TWI845124B - Method for estimating residual thickness of firebrick of ladles - Google Patents

Abstract

A method for estimating residual thickness of firebrick of ladles, including: providing a plurality of historical production data of a plurality of historical ladles, wherein each of the plurality of historical production data includes at least one use time of the historical ladles and at least one corresponding residual thickness value of the historical ladles; determining the normal and abnormal boundary line according to an original thickness value, a preset safe thickness value, an initial use times and a target use times, thereby distinguishing the plurality of historical production data of the plurality of historical ladles into a normal area or an abnormal area; providing a using time prediction intervals to determine a desired predict range; stretching the plurality of historical production data to have a same use times according to the using time prediction intervals, thereby obtaining a complete historical production training data, wherein each use times of the at least one ladle in the historical production data that is less than the same use times is filled by a filling value; establishing a residual thickness estimating model by using the complete historical production training data; substituting an actual production data of a ladle to be predicted into the residual thickness estimating model to obtain an estimated residual thickness value.

Description

Method for estimating the residual thickness of refractory brick lining in steel ladle

The present invention relates to a method for estimating thickness, and in particular to a method for estimating the residual thickness of a refractory brick lining of a steel ladle.

Steel drums are the main equipment for loading molten steel in steel mills. They must be monitored and maintained to stabilize the number of steel drums available online. However, the current maintenance management of refractory brick linings of steel drums is based on manual visual inspection of defects and the number of times they are used as the basis for judging offline. Misjudgment may lead to steel leak accidents, delayed scheduling, and personal safety incidents.

The purpose of the present invention is to provide a method for estimating the residual thickness of refractory brick lining of a steel ladle, which is applicable to estimating the residual thickness of refractory brick lining of a steel ladle that is continuously used online, thereby assisting the steel ladle maintenance unit to judge and analyze the maintenance quality.

The method for estimating the residual thickness of refractory brick lining of a steel ladle comprises: a model building stage and an online estimation stage. In the model building stage, firstly, a plurality of historical production data of a plurality of historical steel ladle are provided, wherein each of the historical production data comprises at least one steel ladle usage count and at least one corresponding steel ladle residual thickness value, and wherein each of the historical steel ladle corresponds to one historical production data; a positive and abnormal boundary line is determined according to the original thickness value, the preset safety thickness value, the initial usage count and the target usage count, so as to divide the historical production data of the historical steel ladle into a normal area and an abnormal area. and abnormal area; provide a prediction interval to determine the range to be predicted; according to the prediction interval, the historical production data is stretched to have the same number of times used, so that the historical production data becomes a complete historical production training set, where the same number of times used is the maximum value of the prediction interval, and each time the number of times the steel barrel in the historical production data is used is less than the same number of times used, it is filled in by filling in the value; finally, the complete historical production training set is used to establish a residual thickness estimation model. In the online estimation stage, the actual production data of the steel barrel to be predicted is first input into the residual thickness estimation model, and the residual thickness estimation value of the steel barrel to be predicted is obtained according to the residual thickness estimation model.

In some embodiments, the historical production data includes historical steel ladle production parameters.

In some embodiments, the model building stage further includes a positive-abnormal labeling step for the historical production data, wherein the historical production data in the normal zone is marked as 0, and the historical production data in the abnormal zone is marked as 1.

In some embodiments, the residual thickness estimation model is one of the common regression models of Random Forest, Extreme Gradient Boosting (XGBoost), Lightweight Gradient Boosting Machine (LightGBM), and CatBoost.

In some embodiments, the number of prediction intervals is determined based on operator demand.

In some embodiments, the maximum value of the prediction interval of the number of times may be determined according to the maximum number of times the steel ladle has been used in history.

In some embodiments, the online prediction stage further includes predicting the erosion curvature of the steel ladle to be predicted.

In some embodiments, the step of predicting the erosion curvature includes: determining whether the ladle to be predicted is located in a normal zone or an abnormal zone; when the ladle to be predicted is located in the normal zone, performing a first erosion curvature prediction on the ladle to be predicted; and when the ladle to be predicted is located in the abnormal zone, performing a second erosion curvature prediction on the ladle to be predicted.

In some embodiments, the minimum value of the prediction interval is 0.

In some embodiments, the minimum value of the prediction interval is a positive integer greater than 0.

The following is a detailed description of the implementation with accompanying drawings, but the implementation provided is not intended to limit the scope of the invention, and the description of the structure and operation is not intended to limit the order of its execution. Any structure reassembled from the components to produce a device with equal functions is within the scope of the invention. In addition, the drawings are for illustration purposes only and are not drawn according to the original size.

The terms “first,” “second,” etc. used in this document do not particularly refer to order or sequence, but are only used to distinguish elements or operations described with the same technical terminology.

Please refer to FIG. 1 , which is a schematic flow chart of a method 100 for estimating the residual thickness of a refractory brick lining of a steel ladle according to an embodiment of the present invention. The method 100 for estimating the residual thickness of a refractory brick lining of a steel ladle is applicable to estimating the residual thickness of a refractory brick lining of a steel ladle to assist a steel ladle maintenance unit in judging, analyzing and maintaining quality. The method 100 for estimating the residual thickness of a refractory brick lining of a steel ladle comprises a model building stage 110 and an online estimation stage 120. In the model building stage 110, steps 111 to 115 are performed to pre-process the historical production data of the historical steel ladle, thereby obtaining a complete historical production training set, and a model capable of estimating the residual thickness of the refractory brick lining of the steel ladle is established through the complete historical production training set. In the online estimation stage 120, steps 121 to 123 are performed to obtain the estimated residual thickness value and erosion curvature of the refractory brick lining of the online steel ladle using the residual thickness estimation model established above.

In the model building stage 110, step 111 is first performed to provide a plurality of historical production data of a plurality of historical steel ladle. One historical steel ladle corresponds to one historical production data, and each historical production data includes but is not limited to the number of times the historical steel ladle is used and the corresponding steel ladle residual thickness value and historical production parameters, wherein the historical production parameters also include the steel grade characteristics of the historical steel ladle and a plurality of working hours of the historical steel ladle at each production station, for example, the working time of a historical steel ladle at workstation A is 30 minutes, the working time at workstation B is 1 hour, and/or the working time at workstation C is 2 hours, etc. and other related process information.

After step 111, step 112 is then performed to determine the positive-abnormal boundary line 200 according to the original thickness value, the preset safety thickness value, the initial number of times used, and the target number of times used, and then the historical production data of the historical steel-filling barrel is divided into a normal area and an abnormal area. As shown in FIG2, the positive-abnormal boundary line 200 is a dividing line defined by the initial number of times used (the icon is set to 0), the target number of times used, the original thickness value, and the preset safety thickness value, and the historical production data of the historical steel-filling barrel is classified by this. When the historical production data point falls below the positive-abnormal boundary line 200, it is determined to be in the abnormal area; when the historical production data point falls above the positive-abnormal boundary line 200, it is determined to be in the normal area. Specifically, different locations (east, south, west and north) of the steel ladle usually have different settings for the initial thickness, safety thickness and target number of times of use of the refractory brick lining. When these three parameters are set in advance, the positive-abnormal boundary line 200 shown in FIG. 2 can be obtained. In the embodiment of the present invention, although the positive-abnormal boundary line 200 shown in FIG. 2 is an oblique straight line, in fact, the positive-abnormal boundary line 200 can also be a curve or a plurality of turning line segments, etc., including at least two set points.

In the embodiment of the present invention, step 112 further includes a step of performing positive-abnormal labeling on the historical production data, marking the historical production data that is distinguished by the positive-abnormal boundary line 200 as 0 in the normal zone, and marking the historical production data that is distinguished by the positive-abnormal boundary line 200 as 1 in the abnormal zone.

Please return to FIG. 1. After step 112, step 113 is then performed to provide a number prediction interval to determine the range to be predicted. The number prediction interval uses the number of times used in the historical production data of the historical steel ladle as the setting value of the prediction range. In the embodiment of the present invention, the number prediction interval can adopt the maximum number of times actually used in the historical steel ladle. For example, among a plurality of historical steel ladles, the maximum actual number of times used is 40, and the maximum value of the number prediction interval can be set to 40. After that, it is applicable to the residual thickness estimation for the online steel ladle with a number of times used less than 40 times, and the steel ladle with a number of times used more than 41 times is not within the estimation range. In another embodiment of the present invention, the interval of number of times prediction can be determined by the operator according to the needs. For example, when the operator decides that the steel ladle to be predicted is a steel ladle with a usage number of 0 to 35 times, the maximum value of the interval of number of times prediction can be set to 35, so the residual thickness estimation model established is applicable to the steel ladle with a usage number of less than 35 times for residual thickness estimation, and the steel ladle with a usage number of more than 36 times is not within the estimation range. It should be understood that the interval of number of times prediction for the steel ladle does not need to start from 0 times. In fact, the interval of number of times can be set by oneself, and the residual thickness estimation is performed for the steel ladle with a usage number falling within the interval of number of times prediction.

Please go back to Figure 1. After step 113, step 114 is then performed to expand the historical production data to have the same number of times of use according to the number of times prediction interval, so that the historical production data becomes a complete historical production training set. Among them, the same number of times of use is the maximum value of the number of times prediction interval, and each time the number of times of use of the steel barrel in the historical production data is less than the same number of times of use is filled by padding. Specifically, since not all historical steel ladle ladles were taken offline for maintenance at the same number of times of use, the same refractory brick lining residual thickness corresponds to multiple historical production data with different numbers of times of use. In the embodiment of the present invention, a method of filling data is adopted to stretch all historical production data to the same number of times of use (using the maximum value of the number of times prediction interval set in step 113). For example, the maximum value of the number of times prediction interval is set to 40 times (number of times of use), and when the number of times a steel ladle is taken offline for maintenance is 35 times, the remaining 5 times are filled with 0 data to make the historical production data a complete historical production training set.

Please return to FIG. 1. After performing step 114, step 115 is then performed to establish a residual thickness estimation model based on the complete historical production training set generated in step 114. In the embodiment of the present invention, the residual thickness estimation model adopts an artificial intelligence (AI) regression prediction model, and four common regression models are established using the complete historical production training set, including random forest algorithm (Random Forest), extreme gradient boosting algorithm (XGBoost), lightweight gradient boosting machine (LightGBM) and CatBoost, and the four regression prediction models are used to estimate the residual thickness of the online steel ladle respectively, as a reference for the prediction accuracy. It should be understood that any one of the above regression models can be used as an embodiment of the method 100 for estimating the residual thickness of the refractory brick lining of the steel ladle of the present invention.

Please continue to refer to FIG. 1. After completing the model building stage 110, the online estimation stage 120 is then performed. In the steps of the online estimation stage 120, firstly, step 121 is performed to input the actual production parameters and maintenance data of the steel ladle to be predicted into the residual thickness estimation model, wherein the actual production parameters include steel type information and the working time of the steel ladle to be predicted at each production station, for example, the working time of the steel ladle to be predicted at production station A is 1 hour, the working time at production station B is 2 hours, and the working time at production station C is 3 hours and other related process data.

After step 121, step 122 is then performed to obtain the estimated residual thickness value of the steel ladle to be predicted according to the residual thickness estimation model and the input actual production parameters and maintenance data. In the embodiment of the present invention, the residual thickness of the main slag line bricks of the steel ladle to be predicted is estimated in four directions, and finally the estimated residual thickness values of the four directions of the east, west, south and north of the steel ladle to be predicted are obtained, so that the maintenance personnel can further evaluate and maintain the steel ladle to be predicted according to the estimated residual thickness value of the steel ladle to be predicted.

In the online estimation stage 120, step 123 is further included to obtain the erosion curvature of the ladle to be predicted according to the residual thickness estimation model. Since the residual thickness erosion curvature of the ladle in the normal area and the abnormal area is greatly different, in order to make the overall prediction more in line with the production application, in the embodiment of the present invention, the erosion curvature of the ladle in the normal area and the ladle in the abnormal area are predicted respectively. As shown in FIG3, step 123 further includes performing steps 1231 and 1232 to respectively predict the erosion curvature of the ladle to be predicted that is determined to be normal or abnormal.

Please continue to refer to FIG. 3 , firstly, step 1231 is performed to determine whether the steel ladle to be predicted is abnormal. If the result is determined to be normal, step 1232 is entered to determine it as a normal zone, and the first erosion curvature prediction value of the steel ladle to be predicted is obtained; if the result is determined to be abnormal, step 1233 is entered to determine it as an abnormal zone, and the second erosion curvature value of the steel ladle to be predicted is obtained.

In an embodiment of the present invention, a residual thickness estimation model for a refractory brick lining of a steel ladle is provided, so as to obtain a residual thickness estimation value of the steel ladle according to the input actual production parameters and maintenance data of the steel ladle to be predicted, and also obtain a determination result of a positive abnormal region of the steel ladle to be predicted, and predict the erosion curvature of the steel ladle to be predicted according to the determination result of the positive abnormal region. With the assistance of the residual thickness prediction model, the maintenance personnel can be warned in advance of the current state of the steel ladle to be predicted, so as to facilitate the maintenance personnel to make further evaluation and maintenance, and also to prevent accidents.

Although the present invention has been disclosed as above with several embodiments, they are not intended to limit the present invention. Anyone with ordinary knowledge in the technical field to which the present invention belongs can make various changes and modifications without departing from the spirit and scope of the present invention. Therefore, the protection scope of the present invention shall be defined by the scope of the attached patent application.

100: Method for estimating the residual thickness of refractory brick lining in steel ladle

110: Model building phase

120: Online estimation stage

111, 112, 113, 114, 115: Steps

121, 122, 123: Steps

200: Positive abnormality boundary

1231, 1232, 1233: Steps

In order to make the above and other purposes, features, advantages and embodiments of the present invention more clearly understandable, the detailed description of the attached figures is as follows: Figure 1 is a schematic diagram of the process of estimating the residual thickness of the refractory brick lining of the steel ladle according to the embodiment of the present invention; Figure 2 is a schematic diagram of the boundary line of the positive abnormal region according to the embodiment of the present invention; Figure 3 is a schematic diagram of the process of predicting the erosion curvature of the positive abnormal region according to the embodiment of the present invention.

without

100: Method for estimating the residual thickness of refractory brick lining in steel ladle

110: Model building stage

120: Online estimation stage

111,112,113,114,115: Steps

121,122,123: Steps

Claims (10)

A method for estimating residual thickness of refractory brick lining of a steel ladle, the method comprising: a model building stage, comprising: providing a plurality of historical production data of a plurality of historical steel ladle, wherein each of the historical production data comprises at least one steel ladle usage count and at least one corresponding steel ladle residual thickness value, and wherein each of the historical steel ladle corresponds to a piece of historical production data; determining a positive-abnormal boundary line according to an original thickness value, a preset safety thickness value, an initial usage count, and a target usage count, so as to divide the historical production data of the historical steel ladle into a normal area and an abnormal area; providing a count prediction interval, so as to determine the desired prediction range; according to the number prediction interval, the historical production data are stretched to have the same number of times of use, so that the historical production data become a complete historical production training set, wherein the same number of times of use is the maximum value of the number prediction interval, and each time the number of times of use of at least one steel ladle in the historical production data is less than the same number of times of use is filled by a filling value; and a residual thickness estimation model is established using the complete historical production training set; and an online estimation stage, including: substituting an actual production data of a steel ladle to be predicted into the residual thickness estimation model; and obtaining a residual thickness estimation value of the steel ladle to be predicted. A method for estimating the residual thickness of refractory brick lining of a steel ladle as described in claim 1, wherein the historical production data also includes a production parameter of the historical steel ladle. The method for estimating the residual thickness of refractory brick lining in steel ladle as described in claim 1, wherein the model building stage further includes a positive abnormality labeling step for the historical production data, wherein the historical production data located in the normal area are marked as 0, and the historical production data located in the abnormal area are marked as 1. A method for estimating the residual thickness of refractory brick lining of a steel ladle as described in claim 1, wherein the residual thickness estimation model is one of the common regression models of Random Forest, Extreme Gradient Boosting (XGBoost), Lightweight Gradient Boosting Machine (LightGBM) and CatBoost. A method for estimating the residual thickness of refractory brick lining in a steel ladle as described in claim 1, wherein the number of prediction intervals is determined based on the needs of an operator. A method for estimating the residual thickness of refractory brick lining of a steel ladle as described in claim 1, wherein the maximum value of the number of prediction intervals can be determined based on the maximum number of times the steel ladle has been used in the historical steel ladle. A method for estimating the residual thickness of refractory brick lining of a steel ladle as described in claim 1, wherein the online estimation stage further includes predicting a erosion curvature of the steel ladle to be predicted. The method for estimating the residual thickness of the refractory brick lining of the steel ladle as described in claim 7, wherein the step of predicting the erosion curvature includes: determining whether the steel ladle to be predicted is located in the normal zone or the abnormal zone; when the steel ladle to be predicted is located in the normal zone, performing a first erosion curvature prediction on the steel ladle to be predicted; and when the steel ladle to be predicted is located in the abnormal zone, performing a second erosion curvature prediction on the steel ladle to be predicted. A method for estimating the residual thickness of refractory brick lining in a steel ladle as described in claim 1, wherein the minimum value of the prediction interval is 0. A method for estimating the residual thickness of refractory brick lining in a steel ladle as described in claim 1, wherein the minimum value of the prediction interval is a positive integer greater than 0.

Images