Pipe thinning prediction system, pipe integrity evaluation system, and method

Abstract

To provide a pipe wall thinning prediction system, a pipe soundness evaluation system, and a method that are able to evaluate an amount of wall thinning, a thinning rate, and a remaining life highly speedily from a limited actual measurement result and are able to specify a monitoring-required part of a pipe in a plant.SOLUTION: A pipe wall thinning prediction system includes a first AI evaluation method for evaluating a fluid characteristic value of a pipe from an input value including information on a pipe shape and information on a fluid in the pipe and a second AI evaluation method for evaluating an index regarding pipe wall thinning from the input value including the fluid characteristic value and predicts the index regarding the pipe wall thinning.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present invention relates to a technology for predicting wall thinning events in pipes and the like and optimizing a pipe wall thickness measurement plan using the prediction. In particular, the present invention relates to a technology for predicting wall thinning events in pipes and the like and optimizing a pipe wall thickness measurement plan using the prediction. droplet This invention relates to a pipe thinning prediction system, a pipe integrity evaluation system, and a method suitable for various plants such as nuclear plants, thermal power plants, oil refining plants, and chemical plants, where pipe thickness measurement is required at numerous locations in order to prevent thinning phenomena caused by complex fluid flows in the pipes, such as impact erosion. [Background technology]

[0002] It is known that in various plants, thinning of pipes and other components can occur due to flow-accelerated corrosion, droplet impact erosion, and other factors associated with the flow and conditions of internal fluids in pipes and other components. Therefore, in various plants, pipe thickness measurements are performed during regular plant inspections to determine whether or not thinning has occurred and its status. However, since it is difficult to predict where thinning will occur, pipe thickness is measured at numerous locations, which is an obstacle to shortening regular plant inspection periods and improving plant operating rates. To solve this problem, evaluation techniques for predicting where thinning will occur are being developed.

[0003] Non-Patent Document 1 introduces prediction models and prediction codes for flow accelerated corrosion that have been developed in Japan and overseas. Although the prediction models differ depending on the developer, all prediction models use a method to calculate the wall thinning rate and remaining life based on the fluid conditions inside the pipe. The evaluation methods in these documents are based on physical phenomena. [Prior art documents] [Non-patent literature]

[0004] [Non-Patent Document 1] "Current status of research on flow-accelerated corrosion", Journal of the Atomic Energy Society of Japan, Vol. 59 Summary of the Invention [Problem to be solved by the invention]

[0005] The method shown in the above Non-Patent Document 1 is constructed based on a physical formula. Phenomena such as flow accelerated corrosion are mainly caused by the flow of fluid in the pipes. In the various pipe shapes of a plant, the flow mechanism is complex, so it is not easy to elucidate all phenomena and reflect them in the physical formula. In addition, in order to improve the prediction accuracy, calculations that reflect the complex pipe shapes are required. Therefore, the calculation time including the input work becomes long, and it is not practical to evaluate the target parts of all the pipes in a huge plant, which number tens of thousands, and narrow down the parts with a high risk of thinning.

[0006] On the other hand, one method that can be considered for evaluating thinning quickly is to use a method that uses artificial intelligence (AI). Although AI allows for high-speed evaluation, due to the diversity of piping shapes used in plants, a very large number of actual plant measurement results (learning data) is required. If the amount of this learning data is small, there is a concern that it may cause incorrect evaluation of thinning or a decrease in prediction accuracy.

[0007] In view of the above, the present invention aims to provide a pipe thinning prediction system, a pipe integrity evaluation system, and a method that are capable of quickly evaluating the amount of thinning, the thinning rate, and the remaining life from limited actual measurement results (learning data), and of identifying areas in the piping within a plant that require monitoring. [Means for solving the problem]

[0008] In order to solve the above problems, the present invention provides a The present invention provides a plant that includes a first AI evaluation method that learns from a result of a computational fluid analysis as training data, and a second AI evaluation method that learns from a mass transfer coefficient output by the first AI evaluation method and actual measurement data of information on a fluid in a pipe, which is a plant operating condition, as training data. The first AI evaluation method is Pipe shape information, Obtained by sensors installed in the plant The pipe's characteristics are calculated from input values ​​including information on the fluid in the pipe. The mass transfer coefficient is evaluation death , The second AI evaluation method is to use the mass transfer coefficient and information on the fluid in the pipe. Evaluate pipe thinning indicators from input values ​​including death, A pipe thinning prediction system characterized by predicting indicators related to pipe thinning.

[0009] In addition, in the present invention, "pipe shape information, Fluid and heat information in pipes, including flow rate (velocity) and temperature From input values ​​including Mass Transfer Coefficient The first AI evaluation method evaluates the Mass Transfer Coefficient The second AI evaluation method evaluates the indices that can evaluate the thermal health of the piping from input values ​​including the above, and predicts the indices that can evaluate the thermal health of the piping. The first AI evaluation method is trained using the results of computational fluid analysis using information on fluid and heat inside the piping, which are plant operating conditions, as training data, while the second AI evaluation method is trained using actual measurement data on information on fluid and heat inside the piping, which are plant operating conditions, as training data. The present invention relates to a piping integrity evaluation system characterized by the above.

[0010] In addition, in the present invention, "the pipe shape information and the input value including the information of the fluid in the pipe are used to calculate the pipe shape. Mass Transfer Coefficient is evaluated by a first AI evaluation method, Mass Transfer Coefficient The second AI evaluation method is used to evaluate the pipe thinning index from the input values ​​including the above, and predict the pipe thinning index. The first AI evaluation method is trained using the results of computational fluid analysis using information on fluid and heat inside the piping, which are plant operating conditions, as training data, while the second AI evaluation method is trained using actual measurement data on information on fluid and heat inside the piping, which are plant operating conditions, as training data. The method is a pipe thinning prediction method characterized by the above.

[0011] In addition, in the present invention, "pipe shape information, Fluid in the pipe, including flow rate (velocity) and temperature Pipe temperature is calculated from input values ​​including thermal information. Mass Transfer Coefficient is evaluated by a first AI evaluation method, Mass Transfer Coefficient The second AI evaluation method is used to evaluate the indexes capable of evaluating the heat-related soundness of the piping from the input values ​​including the above, and predict the indexes capable of evaluating the heat-related soundness of the piping. The first AI evaluation method is trained using the results of computational fluid analysis using information on fluid and heat inside the piping, which are plant operating conditions, as training data, while the second AI evaluation method is trained using actual measurement data on information on fluid and heat inside the piping, which are plant operating conditions, as training data. The present invention relates to a piping integrity evaluation method, characterized in that Effect of the Invention

[0012] The present invention makes it possible to provide a pipe thinning prediction system and a pipe thinning prediction method that can evaluate the amount of thinning, thinning rate, and remaining life of piping in a plant quickly and accurately using limited actual plant measurement results (learning data), and identify areas in the piping in the plant that require monitoring.

[0013] More specifically, according to the embodiment of the present invention, a method can be realized that identifies and screens locations with a high risk of thinning in a huge amount of plant piping in a short time, evaluates the amount of thinning, thinning rate, and remaining life of the high risk locations, and revises or optimizes the pipe wall thickness measurement plan based on the results. This can also contribute to shortening the periodic plant inspection period and preventing leakage events due to thinning, thereby improving the plant's availability.

[0014] Problems, configurations and effects other than those described above will become apparent from the following description of the embodiments. [Brief description of the drawings]

[0015] [Figure 1] FIG. 2 is a diagram showing a series of processing steps for predicting pipe thinning according to the first embodiment of the present invention. [Diagram 2] FIG. 11 is a diagram showing a series of processing steps for predicting pipe thinning according to a second embodiment of the present invention. [Diagram 3] FIG. 11 is a diagram showing a series of processing steps for predicting various issues according to the third embodiment of the present invention. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0016] Hereinafter, an embodiment of the present invention will be described with reference to the drawings. In each drawing, the same components are given the same reference numerals, and detailed description of overlapping parts will be omitted. In the following, a pipe thinning prediction system and method will be described in Example 1, and a pipe integrity evaluation system and method will be described in Example 2. EXAMPLES

[0017] 1 shows a system and method for predicting pipe wall thinning according to a first embodiment of the present invention. Here, the pipe wall thinning prediction system 1 is realized using a computer, and the computer 1 receives inputs of pipe shape information D1 from a piping CAD 6 and plant operating condition information D2 from a sensor 7 installed in the plant.

[0018] The series of processing functions in the pipe thinning prediction system 1 of the present invention are composed of a first-stage AI evaluation method AI1 that evaluates a fluid characteristic value D3 of the pipe from input values ​​including pipe shape information D1 and plant operating condition information D2, and a second-stage AI evaluation method AI2 that evaluates thinning information 4 regarding the amount of thinning, thinning rate, and remaining life from input values ​​including the fluid characteristic value D3 and plant operating condition information D2.

[0019] The flow of this processing stage involves creating multiple AI evaluation methods AI1, AI2 in advance from learning data, and using the multiple created AI evaluation methods to predict wall thinning information 4 (wall thinning amount, wall thinning rate, remaining life) from pipe shape information D1 of a piping CAD 6 and plant operating condition information D2 from a sensor 7 installed in the plant. The plant operating conditions D2 include water quality, temperature, pressure, flow rate, etc. Deep learning (neural network) etc. is used as the AI.

[0020] In this way, the pipe thinning prediction system 1 of the present invention employs a two-stage AI evaluation method AI, and obtains a fluid characteristic value D3 in the first stage and thinning information D4 in the second stage. In this regard, in order to construct a conventional simple single-stage AI evaluation method, if the pipe shape is input, a very large number of plant actual measurement results (learning data) is required due to the diversity of pipe shapes. The piping of a plant is made up of various components such as bent pipes, T-shaped pipes, reduction pipes, expansion pipes, orifices, and valves, which are intricately connected. Usually, actual measurement values ​​measured at the plant are used as learning data. However, since it takes several years to evaluate the actual measurement data, the number of measurement data at the plant is limited. Therefore, it is difficult to secure sufficient learning data to enable conventional AI evaluation. On the other hand, the actual measurement results of the plant are the most reliable quantitative numerical data, and in order to improve the prediction accuracy, it is necessary to effectively use the limited actual measurement results.

[0021] Therefore, in order to solve the above problem, the present invention has decided to install AI evaluation methods in multiple stages. In the first stage, an AI evaluation method AI1 is installed that uses input including complex piping shapes obtained from a piping CAD 6 to output a fluid characteristic value D3. This first stage AI evaluation method AI1 also requires AI learning to be performed in advance.

[0022] Therefore, the output of this first-stage AI evaluation method AI1 is a fluid characteristic value D3. This means that learning data for this fluid characteristic value D3 can be created in advance by computational fluid analysis using the plant operating conditions D2, such as water quality, temperature, pressure, and flow rate. In other words, the first-stage AI evaluation method D3 (weights and biases of the neural network) can be learned using the results of computational fluid analysis as training data. Note that the fluid characteristic values ​​do not have to be numerical values, and may be classified by category.

[0023] On the other hand, when we look at the piping shape, which is the input to the first-stage AI evaluation method AI1, the piping shape is complex and requires the results of a large number of numerical fluid analyses; however, by performing simulation calculations, it is possible to obtain results for a large number of different piping shapes.

[0024] Next, we will explain specific examples of fluid characteristic values ​​D3 obtained in the first stage AI evaluation method AI1. Fluid characteristic values ​​D3 in piping include wall shear force due to flow, mass transfer coefficient, heat transfer coefficient, turbulence strength (turbulence energy, turbulence intensity), swirling flow strength (swirl number), eddy viscosity coefficient, dissipation rate, flow velocity distribution, pressure distribution, vorticity, etc., and these various fluid characteristic values ​​D3 can be obtained from computational fluid analysis.

[0025] Figure 2 shows a specific example of the configuration of the first-stage AI evaluation method AI1, in which the mass transfer coefficient, the strength of the swirl flow, and the turbulent energy are determined in AI evaluation method AI1a, AI evaluation method AI1b, and AI evaluation method AI1c, respectively, as representatives of the above-mentioned fluid characteristic values ​​D3. By determining one or more of these fluid characteristic values ​​D3 in the first-stage AI evaluation method AI1 and inputting them into the second AI evaluation method AI2, it is possible to evaluate the amount of thinning, the thinning rate, and the lifespan. By using multiple fluid characteristic values ​​D3, it is expected that accuracy will be improved.

[0026] In Figure 1, a second-stage AI evaluation method AI2 is further installed, which can evaluate the amount of wall thinning as an output from an input layer that includes the fluid characteristic value D3, which is the output of the first-stage AI evaluation method AI1. This second-stage AI2 also requires AI learning to be performed in advance. Note that the amount of wall thinning here is an example of an index related to pipe wall thinning, and it may be the wall thinning rate or remaining life other than the amount of wall thinning.

[0027] However, in this case, the input fluid characteristic values ​​D3 can be given as simple numerical values ​​or categories, and furthermore, the second-stage AI evaluation method AI2 does not include complex piping shapes in the input, so this second-stage AI evaluation method AI2 can be constructed even with a small amount of actual plant measurement data. In other words, the second-stage AI evaluation method AI2 (neural network weights and biases) can learn from limited plant measurement data (plant operating condition information D2) as teacher data.

[0028] As shown in the above discussion, the first-stage AI evaluation method AI1, which is trained using the computational fluid analysis results as training data, and the second-stage AI evaluation method AI2, which is trained using the plant's actual measurement data D2 as training data, can be realistically constructed even with a small amount of training data.

[0029] After the learning is completed, the actual operation and the amount of thinning are evaluated. This multi-stage AI evaluation method is built in a computer, and when the plant operating conditions, piping shape conditions, etc. are input, the amount of thinning at each location is evaluated at high speed by the multi-stage AI method. This amount of thinning is evaluated numerically when neural network regression is used, and by category when classification is used.

[0030] As described above, the multi-stage AI evaluation method can quickly evaluate the amount of thinning in complex and diverse plant piping using a small amount of data. Here, two stages are explained, but the number of stages can be increased.

[0031] In addition, droplet impact erosion can be evaluated in a similar manner by using values ​​related to droplets in the pipe as fluid characteristic values. For this issue as well, the system is composed of a first-stage AI evaluation method AI1 that evaluates the fluid characteristic values ​​of the pipe from input values ​​including pipe shape information, and a second-stage AI evaluation method AI2 that evaluates the amount of wall thinning from input values ​​including the fluid characteristic values.

[0032] Furthermore, it is preferable for the system in FIG. 1 to have a function for predicting indicators related to pipe thinning and for providing reference information for determining locations with a high risk of pipe thinning as locations for measuring pipe thickness. EXAMPLES

[0033] The evaluation method using the two-stage AI processing described in the first embodiment can be expanded to general soundness evaluation indexes as it is. Figure 3 shows a series of processing steps for predicting various problems according to the present invention.

[0034] The configuration in Figure 3 is basically the same as that in Figure 1, but the detection information D2' from the sensor 7 includes information on the heat held by the fluid and information on the support conditions of the piping, the first-stage AI method AI1' handles thermal fluid characteristic values ​​D3', and the second-stage AI method AI2' obtains indicators D4' (thermal fatigue, thermal stress, SCC) that can evaluate heat-related soundness in addition to thinning.

[0035] Using the second embodiment of the present invention, various issues such as pipe vibration, earthquake resistance, thermal fatigue, thermal stress, and stress corrosion cracking (SCC) can be evaluated in the same manner as in the first embodiment, in addition to wall thinning. These issues may be evaluated simultaneously with wall thinning. For vibration, earthquake resistance, and thermal stress, it is necessary to add support conditions for the pipe such as supports to the input. For thermal fatigue and thermal stress, input of thermal conditions, temperature conditions, etc. is important. For these issues, the method is composed of a first stage AI evaluation method that evaluates the thermal fluid characteristic values ​​of the pipe from input values ​​including pipe shape information, and a second stage AI evaluation method that evaluates the amount of wall thinning from input values ​​including the thermal fluid characteristic values. In addition, in the case of vibration and earthquake resistance, it is possible to use vibration characteristic values ​​in addition to thermal fluid characteristic values.

[0036] As described above, the present invention used to solve the problem of wall thinning can also be used to solve various problems within plants. [Explanation of symbols]

[0037] D1: Pipe shape D2: Plant operating condition information D3: Fluid characteristic value AI1: First stage AI evaluation method to evaluate fluid property values D4: Thickness reduction amount, thickness reduction rate, remaining life AI2: The second stage of AI evaluation method to evaluate the amount of thinning, etc. 6: Piping CAD 7: Plant sensors

Claims

1. A first AI evaluation method that learns the results of computational fluid analysis as training data; A second AI evaluation method that learns the mass transfer coefficient output by the first AI evaluation method and actual measurement data of information on a fluid in a pipe, which is a plant operating condition, as teacher data; The first AI evaluation method evaluates a mass transfer coefficient of a pipe from an input value including information on a pipe shape of a plant and information on a fluid in the pipe acquired by a sensor installed in the plant; The second AI evaluation method is a pipe thinning prediction system that evaluates an index related to pipe thinning from input values ​​including the mass transfer coefficient and information on the fluid in the pipe, and predicts the index related to pipe thinning.

2. The pipe thinning prediction system according to claim 1, A pipe thinning prediction system that includes an AI evaluation method for evaluating the mass transfer coefficient related to droplets and evaluates the amount of thinning caused by droplet impact erosion.

3. The pipe thinning prediction system according to claim 1, A pipe thinning prediction system that predicts indicators of pipe thinning and presents reference information for determining locations with a high risk of thinning as locations for measuring pipe thickness.

4. The pipe thinning prediction system according to claim 1, A pipe thinning prediction system, characterized in that the index related to the pipe thinning is any one of the amount of thinning, the thinning rate, and the remaining life of the pipe.

5. The present invention is provided with a first AI evaluation method for evaluating a mass transfer coefficient of a pipe from input values ​​including information on the pipe shape and information on the fluid and heat in the pipe, including the flow rate (flow velocity) and temperature, and a second AI evaluation method for evaluating an index capable of evaluating the heat-related soundness of the pipe from the input values ​​including the mass transfer coefficient, and predicts an index capable of evaluating the heat-related soundness of the pipe, and The first AI evaluation method learns by using as training data the results of a computational fluid analysis using information on fluid and heat in piping, which is a plant operating condition; A piping health evaluation system, characterized in that the second AI evaluation method is learned using actual measured data of fluid and heat information in the piping, which are plant operating conditions, as teaching data.

6. The piping integrity evaluation system according to claim 5, A piping integrity evaluation system that evaluates one or more indicators including thermal fatigue, thermal stress, and stress corrosion cracking of piping as indicators by which the heat-related integrity of piping can be evaluated.

7. A mass transfer coefficient of the pipe is evaluated by a first AI evaluation method from input values ​​including information on the pipe shape and information on the fluid in the pipe, and an index related to wall thinning of the pipe is evaluated by a second AI evaluation method from the input values ​​including the mass transfer coefficient, and an index related to wall thinning of the pipe is predicted; The first AI evaluation method learns by using as training data the results of a computational fluid analysis using information on fluid and heat in piping, which is a plant operating condition; The second AI evaluation method is a pipe thinning prediction method characterized in that it learns using actual measured data of fluid and heat information in the pipes, which are plant operating conditions, as training data.

8. A mass transfer coefficient of the pipe is evaluated by a first AI evaluation method from input values ​​including information on the pipe shape and information on the fluid and heat in the pipe including the flow rate (flow velocity) and temperature, and an index capable of evaluating the thermal-related soundness of the pipe is evaluated by a second AI evaluation method from the input values ​​including the mass transfer coefficient, and the index capable of evaluating the thermal-related soundness of the pipe is predicted; The first AI evaluation method learns by using as training data the results of a computational fluid analysis using information on fluid and heat in piping, which is a plant operating condition; The second AI evaluation method is a piping integrity evaluation method characterized in that it learns using actual measured data of fluid and heat information in the piping, which are plant operating conditions, as teaching data.

Images