surface of a casting blank based on infrared images. Background Art

[0002] In the steel production process, ingot production is a critical step. During the continuous casting cooling process, the uniformity of the ingot's surface and internal temperatures directly impacts the subsequent rolling process and the quality of the final product. Abnormal temperatures during the cooling process can lead to uneven internal structure, thus affecting performance.

[0003] Traditional methods for measuring slab temperature primarily rely on thermocouples or infrared thermometers. However, these methods are often limited by measurement distance, environmental factors, and the equipment itself, making them incapable of accurately and real-timely acquiring comprehensive slab temperature information. Furthermore, these methods typically only measure temperature at specific points, making it difficult to achieve continuous temperature field monitoring across the entire slab surface.

[0004] In recent years, with the development of computer vision and deep learning technologies, infrared image processing has been widely used in the industrial field. Infrared images can capture the temperature distribution of an object's surface, providing new possibilities for identifying temperature anomalies in cast billets. However, most existing deep learning and computer vision detection methods based on infrared images have not yet been applied to the cast billet field or are not specifically optimized for the characteristics of cast billets, resulting in challenges in accuracy and real-time performance. Summary of the Invention

[0005] The purpose of the present invention is to overcome the shortcomings of the existing technology and provide a method and device for detecting abnormalities in the real-time temperature field information on the surface of the billet. The target detection algorithm model is used to perform real-time and accurate detection of the surface temperature of the billet, which can accurately and reliably realize the temperature abnormality detection of the billet.

[0006] In order to achieve the above object, the technical solution adopted by the present invention is: a method for detecting abnormalities in real-time temperature field information on the surface of a casting billet, comprising:

[0007] S1, collecting infrared image data of the casting surface;

[0008] S2. Sending the infrared image data to the target detection model for processing, and analyzing to obtain the area with abnormal temperature on the surface of the ingot;

[0009] S3. Extract temperature data from the temperature abnormality area on the surface of the ingot, and then use the trained K-means clustering algorithm to perform a secondary judgment on the abnormal area to locate the temperature abnormality status result.

[0010] The method further comprises step S4: when the secondary determination detects a temperature abnormality, an alarm signal is issued or a production line interlocking control signal is issued to the production line control system to control the operation state of the production line.

[0011] In step S1, an infrared spectrum detection device is set up in front of the billet transmission track to collect infrared temperature image data of the billet.

[0012] The target detection model in step S2 adopts the YOLOv5 network model, which is pre-trained to identify areas with abnormal surface temperature of the ingot.

[0013] Before training the YOLOv5 network model, an infrared camera is used to capture infrared images of the billets on the billet production line, and the temperature abnormalities in the images are marked as labels for the training data to form an infrared image dataset, which is then divided into a training set and a validation set. The infrared image training dataset is used to train the YOLOv5 network model, and after the training is completed, the validation set is used to verify the model indicators of the trained YOLOv5 network model.

[0014] In step S3, in the real-time production line image, the trained YOLOv5 model is used to perform preliminary detection of the temperature abnormal area of ​​the ingot to obtain the potential abnormal area, and then K-means cluster analysis is performed on the potential abnormal area detected by YOLOv5 to confirm and locate the temperature abnormal area.

[0015] Before clustering the temperature data using the K-means algorithm, feature extraction is performed on the surface area temperature data of the steel section detected by the YOLOv5 model and the extracted features are applied to the K-means algorithm.

[0016] The temperature features extracted from the steel detection area are:

[0017]

[0018] Where T = {t1, t2, ..., t n} represents the temperature set of the steel detection area, μ(T) represents the average temperature, σ(T) represents the temperature standard deviation, Represents the average value of temperature gradient, Peaks(T) represents the number of local peaks, if t i is a local peak if and only if t i >t i-1 And t i >t i+1 , the number of local peaks is:

[0019]

[0020] Among them, 1(·) is the indicator function; then the temperature feature vectors of each steel temperature field detection area in the data set are extracted, and these vectors are input into the K-means clustering algorithm.

[0021] Select k = 2 to represent normal temperature features and abnormal temperature features. The goal of K-means is to assign each sample to the nearest cluster center to minimize the total internal variance:

[0022]

[0023] Among them, F i is the eigenvector of the i-th sample, C1 and C2 represent the cluster centers of the normal eigenvector and the abnormal eigenvector, respectively. Then, the temperature eigenvector of the area after the network model detection is extracted. If it is close to the abnormal cluster center, the area is judged to have a temperature anomaly; otherwise, it is judged to be normal.

[0024] A device for detecting anomalies in the real-time temperature field information on the surface of a casting blank, comprising an infrared image detection device and a server. The infrared image detection device is used to collect infrared image data of the casting blank surface and send the data to the server; the server runs the method for detecting anomalies in the real-time temperature field information on the surface of the casting blank to perform real-time temperature field information anomaly detection on the surface of the casting blank.

[0025] The advantages of the present invention are that it uses a target detection algorithm model to accurately detect the surface temperature of the billet in real time, enabling accurate and reliable detection of temperature anomalies in the billet. It uses infrared images to capture the temperature distribution of the billet, providing a continuous, contactless temperature measurement method suitable for real-time monitoring in billet production environments. It combines YOLOv5 with a K-means-based temperature clustering evaluation method. Through dual judgment and complementary strategies, this method can accurately locate abnormal temperature areas on the billet surface, improving the accuracy and reliability of temperature anomaly judgment. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] The following is a brief description of the contents and symbols in the drawings of the present invention:

[0027] Figure 1 Schematic diagram of the method for detecting abnormalities in the real-time temperature field information on the surface of a casting billet based on infrared images according to the present invention.

[0028] The marks in the above figures are: 1-slab transmission track; 2-infrared spectrum detection device; 3-server. DETAILED DESCRIPTION

[0029] The specific implementation of the present invention will be further explained in detail below by describing the best embodiment with reference to the accompanying drawings.

[0030] In view of the insufficient accuracy and reliability of the detection of temperature anomalies on the surface of the ingot in the prior art, this embodiment develops a method for detecting anomalies in the real-time temperature field information on the surface of the ingot, which can realize real-time monitoring and has high accuracy and reliability.

[0031] In response to the limitations and shortcomings of existing technologies in detecting anomalies in the real-time temperature field information on the surface of billets, this solution combines infrared imaging technology, deep learning and cluster analysis methods to achieve more accurate, rapid and efficient detection of anomalies in the surface temperature of billets, ensuring stable and continuous high-quality billet temperature monitoring in complex production environments.

[0032] The detection method in this embodiment is composed of an infrared spectrum detection device, a memory, a processor and a method for detecting abnormal surface temperature of a casting blank. The infrared spectrum detection device is connected to the memory and the processor and stores software corresponding to the method for detecting abnormal surface temperature of a casting blank. Figure 1 This is a schematic diagram of a method for detecting anomalies in the real-time temperature field information on the surface of a slab based on infrared images. The infrared spectrum detection device is used to collect infrared image data, which is uploaded to a memory and processor for processing via a communication cable or wireless communication. The processor is pre-integrated with corresponding software developed based on the slab surface temperature anomaly detection method. Anomaly detection is achieved through the operation of the software, and the memory can store the software data corresponding to the detection method. The specific steps of the anomaly detection method are as follows:

[0033] 1. Use infrared spectrum detection equipment and infrared thermal imaging technology to monitor the moving billets on the production line in real time and collect infrared image data of their surface temperature.

[0034] 2. Collect a large number of infrared images from the casting process, including scenes with both normal and abnormal temperatures. Build a dataset of infrared images specifically targeting abnormal casting temperatures.

[0035] 3. Use the trained YOLOv5 object detection network to process the infrared image and determine whether there is an abnormal temperature condition in the monitored area of ​​the ingot surface. If the object detection model does not detect an abnormal temperature area, return to step 3 and continue to determine the abnormal temperature condition. Process the collected infrared image. If an abnormal temperature area is detected, proceed to step 4.

[0036] 4. Extract temperature data from the abnormal areas marked by the YOLOv5 detection network, and then use the trained K-means clustering algorithm to perform a secondary judgment on the abnormal areas. Secondary judgment refers to the re-judgment of the temperature anomaly areas detected by the YOLOv5 model. The corresponding temperature anomaly areas are selected and further confirmed using the K-means method. The temperature anomaly area detection using the YOLOv5 model is a preliminary determination of the existence of temperature anomalies. The second step is to further refine the location of the abnormal area through the K-means algorithm to accurately locate the location of the temperature anomaly.

[0037] An electronic device for infrared thermal image acquisition includes: at least one infrared spectrum detection device, a memory communicatively connected to the detection device, and software stored in the memory and recognized by the infrared spectrum detection device; the software is used to execute an infrared thermal image acquisition method.

[0038] An electronic device for calculating a method for detecting abnormal surface temperature of a casting billet includes: at least one processor, a memory connected to the processor, and software stored in the memory and recognized by the processor; the software is used to execute the method described.

[0039] like Figure 1 As shown, the memory can be set to run in the server, and the collected data is transmitted to the server via the network or communication cable. The memory in the server stores the data, and the processor in the server runs the corresponding software to realize temperature anomaly detection. The software is the development and implementation of the temperature anomaly detection method. The temperature anomaly detection method is described as follows:

[0040] Embodiments of the present invention provide a method, device, and electronic device for detecting anomalies in the real-time temperature field information on the surface of a casting billet based on infrared images. The method collects infrared thermal imaging data to establish a casting billet infrared image dataset. By combining the YOLOv5 network model and the K-means clustering evaluation algorithm, a method for detecting anomalies in the casting billet surface temperature is ultimately obtained. Because it is a trained model, the method as a whole has the advantages of extremely low computational complexity and strong real-time performance. Based on this, the present invention has good application in the fields of casting technology and quality control. In order to better explain the overall technical solution of the present invention, the present invention is further explained in conjunction with the accompanying drawings.

[0041] like Figure 1 As shown, an infrared spectroscopy device is installed directly in front of the billet transport track (direction 1). When hot billet appears, infrared spectroscopy device 2 captures the rolled material and begins collecting, calculating, recording, and analyzing data. The infrared spectroscopy device's detection viewport includes a pre-defined detection area. The infrared spectroscopy device's temperature collection area is set based on the billet's transport position on the track.

[0042] Use an infrared camera to capture a large number of infrared images of ingots on the ingot casting production line. Based on the staff's experience and knowledge, manually mark the areas with abnormal temperatures in the images as labels for the training data. Use the preprocessed infrared image data to train the YOLOv5 model. Test the model's accuracy on the validation set to ensure that it meets the required performance indicators. Optimize and fine-tune the model until satisfactory detection results are achieved.

[0043] Extract the temperature data of the ingot from the infrared image. Use the K-means algorithm to cluster the temperature data. Before implementing clustering, it is necessary to first perform feature extraction on the temperature data of the steel surface area detected by the YOLOv5 model. Assume that the temperature features are extracted from the steel detection area:

[0044]

[0045] Where T = {t1, t2, ..., t n} represents the temperature set of the steel detection area, tn represents the temperature of the nth monitoring area point, μ(T) represents the average temperature, σ(T) represents the temperature standard deviation, Represents the average value of temperature gradient. Peaks(T) represents the number of local peaks, assuming ti is a local peak if and only if t i >t i-1 And t i >t i+1 , the number of local peaks is:

[0046]

[0047] Among them, 1(·) is the indicator function. Next, the temperature feature vectors of each section steel temperature field detection area in the data set are extracted and these vectors are input into the K-means clustering algorithm.

[0048] Assume that k = 2 is selected to represent normal temperature features and abnormal temperature features. The goal of K-means is to assign each sample to the nearest cluster center, thereby minimizing the total internal variance:

[0049]

[0050] Among them, F iis the eigenvector of the i-th sample, C1 and C2 represent the cluster centers of the normal and abnormal eigenvectors, respectively. Parameter n is the number of pixels in the infrared image, or in other words, the number of temperature measurement points. Setting the resolution of the infrared image fixes parameter n. Next, the temperature eigenvector of the region detected by the network model is extracted. If it is close to the abnormal cluster center, the region is judged to have a temperature anomaly; otherwise, it is judged to be normal. Whether a region's eigenvector is close to the cluster center is determined by taking the difference between the region's eigenvector and the cluster center vector. This difference is then compared with a set threshold. The difference is then used to determine whether it is close to the abnormal cluster center. If the difference is less than the set threshold, the region is considered close to the cluster center; if it is greater than the threshold, the region is not close to the cluster center. This allows for a determination of proximity. Setting the threshold requires on-site testing and debugging.

[0051] In real-time production line images, the trained YOLOv5 model is first used to perform preliminary detection of areas with abnormal slab temperatures. Next, K-means clustering analysis is performed on potential abnormal areas detected by YOLOv5 to further confirm and locate temperature anomalies. When an anomaly is detected, the system can issue an alarm or integrate with the production line control system to automatically take countermeasures.

[0052] The above method process can be executed in a processor or stored in a memory (or computer-readable medium). Computer-readable media includes permanent and non-permanent, removable and non-removable media that can implement information storage by any method or technology. Information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer-readable media does not include transitory computer-readable media (transitory media), such as modulated data signals and carrier waves.

[0053] This invention provides a method, device, and electronic device for detecting real-time temperature field information on the surface of a casting billet based on infrared images. This technology combines YOLOv5 target detection technology and cluster evaluation to determine abnormal temperature conditions during high-temperature casting billet production. Specific innovations are as follows:

[0054] 1. Unlike traditional temperature measurement methods, this technology uses infrared images to capture the temperature distribution of the billet, providing a continuous, non-contact temperature measurement method suitable for real-time monitoring in the billet production environment.

[0055] 2. This technology combines YOLOv5 with a K-means-based temperature clustering evaluation method. Through dual judgment and complementary strategies, this method can accurately locate abnormal temperature areas on the surface of the ingot.

[0056] After implementation, the solution of this embodiment collects thermal imaging data during the movement of the billet and applies it to the detection of abnormal conditions in the real-time temperature field of the billet surface. This technology improves the accuracy and stability of billet surface temperature detection. The device of the present invention has a simple structure, is easy to install and maintain, and has high practicality and cost-effectiveness.

[0057] Obviously, the specific implementation of the present invention is not limited to the above-mentioned methods. As long as various non-substantial improvements are made using the method concept and technical solution of the present invention, they are all within the scope of protection of the present invention.

Claims

1. A method for detecting abnormality of real-time temperature field information on the surface of a casting billet, characterized in that: include: S1, collecting infrared image data of the surface of the ingot; S2, sending the infrared image data to the target detection model for processing, and analyzing to obtain the abnormal temperature area on the surface of the ingot; S3. Extract temperature data from the temperature abnormality area on the surface of the ingot, and then use the trained K-means clustering algorithm to perform secondary judgment on the abnormal area to locate the temperature abnormality status result.

2. A method for detecting abnormality of real-time temperature field information on the surface of a casting billet according to claim 1, characterized in that: The method further comprises step S4: when the secondary determination detects that the temperature is abnormal, an alarm signal is issued or a production line interlocking control signal is issued to the production line control system to control the operation state of the production line.

3. The method for detecting abnormality of real-time temperature field information on the surface of a casting billet according to claim 1, characterized in that: In step S1, an infrared spectrum detection device is set up in front of the billet transmission track to collect infrared temperature image data of the billet.

4. The method for detecting abnormality of real-time temperature field information on the surface of a casting billet according to claim 1, characterized in that: The target detection model in step S2 adopts the YOLOv5 network model, and the YOLOv5 network model is used to identify the abnormal temperature area on the surface of the casting billet after pre-training.

5. A method for detecting abnormality of real-time temperature field information on the surface of a casting billet according to claim 4, characterized in that: Before training, the YOLOv5 network model uses an infrared camera to capture infrared images of the ingots on the ingot production line, marks the temperature abnormality areas in the images as labels for training data, and forms an infrared image dataset. The infrared image dataset is then divided into a training set and a validation set. The infrared image training dataset is used to train the YOLOv5 network model. After the training is completed, the validation set is used to verify the model indicators of the trained YOLOv5 network model.

6. A method for detecting abnormality of real-time temperature field information on the surface of a casting billet according to claim 1, characterized in that: In step S3, in the real-time production line image, the trained YOLOv5 model is used to perform preliminary detection of the temperature abnormal area of ​​the ingot to obtain the potential abnormal area, and then K-means clustering analysis is performed on the potential abnormal area detected by YOLOv5 to confirm and locate the temperature abnormal area.

7. A method for detecting abnormality of real-time temperature field information on the surface of a casting billet according to claim 6, characterized in that: Before using the K-means algorithm to cluster the temperature data, feature extraction is performed on the surface area temperature data of the steel section detected by the YOLOv5 model and the extracted features are applied to the K-means algorithm.

8. A method for detecting abnormality of real-time temperature field information on the surface of a casting billet according to claim 7, characterized in that: The temperature features extracted from the steel detection area are: Where T = {t1, t2, ..., t n } represents the temperature set of the steel detection area, μ(T) represents the average temperature, σ(T) represents the temperature standard deviation, represents the average value of temperature gradient, Peaks(T) represents the number of local peaks, if t i is a local peak if and only if t i >t i-1 And t i >t i+1 , the number of local peaks is: Among them, 1(·) is the indicator function; then the temperature feature vectors of each steel temperature field detection area in the data set are extracted, and these vectors are input into the K-means clustering algorithm.

9. A method for detecting abnormality of real-time temperature field information on the surface of a casting billet according to claim 8, characterized in that: Select k = 2 to represent normal temperature features and abnormal temperature features. The goal of K-means is to assign each sample to the nearest cluster center to minimize the total internal variance: Among them, F i is the feature vector of the i-th sample, C1 and C2 represent the cluster centers of the normal feature vector and the abnormal feature vector respectively; then, the temperature feature vector of the area after the network model detection is extracted, if it is close to the abnormal cluster center, the area is judged to have temperature anomaly; otherwise, it is judged to be normal.

10. A device for detecting abnormal temperature field information on the surface of a casting billet in real time, characterized in that: It includes an infrared image detection device and a server, wherein the infrared image detection device is used to collect infrared image data of the surface of the casting billet and send it to the server; The server runs the method for detecting abnormalities in the real-time temperature field information on the surface of the ingot as described in any one of claims 1 to 9, for detecting abnormalities in the real-time temperature field information on the surface of the ingot.

Images