Method for predicting life of refractory

Abstract

A method for predicting the service life of a refractory that lines the interior of a molten steel vessel and contacts molten metal in a state in which deposits are present on the surface comprises: a step (S22) for acquiring measurement data indicating the surface state of the refractory as measured by a sensor; a step (S23) for calculating thickness data from the difference between the measurement data and surface reference data indicating a reference surface state; a step (S24) for comparing the thickness data with minimum thickness data that is calculated on the basis of prior thickness data and that indicates the minimum thickness, and calculating deposition thickness data indicating the thickness of the deposit; and a step (S26) for inputting the deposited thickness data into a service life estimation model to predict the service life of the refractory.

Description

Method for predicting the life of refractory materials

[0001] The present disclosure relates to a method for predicting the life of refractory materials. In particular, the present disclosure relates to a method for predicting the life of refractory materials used for lining a molten steel container that transports or holds molten steel and performs various refining processes, taking into account the peeling of the unshaped refractory materials.

[0002] The refractory materials used in molten steel containers are divided into a lining in contact with the high-temperature melt and an outer lining (permanent lining) on the casing side such as an iron skin. The lining refractory materials are damaged by chemical erosion due to contact with molten steel and slag, the occurrence of cracks due to thermal load, and peeling following the occurrence of cracks. Therefore, before the lining refractory materials are completely consumed, they are replaced with new lining refractory materials. The damage situation is confirmed by visual monitoring of the lining, in addition to monitoring the temperature of the iron skin using thermocouples embedded in the refractory lining or infrared thermography, and measuring the remaining thickness using a three-dimensional laser scanner.

[0003] For example, Patent Document 1 proposes a measurement system for predicting the future state of a refractory lining that is lined on the inner surface of the outer wall of a metallurgical container and is exposed to heat while being exposed to molten metal. This system includes one or more laser scanners and a processor. The laser scanner performs a plurality of laser scans on the refractory lining when the metallurgical container is empty. At least one of the laser scanners laser scans the refractory lining before heating and collects data related to the structural state of the refractory lining before heating. At least one of the laser scanners collects data related to the structural state of the refractory lining after heating. The processor predicts the future state of the lining.

[0004] Japanese Patent Publication No. 2022-550056

[0005] Here, the technology described in Patent Document 1 does not assume a state where there are deposits on the surface of the unshaped refractory materials, and it is difficult to estimate the peeling damage of such refractory materials.

[0006] In view of such circumstances, an object of the present disclosure is to provide a method for predicting the life of refractory materials that takes into account the peeling damage of refractory materials lined on the inner surface of a molten steel container that transports or holds molten metal.

[0007] (1) A method for predicting the lifespan of a refractory material according to one embodiment of the present disclosure is a method for predicting the lifespan of a refractory material that is in contact with molten metal with a surface that has deposits on it and is lined with a molten steel container, and includes the steps of: acquiring measurement data indicating the surface state of the refractory material measured by a sensor; calculating thickness data from the difference between the measurement data and surface reference data indicating a reference surface state; calculating adhesion thickness data indicating the thickness of the deposits by comparing minimum thickness data indicating the minimum thickness calculated based on past thickness data with the thickness data; and predicting the lifespan of the refractory material by inputting the adhesion thickness data into a lifespan estimation model.

[0008] (2) In one embodiment of the present disclosure, in (1), the life estimation model is a model generated by machine learning, with past adhesion thickness data as input and life data indicating the past life of the refractory material as output.

[0009] (3) In one embodiment of the present disclosure, in (2), the lifetime estimation model is a model generated by machine learning, with the surface temperature of the refractory material and operational data from the manufacturing of metal products from the molten metal as further inputs.

[0010] (4) In one embodiment of the present disclosure, in (2), the lifetime estimation model is a model generated by machine learning with the thickness data and the minimum thickness data as further inputs.

[0011] (5) In one embodiment of the present disclosure, in (4), the input image is an image obtained by combining multiple images which are images of the thickness data and the minimum thickness data for multiple charges.

[0012] (6) In one embodiment of the present disclosure, in (2), the lifetime estimation model is a model generated by machine learning, with the mean and standard deviation calculated from the thickness data at each measurement point of the refractory material as further input.

[0013] According to this disclosure, a method for predicting the lifespan of refractories can be provided, which takes into account delamination damage to refractories lining the inner surface of molten steel containers used for transporting or holding molten metal.

[0014] Figure 1 is a schematic diagram of a refractory life prediction system that implements a refractory life prediction method according to one embodiment of the present disclosure. Figure 2 is a diagram showing an example of the configuration of a molten steel container. Figure 3 is a diagram for explaining the temperature gradient of the refractory. Figure 4 is a diagram for explaining delamination damage. Figure 5 is a diagram showing the measurement of the molten steel container with a sensor. Figure 6 is a flowchart illustrating the processing of a method for generating a life estimation model. Figure 7 is a flowchart illustrating the processing of a refractory life prediction method according to one embodiment of the present disclosure.

[0015] A method for predicting the lifespan of a refractory material according to one embodiment of this disclosure will be described below with reference to the drawings. Herein, the drawings are schematic and may differ from the actual shape, size, etc.

[0016] Figure 1 is a schematic diagram of a refractory life prediction system that performs the refractory life prediction method according to this embodiment. The refractory life prediction system comprises a data processing device 10 and a sensor 20. The refractory life prediction system may further include an operation database 21. The operation database 21 stores operation data when manufacturing metal products from molten metal. The operation database 21 also stores various measurement data related to the manufacturing equipment. The data processing device 10 comprises a storage unit 11 and a calculation unit 12. The storage unit 11 stores various data used in the refractory life prediction method, including a life prediction model described later. The calculation unit 12 performs various calculations for the refractory life prediction method. The data processing device 10 predicts the life of the refractory that is in contact with molten metal with deposits on its surface, and which is lined inside a molten steel container 30 (see Figure 2). The components of the refractory life prediction system are described individually below.

[0017] (Structure of the molten steel container 30) Figure 2 shows an example of the configuration of the molten steel container 30. The molten steel container 30 is sometimes called a ladle. The molten steel container 30 has a multi-layered structure with an inner lining and a permanent lining inside the steel shell. As an inner lining, bricks (shaped refractory material) are installed on the upper part of the sides, and refractory material (unshaped refractory material) is installed on the lower part. On the upper part of the sides, the refractory material is damaged (melted away) by contact with molten slag. For this reason, dense bricks with excellent melting resistance are used. On the lower part of the sides, since it is mainly in contact with molten steel, unshaped refractory material with excellent workability is used. In this embodiment, the refractory material that is the subject of life prediction is the unshaped refractory material installed on the lower part of the sides. Hereinafter, unshaped refractory material will be simply referred to as refractory material.

[0018] The molten steel container 30 receives molten steel from the converter, performs secondary refining within the container, and then discharges the molten steel from the bottom of the container 30 for casting. After that, any remaining slag in the container 30 is discharged by tilting the container 30. The container 30 remains empty until the next batch of molten steel is received from the converter. Receiving molten steel from the converter and receiving the next batch of molten steel constitutes one cycle (one charge).

[0019] (Adhesion and Infiltration Phase) The molten steel container 30 receives molten steel from the converter multiple times until the refractory material is dismantled and repaired. Through use, the molten steel container 30 develops deposits on its surface called buildup. These deposits are impurities consisting of slag and steel mixed with the alumina content in the steel that has precipitated on the refractory material.

[0020] Furthermore, during the secondary refining process, slag penetrates into the refractory material through the open vents in the refractory material of the molten steel container 30. The refractory material and slag react chemically to form compounds. Here, Figure 3 is a diagram illustrating the temperature gradient of the refractory material. The refractory material has a temperature gradient in which the working surface side in contact with the molten steel is hotter and the back side (the side away from the molten steel) is hotter. Because the temperature inside the refractory material is lower than the working surface side, the solidification temperature is reached at a predetermined location inside the refractory material. The region of refractory material where slag reacts with the refractory material and solidifies is called the wetting phase. The region of refractory material where slag does not penetrate and retains the original structure is called the sound phase. Here, since the thermal conductivity of the adhering material and the wetting phase is different from that of the sound phase, the temperature gradient is not strictly linear, but in Figure 3, the temperature gradient is shown as a straight line for convenience.

[0021] (Changes in Thickness) The refractory material of the molten steel container 30 decreases in thickness with use. The factors causing the decrease in the thickness of the refractory material are delamination damage and melting. Of these, delamination damage occurs suddenly. In conventional technology, delamination damage was not estimated. When delamination damage occurs, there is a risk of steel leakage due to the rapid decrease in the thickness of the refractory material, so estimating delamination damage is important for life management such as repair planning of the refractory material. In addition, as mentioned above, deposits are formed on the surface of the refractory material. The thickness of these deposits increases with use.

[0022] (Splitting Damage) Figure 4 is a diagram illustrating spalling damage and shows a cross-section of a part of the molten steel container 30. As described above, the refractory material consists of an adhering layer, a wetting phase, and a sound phase. Due to the effects of heating and cooling, as well as differences in physical properties, as shown in Figure 4, a part of the refractory material on the working surface (the surface in contact with the molten steel) side may suffer spalling damage from the interface between the wetting phase and the sound phase.

[0023] (Estimation of delamination damage) In this embodiment, the refractory life prediction system measures the surface of the molten steel container 30 with the sensor 20 before receiving the next batch of molten steel from the converter. Here, the surface of the molten steel container 30 is the working surface of the refractory material of the molten steel container 30 (see Figure 4). The sensor 20 is not limited to any specific device as long as it is capable of measuring the surface shape, but may be, for example, a three-dimensional laser scanner or an infrared thermograph. In this embodiment, the sensor 20 will be described as a three-dimensional laser scanner.

[0024] Figure 5 shows the process of measuring the molten steel vessel 30 with a three-dimensional laser scanner. The three-dimensional laser scanner measures the surface condition of the refractory material at each charge before receiving the next batch of molten steel from the converter. As shown in Figure 5, the surface condition is measured at multiple measurement points on the refractory material. In this embodiment, the three-dimensional laser scanner also measures the molten steel vessel 30 before the refractory material is applied (only permanent lining). The measurement data of the molten steel vessel 30 before the refractory material is applied is used as surface reference data indicating a standard surface condition. In this embodiment, all data measured by the three-dimensional laser scanner is stored in the operation database 21.

[0025] The calculation unit 12 of the data processing device 10 acquires measurement data indicating the surface state of the measured refractory material from the operation database 21 or a three-dimensional laser scanner. The calculation unit 12 also acquires surface reference data indicating the reference surface state of the molten steel container 30. The calculation unit 12 can calculate thickness data from the difference between the measurement data and the surface reference data. The thickness data indicates the thickness of the refractory material at each measurement point. Here, as shown in Figure 4, the calculated thickness is the total thickness of the refractory material and the adhering material. Hereafter, the thickness of the adhering material alone will be referred to as the adhering thickness.

[0026] As described above, the thickness of the refractory material may become thinner than its initial thickness due to sudden delamination damage. Therefore, if the adhesion thickness is calculated using the initial thickness of the refractory material after the thickness has changed due to delamination damage, an accurate value cannot be obtained. In this embodiment, therefore, minimum thickness data, which is calculated based on past thickness data and indicates the minimum thickness, is used. More specifically, the thickness data of the refractory material after its construction is stored in the storage unit 11 as a measurement history in chronological order, and the calculation unit 12 identifies the minimum thickness of the refractory material (minimum thickness) by comparing it with past thickness data. For example, in the example in Figure 4, suppose the initial thickness of the refractory material was the minimum thickness (left figure), but delamination damage occurs. In this case, the minimum thickness is updated by the calculation unit 12 (right figure). By updating the minimum thickness in this way, even if delamination damage occurs, the calculation unit 12 can accurately calculate the adhesion thickness data by comparing the minimum thickness data with the thickness data.

[0027] The calculation unit 12 predicts the lifespan of the refractory by inputting the adhesion thickness data into the lifespan estimation model. Before executing the process of predicting the lifespan of the refractory, the calculation unit 12 generates a lifespan estimation model and stores it in the storage unit 11. In this embodiment, the lifespan estimation model is a model generated by machine learning, with past adhesion thickness data as input and lifespan data indicating the lifespan of past refractory materials as output. In other words, the lifespan estimation model in this embodiment is a trained model generated by machine learning using training data in which the accumulated past adhesion thickness data is used as an explanatory variable and lifespan data indicating the lifespan of refractory materials corresponding to that adhesion thickness data is used as the objective variable. The lifespan data may be classified (labeled), for example, by whether or not there is delamination after one charge. Here, the method for generating the lifespan estimation model is not limited to a specific method, and for example, any of the following first to fourth methods may be executed.

[0028] (First Method) As explanatory variables, unfolded diagrams obtained by imaging thickness data and minimum thickness data in grayscale may be used. In other words, a lifetime estimation model may be generated by machine learning using images obtained by imaging thickness data and minimum thickness data as further input. In this case, the model is designed to distinguish between the thickness of the invasive phase and the healthy phase, which is thought to improve the accuracy of lifetime prediction. Here, a convolutional neural network may be used as the machine learning method.

[0029] (Second Method) As explanatory variables, a combination of multiple (e.g., three) grayscale images of unfolded diagrams of thickness data and minimum thickness data may be used. In other words, the input image may be an image created by combining multiple images of thickness data and minimum thickness data for multiple charges. Here, a convolutional neural network may be used as the machine learning method.

[0030] (Third Method) Surface temperature data of refractories and operational data used in steel manufacturing may be used as explanatory variables. In other words, a life estimation model may be generated by machine learning using the surface temperature of refractories and operational data used in manufacturing metal products from molten metal as further inputs. It is thought that the accuracy of life prediction will improve by inputting, for example, the highest temperature as the surface temperature data of the refractories. It is also thought that the accuracy of life prediction will improve by inputting, for example, the components of molten steel that are greatly involved in the chemical reaction with the refractories as operational data. Here, a convolutional neural network may be used as the machine learning method.

[0031] (Fourth Method) The mean and standard deviation calculated from the thickness of the refractory material at each measurement point may be used as explanatory variables. In other words, a lifetime estimation model may be generated by machine learning using the mean and standard deviation calculated from the thickness data at each measurement point of the refractory material as input. Here, a gradient boosting tree may be used as the machine learning method.

[0032] Here, the data processing device 10 may acquire a lifetime estimation model generated by another device and store it in the storage unit 11, but it is preferable that the calculation unit 12 has a function (learning function) to generate lifetime estimation models, as in this embodiment. By having a learning function in the calculation unit 12, the lifetime estimation model stored in the storage unit 11 can be updated, thereby improving performance.

[0033] The performance of the lifetime estimation model may be evaluated by the error in the predicted values, for example, using test data. For the evaluation, the mean absolute error (MAE) may be used as the error in the predicted values. If the magnitude of the error in the predicted values ​​is greater than a predetermined threshold, the calculation unit 12 may determine that the performance is insufficient and perform retraining of the lifetime estimation model. The threshold may be determined according to the required prediction performance. Furthermore, the lifetime estimation model may be evaluated by recall and precision. Here, recall is an index representing the proportion of events that actually delaminated that the model successfully determined to delaminate. Precision is an index representing the proportion of events that actually delaminated that the model predicted to delaminate.

[0034] The machine learning methods are not limited to convolutional neural networks and gradient boosting trees. Furthermore, the lifetime estimation model may be a combination of multiple models. Additionally, data from various sensors may be added as explanatory variables.

[0035] The data processing device 10 may be implemented as, for example, a computer. The computer may include a storage device such as memory, a control device such as a CPU (central processing unit), and other processors. In this case, the storage unit 11 may be implemented as the storage device of the computer. The arithmetic unit 12 may be implemented by the control device of the computer by executing a program. The program may be stored in the storage device and read by the control device when it is to be executed.

[0036] (Method for generating a lifespan estimation model) Figure 6 is a flowchart showing the process for generating a lifespan estimation model. The generation of the lifespan estimation model is performed before the execution of the lifespan prediction method for refractories. The calculation unit 12 acquires surface reference data (step S11). The calculation unit 12 also acquires past measurement data (step S12). The calculation unit 12 calculates thickness data from the difference between the measurement data and the surface reference data (step S13). The calculation unit 12 compares the minimum thickness data with the thickness data to calculate the adhesion thickness data (step S14). As described above, the minimum thickness data is updated when delamination damage occurs by comparison with past thickness data. The calculation unit 12 also acquires lifespan data indicating the past lifespan of refractories corresponding to past measurement data (step S15). The calculation unit 12 generates a lifespan estimation model by machine learning (step S16). The calculation unit 12 evaluates the generated lifespan estimation model using test data and calculates the error in the predicted value (step S17). If the magnitude of the error is greater than a predetermined threshold (Yes in step S18), the calculation unit 12 determines that the performance is insufficient and returns to the process in step S16 to retrain the lifetime estimation model. If the magnitude of the error is less than or equal to a predetermined threshold (No in step S18), the calculation unit 12 determines that the performance is sufficient and stores the generated lifetime estimation model in the storage unit 11, ending the series of processes.

[0037] (Method for predicting the lifespan of refractory materials) Figure 7 is a flowchart showing the processing of the method for predicting the lifespan of refractory materials according to this embodiment. The calculation unit 12 acquires surface reference data (step S21). The calculation unit 12 also acquires measurement data (step S22). The calculation unit 12 calculates thickness data from the difference between the measurement data and the surface reference data (step S23). The calculation unit 12 compares the minimum thickness data with the thickness data to calculate the adhesion thickness data (step S24). As described above, the minimum thickness data is updated when delamination damage occurs by comparing it with past thickness data. The calculation unit 12 also acquires a lifespan estimation model from the storage unit 11 (step S25). The calculation unit 12 predicts the lifespan of the refractory material by inputting the adhesion thickness data into the lifespan estimation model (step S26).

[0038] Here, the processing for generating the lifetime estimation model in Figure 6 and the processing for predicting the lifetime in Figure 7 are examples. For example, when the first method described above is adopted, in the processing for generating the lifetime estimation model and predicting the lifetime, the calculation unit 12 further uses thickness data and minimum thickness data as inputs (explanatory variables) to the lifetime estimation model.

[0039] (Example) The molten steel container 30 has a cylindrical inner surface, a circular opening with a diameter of 4.5 m, and a height of 4 m. A 3D laser scanner was installed at a height of 2.8 m from the opening plane of the molten steel container 30 to measure the surface shape of the refractory. The data processing device 10 calculated the thickness of the refractory as described above, and the lifespan of the refractory was predicted. In generating the lifespan estimation model, 20,000 images of the refractory thickness data were prepared and classified into 14,000 training data and 6,000 test data. Images were labeled based on whether or not delamination occurred after one charge. When the lifespan estimation model was evaluated using the test data, it was confirmed by cross-validation that it was not overfitting, and its usefulness was confirmed.

[0040] As described above, the refractory life prediction method according to this embodiment can take into account (estimate) delamination damage to the refractory material used for lining the molten steel container 30. By estimating delamination damage to the refractory material, the risk of steel leakage is reduced, and it becomes possible to accurately set the timing for repairing the refractory material.

[0041] While embodiments relating to this disclosure have been described based on the drawings and examples, the above embodiments are illustrative of methods for realizing the technical idea of ​​this disclosure and do not specify a particular configuration. Those skilled in the art should note that it will be easy to make various modifications or alterations based on this disclosure. Therefore, it should be noted that these modifications or alterations are within the scope of this disclosure.

[0042] 10 Data processing device 11 Storage unit 12 Calculation unit 20 Sensor 21 Operation database 30 Molten steel container

Claims

1. A method for predicting the lifespan of a refractory material that is in contact with molten metal with a surface containing deposits, comprising: acquiring measurement data indicating the surface condition of the refractory material measured by a sensor; calculating thickness data from the difference between the measurement data and surface reference data indicating a reference surface condition; calculating deposit thickness data indicating the thickness of the deposits by comparing minimum thickness data indicating the minimum thickness calculated based on past thickness data with the thickness data; and predicting the lifespan of the refractory material by inputting the deposit thickness data into a lifespan estimation model.

2. The method for predicting the lifespan of a refractory material according to claim 1, wherein the lifespan estimation model is a model generated by machine learning, with past adhesion thickness data as input and lifespan data indicating the lifespan of the refractory material in the past as output.

3. The method for predicting the lifespan of a refractory material according to claim 2, wherein the lifespan estimation model is a model generated by machine learning, with the surface temperature of the refractory material and operational data from the manufacturing of metal products from the molten metal as further inputs.

4. The method for predicting the lifespan of a refractory material according to claim 2, wherein the lifespan estimation model is a model generated by machine learning using the thickness data and the minimum thickness data as further inputs.

5. The method for predicting the lifespan of a refractory material according to claim 4, wherein the input image is an image obtained by combining multiple images of the thickness data and the minimum thickness data for multiple charges.

6. The method for predicting the lifespan of a refractory material according to claim 2, wherein the lifespan estimation model is a model generated by machine learning, with the mean and standard deviation calculated from the thickness data at each measurement point of the refractory material as further inputs.

Images