TWI737562B - Container wall thickness estimation modeling method, system, computer program product, and computer-readable recording medium - Google Patents

Abstract

A container wall thickness estimation modeling method, a system, a computer program product, and a computer-readable recording medium are disclosed. The method includes steps of: reading several outer-wall temperatures and corresponding thicknesses acquired in a discrete-time manner from a wall of a container; finding several wave-peaks formed by temperature rapidly raised in at least one slope greater than a temperature-raised threshold to several peak-values and then dropped in a relationship curve formed by the several outer-wall temperatures and time, and generating several slope-accumulating values by accumulating the slope corresponding to each of the several wave-peaks; and generating a training set by the thicknesses and the peak-values, the slopes, and the slope-accumulating values derived from the several outer-wall temperatures, training a learning model based on the training set, until a result of training meets a stopping condition, to generate a thickness estimation model based on features of the trained learning model.

Description

Modeling method, system, computer program product and computer readable recording medium for estimating container wall thickness

The present invention relates to a vessel thickness estimation technology, in particular to a modeling method, system, computer program product and computer program for estimating vessel wall thickness for establishing a model based on machine learning technology for estimating the residual thickness of blast furnace copper stave Read the recording medium.

In some application places (such as factories) that require high-temperature containers, it is difficult to immediately evaluate changes in container characteristics due to extremely high temperatures during the use of the containers.

Take the steelmaking blast furnace as an example. The conventional copper stave residual thickness monitoring method is to measure the residual thickness of the copper wall during regular maintenance, but this method is not to measure the residual thickness of the copper wall during high-temperature line operation; in addition, the regular maintenance cycle of the blast furnace is long, and There may not be enough time to measure the copper wall thickness of all parts as reference data.

Although some improvement techniques have been proposed in the past, such as using the inner and outer temperature of the copper wall to estimate the thickness of the copper wall, it is difficult to obtain the high temperature inside the copper wall, and the estimation process is too complicated and still needs to be improved.

In view of this, it is necessary to provide a technical solution different from the past to solve the problems existing in the conventional technology.

An object of the present invention is to provide a modeling method for estimating the wall thickness of a container, which builds a model based on the external temperature of the container and the corresponding thickness, so as to be used as a basis for real-time estimation of the wall thickness of the container.

The second purpose of the present invention is to provide a modeling system for estimating the wall thickness of a container, which builds a model based on the external temperature of the container and the corresponding thickness, so as to be used as a basis for real-time estimation of the wall thickness of the container.

Another object of the present invention is to provide a computer program product that builds a model based on the external temperature of the container and the corresponding thickness, so as to be used as a basis for real-time estimation of the container wall thickness.

Another object of the present invention is to provide a computer-readable recording medium with a program stored therein, and to establish a model based on the external temperature of the container and the corresponding thickness, so as to be used as a basis for real-time estimation of the container wall thickness.

To achieve the above objective, one aspect of the present invention provides a modeling method for estimating the wall thickness of a container, which is configured to be executed by a processor coupled to a memory to execute instructions stored in the memory. The modeling method It includes the steps of: reading several outer wall temperatures and corresponding several thicknesses obtained from a container wall in discrete time; searching for at least one temperature greater than a temperature rise threshold in a relationship curve formed by the several outer wall temperatures and time A slope rises sharply to a few peaks and then drops to form a number of peaks, accumulate according to the corresponding slopes of the peaks to generate a number of slope cumulative values; and according to the number of thicknesses and the temperature of the outer wall The derived peak, the slope, and the cumulative value of the slope generate a training set, and a learning model is trained based on the training set until the training result of the learning model meets a stopping condition, and an estimate is generated based on the characteristics of the learning model after being trained. Thick model.

In an embodiment of the present invention, the modeling method for estimating the thickness of the container wall further includes the step of generating a corresponding thickness according to the estimated thickness model based on a test temperature.

In an embodiment of the present invention, the peaks corresponding to the several wave crests are sorted and simplified into one of several peak levels respectively, and the slopes corresponding to the several wave crests are sorted and simplified into one of several slope levels respectively, and The slope accumulation values are sorted and simplified into one of the accumulation levels, based on the thickness and the peak levels, the slope levels, and the accumulation levels derived from the outer wall temperatures Generate the training set.

In an embodiment of the present invention, any one of the several peak levels is a 25% peak level characterization, a 50% peak level characterization, a 75% peak level characterization, or a highest peak level characterization; the several slopes Any one of the grades is a 25% slope grade characterization, a 50% slope grade characterization, a 75% slope grade characterization, or a highest slope grade characterization; and any one of the several cumulative grades is a 25% cumulative grade characterization , A 50% cumulative grade characterization, a 75% cumulative grade characterization, or a highest cumulative grade characterization.

In an embodiment of the present invention, the temperature rise threshold is 2 degrees Celsius per second.

In an embodiment of the present invention, the learning model is a regression model.

In an embodiment of the present invention, the container wall is a steel-making blast furnace side wall or a boiler side wall.

To achieve the above objective, another aspect of the present invention provides a modeling system for estimating the wall thickness of a container, which includes a processor and a memory, the processor is coupled to the memory, and the memory stores at least one instruction, The processor executes the instruction to execute the modeling method for estimating the wall thickness of the container as described above.

To achieve the above objective, another aspect of the present invention provides a computer program product. After the computer program is loaded and executed, the computer can execute the above-mentioned modeling method for estimating the wall thickness of the container.

In order to achieve the above objective, another aspect of the present invention provides a computer-readable recording medium. The computer can read a program stored in the recording medium. After the computer loads and executes the program, the computer can complete the above-mentioned A modeling method for estimating the wall thickness of a container.

The modeling method, system, computer program product, and computer readable recording medium for estimating container wall thickness of the present invention establish a model based on the external temperature of the container and the corresponding thickness, so as to be used as a basis for real-time estimation of container wall thickness, which can be effective Simplify the parameters required for estimating the wall thickness of the vessel, and improve the complicated calculation process of the conventional technology, which is conducive to real-time monitoring of the vessel wall thickness (such as the residual thickness of the blast furnace copper stave), and is convenient for relevant personnel to use, repair, design or use the vessel. Research and other references.

In order to make the above and other objectives, features, and advantages of the present invention more obvious and understandable, the preferred embodiments of the present invention will be described in detail below in conjunction with the accompanying drawings. Furthermore, the directional terms mentioned in the present invention, such as up, down, top, bottom, front, back, left, right, inside, outside, side, surrounding, center, horizontal, horizontal, vertical, vertical, axial, The radial direction, the uppermost layer or the lowermost layer, etc., are only the direction of reference to the attached drawings. Therefore, the directional terms used are used to describe and understand the present invention, rather than to limit the present invention.

One aspect of the present invention provides a modeling method for estimating the wall thickness of a container. The modeling method for estimating the wall thickness of a container can be performed, for example, using a modeling system for estimating the wall thickness of the container by using a temperature to build a model based on Temperature estimation of vessel wall thickness, such as the residual thickness of the copper cooling wall of a steelmaking blast furnace, but not limited to this, it can also be applied to estimate the wall thickness of other vessels based on temperature, such as the thickness of a boiler used for heating. The following examples illustrate related implementation aspects, but are not limited to this.

For example, referring to FIG. 1, the modeling system for estimating the wall thickness of the container may include a database 1, a training unit 2 and a testing unit 3, for example. The training unit 2 is coupled to the database 1 and the test unit 3. The coupling method can be a wired connection (such as an electrical connection or a network) that can be used to transmit data carriers (such as electricity, light, magnetism, or a combination thereof). Connection, etc.) or wireless coupling (such as photoelectric coupling or electromagnetic coupling, etc.), so that the two objects that are coupled can transfer data to each other to facilitate the content disclosed here, and they have common knowledge in the technical field. It is understandable and will not be repeated here.

For example, as shown in Figure 1, the database 1 can be a storage carrier with a data storage function, for example: it can be selected from memory, disk storage carriers (such as optical disks or hard disks, etc.) and data servers The group, but not limited to this, can also be configured as a cloud storage medium (such as Google Drive, etc.) to store a data set, for example: the data in the data set can include reading a container wall (For example, steelmaking blast furnace side wall or boiler side wall, etc.)Several outer wall temperatures continuously obtained in a discrete time manner (such as temperature sensing elements to measure the surface temperature of copper staves, etc.) and corresponding thicknesses (such as ultrasonic thickness measurement data) Etc.) as the basis for model training.

Please refer to Figure 1 again. The training unit 2 and the testing unit 3 can be used for model training and testing, respectively. For example, the training unit 2 and the testing unit 3 can be configured as a software-hardware cooperative operation device, such as a server, a processor, or a special application integrated circuit. The training unit 2 can train a learning model based on the data in the training set, for example, a regression model that can predict a specific value (such as the residual thickness of the blast furnace copper stave, etc.) to generate a thickness estimation model to estimate the container based on temperature Wall thickness; the test unit 3 can test the trained thickness estimation model according to the data in the test set.

For example, the method for modeling the wall thickness of the temperature estimation container can be configured to be executed by a processor coupled to a memory, and the processor executes instructions stored in the memory. For example, the memory can be Configured to have the function of the aforementioned database, the processor can be configured to have the functions of the aforementioned training unit and test unit for executing the modeling method for estimating the wall thickness of the container. The following examples illustrate the implementation of the aforementioned embodiment of the present invention State, but not limited by this.

For example, as shown in Figure 2, the modeling method for estimating the wall thickness of a vessel may include steps: an acquisition step S1, an extraction step S2, and a training step S3. The following only uses the copper stave of a steelmaking blast furnace The residual thickness is described as an example, but it is not limited to this, and it can also be applied to estimate the wall thickness of other vessels based on temperature, such as the thickness of a boiler for heating.

The obtaining step S1 can read several outer wall temperatures and corresponding thicknesses obtained from a container wall in a discrete time manner; the extracting step S2 can be in a relationship curve formed by the several outer wall temperatures and time Looking for a number of peaks formed by the temperature rising sharply to a number of peaks and then falling with at least a slope greater than a temperature rise threshold, and accumulating operations are performed according to the corresponding slopes of the peaks to generate a number of cumulative slope values; the training step S3, A training set can be generated according to the thickness and the peak value derived from the outer wall temperature, the slope, and the slope cumulative value, and a learning model is trained according to the training set until the training result of the learning model satisfies a stop Condition, based on the characteristics of the learning model after being trained, an estimation model is generated.

For example, a part of the data set can be used as the basis for generating the training set. For example, the processor can obtain several discrete-time data from the vessel wall (such as the side wall of a steelmaking blast furnace or a boiler side wall). The temperature of the outer wall and its corresponding thicknesses, such as multiple sets of temperatures and thicknesses obtained by multiple measurements of the outer wall, are used as the basis for subsequent training models.

(1)

(2) 其中，T _h為該高爐容物H的溫度；T _w為該冷卻水層W的溫度；q為單位熱傳量；d為該銅冷卻壁B的厚度；k為該銅冷卻壁B的熱傳導係數；T _b為該銅冷卻壁B的溫度，h為該銅冷卻壁B的冷端熱對流係數。從而，藉由如上公式(1)及(2)可解出該銅冷卻壁B的厚度d。 In an example, the container wall is taken as the side wall of a steelmaking blast furnace. The first thing that can be explored is the effect of heat conduction. As shown in Figure 3, a blast furnace side wall can be simplified to include a blast furnace content H, a copper stave B, and a cooling water layer W. The blast furnace content is, for example, molten iron with a high temperature of 1,000 degrees Celsius. The copper stave B is located between the blast furnace content H and the cooling water layer W. Assuming that the relevant heat conduction parameters such as the temperature of the blast furnace content H is _Th , the temperature and thickness of the copper _{stave B are T b} and d, respectively, and the temperature of the cooling water layer W is T _w , it can be seen that the heat conduction relationship is as follows (1 ) And (2) show:

(1)

(2) Among them, T _h is the temperature of the blast furnace content H; T _{w is} the temperature of the cooling water layer W; q is the unit heat transfer; d is the thickness of the copper stave B; k is the copper stave The thermal conductivity of B; T _{b is} the temperature of the copper stave B, and h is the heat convection coefficient of the cold end of the copper stave B. Therefore, the thickness d of the copper stave B can be solved by the above formulas (1) and (2).

It should be understood that in the actual blast furnace application process, it is extremely difficult to obtain all the parameters required by the above formula completely. For example: the temperature T _{w of} the cooling water layer W can only be measured from the water inlet and the water outlet, but the temperature cannot be measured for each of the many copper staves in the blast furnace; in addition, the copper stave B’s cold-end thermal convection coefficient h is complicated to calculate, and it needs to be verified by experiments to obtain the correct data. In addition, when the blast furnace is used at a temperature of up to 1,000 degrees, the inside of the blast furnace is like a black box and cannot be directly _{Measure the temperature Th of the} content H of the blast furnace. In fact, the only data that _{can be easily obtained are the temperature T b of} the copper stave B and the ultrasonic thickness measurement data that can be measured regularly, such as the thickness d of the copper stave B. Therefore, a feasible way to solve the related heat conduction is to use the temperature T _{b of} the copper stave B to obtain the thickness d of the copper stave B.

For example, from the above equations (1) and (2), it can be seen that the thermal conductivity k of the copper stave B and the cold end heat convection coefficient h of the copper stave B can be regarded as constants, and the remaining unknowns still have the cooling water layer W The temperature T _w of the blast furnace and the temperature T _{h of the contents of the blast furnace H.} Wherein, since the temperature of the cooling water layer W is usually set to maintain between 30 and 40 degrees Celsius (°C), compared to the temperature of the blast furnace content H, which is usually as high as thousands of degrees, the temperature of the cooling water layer W T _w can be assumed to be regarded as a constant; in addition, since the molten iron in the blast furnace is the main material reflecting the internal temperature of the blast furnace, the temperature of the molten iron in the blast furnace can be compared to the temperature T _{h of the} content H of the blast furnace.

As shown in Figure 4, it is a schematic diagram of the temperature of molten iron in a blast furnace (for example, the unit is °C) during an exemplary time period (for example, 0:00:00 on a certain day to 0:00:00 on the next day). It can be seen from the figure that there is a range of temperature changes inside the blast furnace. Therefore, the temperature _{Th of the} contents of the blast furnace H can be regarded as a constant that changes within a range. In order to reduce the impact on model training caused by its changes, you can also choose to simplify it into several interval representative values during the data collection process. For example, taking the quartile method as an example, many data can be sorted and divided into four equal parts. Levels, such as 25% level, 50% level, 75% level and the highest level. In this way, the variation range can be simplified to one of four levels, so as to reduce the impact caused by the numerical value variation (such as computational complexity, etc.). Therefore, the thickness can be simplified as d=(K×T _b ), where K can be regarded as a set of model features (such as several parameters, etc.), and further machine learning and deep learning methods can be used to measure the thickness by ultrasound. The data and the temperature data of the copper stave are used for model training and solution process.

It should be noted that the hidden physical meaning behind the temperature data of the copper stave, for example: molten iron will wear the copper stave, causing the thickness of the copper stave to become thinner, and more importantly, there will be adhesions on the copper stave There is a slag skin, which causes the thermal conductivity to change, causing the temperature of the copper stave with the slag skin to be lower. In order to avoid excessive distortion of the temperature of the copper stave, the temperature change of the copper stave before and after the slag skin falls off can also be observed, which can be used as a basis for training to be more suitable for the characteristic model of the copper stave.

For example, as shown in Figure 5, it is a schematic diagram of the furnace wall temperature during the process of peeling off the slag skin attached to the copper cooling stave. Among them, when the slag skin falls off the copper stave, because the part of the copper stave on the furnace wall suddenly loses the slag skin but still bears the internal high temperature energy, the temperature of the furnace wall will increase rapidly after the slag skin falls off, such as a temperature curve C The temperature rises rapidly from the leftmost time starting point (marked as 0), and reaches a local high point (which is regarded as the actual furnace wall temperature with higher reliability) at about the 0.9th minute. After that, The temperature gradually decreased until the 23.5 minute point, the temperature appeared to ease. During this process, it can be observed that before and after the slag attached to the copper stave falls off, the temperature curve C measured by the copper stave will rise sharply to a peak and then drop at a slope greater than a temperature rise threshold. A crest. In an example of a blast furnace, the temperature rise threshold can be 2 degrees Celsius per second, but it is not limited to this. The temperature rise threshold can be fine-tuned according to the actual situation; in addition, if the slag crust repeatedly adheres and falls off over time , It can be inferred that the temperature curve C will have several peaks formed by steeply increasing to several peaks and then decreasing with at least one slope greater than the temperature rising threshold.

Therefore, in order to train the characteristic model of copper staves more suitable for steelmaking blast furnaces, the peak characteristics of the temperature curve observed during the slag peeling process can also be added, so that the model training results can be closer to the real copper stave characteristics.

For example, since the temperature curve will have a peak formed by a steep rise to the peak and then drop with a slope greater than the temperature rise threshold, in order to simplify the amount of data that constitutes the peak, in the temperature curve waveform with several peaks , You can take the corresponding peak values and slopes of the several peaks of the temperature curve, and generate several cumulative slope values according to the corresponding slopes of the several peaks, which respectively represent the temperature characteristics of the slag peeling process. For example, the peak and the slag peeling are indeed No, the slope is related to the size of the slag peeling range, and the cumulative value of the slope is related to the temperature change trend.

Optionally, in an embodiment, the peaks corresponding to the several wave crests can be sorted and simplified into one of several peak levels respectively, and the slopes corresponding to the several wave crests can be sorted and simplified into one of several slope levels respectively. , The several slope accumulation values are sorted and simplified into one of several accumulation levels respectively, based on the several thicknesses and the several peak levels derived from the several outer wall temperatures, the several slope levels and the several Accumulate the levels to generate the training set. In this way, it can be effectively applied to the training model close to the true temperature characteristics of the copper stave, and the amount of data can be further streamlined to reduce the impact of numerical changes (such as computational complexity, etc.).

For example, in order to further simplify the amount of data, the number of peaks, the slopes, and the cumulative values of the slopes derived from the peaks of the temperature curve can be simplified into representative values of intervals, for example, the quartile method is For example, a lot of data can be sorted and divided into four equal grades, such as 25% grade, 50% grade, 75% grade and the highest grade, but it is not limited to this.

As shown in Fig. 6, it is a schematic diagram of temperature and temperature difference (for example, the unit is °C) during an example period (for example, a certain day). Among them, the upper part represents a temperature curve, and the lower part represents a temperature difference curve. In the upper temperature curve, there are also four peak level representations such as 25% level, 50% level, 75% level and the highest level. limit.

Optionally, in an embodiment, any one of the several peak levels may be represented as a 25% peak level characterization, a 50% peak level characterization, a 75% peak level characterization, or a highest peak level characterization. In this way, the amount of data can be further simplified, and the range of variation can be simplified to one of four levels, so as to further reduce the influence caused by the variation of the temperature of the molten iron (such as computational complexity).

Optionally, in an embodiment, any one of the several slope levels may be a 25% slope level characterization, a 50% slope level characterization, a 75% slope level characterization, or a highest slope level characterization. In this way, the amount of data can be further simplified, and the range of variation can be simplified to one of four levels, so as to further reduce the influence caused by the variation of the temperature of the molten iron (such as computational complexity).

Optionally, in an embodiment, any one of the several accumulation levels may be a 25% accumulation level characterization, a 50% accumulation level characterization, a 75% accumulation level characterization, or a highest accumulation level characterization. In this way, the amount of data can be further simplified, and the range of variation can be simplified to one of four levels, so as to further reduce the influence caused by the variation of the temperature of the molten iron (such as computational complexity).

The above describes the example of data pre-processing before model training, and then discusses the example of using data for model training, which is used as an example to illustrate the available implementation process and not as a necessary limitation of the present invention.

For example, a regression model can be used as the learning model to predict the wall thickness of the container (such as the residual thickness of the copper stave, etc.), for example, support vector machine (support vector machine, SVM) or gradient boost regression ( Gradient Boosting regression, GBR) and other regression models.

For example, taking SVM as an example, it is a supervised learning model. The principle of statistical risk minimization can be used to estimate a hyperplane for classification, so that a hyperplane can perfectly separate two types of different data, such as finding Draw a straight line to maximize the distance between the straight line and the two types of data; but not limited to this. Based on the deformation of SVM, it can be used to find a straight line. The straight line can include all two types of data. The formula for this straight line can be calculated, and Calculate y corresponding to different x. For example: Suppose the data is expressed as (x ₁ , y ₁ ),..., (X _i , y _i ) ∈ R ^d ×R, where x _1...i represents the input process parameters; y _1...i represents the The corresponding regression value, such as the thickness of the zinc layer of the copper stave, R is the set of real numbers, and ^Rd is the set of real vectors. Let f(x) = w · x + b, w ∈ R ^d , b ∈ R, if a stopping condition is f(x)≈y, then the stopping condition is satisfied, we know that f(x) can be accurately predicted from x y, this w is the hyperplane that SVR is looking for. Therefore, it can be seen that the model training process can be based on the collected copper stave temperature and the corresponding copper stave thickness, plus a certain error range to find out w, then the copper stave temperature and the copper stave thickness corresponding to the copper cooling can be trained The wall thickness prediction model, as the thickness estimation model, the calculation process of which can be understood by those with ordinary knowledge in the technical field, and will not be repeated.

For example, taking GBR as an example, it can be classified by many weak classifiers, and finally the classification results of all weak classifiers are aggregated into answers. Among them, the weak classifier is a simple classifier that can first distinguish different categories with different conditions as the output , And finally gather the output of all weak classifiers, and vote for the best output result. For example: assuming the data is (X ₁ , Y ₁ ),..., (X _n , Y _n ) ∈ R ^d ×R, X _1...n represents the input process parameter; Y _1...n represents the corresponding regression value , R is the set of real numbers, and R ^d is the set of real number vectors. _{First calculate Y pred1} with a simple formula (which can be a linear or polynomial formula), the goal is Y _pred1 -Y _n = 0, if it is not 0, continue to bring Y _pred1 -Y _n into the second weak category The device calculates and obtains Y _pred2 until a stopping condition of Y _n -∑Y _pred1...n ≈0 is _{met, and the} sum of Y pred1...n is calculated to complete the training required model. Therefore, it can be known that the copper stave thickness prediction model can be established by adding the prediction model to the formula in the form of parameter input. As the thickness estimation model, the calculation process can be understood by those with ordinary knowledge in the technical field and will not be repeated.

In addition, as shown in Figure 2, the modeling method may further include the following steps: a test step S4, generating a corresponding thickness according to the estimated thickness model based on a test temperature. For example: the test temperature can be the temperature of the copper stave, the corresponding thickness can be the thickness of the copper stave, the corresponding thickness can be generated based on the test temperature according to the estimated thickness model, and the corresponding thickness can be used as a further optimization of the estimated thickness model. It can also be used as a basis for testing the estimated thickness model to output the estimated copper stave thickness. The copper stave thickness can be used as a high-confidence reference value for the residual thickness of the blast furnace copper stave for relevant personnel as a container (Such as blast furnace) application, maintenance, design or research, etc.

For example, as shown in Figures 7 and 8, which are schematic diagrams of two types of copper stave thickness estimation, for example, it can represent the thickness estimation results of two copper staves. Among them, ◆ represents the ultrasonic thickness measurement data actually measured, and the dashed line and the solid line represent the two estimation results. Because the ultrasonic thickness measurement data will have some slight errors, which will cause the actual measurement data to be unstable. Therefore, the solid line represents the estimated result of the thickness update based on each measurement data, and the dashed line represents the initial thickness from 130 millimeters (mm ), regardless of the measurement data to update the thickness estimation result. It can be seen from the figure that the prediction errors between the prediction results and the actual measurement results are all less than 2 mm. Therefore, when the modeling method for estimating the thickness of the vessel wall of the embodiment of the present invention is applied to a steelmaking blast furnace, the corresponding thickness generated by the estimating model can indeed be regarded as a highly reliable copper stave thickness estimation result. .

In another aspect, the present invention also provides a modeling system for estimating the wall thickness of a container, including a processor and a memory, the processor is coupled to the memory, the memory stores at least one instruction, and the processor executes the Command to execute the modeling method for estimating the wall thickness of the container as described above.

For example, the modeling system for estimating container wall thickness can be configured as an electronic device with data processing functions, such as cloud platform machines, servers, desktop computers, laptops, tablets, or smartphones Etc., but not limited to this, it is used to implement the modeling method for estimating the wall thickness of the container as described above. The implementation mode has been described above and will not be repeated.

On the other hand, the present invention also provides a computer program product. After the computer program is loaded and executed, the computer can execute the modeling method for estimating the wall thickness of the container as described above. For example: the computer program product may contain several program instructions, which can be implemented using existing programming languages to execute the modeling method for estimating the wall thickness of the container as described above, for example: C, C++, Programming languages such as Labview, Python, R, or combinations thereof, but not limited to these.

On the other hand, the present invention also provides a computer-readable recording medium, such as an optical disc, a flash drive, or a hard disk. The computer can read a program stored in the recording medium (such as the above-mentioned computer program). After execution, the computer can complete the modeling method for estimating the wall thickness of the container as described above.

The modeling method, system, computer program product, and computer readable recording medium for estimating container wall thickness in the above embodiments of the present invention establish a model based on the external temperature of the container and the corresponding thickness, so as to be used as a basis for real-time estimation of container wall thickness , It can effectively simplify the parameters needed to estimate the wall thickness of the vessel, improve the complicated calculation process of the conventional technology and other problems, which is conducive to the real-time monitoring of the vessel wall thickness (such as the residual thickness of the blast furnace copper cooling stave), and is convenient for relevant personnel to use and maintain the vessel. , Design or research, etc.

Although the present invention has been disclosed in preferred embodiments, it is not intended to limit the present invention. Anyone familiar with the art can make various changes and modifications without departing from the spirit and scope of the present invention. Therefore, the present invention The scope of protection shall be subject to the scope of the attached patent application.

1: Database 2: Training Unit 3: Test Unit B: Copper Stave C: Temperature Curve H: Blast Furnace Content S1: Acquisition Step S2: Extraction Step S3: Training Step S4: Test Step T _b : Copper Stave Temperature T _h : the temperature of the blast furnace contents T _w : the temperature of the cooling water layer W: the cooling water layer d: the thickness of the copper stave

[Figure 1]: A block diagram of a modeling system for estimating the wall thickness of a container according to an embodiment of the present invention. [Figure 2]: A schematic flow diagram of a modeling method for estimating the wall thickness of a container according to an embodiment of the present invention. [Figure 3]: Schematic diagram of the heat conduction parameters of the copper stave of the steel-making blast furnace applied to the embodiment of the present invention. [Figure 4]: Schematic diagram of the temperature of molten iron in a blast furnace where the embodiment of the present invention is applied to a steelmaking blast furnace. [Figure 5]: Schematic diagram of the furnace wall temperature of the slag skin shedding process of the steelmaking blast furnace that the embodiment of the present invention is applied to. [Figure 6]: A schematic diagram of the daily temperature and temperature difference of the copper stave of the steel-making blast furnace applied to the embodiment of the present invention. [Figure 7]: The embodiment of the present invention is applied to a schematic diagram of the thickness estimation of a copper stave in a steelmaking blast furnace. [Figure 8]: The embodiment of the present invention is applied to another schematic diagram of estimating the thickness of copper stave in a steelmaking blast furnace.

S1: Obtaining steps

S2: extraction step

S3: training steps

S4: Test steps

Claims (10)

A modeling method for estimating the wall thickness of a container is configured to be executed by a processor coupled to a memory to execute instructions stored in the memory. The modeling method includes the steps: Read several outer wall temperatures and corresponding thicknesses obtained from a container wall in discrete time; Find the several peaks formed by the temperature rising to several peaks and then falling at at least a slope greater than a temperature rise threshold in a relationship curve formed by the several outer wall temperatures and time, and accumulate according to the corresponding slopes of the several peaks Operate to generate several accumulated slope values; and Generate a training set based on the thickness and the peak value derived from the outer wall temperature, the slope, and the slope cumulative value, and train a learning model based on the training set until the training result of the learning model meets a stopping condition , Based on the trained characteristics of the learning model to generate an estimation model. According to the modeling method for estimating the wall thickness of a container as described in claim 1, the modeling method further includes the step of generating a corresponding thickness according to the estimated thickness model based on a test temperature. The modeling method for estimating the wall thickness of a container as described in claim 1, wherein the corresponding peaks of the several wave crests are sorted and simplified into one of several peak levels, and the corresponding slopes of the several wave crests are sorted and simplified respectively Into one of several slope levels, the several slope accumulation values are sorted and simplified into one of several accumulation levels respectively, according to the several thicknesses and the several peak levels derived from the several outer wall temperatures, the number A slope level and the plurality of accumulation levels generate the training set. The modeling method for estimating the wall thickness of a container as described in claim 3, wherein any one of the several peak levels is a 25% peak level characterization, a 50% peak level characterization, a 75% peak level characterization, or a The highest peak level characterization; any one of the several slope levels is a 25% slope level characterization, a 50% slope level characterization, a 75% slope level characterization, or a highest slope level characterization; and one of the several cumulative levels Either one is a 25% cumulative grade characterization, a 50% cumulative grade characterization, a 75% cumulative grade characterization, or a highest cumulative grade characterization. The modeling method for estimating the wall thickness of a container as described in claim 1, wherein the temperature rise threshold is a temperature rise of 2 degrees Celsius per second. The modeling method for estimating the wall thickness of a container as described in claim 1, wherein the learning model is a regression model. The modeling method for estimating the wall thickness of a container as described in claim 1, wherein the container wall is a side wall of a steel-making blast furnace or a side wall of a boiler. A modeling system for estimating the wall thickness of a container includes a processor and a memory, the processor is coupled to the memory, the memory stores at least one instruction, and the processor executes the instruction to execute the request item 1 The modeling method for estimating the wall thickness of the container described in any one of to 7. A computer program product, when the computer program is loaded and executed, the computer can execute the modeling method for estimating the wall thickness of a container as described in any one of claim items 1 to 7. A computer-readable recording medium. The computer can read a program stored in the recording medium. After the computer loads and executes the program, the computer can complete the estimation of the container wall thickness as described in any one of claims 1 to 7 The modeling method.

Images