A method for fault diagnosis of secondary cooling system in continuous casting
By building a multi-objective detection and fault continuity diagnosis model, using the improved YOLO-v5 and IAFSA-Elman algorithms, the problems of high labor and hardware costs and low accuracy in the fault diagnosis of continuous casting second-cold system are solved, achieving high efficiency and accuracy The fault detection and diagnosis improves the reference of production efficiency and diagnostic results.
Patent Information
- Application Number
- CN202311837120.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-12-28
- Publication Date
- 2024-11-01
- Estimated Expiration
- 2043-12-28
AI Technical Summary
In the prior art, the fault diagnosis of continuous casting second-cold system has problems such as large labor cost and hardware cost investment, low accuracy and efficiency, and small reference for fault diagnosis results.
Build a multi-object detection model of the second cold nozzle, a fault continuity diagnosis model and a fault continuity diagnosis model of the continuous casting second cold system. Using the improved YOLO-v5 algorithm and IAFSA-Elman algorithm, the second cold nozzle can be realized through video detection and data analysis. Real-time detection and fault diagnosis of blockage situations.
It improves detection efficiency and accuracy, reduces labor and hardware costs, enhances the reference of fault diagnosis results, reduces downtime, and improves production efficiency.
Smart Images

Figure CN117900407B_ABST
Abstract
Description
Technical Field
[0001] The invention belongs to the technical field of continuous casting production, and in particular relates to a fault diagnosis method for a continuous casting secondary cooling system. Background Art
[0002] The design and operation of water spray cooling in the secondary cooling zone of continuous casting have a great influence on the output and quality of the ingot. Many defects on the surface and inside of the ingot are caused by improper water spray cooling. As an important part of the secondary cooling system of continuous casting, the normal operation of the secondary cooling nozzle is an important basis for ensuring the stable operation of the secondary cooling system of continuous casting. Therefore, the fault diagnosis of the secondary cooling system of continuous casting focuses on mastering the blockage of the secondary cooling nozzle in the system. At present, how to detect the blockage of the secondary cooling nozzle in real time and find the fault of the secondary cooling system of continuous casting has become a problem that must be solved by major steel plants to improve the quality of ingots, improve the operation rate of casting machines, and carry out continuous production. In the conventional production process, the diagnostic operation of the secondary cooling nozzle in the secondary cooling system of continuous casting still uses the process of full manual on-site recording - manual drawing of record tables - expert analysis - manual transmission of results. The labor cost investment is large, and the number of secondary cooling nozzles is large. The accuracy and efficiency of detection with the help of manpower are low, the detection time is long, and generally a long downtime is required, which affects the production of the continuous casting machine. Moreover, with the development of online detection technology, the prior art has adopted a solution of using sensor technology to detect the blockage of the secondary cooling nozzle. For example, a vibration sensor is used to collect the vibration data of the secondary cooling nozzle for judgment. However, this solution requires a large number of sensors and has high hardware cost. In addition, due to the mutual influence between the secondary cooling nozzles, the detection accuracy is low, resulting in the final continuous casting secondary cooling system fault diagnosis results having little reference value for actual production. Summary of the invention
[0003] In order to solve the problems in the prior art of high labor and hardware costs, low accuracy and efficiency, and low reference value of fault diagnosis results, the present invention aims to provide a continuous casting secondary cooling system fault diagnosis method.
[0004] The technical solution adopted by the present invention is:
[0005] A method for diagnosing faults in a continuous casting secondary cooling system comprises the following steps:
[0006] Construct a multi-target detection model for secondary cooling nozzles, a continuity diagnosis model for secondary cooling nozzle faults, and a continuity diagnosis model for continuous casting secondary cooling system faults;
[0007] According to the pressure data of the water pipe of the current fan-shaped segment, determine whether the fan-shaped segment has a secondary cooling nozzle blockage. If so, collect the secondary cooling nozzle test video of the corresponding fan-shaped segment and proceed to the next step, otherwise repeat this step;
[0008] Input the secondary cooling nozzle test video of the current fan-shaped segment into the secondary cooling nozzle multi-target detection model, perform multi-target detection, and obtain the secondary cooling nozzle distribution and fault data of the fan-shaped segment;
[0009] Input the secondary cooling nozzle distribution and fault data of the current fan-shaped segment into the secondary cooling nozzle fault continuity diagnosis model, perform secondary cooling nozzle fault continuity diagnosis, and obtain the secondary cooling nozzle fault diagnosis result of the fan-shaped segment;
[0010] According to the fault diagnosis results of the secondary cooling nozzles of all the sectors in the current continuous casting secondary cooling system, the sector distribution and fault data of the continuous casting secondary cooling system are constructed;
[0011] The current sector distribution and fault data of the continuous casting secondary cooling system are input into the continuous casting secondary cooling system fault continuity diagnosis model, and the continuous casting secondary cooling system fault continuity diagnosis is performed to obtain the continuous casting secondary cooling system fault diagnosis results.
[0012] Furthermore, the secondary cooling nozzle multi-target detection model is established based on the improved YOLO-v5 algorithm.
[0013] Furthermore, a multi-target detection model for the secondary cooling nozzle is constructed, including the following steps:
[0014] Collecting several historical secondary cooling nozzle images in different working states to form a secondary cooling nozzle image training data set;
[0015] Based on the improved YOLO-v5 algorithm, the initial secondary cooling nozzle multi-target detection model was constructed;
[0016] The secondary cooling nozzle image training data set is input into the initial secondary cooling nozzle multi-target detection model for optimization training to obtain the optimal secondary cooling nozzle multi-target detection model.
[0017] Furthermore, the secondary cooling nozzle fault continuity diagnosis model and the continuous casting secondary cooling system fault continuity diagnosis model are both established based on the IAFSA-Elman algorithm.
[0018] Furthermore, a secondary cooling nozzle fault continuity diagnosis model and a continuous casting secondary cooling system fault continuity diagnosis model are constructed, including the following steps:
[0019] Collect the historical secondary cooling nozzle distribution and fault data of several fan-shaped segments to form a secondary cooling nozzle fault continuity diagnosis training data set with secondary cooling nozzle fault level labels;
[0020] Collecting historical fan-shaped segment distribution and fault data of several continuous casting secondary cooling systems to form a continuous casting secondary cooling system fault continuity diagnosis training data set with continuous casting secondary cooling system fault level labels;
[0021] Based on the IAFSA-Elman algorithm, the initial fault continuity diagnosis model of the secondary cooling nozzle and the initial fault continuity diagnosis model of the continuous casting secondary cooling system are constructed respectively.
[0022] The secondary cooling nozzle fault continuity diagnosis training data set is input into the initial secondary cooling nozzle fault continuity diagnosis model, and optimization training is performed to obtain the optimal secondary cooling nozzle fault continuity diagnosis model;
[0023] The continuous casting secondary cooling system fault continuity diagnosis training data set is input into the initial continuous casting secondary cooling system fault continuity diagnosis model for optimization training to obtain the optimal continuous casting secondary cooling system fault continuity diagnosis model.
[0024] Furthermore, the secondary cooling nozzle image of the current fan-shaped segment is input into the secondary cooling nozzle multi-target detection model to perform multi-target detection, including the following steps:
[0025] Frame capture is performed on the secondary cooling nozzle test video to obtain a secondary cooling nozzle test image of continuous frames that changes from a closed state to a working state;
[0026] Preprocessing the continuous frames of the secondary cooling nozzle test images to obtain the continuous frames of preprocessed images;
[0027] Based on the secondary cooling nozzle multi-target detection model, the pre-processed images of the continuous frames are gridded, and the initial prediction frame of each grid is obtained to obtain the images to be detected of the continuous frames with the initial prediction frame;
[0028] According to a preset intersection-and-union ratio and a preset confidence level, non-maximum suppression screening is performed on the initial prediction frames of the images to be detected of the continuous frames to obtain the final prediction frames of the images to be detected of the continuous frames;
[0029] Perform multi-target detection on the final prediction boxes of the images to be detected in all frames to obtain the corresponding target individuals;
[0030] The SORT algorithm is used to track the same target individual in the images to be detected in consecutive frames;
[0031] If any target individual remains in the closed state, the target individual is blocked and its blocking indicator value is set to 1, otherwise it is set to 0;
[0032] According to the position numbers and blockage indication values of all target individuals, the secondary cooling nozzle distribution and fault data of the current fan segment are constructed.
[0033] Furthermore, after obtaining the secondary cooling nozzle distribution and fault data of the current fan-shaped segment, the following steps are also included:
[0034] According to the pressure data of the water pipe of the current sector segment, the predicted range of the number of blocked secondary cooling nozzles is obtained. If the number of secondary cooling nozzles with a blockage indication value of 1 in the secondary cooling nozzle distribution and fault data is within the predicted range, the secondary cooling nozzle distribution and fault data are accurate, and the next step is entered. Otherwise, the secondary cooling nozzle test video of the sector segment is re-collected, and the secondary cooling nozzle test video is input into the secondary cooling nozzle multi-target detection model for multi-target detection.
[0035] Furthermore, the secondary cooling nozzle distribution and fault data of the current fan-shaped segment are input into the secondary cooling nozzle fault continuity diagnosis model to perform secondary cooling nozzle fault continuity diagnosis, including the following steps:
[0036] Input the secondary cooling nozzle distribution and fault data of the current fan-shaped segment into the secondary cooling nozzle fault continuity diagnosis model;
[0037] Based on the fault continuity diagnosis model of secondary cooling nozzles, the distribution characteristics of secondary cooling nozzles with continuous position numbers and blockage indication values of 1 in the secondary cooling nozzle distribution and fault data, as well as the number characteristics of secondary cooling nozzles with a blockage indication value of 1, are extracted;
[0038] According to the secondary cooling nozzle distribution characteristics and the secondary cooling nozzle quantity characteristics, the corresponding secondary cooling nozzle fault level is output;
[0039] According to the equipment number and the secondary cooling nozzle fault level, the secondary cooling nozzle fault diagnosis result of the current fan-shaped segment is constructed.
[0040] Furthermore, before constructing the fan-shaped segment distribution and fault data of the continuous casting secondary cooling system, the following steps are also included:
[0041] If there is no secondary cooling nozzle blockage in any sector of the current continuous casting secondary cooling system, the secondary cooling nozzle fault level of the sector is set to 0, and the secondary cooling nozzle fault diagnosis result of the sector is constructed according to the equipment number and the secondary cooling nozzle fault level.
[0042] Furthermore, the current sector distribution and fault data of the continuous casting secondary cooling system are input into the continuous casting secondary cooling system fault continuity diagnosis model to perform continuous casting secondary cooling system fault continuity diagnosis, including the following steps:
[0043] Input the current fan-shaped segment distribution and fault data of the continuous casting secondary cooling system into the continuous casting secondary cooling system fault continuity diagnosis model;
[0044] Based on the fault continuity diagnosis model of the continuous casting secondary cooling system, the distribution characteristics of the fan segments with the position equipment number and the secondary cooling nozzle fault level exceeding the fault level threshold in the fan segment distribution and fault data, as well as the number characteristics of the fan segments with the secondary cooling nozzle fault level exceeding the fault level threshold are extracted;
[0045] According to the characteristics of the fan-shaped segment distribution and the number of fan-shaped segments, the corresponding continuous casting secondary cooling system fault level is output, that is, the continuous casting secondary cooling system fault diagnosis result.
[0046] The beneficial effects of the present invention are:
[0047] The present invention provides a continuous casting secondary cooling system fault diagnosis method, which improves the inspection efficiency and accuracy by constructing a multi-target detection model for secondary cooling nozzles to detect the blockage of secondary cooling nozzles, avoids manual inspection or external hardware sensors, reduces labor costs, hardware costs and time costs, and constructs a secondary cooling nozzle fault continuity diagnosis model to perform fault diagnosis on the continuity distribution and overall number of secondary cooling nozzle blockages, thereby improving the accuracy of sector fault diagnosis, and performs system fault diagnosis based on the continuity distribution and overall number of sector faults, thereby improving the reference of fault diagnosis results to actual production.
[0048] Other beneficial effects of the present invention will be further described in the specific embodiments. BRIEF DESCRIPTION OF THE DRAWINGS
[0049] Figure 1 It is a flowchart of the continuous casting secondary cooling system fault diagnosis method of the present invention. DETAILED DESCRIPTION
[0050] The present invention will be further explained below in conjunction with the accompanying drawings and specific embodiments.
[0051] Embodiment 1:
[0052] like Figure 1 As shown, this embodiment provides a method for diagnosing faults in a continuous casting secondary cooling system, comprising the following steps:
[0053] Construct a multi-target detection model for secondary cooling nozzles, a continuity diagnosis model for secondary cooling nozzle faults, and a continuity diagnosis model for continuous casting secondary cooling system faults;
[0054] The multi-target detection model of the secondary cooling nozzle is established based on the improved YOLO-v5 algorithm. The improved YOLO-v5 algorithm is as follows: the CSP-Bottleneck module of the CSP-Darknet53 backbone network of the existing YOLO-v5 algorithm is replaced with a Transformer module, the FPN structure of the Neck network is replaced with a BiFPN structure, and the shallow feature output and the convolution output are used as the input of the Head network to obtain the improved YOLO-v5 algorithm;
[0055] The improved YOLO-v5 algorithm can enhance the feature expression of the secondary cooling nozzle test image, so that the detection of the secondary cooling nozzle test images with different backgrounds, different light and different overlapping parts has good results, and is suitable for small volume targets such as secondary cooling nozzles;
[0056] Constructing a multi-target detection model for secondary cooling nozzles includes the following steps:
[0057] Collecting several historical secondary cooling nozzle images in different working states to form a secondary cooling nozzle image training data set;
[0058] Based on the improved YOLO-v5 algorithm, the initial secondary cooling nozzle multi-target detection model was constructed;
[0059] The secondary cooling nozzle image training data set is input into the initial secondary cooling nozzle multi-target detection model for optimization training to obtain the optimal secondary cooling nozzle multi-target detection model;
[0060] The fault continuity diagnosis model of the secondary cooling nozzle and the fault continuity diagnosis model of the continuous casting secondary cooling system are both established based on the IAFSA-Elman algorithm. The IAFSA optimization algorithm is: the traditional artificial fish swarm algorithm is improved by introducing a dynamic reverse learning strategy and an adaptive artificial fish field of view based on Cauchy distribution;
[0061] The formula of the dynamic reverse learning strategy is:
[0062] x' ij (t) = k(a j (t)+b j (t))-x ij (t)
[0063] In the formula, x' ij (t), x ij (t) are the reverse position and forward position of the j-th dimension of the ith artificial fish; a j (t), b j (t) are the upper and lower bounds of the jth dimension of the current artificial fish swarm; k is the decreasing inertia factor, k = 0.9-0.5D / D max ; D, D max are the current number of iterations and the maximum number of iterations respectively; t is the time indication;
[0064] The formula for updating the field of view of the adaptive artificial fish based on Cauchy distribution is:
[0065]
[0066] In the formula, v(x i ) is the visual field update function of the artificial fish; v is the original visual field of the artificial fish; v C (x i ) is the visual field of the artificial fish after the Cauchy distribution transformation in the second half of the iteration; x i (t) is the position of the ith artificial fish; D, D maxare the current number of iterations and the maximum number of iterations respectively;
[0067] The formula for the field of view after the Cauchy distribution transformation is:
[0068]
[0069] In the formula, v C (xi) is the visual field of the artificial fish after the Cauchy distribution transformation in the second half of the iteration; f(x i ) is the fitness value of the ith artificial fish; f worst is the worst fitness value of the artificial fish; γ is the scale parameter, the smaller its value, the steeper the Cauchy distribution probability density curve, γ = 1 / (π·v origin );v origin is the initial field of view value;
[0070] Constructing the fault continuity diagnosis model of the secondary cooling nozzle and the fault continuity diagnosis model of the continuous casting secondary cooling system includes the following steps:
[0071] Collect the historical secondary cooling nozzle distribution and fault data of several fan-shaped segments to form a secondary cooling nozzle fault continuity diagnosis training data set with secondary cooling nozzle fault level labels;
[0072] Collecting historical fan-shaped segment distribution and fault data of several continuous casting secondary cooling systems to form a continuous casting secondary cooling system fault continuity diagnosis training data set with continuous casting secondary cooling system fault level labels;
[0073] Based on the IAFSA-Elman algorithm, the initial fault continuity diagnosis model of the secondary cooling nozzle and the initial fault continuity diagnosis model of the continuous casting secondary cooling system are constructed respectively.
[0074] The secondary cooling nozzle fault continuity diagnosis training data set is input into the initial secondary cooling nozzle fault continuity diagnosis model, and optimization training is performed to obtain the optimal secondary cooling nozzle fault continuity diagnosis model;
[0075] Input the continuous casting secondary cooling system fault continuity diagnosis training data set into the initial continuous casting secondary cooling system fault continuity diagnosis model, perform optimization training, and obtain the optimal continuous casting secondary cooling system fault continuity diagnosis model;
[0076] The secondary cooling nozzle fault continuity diagnosis model and the continuous casting secondary cooling system fault continuity diagnosis model built based on machine learning can fully learn the data features of the training data set, and quickly and accurately extract data features for prediction when new data is input, and output the corresponding fault level label; automated diagnosis avoids the limitations of manual judgment and improves accuracy;
[0077] According to the pressure data of the water pipe of the current fan-shaped segment, it is judged whether the fan-shaped segment has a secondary cooling nozzle blockage. If so, the secondary cooling nozzle test video of the corresponding fan-shaped segment is collected and the next step is entered. Otherwise, this step is repeated. In actual production, only a short shutdown time is required to collect a secondary cooling nozzle test video for one test, thus avoiding long-term shutdown affecting production.
[0078] Input the secondary cooling nozzle test video of the current fan-shaped segment into the secondary cooling nozzle multi-target detection model, perform multi-target detection, and obtain the secondary cooling nozzle distribution and fault data of the fan-shaped segment, including the following steps:
[0079] Frame capture is performed on the secondary cooling nozzle test video to obtain a secondary cooling nozzle test image of continuous frames that changes from a closed state to a working state;
[0080] Preprocessing the continuous frames of the secondary cooling nozzle test images to obtain the continuous frames of preprocessed images;
[0081] Based on the secondary cooling nozzle multi-target detection model, the pre-processed images of the continuous frames are gridded, and the initial prediction frame of each grid is obtained to obtain the images to be detected of the continuous frames with the initial prediction frame;
[0082] According to a preset intersection-and-union ratio and a preset confidence level, non-maximum suppression screening is performed on the initial prediction frames of the images to be detected of the continuous frames to obtain the final prediction frames of the images to be detected of the continuous frames;
[0083] Perform multi-target detection on the final prediction boxes of the images to be detected in all frames to obtain the corresponding target individuals;
[0084] The single target tracking SORT algorithm is used to track the same target individual in the continuous frames of the image to be detected; the target individual in the final prediction frame in the image to be detected is numbered and tracked to ensure the capture of the process of the secondary cooling nozzle switching from the closed state to the working state;
[0085] If any target individual remains in the closed state, the target individual is blocked and its blocking indicator value is set to 1, otherwise it is set to 0;
[0086] According to the position numbers and blockage indication values of all target individuals, the secondary cooling nozzle distribution and fault data of the current fan-shaped segment are formed;
[0087] According to the pressure data of the water pipe of the current fan-shaped segment, the predicted range of the number of blocked secondary cooling nozzles is obtained. If the number of secondary cooling nozzles with a blockage indication value of 1 in the secondary cooling nozzle distribution and fault data is within the predicted range, the secondary cooling nozzle distribution and fault data are accurate, and the next step is entered. Otherwise, the secondary cooling nozzle test video of the fan-shaped segment is collected again, and the secondary cooling nozzle test video is input into the secondary cooling nozzle multi-target detection model for multi-target detection.
[0088] Inputting the secondary cooling nozzle distribution and fault data of the current fan-shaped segment into the secondary cooling nozzle fault continuity diagnosis model, performing secondary cooling nozzle fault continuity diagnosis, and obtaining the secondary cooling nozzle fault diagnosis result of the fan-shaped segment, including the following steps:
[0089] Input the secondary cooling nozzle distribution and fault data of the current fan-shaped segment into the secondary cooling nozzle fault continuity diagnosis model;
[0090] Based on the fault continuity diagnosis model of secondary cooling nozzles, the distribution characteristics of secondary cooling nozzles with continuous position numbers and blockage indication values of 1 in the secondary cooling nozzle distribution and fault data, as well as the number characteristics of secondary cooling nozzles with a blockage indication value of 1, are extracted;
[0091] According to the secondary cooling nozzle distribution characteristics and the secondary cooling nozzle quantity characteristics, the corresponding secondary cooling nozzle fault level is output;
[0092] According to the equipment number and the secondary cooling nozzle fault level, the secondary cooling nozzle fault diagnosis result of the current fan-shaped segment is formed;
[0093] If there is no secondary cooling nozzle blockage in any sector of the current continuous casting secondary cooling system, the secondary cooling nozzle fault level of the sector is set to 0, and the secondary cooling nozzle fault diagnosis result of the sector is formed according to the equipment number and the secondary cooling nozzle fault level;
[0094] According to the fault diagnosis results of the secondary cooling nozzles of all the sectors in the current continuous casting secondary cooling system, the sector distribution and fault data of the continuous casting secondary cooling system are constructed;
[0095] Inputting the current sector distribution and fault data of the continuous casting secondary cooling system into the continuous casting secondary cooling system fault continuity diagnosis model, performing continuous casting secondary cooling system fault continuity diagnosis, and obtaining the continuous casting secondary cooling system fault diagnosis result, including the following steps:
[0096] Input the current fan-shaped segment distribution and fault data of the continuous casting secondary cooling system into the continuous casting secondary cooling system fault continuity diagnosis model;
[0097] Based on the fault continuity diagnosis model of the continuous casting secondary cooling system, the distribution characteristics of the fan segments with the position equipment number and the secondary cooling nozzle fault level exceeding the fault level threshold in the fan segment distribution and fault data, as well as the number characteristics of the fan segments with the secondary cooling nozzle fault level exceeding the fault level threshold are extracted;
[0098] According to the distribution characteristics and quantity characteristics of the fan-shaped segments, the corresponding continuous casting secondary cooling system fault level is output, that is, the continuous casting secondary cooling system fault diagnosis result. The continuous casting secondary cooling system fault diagnosis result combines the distribution and quantity of faulty fan-shaped segments, as well as the distribution and quantity of blocked secondary cooling nozzles, reflecting the actual status of the continuous casting machine production line, and can provide a great reference for subsequent production of the production line.
[0099] The present invention provides a continuous casting secondary cooling system fault diagnosis method, which improves the inspection efficiency and accuracy by constructing a multi-target detection model for secondary cooling nozzles to detect the blockage of secondary cooling nozzles, avoids manual inspection or external hardware sensors, reduces labor costs, hardware costs and time costs, and constructs a secondary cooling nozzle fault continuity diagnosis model to perform fault diagnosis on the continuity distribution and overall number of secondary cooling nozzle blockages, thereby improving the accuracy of sector fault diagnosis, and performs system fault diagnosis based on the continuity distribution and overall number of sector faults, thereby improving the reference of fault diagnosis results to actual production.
[0100] The present invention is not limited to the above optional implementations, and anyone can derive other various forms of products under the enlightenment of the present invention. The above specific implementations should not be understood as limiting the scope of protection of the present invention. The scope of protection of the present invention should be based on the definition in the claims, and the description can be used to interpret the claims.
Claims
1. A method for fault diagnosis of continuous casting secondary cooling system, Features: The steps include: Construct a multi-target detection model for secondary cooling nozzles, a continuity diagnosis model for secondary cooling nozzle faults, and a continuity diagnosis model for continuous casting secondary cooling system faults; The secondary cooling nozzle multi-target detection model is established based on the improved YOLO-v5 algorithm. The improved YOLO-v5 algorithm is as follows: the CSP-Bottleneck module of the CSP-Darknet53 backbone network of the existing YOLO-v5 algorithm is replaced with a Transformer module, the FPN structure of the Neck network is replaced with a BiFPN structure, and the shallow feature output and the convolution output are used as the input of the Head network to obtain the improved YOLO-v5 algorithm; The secondary cooling nozzle fault continuity diagnosis model and the continuous casting secondary cooling system fault continuity diagnosis model are both established based on the IAFSA-Elman algorithm, which introduces a dynamic reverse learning strategy and an adaptive artificial fish field of view based on Cauchy distribution to improve the traditional artificial fish algorithm; According to the pressure data of the water pipe of the current fan-shaped segment, determine whether the fan-shaped segment has a secondary cooling nozzle blockage. If so, collect the secondary cooling nozzle test video of the corresponding fan-shaped segment and proceed to the next step, otherwise repeat this step; Input the secondary cooling nozzle test video of the current fan-shaped segment into the secondary cooling nozzle multi-target detection model, perform multi-target detection, and obtain the secondary cooling nozzle distribution and fault data of the fan-shaped segment; Input the secondary cooling nozzle distribution and fault data of the current fan-shaped segment into the secondary cooling nozzle fault continuity diagnosis model, perform secondary cooling nozzle fault continuity diagnosis, and obtain the secondary cooling nozzle fault diagnosis result of the fan-shaped segment; According to the fault diagnosis results of the secondary cooling nozzles of all the sectors in the current continuous casting secondary cooling system, the sector distribution and fault data of the continuous casting secondary cooling system are constructed; The current sector distribution and fault data of the continuous casting secondary cooling system are input into the continuous casting secondary cooling system fault continuity diagnosis model, and the continuous casting secondary cooling system fault continuity diagnosis is performed to obtain the continuous casting secondary cooling system fault diagnosis results.
2. A method for diagnosing faults in a continuous casting secondary cooling system according to claim 1, Features: Constructing a multi-target detection model for secondary cooling nozzles includes the following steps: Collecting several historical secondary cooling nozzle images in different working states to form a secondary cooling nozzle image training data set; Based on the improved YOLO-v5 algorithm, the initial secondary cooling nozzle multi-target detection model was constructed; The secondary cooling nozzle image training data set is input into the initial secondary cooling nozzle multi-target detection model for optimization training to obtain the optimal secondary cooling nozzle multi-target detection model.
3. A method for diagnosing faults in a continuous casting secondary cooling system according to claim 1, Features: Constructing the fault continuity diagnosis model of the secondary cooling nozzle and the fault continuity diagnosis model of the continuous casting secondary cooling system includes the following steps: Collect the historical secondary cooling nozzle distribution and fault data of several fan-shaped segments to form a secondary cooling nozzle fault continuity diagnosis training data set with secondary cooling nozzle fault level labels; Collect the historical fan-shaped segment distribution and fault data of several continuous casting secondary cooling systems to form a continuous casting secondary cooling system fault continuity diagnosis training data set with continuous casting secondary cooling system fault level labels; Based on the IAFSA-Elman algorithm, the initial fault continuity diagnosis model of the secondary cooling nozzle and the initial fault continuity diagnosis model of the continuous casting secondary cooling system are constructed respectively. The secondary cooling nozzle fault continuity diagnosis training data set is input into the initial secondary cooling nozzle fault continuity diagnosis model, and optimization training is performed to obtain the optimal secondary cooling nozzle fault continuity diagnosis model; The continuous casting secondary cooling system fault continuity diagnosis training data set is input into the initial continuous casting secondary cooling system fault continuity diagnosis model for optimization training to obtain the optimal continuous casting secondary cooling system fault continuity diagnosis model.
4. A method for diagnosing faults in a continuous casting secondary cooling system according to claim 1, Features: The secondary cooling nozzle image of the current fan-shaped segment is input into the secondary cooling nozzle multi-target detection model to perform multi-target detection, including the following steps: Frame capture is performed on the secondary cooling nozzle test video to obtain a secondary cooling nozzle test image of continuous frames that changes from a closed state to a working state; Preprocessing the continuous frames of the secondary cooling nozzle test images to obtain the continuous frames of preprocessed images; Based on the secondary cooling nozzle multi-target detection model, the pre-processed images of the continuous frames are gridded, and the initial prediction frame of each grid is obtained to obtain the images to be detected of the continuous frames with the initial prediction frame; According to a preset intersection-and-union ratio and a preset confidence level, non-maximum suppression screening is performed on the initial prediction frames of the images to be detected of the continuous frames to obtain the final prediction frames of the images to be detected of the continuous frames; Perform multi-target detection on the final prediction boxes of the images to be detected in all frames to obtain the corresponding target individuals; The SORT algorithm is used to track the same target individual in the images to be detected in consecutive frames; If any target individual remains in the closed state, the target individual is blocked and its blocking indicator value is set to 1, otherwise it is set to 0; According to the position numbers and blockage indication values of all target individuals, the secondary cooling nozzle distribution and fault data of the current fan segment are constructed.
5. A method for diagnosing faults in a continuous casting secondary cooling system according to claim 4, Features: After obtaining the secondary cooling nozzle distribution and fault data of the current fan segment, the following steps are also included: According to the pressure data of the water pipe of the current sector segment, the predicted range of the number of blocked secondary cooling nozzles is obtained. If the number of secondary cooling nozzles with a blockage indication value of 1 in the secondary cooling nozzle distribution and fault data is within the predicted range, the secondary cooling nozzle distribution and fault data are accurate, and the next step is entered. Otherwise, the secondary cooling nozzle test video of the sector segment is re-collected, and the secondary cooling nozzle test video is input into the secondary cooling nozzle multi-target detection model for multi-target detection.
6. A method for diagnosing faults in a continuous casting secondary cooling system according to claim 4, Features: Input the secondary cooling nozzle distribution and fault data of the current fan-shaped segment into the secondary cooling nozzle fault continuity diagnosis model to perform secondary cooling nozzle fault continuity diagnosis, including the following steps: Input the secondary cooling nozzle distribution and fault data of the current fan-shaped segment into the secondary cooling nozzle fault continuity diagnosis model; Based on the fault continuity diagnosis model of secondary cooling nozzles, the distribution characteristics of secondary cooling nozzles with continuous position numbers and blockage indication values of 1 in the secondary cooling nozzle distribution and fault data, as well as the number characteristics of secondary cooling nozzles with a blockage indication value of 1, are extracted; According to the secondary cooling nozzle distribution characteristics and the secondary cooling nozzle quantity characteristics, the corresponding secondary cooling nozzle fault level is output; According to the equipment number and the secondary cooling nozzle fault level, the secondary cooling nozzle fault diagnosis result of the current fan-shaped segment is constructed.
7. A method for diagnosing faults in a continuous casting secondary cooling system according to claim 6, Features: Before constructing the fan segment distribution and fault data of the continuous casting secondary cooling system, the following steps are also included: If there is no secondary cooling nozzle blockage in any sector of the current continuous casting secondary cooling system, the secondary cooling nozzle fault level of the sector is set to 0, and the secondary cooling nozzle fault diagnosis result of the sector is constructed according to the equipment number and the secondary cooling nozzle fault level.
8. A method for diagnosing faults in a continuous casting secondary cooling system according to claim 7, Features: Input the current sector distribution and fault data of the continuous casting secondary cooling system into the continuous casting secondary cooling system fault continuity diagnosis model to perform the continuous casting secondary cooling system fault continuity diagnosis, including the following steps: Input the current fan-shaped segment distribution and fault data of the continuous casting secondary cooling system into the continuous casting secondary cooling system fault continuity diagnosis model; Based on the fault continuity diagnosis model of the continuous casting secondary cooling system, the distribution characteristics of the fan segments with the position equipment number and the secondary cooling nozzle fault level exceeding the fault level threshold in the fan segment distribution and fault data, as well as the number characteristics of the fan segments with the secondary cooling nozzle fault level exceeding the fault level threshold are extracted; According to the characteristics of the fan-shaped segment distribution and the number of fan-shaped segments, the corresponding continuous casting secondary cooling system fault level is output, that is, the continuous casting secondary cooling system fault diagnosis result.
