Meat loss prediction system and meat loss prediction method

Abstract

This technology improves the accuracy of predicting material thinning by considering multiple different flow conditions. [Solution] The thinning prediction system comprises a fluid simulator 3 that calculates fluid analysis values ​​by fluid simulation using the amount of thinning in the test piece 10 as a result of a water flow corrosion test and the shape of the test piece 10 for a test piece 10 used in a water flow corrosion test; a multiple regression equation generation unit 4 that generates a thinning prediction equation with the amount of thinning as the objective variable and the fluid analysis values ​​as the explanatory variables; and a thinning amount calculation unit 5 that calculates the amount of thinning in the target pipe based on the shape data of the target pipe and the thinning prediction equation, wherein the thinning prediction equation is a combination of multiple provisional prediction equations generated under different flow conditions in the test piece 10.

Description

[Technical Field]

[0001] This invention relates to a pipe wall thinning prediction system and method, which predict pipe wall thinning due to corrosion, particularly flow-accelerated corrosion. [Background technology]

[0002] Flow-accelerated corrosion (FAC) is a type of corrosion in which the rate at which the coating on pipes dissolves into water is accelerated by the water flow, resulting in increased pipe thinning. As a method for predicting the thinning rate due to FAC, a method has been proposed in which a thinning rate model is created according to the fluid velocity, temperature, pressure, pH, etc., and this model is used to predict the thinning rate for the actual pipes.

[0003] For example, Patent Document 1 discloses a plant diagnostic system that calculates a second analysis result based on a first analysis result obtained by simulation using a predetermined model including a plant model, an equipment / piping model, and a material model, and the results of statistical processing of predetermined operating data, and outputs predetermined analysis result information based on the first and second analysis results. Specifically, in this equipment / piping model, the temperature, flow rate, pressure, etc. of the fluid actually flowing in the piping are determined based on sensor information obtained from the plant model and calculation results from a chemical process simulator, and also taking into account the actual layout of the equipment and piping and the configuration information of the equipment. Furthermore, in order to evaluate material properties, a function is used to derive the thinning rate using flow velocity, temperature, pressure, pH, and dissolved oxygen content as parameters.

[0004] Furthermore, Patent Document 2 discloses a pipe thinning prediction system that includes a first AI evaluation method for evaluating the fluid characteristics of a pipe from input values ​​including information on pipe shape and information on the fluid inside the pipe, and a second AI evaluation method for evaluating an index related to pipe thinning from input values ​​including fluid characteristics, thereby predicting an index related to pipe thinning. The fluid characteristics in this system are obtained by numerical fluid analysis from at least one of the following: wall shear force due to flow, heat transfer coefficient, mass transfer coefficient, turbulence intensity, turbulence energy, turbulence strength, swirling flow intensity, swirl number, eddy viscosity coefficient, dissipation rate, flow velocity distribution, pressure distribution, and vorticity, and the amount of thinning is evaluated using the AI ​​evaluation method with these fluid characteristics. [Prior art documents] [Patent Documents]

[0005] [Patent Document 1] Patent No. 7161872 [Patent Document 2] Japanese Patent Publication No. 2022-148200 [Overview of the project] [Problems that the invention aims to solve]

[0006] In the plant diagnostic system described in Patent Document 1, flow velocity is used as the analytical value obtained from fluid simulation. However, simply using flow velocity as an indicator makes it difficult to accurately predict the amount of pipe wall thinning. In the pipe wall thinning prediction system described in Patent Document 2, the amount of wall thinning is evaluated using an AI evaluation method that utilizes fluid characteristic values ​​other than flow velocity (such as wall shear stress and turbulence intensity). However, even in this system, the trend of wall thinning due to FAC differs between regions where the flow is turbulent and regions where it is not. Therefore, simply dealing with fluid characteristic values ​​makes it difficult to accurately evaluate the amount of wall thinning in areas where the flow conditions differ locally.

[0007] In view of the above circumstances, the object of the present invention is to provide a technology that improves the accuracy of predicting the amount of material loss by considering multiple different flow conditions. [Means for solving the problem]

[0008] The wall thinning prediction system according to the present invention comprises a fluid simulator that calculates fluid analysis values ​​by fluid simulation using the amount of wall thinning in a test piece as a result of a water flow corrosion test and the shape of the test piece for a test piece used in a water flow corrosion test; a multiple regression equation generation unit that generates a wall thinning prediction equation with the amount of wall thinning as the objective variable and the fluid analysis values ​​as the explanatory variables; and a wall thinning amount calculation unit that calculates the amount of wall thinning in the target pipe based on the shape data of the target pipe and the wall thinning prediction equation, wherein the wall thinning prediction equation is a combination of a plurality of provisional prediction equations generated under different flow conditions in the test piece.

[0009] In this configuration, a thinning prediction formula is generated with the thinning amount as the dependent variable, using fluid analysis values ​​calculated by fluid simulation using the amount of thinning in the test specimen as a result of a water flow corrosion test and the shape of the test specimen as explanatory variables. At that time, the thinning prediction formula is generated by combining multiple provisional prediction formulas generated under different flow conditions in the test specimen. As a result, this thinning prediction formula takes into account multiple different flow conditions, improving the accuracy of the thinning amount prediction.

[0010] In this invention, it is proposed that the fluid analysis values ​​include at least two of the following: flow velocity, turbulent energy, wall shear stress, and mass transfer coefficient. This feature configuration, along with the feature configuration in which a final thinning prediction formula is generated from a combination of multiple provisional prediction formulas for different flow states described above, allows for consideration of various flow states and improves the accuracy of thinning prediction.

[0011] In order to specifically realize a feature configuration in which a final thinning prediction formula is generated from a combination of multiple provisional prediction formulas for different flow states, it is preferable to generate the final thinning prediction formula by combining a provisional prediction formula generated for one flow state with a provisional prediction formula generated for another different flow state. Therefore, in this invention, it is proposed that the provisional prediction formula consists of a first provisional prediction formula for a first region where the flow state is a first state, and a second provisional prediction formula for a second region where the flow state is a second state that is more turbulent than the first state.

[0012] Once a first provisional prediction formula for the first region, which is the first state, and a second provisional prediction formula for the second region, which is the second state, are generated, the amount of material loss can be easily and accurately calculated by substituting values ​​corresponding to the explanatory variables in these prediction formulas. Therefore, in this invention, it is proposed that the amount of material loss is calculated by substituting data relating to the fluid analysis values ​​in the first region into the first provisional prediction formula, and values ​​that are the fluid analysis values ​​in the second region into the second provisional prediction formula.

[0013] Depending on the flow path, the correlation between the amount of wall thinning and the mass transfer coefficient may be positive and negative in two different regions. In this case, it is preferable to clearly distinguish between these two different regions and generate a wall thinning prediction formula accordingly. Therefore, in this invention, it is proposed that the first region and the second region are determined based on the flow state, which is distinguished using the relationship between the amount of wall thinning in the water flow corrosion test and the distribution of the mass transfer coefficient calculated by the fluid simulator.

[0014] It is important that the test specimen of the present invention has a flow path that exhibits different flow states. Therefore, in the present invention, it is proposed that the test specimen is a T-shaped flow path, in which the fluid flows in from the lower end of the longitudinal flow path of the T-shaped flow path and flows out from the left and right ends of the transverse flow path having a rectangular cross-section of the T-shaped flow path, and that the fluid simulator determines the flow velocity, turbulent energy, wall shear stress, and mass transfer coefficient on the upper surface of the transverse flow path.

[0015] The present invention is not only directed to a weight loss prediction system, but also to a weight loss prediction method. Such a weight loss prediction method includes a fluid simulation step of calculating a fluid analysis value by fluid simulation using the amount of weight loss in the test piece and the shape of the test piece as the results of a water flow corrosion test for the test piece used in the water flow corrosion test, a multiple regression equation generation step of generating a weight loss prediction equation with the amount of weight loss as the objective variable and the fluid analysis value as the explanatory variable, and a weight loss amount calculation step of calculating the amount of weight loss in the prediction target pipe based on the shape data of the prediction target pipe and the weight loss prediction equation. The weight loss prediction equation is a combination of a plurality of provisional prediction equations generated under different flow states in the test piece. Such a weight loss prediction method also has the operations and effects in the various embodiments described above.

[0016] Other features, operations, and effects of the present invention will be clarified by the description of the present invention using the following drawings.

Brief Description of the Drawings

[0017] [Figure 1] It is a schematic diagram for explaining the schematic configuration of the weight loss prediction system. [Figure 2] It is a schematic diagram showing an example of a test piece used in the water flow corrosion test. [Figure 3] It is a flowchart for explaining a method of generating a weight loss prediction equation. [Figure 4] It is a distribution diagram showing a part of the results of the fluid simulation. [Figure 5] It is a distribution diagram showing the relationship between the mass transfer coefficient and the amount of weight loss. [Figure 6] It is a distribution diagram showing the relationship between the predicted weight loss value and the measured weight loss value. [Figure 7] It is a table showing the values of the partial regression coefficients obtained by multiple regression analysis. [Figure 8] It is a distribution diagram showing the relationship between the measured weight loss depth and the flow velocity.

Embodiments for Carrying Out the Invention

[0018] Figure 1 is a schematic diagram showing the general configuration of the meat loss prediction system according to the present invention. The basic functions of the meat loss prediction system will be explained using Figure 1.

[0019] In this embodiment, the wall thinning prediction system performs a water flow corrosion test using a test specimen 10 in laboratory 1, and provides the test results obtained from this water flow corrosion test, along with test specimen data such as the shape of the test specimen 10 and the fluid flowing through the test specimen 10, to the fluid simulator 3.

[0020] The test specimen 10 used in this embodiment is shown in Figure 2. This test specimen 10 is a T-shaped channel body consisting of a vertical channel 11 and a horizontal channel 12, both having a rectangular cross-section. The fluid flows in from the lower end of the vertical channel 11 and flows out from the left and right ends of the horizontal channel 12.

[0021] Fluid simulator 3 assigns the test results, including the amount of wall thinning, and the test specimen data to a wall thinning prediction model equation, performs a fluid simulation, and calculates fluid analysis values. The fluid analysis values ​​include flow velocity, turbulent energy, wall shear stress, and mass transfer coefficient. Fluid simulator 3 calculates and outputs the flow velocity, turbulent energy, wall shear stress, and mass transfer coefficient as fluid analysis values ​​for the upper surface of the lateral channel 12 in the test specimen 10. The fluid model used in the fluid simulation of this fluid simulator 3 consists of the mass conservation equation, the equation of motion, the turbulent energy transport equation, the turbulent dissipation rate transport equation, and so on.

[0022] The multiple regression equation generation unit 4 generates a thinning prediction equation with the test result, the amount of thinning, as the dependent variable and the fluid analysis value as the independent variable. In this invention, the method for generating the thinning prediction equation will be explained in detail later, but it is basically generated in two steps. That is, the thinning prediction equation is generated based on a combination of multiple (in this case, two) provisional prediction equations (two distinct relative relationships) generated under different flow conditions in the test piece. More specifically, the provisional prediction equation consists of a first provisional prediction equation (first relative relationship) for a first state where the flow condition is good (low mass transfer coefficient), and a second provisional prediction equation (second relative relationship) for a second state where the flow condition is more turbulent than in the first state (high mass transfer coefficient). The boundary between the first and second states with respect to the mass transfer coefficient is defined as the region where the slope of the relationship between the mass transfer coefficient and the amount of thinning is approximately zero.

[0023] The thinning prediction formula generated by the multiple regression equation generation unit 4 is used by the thinning amount calculation unit 5. The thinning amount calculation unit 5 assigns the shape data (actual equipment data) of the target piping for which thinning is to be predicted, for example, the boiler 20 of the plant equipment 2, to the thinning prediction formula and calculates the amount of thinning in the target piping.

[0024] Next, we will explain the specific method for generating the meat loss prediction formula using the flowchart in Figure 3.

[0025] First, a water flow corrosion test is conducted using specimen 10 in laboratory 1 (#01). The amount of wall thinning is measured through the water flow corrosion test (#02). Using the specimen data, including the measured amount of wall thinning and water flow conditions, a fluid simulation is performed using fluid simulator 3 (#03). This fluid simulation calculates fluid analysis values ​​such as flow velocity, turbulent energy, wall shear stress, and mass transfer coefficient. Desired fluid analysis values ​​are extracted from the calculated fluid analysis values ​​(#04).

[0026] An example of the result of the fluid simulation is shown by the distribution diagrams (graph diagrams) of (a), (b), (c), and (d) in FIG. 4. In these graphs, the horizontal axis represents the coordinate value of the cross-flow path 12 of the test piece 10. The vertical axis in the graph of (a) in FIG. 4 is the flow velocity, the vertical axis in the graph of (b) in FIG. 4 is the turbulent energy, the vertical axis in the graph of (c) in FIG. 4 is the wall shear stress, and the vertical axis in the graph of (d) in FIG. 4 is the mass transfer coefficient.

[0027] Focusing on the relationship between the mass transfer coefficient and the depth of material removal (amount of material removal), an approximate formula (prediction formula) for the depth of material removal (depth of material removal) is created using a sixth-order polynomial of the mass transfer coefficient. An example of the prediction formula is y = a·X 6 +b·X 5 +c·X 4 +d·X 3 +e·X 2 +f·X - 9.6·10 2 Here, a = -5.3·10 9 、b = +2.9·10 9 、c = -6.2·10 8 、 d = +6.6·10 7 、e = -3.6·10 6 、f = +9.7·10 4 、y is the depth of material removal (μm), and X is the mass transfer coefficient (m / s).

[0028] The relationship between the mass transfer coefficient and the depth of material removal using the above prediction formula is shown in FIG. 5. Focusing on the relationship between the mass transfer coefficient and the depth of material removal here, it can be seen that there is a positive correlation between the depth of material removal and the mass transfer coefficient up to a certain magnitude of the mass transfer coefficient, but there is a negative correlation in the region of the mass transfer coefficient above a certain magnitude.

[0029] Furthermore, Figure 5 shows not only the distribution of the relationship between the mass transfer coefficient and the amount of meat loss, but also the approximation curve 6 based on the prediction formula. This approximation curve 6 shows that the slope of approximation curve 6 is positive (increasing) in the range where the mass transfer coefficient is approximately less than 0.1, while the slope of approximation curve 6 is negative (decreasing) in the range where the mass transfer coefficient is approximately 0.1 or greater. In the range where the mass transfer coefficient is approximately less than 0.1, approximation curve 6 can be represented by a positive correlation 6a, and in the range where the mass transfer coefficient is approximately 0.1 or greater, approximation curve 6 can be represented by a negative correlation 6b. Therefore, the positive correlation 6a of approximation curve 6 can be used to generate a relational equation, which we will call the first hypothesis prediction formula (first correlation), through multiple regression analysis, and the negative correlation 6b of approximation curve 6 can be used to generate a relational equation, which we will call the second hypothesis prediction formula, through multiple regression analysis. This can be verified experimentally.

[0030] In this invention, the "classification of flow states" is performed using the mass transfer coefficient, and the state is classified into a first state region with a low mass transfer coefficient and a second state region with a high mass transfer coefficient. Specifically, the mass transfer coefficient at which the slope of the prediction formula becomes 0 is determined, and in the region exhibiting that mass transfer coefficient, the wall thinning depth is 100 μm or more or the value of the mass transfer coefficient is 0.1 m / s or more (#05).

[0031] Figure 6 illustrates the relationship between the predicted amount of material loss, which is the result of the prediction formula, and the actual amount of material loss. Figure 6 is a distribution map showing the relationship between the predicted amount of material loss and the actual amount of material loss. Here, the distribution where the mass transfer coefficient is <0.11 m / s is shown as "dense circles," and the distribution where the mass transfer coefficient is >=0.11 m / s is shown as "slight circles." In this way, the relationship between the mass transfer coefficient and the amount of material loss is classified by the value of the mass transfer coefficient, which indicates the flow state (Figure 3, #05), and multiple regression analysis is performed for each flow state (Figure 3, #06). In other words, the obtained mass transfer coefficient is used as a boundary to classify the data used to generate the material loss prediction formula, and multiple regression analysis is performed on the data classified by the value of the mass transfer coefficient (by different flow states).

[0032] The multiple regression equation used in the above multiple regression analysis is: t=a0+a1·v+a2·kT+a3·σ+a4·kM Here, t: amount of material removed (μm), v: flow velocity (m / s), kT: turbulent energy (m 2 / s 2 ), σ: wall shear stress (Pa), kM: mass transfer coefficient (m / s), a0~a4: partial regression constants.

[0033] Figure 7 shows the values ​​of the partial regression coefficients obtained from the multiple regression analysis. In this analysis, data from the first region, where the mass transfer coefficient is less than 0.11 m / s, is substituted into the first prediction formula, and data from the second region, where the mass transfer coefficient is 0.11 m / s or greater, is substituted into the second prediction formula. The regression equation using the partial regression coefficients shown in Figure 7 becomes the meat loss prediction formula. This meat loss prediction formula is a combination of a first provisional prediction formula using the values ​​of the partial regression coefficients in the range where the mass transfer coefficient is approximately less than 0.1, and a second provisional prediction formula using the values ​​of the partial regression coefficients in the range where the mass transfer coefficient is approximately 0.1 or greater. By combining the first and second provisional prediction formulas, a meat loss prediction formula for the entire range of mass transfer coefficients is generated (Figure 3, #07). In order to obtain the partial regression coefficients, probabilistic statistical methods such as principal component analysis are used, and furthermore, machine learning methods such as deep learning are also used.

[0034] Figure 8 shows the relationship between the measured wall thinning depth and the flow velocity. Verification results by the inventors showed that when predicting the amount of wall thinning using a linear relationship between flow velocity and wall thinning depth, the prediction error was up to 25%, whereas when using the wall thinning amount prediction formula generated through the procedure shown in Figure 3, the prediction error was reduced to a maximum of 20%, confirming that predictions can be made with greater accuracy.

[0035] [Another embodiment] (1) In the embodiment described above, the test piece 10 was a T-shaped channel body consisting of a vertical channel 11 and a horizontal channel 12, but a test piece 10 having a different shape may be used depending on the shape of the piping to be predicted.

[0036] (2) In the embodiments described above, mass transfer coefficients were used as different flow states used to classify the test data, but other fluid analysis values, such as turbulent energy or the direction of flow near the wall of the channel (direction relative to the wall), may be used instead.

[0037] (3) In the embodiments described above, the fluid analysis values ​​used in the multiple regression analysis were flow velocity, turbulent energy, wall shear stress, and mass transfer coefficient. However, at least two of these may be used, or other fluid analysis values ​​may be used.

[0038] (4) Each functional unit shown in the schematic configuration of the thinning prediction system in Figure 1 may be integrated or divided into multiple units. Furthermore, the functional units may be distributed across multiple computer systems (such as cloud service systems or edge computer systems). For example, the fluid simulator 3 and the multiple regression equation generation unit 4 may be built on a cloud service system.

[0039] (5) In fluid analysis, pH concentration may be considered separately from the prediction formula described above.

[0040] Furthermore, the configurations disclosed in the above embodiments (including other embodiments, the same applies hereinafter) can be applied in combination with configurations disclosed in other embodiments, as long as no inconsistencies arise. Moreover, the embodiments disclosed herein are illustrative, and the embodiments of the present invention are not limited thereto, and can be modified as appropriate without departing from the object of the present invention. [Industrial applicability]

[0041] This invention is applicable to pipe wall thinning prediction, which predicts pipe wall thinning due to corrosion, particularly flow-accelerated corrosion. [Explanation of symbols]

[0042] 1: Laboratory 2: Plant equipment 3: Fluid simulator 4: Multiple regression equation generation unit 5: Thickness reduction amount calculation section 6:Approximate curve 6a: Correlation 6b: Correlation 10: Test specimen 11: Longitudinal channel 12: Horizontal flow path 20: Boiler

Claims

1. A fluid simulator that calculates fluid analysis values ​​by fluid simulation using the amount of wall thinning in a test specimen as a result of a water corrosion test and the shape of the test specimen, for a test specimen used in a water corrosion test, A multiple regression equation generation unit generates a thinning prediction equation with the thinning amount as the dependent variable and the fluid analysis value as the explanatory variable, A wall thinning amount calculation unit calculates the amount of wall thinning in the pipe to be predicted based on the shape data of the pipe to be predicted and the wall thinning prediction formula, Equipped with, A wall thinning prediction system in which the wall thinning prediction formula is a combination of multiple provisional prediction formulas generated under different flow conditions in the test specimen.

2. The wall thinning prediction system according to claim 1, wherein the fluid analysis values ​​include at least two of the following: flow velocity, turbulent energy, wall shear stress, and mass transfer coefficient.

3. The thinning prediction system according to claim 1, wherein the provisional prediction formula comprises a first provisional prediction formula for a first region where the flow state is a first state, and a second provisional prediction formula for a second region where the flow state is a second state that is more turbulent than the first state.

4. The first region and the second region are determined based on the flow state, which is divided using the relationship between the amount of wall thinning in the water flow corrosion test and the distribution of mass transfer coefficients calculated by the fluid simulator, according to claim 3.

5. The thinning amount prediction system according to claim 4, wherein the amount of thinning is calculated by substituting data relating to the fluid analysis values ​​in the first region into the first provisional prediction formula and a value which is the fluid analysis value in the second region into the second provisional prediction formula.

6. The wall thinning prediction system according to claim 1, wherein the test specimen is a T-shaped channel, the fluid flows in from the lower end of the longitudinal channel of the T-shaped channel and flows out from the left and right ends of the transverse channel having a rectangular cross-section of the T-shaped channel, and the fluid simulator determines the flow velocity, turbulent energy, wall shear stress, and mass transfer coefficient on the upper surface of the transverse channel.

7. A fluid simulation step in which fluid analysis values ​​are calculated by performing a fluid simulation using the amount of wall thinning in the test piece as a result of a water corrosion test and the shape of the test piece for a test piece used in a water corrosion test, A multiple regression equation generation step that generates a thinning prediction equation with the amount of thinning as the dependent variable and the fluid analysis value as the independent variable, A step of calculating the amount of wall thinning in the pipe to be predicted based on the shape data of the pipe to be predicted and the wall thinning prediction formula, Equipped with, A method for predicting wall thinning, wherein the wall thinning prediction formula is a combination of multiple provisional prediction formulas generated under different flow conditions in the test specimen.

Images