TW201445131A - Correction method for water-cooled thermal conductivity - Google Patents

Abstract

A correction method for water-cooled thermal conductivity is adapted for a high temperature surface and an image-capturing device. It comprises an image taking step, a characteristic set taking step, a corrected value operation step, and athermal conductivity operation step. The image taking step utilizes the image-capturing device to obtain an image of the high temperature surface. The characteristic set taking step takes a characteristic set corresponding to quality of the high temperature surface from the image. The corrected value operation step obtains a corrected value in accordance with the characteristic set. The thermal conductivity operation step utilizes the corrected value to correct a thermal conductivity theoretical value to a corrected water-cooled thermal conductivity.

Description

Water cooling heat transfer coefficient correction method

The invention relates to a correction method, in particular to a water cooling heat transfer coefficient correction method.

In order to achieve the requirements of the machine and reduce the addition of alloys, more and more steel mills have rapidly cooled the products to a set temperature range through various types of cooling equipment after thermal processing. The accuracy of the temperature drop process control has become one of the main conditions for product quality stability. One of the methods to accurately control the temperature drop is to try to accurately calculate the heat transfer coefficient of the cooling water.

The heat transfer coefficient of the cooling water on the high temperature surface has been studied in many related ways, such as water flow velocity, water volume, surface temperature and size, as parameters in the calculation. In the study by Shimoi et al., the relative heat transfer coefficient can be found by using the metering method and using the water density and the surface temperature of the steel sheet. However, such a method does not include information on the surface quality. Therefore, in the production process, when the surface quality is mutated, the control system may have an error in estimating the water cooling capacity and affect the stability of the process.

Accordingly, it is an object of the present invention to provide a water cooling heat transfer coefficient correction method.

Therefore, the water cooling heat transfer coefficient correction method of the present invention is applicable to one The high temperature surface, and an image capturing device, comprise an image obtaining step, a feature set taking step, a correction value calculating step, and a heat transfer coefficient calculating step.

The image acquisition step is performed by the image capturing device to obtain an image of the high temperature surface.

The feature set removal step takes out a set of features in the image that are of a quality that should be of a high temperature surface.

The correction value operation step obtains a correction value according to the feature set.

The heat transfer coefficient calculation step uses the correction value to correct the theoretical value of a water-cooled heat transfer coefficient to a corrected water-cooled heat transfer coefficient.

1‧‧‧Image capture device

2‧‧‧Backend host

3‧‧‧Steel

4‧‧‧ conveyor belt

5‧‧‧ classifier

6‧‧‧Water-cooled area

S1‧‧‧Image acquisition steps

S2‧‧‧ Feature Group Removal Steps

S3‧‧‧correction value calculation steps

S4‧‧‧ heat transfer coefficient calculation steps

Other features and effects of the present invention will be apparent from the following description of the drawings, wherein: FIG. 1 is a schematic diagram illustrating an image capturing device, a steel material, and a back-end host in the present invention. Figure 2 is a flow chart illustrating the flow of steps of the water cooling heat transfer coefficient correction method of the present invention; Fig. 3 is a schematic view showing the number of objects, the average area of an object, and the calculation method of the average contour length of an object; 4 is a schematic diagram illustrating the theoretical value of a water-cooled heat transfer coefficient; and FIG. 5 is a schematic diagram illustrating a type of neural network.

Before the present invention is described in detail, it should be noted in the following In the description, similar elements are denoted by the same reference numerals.

Referring to FIG. 1 and FIG. 2, a first preferred embodiment of the water-cooling heat transfer coefficient correction method of the present invention is applied to a high-temperature surface of a steel material 3 to be fed into the water-cooling zone 6 by a conveyor belt 4, and a water-cooling zone 6 The front image capturing device 1 and a back end host 2 connected to the image capturing device 1. The water cooling heat transfer coefficient correction method includes an image acquisition step S1, a feature set extraction step S2, a correction value operation step S3, and a heat transfer coefficient operation step S4.

In the image acquisition step S1, the image capturing device 1 acquires an image of the high temperature surface of the steel material 3 and transmits it to the back end host 2.

In the feature group taking-out step S2, the processor (not shown) of the back-end host 2 takes out a feature set of a pair of high-temperature surfaces in the image. In this case, the plurality of pixels in the image are respectively corresponding to a grayscale value. If the grayscale value is greater than a threshold, the grayscale value is set to 0, otherwise it is set to 1, and thus can be different. The boundary of the grayscale value pixels defines a plurality of closed objects. In this way, by the processor of the backend host 2, the image is converted into a binarized image including a plurality of closed objects. Then, according to the closed objects, a feature set including an object number, an object average area, and an object average contour length is calculated.

The number of objects, the average area of objects, and the calculation method of the average contour length of objects

Referring to FIG. 2 and FIG. 3, in the feature group, the number of the object is The number of such closed objects in the binarized image. The average area of the object is the average of the areas of the enclosed objects in the binarized image. The average contour length of the object is the average of the perimeters of the closed objects in the binarized image.

Taking a 6 by 6 image as an example, the calculation method is as follows. Start from the upper left corner of the image, from top to bottom, then search from left to right. When the grayscale value is found to be 1, mark the pixel as a, then record this as a search starting point, and then continue to Check the adjacent pixels in all directions. If the grayscale value is 1, the same symbol is marked until the pixel grayscale value is 0 or the symbol has been marked, the search is stopped, and the next pixel of the search starting point just recorded is started. Search, search for a pixel with a grayscale value of 1 as a different symbol, such as b, repeat the process until the search to the lower right corner of the image. Then, check how many different symbols are used in total to get the number of objects. In this example, there are four symbols a, b, c, and d, so the number of objects is 4.

For the calculation of the average area of the object and the average contour length of the object, the closed object indicated by a above is taken as an example. It only needs to start from the upper left corner of the image, first from top to bottom, then from left to right to the lower right corner of the image. The number of pixels of a, and the number of pixels respectively labeled as other symbols, and then totaled and divided by the number of objects, is the average area of the object. In this example, the objects a to d each occupy 3, 5, 3, and 1 pixel, respectively. Therefore, in the preferred embodiment, the average area is (3+5+3+1)/4=3. .

Then, the perimeters of the closed objects corresponding to all the marked symbols are calculated separately, and then totaled and divided by the number of objects, which is the average contour length of the object. The first method of obtaining the contour length is based on the coordinate point of the object image. As shown in Fig. 3, it is assumed that there are N _k points on the contour of the object, wherein the i-th point coordinates are (x _i , y _i ). Then, the image object outline length can be based on C _k Calculate, where the % symbol table is divided and the remainder is taken, for example, (N _k +1)% N _k is 1, and the object b is taken as an example. There are 5 pixels in total, and the coordinate positions are (3, 4), (4, 5 respectively). ), (4,6), (3,5), and (2,5), where the distance between (3,4) and (4,5) is 1.4 (taken to the first decimal place), between each pixel The distance is shown in the table below.

After the contour calculation of all the objects is completed, the arithmetic mean value is taken as the average contour length according to the number of objects. In this example, the contour lengths of the four objects are 3.4, 6.2, 3.4, and 0, respectively, and the average contour length of the object is 3.25.

Next, in the correction value calculation step S3, a correction value is obtained based on the feature set. At this time, the correction value is obtained from the feature group using a regression analysis method. In the preferred embodiment, a linear regression method is used, the form of the regression equation is S = w ₁ n + w ₂ a + w ₃ c , where n is the number of known objects and a is the average area of the known object. c is the average contour length of the known object, and S is a known correction value.

In order to obtain a better regression analysis result, all known data are first normalized, and the normalized data is also the training material of the regression analysis method, as shown in the following table:

Next, using the least squares method, it is found that w1, w2, and w3 are 0.41, 0.006, and 0.117, respectively. Then, the number of objects in the feature group, the average area of the object, and the average contour length of the object are substituted into S = w ₁ n + w ₂ a + w ₃ c respectively, and an S value can be obtained. The S value at this time is corresponding. The correction value of the feature group.

It is worth mentioning that the regression analysis method is not limited to a linear regression, but also a quadratic linear regression, or a higher linear regression. When calculated by quadratic linear regression, the regression equation is of the form S = w ₁ n + w ₂ a + w ₃ c + w ₄ na + w ₅ ac + w ₆ nc + w ₇ nn + w ₈ aa + w ₉ cc , After substituting n, a, c, and S, w1, w2, w3, w4, w5, w6, w7, w8, and w9 are respectively 0.056, 0.059, -0.009, 0.002, 0.014, 0.005, 0.008, 0.007, 0.003. Similarly, after substituting the number of objects in the feature group, the average area of the object, and the average contour length of the object into the regression equation, the S value can be obtained, which is the correction value. The correction value at this time will be more relevant to the training data.

Then, referring to FIG. 2 and FIG. 4, in the heat transfer coefficient calculation step S4, the theoretical value of the water-cooling heat transfer coefficient is corrected to a corrected water-cooled heat transfer coefficient by using the correction value. At this time, according to the method of Shimoi et al. (see Figure 4), the theoretical value of the water-cooling heat transfer coefficient is obtained, and then the correction value obtained by the regression analysis method is multiplied by the theoretical value of the water-cooling heat transfer coefficient to obtain the corrected water-cooling heat. Pass coefficient. For example, when the surface temperature is 350 ° C, the water density is 0.8 m ³ /m ² /min. After looking up the table, the theoretical value of the water-cooling heat transfer coefficient is 7000 W/m ² /K. Assuming that the correction value obtained in the correction value calculation step S3 is 0.95, the corrected water-cooling heat transfer coefficient is 6650 W/m ² /K. The above is the first preferred embodiment of the present invention.

Referring to FIG. 2 and FIG. 5, a second preferred embodiment of the water-cooling heat transfer coefficient correction method of the present invention is different from the first preferred embodiment in that in the correction value operation step S3, a classification method is used. The feature group gets the correction value. At this time, the processor of the backend host 2 (see FIG. 1) carries a trained classifier 5, which is a computer program product, which can be implemented after loading the processor, and The information used for training is the normalized material in the first embodiment of the present invention. In the preferred embodiment, the classifier 5 is implemented as a neural network.

The classifier 5 has an input layer, a hidden layer, and an output layer, each layer including a plurality of nodes. In the preferred embodiment, since there are three feature values in the feature group including the number of objects, the average area of the object, and the average contour length of the object, there are three nodes in the input layer. The output layer, from 0.90 to 2.00, corresponds to a node every 0.01 interval.

When the classification method is implemented, the number of objects obtained in step S2, the average area of the object, and the average contour length of the object are extracted from the feature group, and correspondingly input into the three nodes of the input layer, passing through the hidden layer, and correspondingly in the output layer. The node that sorts the result will output 1. Assuming that the node corresponding to 0.95 outputs 1, the correction value is 0.95.

It is worth mentioning that the classifier 5 is not limited to various types of neural networks, and may be a Support Vector Machine (SVM). The design of the nodes in the output layer is also not limited to 0.01, which may be increased or decreased depending on the actual situation.

In summary, the present invention extracts the feature set from the captured image, and cooperates with the regression analysis method or the classification method to correct the theoretical value of the water-cooling heat transfer coefficient obtained by the look-up table, so that the object of the present invention can be achieved.

The above is only the preferred embodiment of the present invention, and the scope of the invention is not limited thereto, that is, the simple equivalent changes and modifications made by the scope of the invention and the description of the invention are All remain within the scope of the invention patent.

S1‧‧‧Image acquisition steps

S2‧‧‧ Feature Group Removal Steps

S3‧‧‧correction value calculation steps

S4‧‧‧ heat transfer coefficient calculation steps

Claims (7)

A method for correcting water-cooling heat transfer coefficient, which is suitable for a high temperature surface, and an image capturing device, comprising the following steps: an image obtaining step, using the image capturing device to obtain an image of the high temperature surface; and a feature group taking step, Extracting a set of features of the image that should be of a high temperature surface; a correction value operation step, obtaining a correction value according to the feature set; and a heat transfer coefficient operation step, using the correction value to calculate a water-cooled heat transfer coefficient theoretical value Corrected to a corrected water-cooling heat transfer coefficient. The water cooling heat transfer coefficient correction method according to claim 1, wherein the feature group extracting step comprises the following substeps: first converting the image into a binarized image including a plurality of closed objects; and further, according to the closed objects The feature set is calculated. The water-cooling heat transfer coefficient correction method according to claim 2, wherein in converting the image into the binarized image, the plurality of pixels in the image are respectively corresponding to a gray scale value, if When the grayscale value is greater than a threshold, the grayscale value is set to 0, otherwise it is set to 1, so that the closed objects can be defined by the boundaries having pixels of different grayscale values. The water cooling heat transfer coefficient correction method according to claim 3, wherein the feature group comprises an object number, an object average area, and an object average contour length. The water-cooling heat transfer coefficient correction method according to claim 1, wherein in the correction value operation step, the correction value is obtained from the feature group using a regression analysis method. The water-cooling heat transfer coefficient correction method according to claim 1, wherein in the correction value operation step, the correction value is obtained from the feature group using a classification method. The water-cooling heat transfer coefficient correction method according to claim 1, wherein in the heat transfer coefficient operation step, the theoretical value of the water-cooled heat transfer coefficient is multiplied by the correction value to obtain the corrected water-cooled heat transfer coefficient.