Image feature extraction method for identifying sticking and breakout in crystallizer

Abstract

The application provides an image feature extraction method for identification of sticking and breakout in a crystallizer, comprising the following steps: step one: collecting temperature data of a plurality of thermocouples in the crystallizer; step two: combining the temperature data of the plurality of thermocouples in sequence in the horizontal direction, performing standardization processing, and obtaining a normal distribution; step three: performing normalization processing, converting into a gray value, and again masking the temperature data less than n to form a temperature gray unit; step four: obtaining a time sequence difference between the thermocouples, and obtaining a sticking feature occurrence area; and step five: gridding the temperature gray unit. The application provides an image feature extraction method for identification of sticking and breakout in a crystallizer, which collects and processes the temperature in the crystallizer, simultaneously integrates time, space and other features, generates a pixel data covering a plurality of sticking features, and serves as an input of a computer vision model, so as to accelerate network model convergence and prediction accuracy.

Description

Technical Field

[0001] The present application relates to the fields of continuous casting technology and computer vision technology, and in particular to an image feature extraction method for identifying bonding breakouts in a crystallizer. Background Art

[0002] In the continuous casting process, the mold is a highly efficient heat transfer device, its primary function being to dissipate heat from the molten steel, allowing it to form a shell of a certain thickness upon exiting the mold. During production, factors such as the speed and thickness of the shell formed in the mold, its geometric dimensions, and the degree of lubrication of the mold wall by the mold slag all have a significant impact on the quality of the continuously cast ingot. Breakout is the most detrimental production accident. It not only directly reduces output and impacts the entire steelmaking production schedule, but also poses a risk to the mold and rollers, indirectly affecting the quality of the ingot. Therefore, breakout has always been a significant factor affecting continuous casting production and the lifespan of its equipment. Among breakout accidents, bonding breakout is the most common.

[0003] If there are factors such as poor lubrication between the mold and the continuous casting billet, uneven mold vibration, unstable drawing speed, and uneven cooling, the billet shell near the meniscus will be broken due to adhesion with the mold copper plate, forming a crack. Molten steel will then enter the crack to fill it, forming a new primary shell. In the next crystallizer vibration cycle that follows, the shell at the high temperature will be broken again due to the existence of billet drawing and adhesion, and a new high-temperature point will be formed below the original high-temperature point. Moreover, since the shell temperature around the high-temperature point is higher and the strength is lower, the crack will easily expand to both sides, causing the crack to become larger and larger. As the above process is repeated as the billet is drawn, the high-temperature point in the crystallizer will continue to move downward, forming a special abnormal area caused by adhesion on the surface of the continuous casting billet, usually in a V-shape.

[0004] After bonding occurs, the temperature field in the mold will change regularly. Therefore, in the prior art, the following methods are usually used to predict mold breakout:

[0005] 1) Breakout Prediction Based on Logic Judgment: A mold breakout prediction method based on logic judgment (patent application number: 2017100346797.4) discloses a mold breakout prediction and control method. Specifically, it uses the temperature changes of the mold thermocouples to prevent sticking breakout during the continuous casting process. The technical solution is: a slab continuous casting breakout prediction and control method, which installs multiple rows of thermocouples in the mold and is characterized by the following steps: A. Determining the typical temperature characteristics of each thermocouple based on mold parameters and the thermocouple's installation position in the mold; B. Determining the breakout probability of each thermocouple based on the typical temperature characteristics of adjacent thermocouples, while also considering breakout temperature transfer and temperature anomaly distribution characteristics; C. Stopping casting when the system predicts sticking breakout. This method is a common breakout prediction logic judgment method, but its drawback is that as equipment ages and materials are replaced, the key characteristic parameters of sticking breakout gradually change, resulting in increasingly inaccurate breakout predictions.

[0006] 2) About the method of using neural network to predict cohesive steel leakage. For example, a method for predicting continuous casting steel leakage based on neural network was invented (patent application number 201010207115.X), which discloses a method for predicting continuous casting steel leakage based on neural network. This method solves the problem of adding thermocouples inside the crystallizer in the prior art, collecting thermocouple temperature data at the continuous casting site online, and inputting the temperature into a single-pair time series model after temperature processing, and using a genetic algorithm module for comparison and judgment. If the temperature exceeds the specified value, an alarm will be issued and recorded. The shortcoming of this system is that both the neural network and the genetic algorithm model require a large amount of data as a training set. In the early stage of the production system, the training samples of the model are insufficient and the data is not perfect, so the entire steel leakage prediction system will have false alarms. At the same time, in step 3 of this scheme, the construction of the group-pair space model from the single-pair time series model is also completely dependent on the neural network method. It can be seen that the group-pair space model does not take into account the spatial variation characteristics of the thermocouple temperature during the propagation of the cohesive V-shaped tear. In the actual production process, when thermocouple failure occurs or the crystallizer temperature fluctuates greatly, the model will have false alarms and missed alarms, and cannot achieve a high alarm rate. Summary of the Invention

[0007] The present application is made in view of the above-mentioned problems, and its purpose is to provide an image feature extraction method for identifying bonding and leakage in a crystallizer. The method collects and processes the temperature in the crystallizer, highlights its changing regularity, and then simultaneously integrates time, space and other features to finally generate a pixel data covering multiple bonding features as input to a computer vision model, thereby accelerating the convergence of the network model and the prediction accuracy.

[0008] Specifically, the first aspect of the present application provides an image feature extraction method for identifying bonding breakout in a crystallizer, comprising the following steps:

[0009] Step 1: Collect temperature data of several thermocouples in the crystallizer;

[0010] Thermocouples are commonly used temperature measuring elements in temperature measuring instruments. They directly measure temperature and convert the temperature signal into a thermoelectric potential signal, which is then converted into the temperature of the measured medium through an electrical instrument (secondary instrument). The appearance of various thermocouples often varies greatly depending on the needs, but their basic structure is roughly the same, usually consisting of a thermode, an insulating protective tube, and a junction box.

[0011] Step 2: Combine the temperature data of several thermocouples horizontally in sequence, standardize the data in the time dimension, and mask the temperature data with a value less than m times the standard deviation to obtain the normal distribution of temperature in the time dimension;

[0012] After the temperature data are combined horizontally, the temperature data of different thermocouples arranged in sequence in the X direction are converted into a distribution with a mean of 0 and a standard deviation of 1 using the z-score normalization method. This method can eliminate the dimensional and order of magnitude differences between different variables, making the data more comparable and reliable.

[0013] The preset threshold is m times the standard deviation. For the standardized data, if a value is less than m times the standard deviation, it is marked as a mask, that is, set to NaN or a specific mark value, indicating that the data point is abnormal or unreliable.

[0014] Step 3: Normalize the obtained data and convert it into grayscale values, and mask the temperature data less than n again to form a temperature grayscale unit;

[0015] Grayscale values ​​are usually between 0 and 255, representing the brightness of the image. Through the linear mapping method and the maximum-minimum normalization method, the minimum value of the temperature data is mapped to 0, the maximum value is mapped to 255, and the intermediate data is proportionally mapped to the corresponding grayscale value.

[0016] The temperature data that is less than another preset threshold n is masked again, that is, the grayscale value of this data is set to 0 or a specific background color. The value of n usually depends on the specific application scenario and requirements to ensure that only key temperature information is retained.

[0017] Step 4: Based on the expansion pattern of the bonding feature between thermocouples, the time difference between the thermocouples is obtained. This is used to align the characteristics of different thermocouples when bonding occurs and obtain the area where the bonding feature occurs.

[0018] Step 5: Grid the temperature grayscale units to form a spatial temperature distribution in the X direction and a time dimension temperature distribution in the Y direction.

[0019] The spatial temperature distribution is formed in the X direction, that is, the temperature distribution in the spatial dimension, which is the temperature distribution of thermocouples at different positions.

[0020] Furthermore, the temperature data of the plurality of thermocouples are distinguished by using rectangular boxes with colors (0, 0, 255) to distinguish different thermocouples.

[0021] Furthermore, the reasonable range of m is 0.3 to 0.6.

[0022] If the temperature data is roughly normally distributed, a smaller m value, such as 1.5 or 2, might be more appropriate because these values ​​give a larger confidence interval that covers most data points. If the data distribution is more skewed, a larger m value might be necessary to ensure that masking doesn't overly impact the validity of the data. In this application, the purpose of masking is to remove outliers or noise, so a larger m value is more appropriate because these values ​​give a smaller confidence interval that effectively removes extreme values.

[0023] Furthermore, the reasonable range of n is 97-127.

[0024] Furthermore, the timing difference between the thermocouples is related to the pulling speed and the distance between the thermocouples.

[0025] Generally speaking, the larger the distance between thermocouples and the faster the pulling speed, the greater the timing difference between thermocouples will be.

[0026] Furthermore, a reasonable range of the timing difference between adjacent thermocouples is 9s to 12s.

[0027] Furthermore, a reasonable range of the timing difference between two thermocouples separated by one thermocouple is 17.5s to 20s.

[0028] Furthermore, the spatial temperature distribution is related to the actual width of the produced ingot and the number of effective thermocouple temperatures selected in the width direction, and is the distribution data of temperature changes with position.

[0029] Furthermore, the cumulative time length of the time-dimensional temperature distribution is related to the drawing speed, and a reasonable range of the cumulative time length is 45s to 60s.

[0030] In the second aspect, the present application also provides a computing device, which has the function of implementing the method described in the first aspect above. The beneficial effects can be found in the description of the first aspect and will not be repeated here. The function can be implemented by hardware or by executing the corresponding software through hardware. The hardware or software includes one or more modules corresponding to the above functions. In one possible design, the structure of the device includes an acquisition module, a training module, and optionally, a construction module. These modules can implement the function of the training node in the method example of the first aspect above. Please refer to the detailed description in the method example for details, which will not be repeated here.

[0031] In a third aspect, the present application further provides a computing device for implementing the functions of the method described in the first aspect above. The beneficial effects can be found in the description of the first aspect and will not be repeated here. The structure of the computing device includes a processor and a memory, and the memory is used to store instructions and / or data. The memory is coupled to the processor, and when the processor executes the program instructions stored in the memory, the functions of the training node in the example of the first aspect above can be implemented. The structure of the computing device also includes a communication interface for communicating with other devices.

[0032] In a fourth aspect, the present application also provides a computer-readable storage medium, which stores instructions. When the computer-readable storage medium is run on a computer, it enables the computer to execute the method in the above-mentioned first aspect and various possible designs of the first aspect.

[0033] In a fifth aspect, the present application also provides a computer program product comprising instructions, which, when executed on a computer, enables the computer to execute the method in the above-mentioned first aspect and various possible designs of the first aspect.

[0034] In a sixth aspect, the present application also provides a computing chip, which is connected to a memory and is used to read and execute software programs stored in the memory, and to execute the methods in the above-mentioned first aspect and various possible implementations of the first aspect. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] In order to more clearly illustrate the embodiments of the present drawings or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present drawings. For ordinary technicians in this field, other drawings can be obtained based on the structures shown in these drawings without paying any creative work.

[0036] Figure 1 A flowchart of the steps of this application;

[0037] Figure 2 This is the temperature distribution map without data processing;

[0038] Figure 3 This is the temperature distribution diagram without timing alignment;

[0039] Figure 4 This is the temperature distribution diagram after timing alignment;

[0040] Figure 5 The final temperature distribution diagram is shown in Figure 2.

[0041] Explanation of the accompanying numbers: 1. Temperature distribution of the first thermocouple; 2. Temperature distribution of the second thermocouple; 3. Temperature distribution of the third thermocouple; 4. Spatial temperature distribution; 5. Temperature distribution in the time dimension; 6. Timing difference between the second thermocouple and the first thermocouple; 7. Timing difference between the third thermocouple and the first thermocouple; 8. Area where bonding features occur; 9. Temperature grayscale unit.

[0042] The purpose, features and advantages of this drawing will be further described with reference to the accompanying drawings in conjunction with the embodiments. DETAILED DESCRIPTION

[0043] In order to make the purpose, technical solutions and advantages of this application more clearly understood, the present application is described and illustrated below in conjunction with the accompanying drawings and examples. It should be understood that the specific embodiments described herein are merely used to explain this application and are not intended to limit this application. Based on the embodiments provided in this application, all other embodiments obtained by those of ordinary skill in the art without making any creative efforts are within the scope of protection of this application.

[0044] Obviously, the drawings described below are merely examples or embodiments of the present application. Those skilled in the art can, without inventive effort, apply the present application to other similar scenarios based on these drawings. Furthermore, it is also understood that, although the effort involved in such a development process may be complex and lengthy, for those skilled in the art related to the content disclosed in this application, changes in design, manufacturing, or production based on the technical content disclosed in this application are merely conventional technical means and should not be construed as an insufficiency of the content disclosed in this application.

[0045] Unless otherwise specified, all embodiments and optional embodiments of the present application can be combined with each other to form a new technical solution.

[0046] Unless otherwise specified, all technical features and optional technical features of this application can be combined with each other to form a new technical solution.

[0047] If not specified otherwise, all steps of the present application can be performed in sequence or randomly, preferably in sequence. For example, the method comprises steps (a) and (b) means that the method can comprise steps (a) and (b) in sequence, or steps (b) and (a) in sequence. For example, the method can further comprise step (c) means that step (c) can be added to the method in any order, for example, the method can comprise steps (a), (b) and (c), or steps (a), (c) and (b), or steps (c), (a) and (b), etc.

[0048] If not specified otherwise, the terms "comprising" and "including" as used in the present application are open-ended. For example, the terms "comprising" and "including" can mean that other components not listed can also be included or can mean that only the listed components are included.

[0049] If not specified otherwise, in the present application, the term "or" is inclusive. For example, the phrase "A or B" means "A, B, or both A and B." More specifically, any of the following satisfy the condition "A or B": A is true (or exists) and B is false (or does not exist); A is false (or does not exist) and B is true (or exists); or both A and B are true (or exist).

[0050] In order to better understand the scheme of the embodiments of the present application, some related terms and concepts that may be involved in the embodiments of the present application are introduced first.

[0051] (1) Thermocouple, is a temperature measuring instrument commonly used in temperature measurement element, it directly measures temperature, and converts the temperature signal into thermoelectric power signal, through the electrical instrument (secondary instrument) into the temperature of the measured medium. The appearance of various thermocouples is often very different due to the need, but their basic structure is roughly the same, usually composed of hot electrode, insulation sleeve protection tube and terminal box and other main parts.

[0052] (2) Standardization, standardization is to convert data into a distribution with mean 0 and variance 1. Specifically, standardization is achieved by subtracting the mean of the data and dividing by the standard deviation, which can scale the data into a standard normal distribution. Standardization preserves the distribution characteristics of the data, including the information of outliers, so it is very useful in scenarios where these information needs to be preserved. For example, in machine learning, many linear models such as logistic regression and support vector machine (SVM) usually use standardization to initialize weights, so that it is easier to learn weights.

[0053] (3) Normalization: Normalization is the process of scaling the data to a fixed range, usually [0, 1]. This is done by subtracting the minimum value and then dividing by the difference between the maximum and minimum values. The main purpose of normalization is to simplify calculations and improve model training efficiency because it limits all feature values ​​to the same scale. Normalization is often used in scenarios that are insensitive to outliers because its output range is fixed.

[0054] In this embodiment, Figure 1 As shown, an image feature extraction method for identifying bonding breakout in a crystallizer includes the following steps:

[0055] Step 1: Collect temperature data from three rows of thermocouples in the crystallizer;

[0056] Thermocouples are commonly used temperature measuring elements in temperature measuring instruments. They directly measure temperature and convert the temperature signal into a thermoelectric potential signal, which is then converted into the temperature of the measured medium through an electrical instrument (secondary instrument). The appearance of various thermocouples often varies greatly depending on the needs, but their basic structure is roughly the same, usually consisting of a thermode, an insulating protective tube, and a junction box.

[0057] The temperature distribution data without data processing is as follows Figure 2 shown.

[0058] Step 2: Combine the temperature data of the three rows of thermocouples horizontally in sequence, standardize the data in the time dimension, and mask the temperature data with a value less than m times the standard deviation to obtain the normal distribution of temperature in the time dimension;

[0059] After the temperature data are combined horizontally, the temperature data of different thermocouples arranged in sequence in the X direction are converted into a distribution with a mean of 0 and a standard deviation of 1 using the z-score normalization method. This method can eliminate the dimensional and order of magnitude differences between different variables, making the data more comparable and reliable.

[0060] The preset threshold is m times the standard deviation. For the standardized data, if a value is less than m times the standard deviation, it is marked as a mask, that is, set to NaN or a specific mark value, indicating that the data point is abnormal or unreliable.

[0061] Step 3: Normalize the obtained data and convert it into grayscale values, and mask the temperature data less than n again to form a temperature grayscale unit;

[0062] Grayscale values ​​are usually between 0 and 255, representing the brightness of the image. Through the linear mapping method and the maximum-minimum normalization method, the minimum value of the temperature data is mapped to 0, the maximum value is mapped to 255, and the intermediate data is proportionally mapped to the corresponding grayscale value.

[0063] The temperature data that is less than another preset threshold n is masked again, that is, the grayscale value of this data is set to 0 or a specific background color. The value of n usually depends on the specific application scenario and requirements to ensure that only key temperature information is retained.

[0064] Step 4: Based on the expansion pattern of the bonding feature between thermocouples, the time difference between the thermocouples is obtained. This is used to align the characteristics of different thermocouples when bonding occurs and obtain the area where the bonding feature occurs.

[0065] Step 5: Grid the temperature grayscale units to form a spatial temperature distribution in the X direction and a time dimension temperature distribution in the Y direction.

[0066] The spatial temperature distribution is formed in the X direction, that is, the temperature distribution in the spatial dimension, which is the temperature distribution of thermocouples at different positions.

[0067] Furthermore, the temperature data of the plurality of thermocouples are distinguished by using rectangular boxes with colors (0, 0, 255) to distinguish different thermocouples.

[0068] Furthermore, the value of m is 0.3.

[0069] If the temperature data is roughly normally distributed, a smaller m value, such as 1.5 or 2, might be more appropriate because these values ​​give a larger confidence interval that covers most data points. If the data distribution is more skewed, a larger m value might be necessary to ensure that masking doesn't overly impact the validity of the data. In this application, the purpose of masking is to remove outliers or noise, so a larger m value is more appropriate because these values ​​give a smaller confidence interval that effectively removes extreme values.

[0070] Furthermore, the value of n is 105.

[0071] Furthermore, the timing difference between thermocouples is related to the pulling speed and the distance between thermocouples.

[0072] Generally speaking, the larger the distance between thermocouples and the faster the pulling speed, the greater the timing difference between thermocouples will be.

[0073] Furthermore, a reasonable range of the timing difference between adjacent thermocouples is 9s to 12s.

[0074] Furthermore, a reasonable range of the timing difference between two thermocouples separated by one thermocouple is 17.5s to 20s.

[0075] In this embodiment, the temperature distribution without timing alignment is shown in Figure 3 The temperature distribution after timing alignment is shown in Figure 4 The timing alignment of thermocouples refers to the process of ensuring that each thermocouple signal is collected and processed at the same time point in a multi-channel thermocouple signal acquisition system. This alignment is to ensure the synchronization of each thermocouple signal, thereby improving the accuracy and reliability of data acquisition.

[0076] Timing alignment can ensure that multiple thermocouples can synchronize data acquisition when measuring the same temperature field, thereby improving the consistency and accuracy of measurement results. In addition, timing alignment helps to reduce errors caused by different synchronization between different thermocouples, which is particularly important for applications requiring high-precision temperature measurement.

[0077] Further, the spatial temperature distribution is related to the actual production of the slab width and the selection of the number of effective thermocouple temperatures in the width direction, which is the distribution data of the temperature changing with position.

[0078] Further, the cumulative time length of the time dimension temperature distribution is related to the pulling speed, and the reasonable range of the cumulative time length is 45s-60s.

[0079] In this embodiment, the timing difference between the second thermocouple and the first thermocouple is 9s, the timing difference between the third thermocouple and the first thermocouple is 18s, and the cumulative time length of the time dimension temperature distribution is 50s.

[0080] In this embodiment, the final temperature distribution is shown in Figure 5 1, 2, and 3 are the temperature distribution of the first thermocouple, the temperature distribution of the second thermocouple, and the temperature distribution of the third thermocouple, respectively, which are the parts within the blue box in the outermost part, and the temperature distributions of the three thermocouples are connected horizontally; 4 is the spatial temperature distribution, which is the horizontal spatial dimension temperature distribution; 5 is the time dimension temperature distribution, which is the vertical temperature distribution; 6 is the timing difference between the second thermocouple and the first thermocouple, which is the part within the green box in the middle of the second thermocouple temperature distribution; 7 is the timing difference between the third thermocouple and the first thermocouple, which is the part within the green box in the third thermocouple temperature distribution; 8 is the bonding feature occurrence area, which is the part within the red box; and 9 is the temperature gray unit after gridding.

[0081] From Figure 5 It can be seen that in the bonding feature occurrence area 8, as the repeated drawing of the billet, the high temperature point in the mold also moves down, and a special abnormal area caused by bonding is formed on the surface of the continuous casting billet, showing a V-shaped temperature distribution characteristic.

[0082] Previously, only the temperature field of one side of the entire crystallizer was recorded in real time, without continuous changes in the time dimension, so the V-shaped features of the bonding area could not be restored. However, the feature extraction method proposed in the present invention can, on the one hand, better restore the V-shaped features of the bonding area, and on the other hand, further extract features from the temperature data, strengthen the bonding expansion law, and improve the accuracy of computer recognition to more than 98%.

[0083] It should be noted that the present application is not limited to the above-mentioned embodiments. The above-mentioned embodiments are merely examples, and any embodiments having substantially the same structure and effect as the technical concept within the scope of the present application are all included in the technical scope of the present application. In addition, without departing from the scope of the present application, any other embodiments that can be conceived by those skilled in the art and that combine some of the constituent elements in the embodiments are also included in the scope of the present application.

Claims

1. An image feature extraction method for identifying bonding breakout in a crystallizer, characterized in that: The following steps are involved: Step 1: Collect temperature data of several thermocouples in the crystallizer; Step 2: Combine the temperature data of several thermocouples horizontally in sequence, standardize the data in the time dimension, and mask the temperature data with a value less than m times the standard deviation to obtain the normal distribution of temperature in the time dimension; Step 3: Normalize the obtained data and convert it into grayscale values, and mask the temperature data less than n again to form a temperature grayscale unit; Step 4: Based on the expansion pattern of the bonding feature between thermocouples, the time difference between the thermocouples is obtained. This is used to align the characteristics of different thermocouples when bonding occurs and obtain the area where the bonding feature occurs. Step 5: Grid the temperature grayscale units to form a spatial temperature distribution in the X direction and a time dimension temperature distribution in the Y direction.

2. The image feature extraction method for identifying bonding breakout in a crystallizer according to claim 1, characterized in that: The temperature data of the several thermocouples are distinguished by rectangular boxes with colors (0, 0, 255) to distinguish different thermocouples.

3. The image feature extraction method for identifying bonding breakout in a crystallizer according to claim 1, characterized in that: The reasonable range of m is 0.3 to 0.

6.

4. The image feature extraction method for identifying bonding breakout in a crystallizer according to claim 1, characterized in that: The reasonable range of n is 97-127.

5. The image feature extraction method for identifying bonding breakout in a crystallizer according to claim 1, characterized in that: The timing difference between the thermocouples is related to the drawing speed and the distance between the thermocouples.

6. The image feature extraction method for identifying bonding breakout in a crystallizer according to claim 5, characterized in that: The reasonable range of the timing difference between adjacent thermocouples is 9s to 12s.

7. The image feature extraction method for identifying bonding breakout in a crystallizer according to claim 5, characterized in that: The reasonable range of the timing difference between two thermocouples separated by one thermocouple is 17.5s to 20s.

8. The image feature extraction method for identifying bonding breakout in a crystallizer according to claim 1, characterized in that: The spatial temperature distribution is related to the actual width of the produced ingot and the number of effective thermocouple temperatures selected in the width direction, and is the distribution data of temperature changes with position.

9. The image feature extraction method for identifying bonding breakout in a crystallizer according to claim 1, characterized in that: The cumulative time length of the time-dimensional temperature distribution is related to the drawing speed, and a reasonable range of the cumulative time length is 45s to 60s.

Images