JPWO2019230975A1 - Maintenance support equipment, maintenance support programs and maintenance support methods - Google Patents

Abstract

A plurality of probability information storage devices that store a plurality of numerical ranges in association with a plurality of probability information indicating the probability that the structure becomes unusable due to corrosion under the heat insulating material, and a plurality of probability information storage devices with reference to the probability information storage device. Probability information acquisition unit (14) that specifies the numerical range including the predicted value of the calculated margin wall thickness from the numerical range of, and acquires the probability index associated with the specified numerical range, and the acquired probability. An output unit (15) for outputting an index is provided.

Description

The present invention relates to a maintenance support device for supporting determination of a maintenance method, inspection specifications, etc. of a plant structure, a maintenance support program, and a maintenance support method.

Conventionally, for example, as a technique for supporting determination of a maintenance method and inspection specifications of a structure of a plant, for example, an estimated value of a corrosion rate according to the type of the structure, a period of use of the structure, and an initial plate are used. There is a technique for estimating the life of a structure from the thickness (see, for example, Non-Patent Document 1).

"Analysis of actual machine inspection results of external surface corrosion under heat insulating materials in chemical plants and examination of probability estimation method", Materials and Environment, 59, 291-297 (2010)

However, in the technique described in Non-Patent Document 1, although the life of the structure can be estimated, there is no viewpoint of calculating the probability of reaching the life, that is, the probability of the structure becoming unusable. Therefore, it has been difficult to appropriately determine the maintenance method and inspection specifications of the structure.
Focusing on the above points, it is an object of the present invention to provide a maintenance support device, a maintenance support program, and a maintenance support method capable of presenting the probability that a plant structure becomes unusable.

In order to solve the above problems, one aspect of the present invention includes (a) an acquisition unit for acquiring data including attributes and initial plate thickness of the plant structure, and (b) attributes included in the data acquired by the acquisition unit. The wall thinning depth prediction unit that calculates the predicted value of the wall thinning depth due to corrosion under the heat insulating material of the structure, and (c) the wall thinning depth prediction unit from the initial plate thickness included in the data acquired by the acquisition unit. The margin wall thickness calculation unit that calculates the predicted value of the margin wall thickness, which is the subtraction result obtained by subtracting the predicted value of the wall thinning depth calculated in (d), (d) a plurality of predetermined numerical ranges, and under the heat insulating material. Refer to the probabilistic information storage device that stores a plurality of probabilistic indexes indicating the probability that the structure becomes unusable due to corrosion in association with each other, and (e) the probabilistic information storage device, and out of a plurality of numerical values. , The probability information acquisition unit that specifies the numerical range including the predicted value of the margin thickness calculated by the margin thickness calculation unit and acquires the probability index associated with the specified numerical range, and (f) the probability information acquisition. The gist is that it is a maintenance support device equipped with an output unit that outputs the probability index acquired by the unit. Here, the corrosion under the heat insulating material occurs in the structure of the plant (pipes, heat exchangers, towers, tanks, etc.), for example, the portion of the structure made of carbon steel or low alloy steel covered with the heat insulating material. It is corrosion.

In addition, another aspect of the present invention is based on (a) an acquisition unit for acquiring data including attributes and initial plate thickness of a plant structure, and (b) attributes included in the data acquired by the acquisition unit. The wall thinning depth prediction unit that calculates the predicted value of the wall thinning depth due to corrosion under the heat insulating material in the part of the structure covered with the heat insulating material, and (c) the initial plate thickness included in the data acquired by the acquisition unit. The margin thickness calculation unit that calculates the prediction value of the margin thickness, which is the subtraction result of subtracting the prediction value of the wall reduction depth calculated by the wall depth prediction unit, and (d) a plurality of predetermined numerical ranges. , Calculate the margin wall thickness from a plurality of numerical ranges by referring to the probability information storage device that stores a plurality of probability indexes indicating the probability that the structure becomes unusable due to corrosion under the heat insulating material. The probability information acquisition unit, (e), and the probability information acquisition unit that specify the numerical range including the predicted value of the margin thickness calculated in the unit and acquire the probability index associated with the specified numerical range. The gist is that it is a maintenance support program to function as an output unit that outputs indicators.

In addition, another aspect of the present invention is to (a) acquire data including the attributes and initial plate thickness of the structure of the plant, and (b) keep the structure warm based on the attributes included in the acquired data. The margin thickness, which is the result of calculating the predicted value of the wall thinning depth due to submaterial corrosion and (c) subtracting the predicted value of the wall thinning depth from the initial plate thickness included in the acquired data. The predicted value of (d) is stored in association with a plurality of predetermined numerical ranges and a plurality of probabilistic indexes indicating the probability that the structure becomes unusable due to corrosion under the heat insulating material. (E) From the plurality of numerical ranges, a numerical range including the predicted value of the margin thickness is specified, and a probability index associated with the specified numerical range is obtained. The gist is that it is a maintenance support method including acquisition and (f) output of the probability index.

According to one aspect of the present invention, it is possible to provide a maintenance support device, a maintenance support program, and a maintenance support method capable of presenting the probability that the structure of the plant becomes unusable.

It is a hardware block diagram of the maintenance support system which concerns on one aspect of this invention. It is a functional block diagram of the maintenance support system which concerns on one aspect of this invention. It is a figure which shows an example of the storage information stored in the probability information storage area. It is a figure which shows the storage information stored in the probability information storage area which concerns on a modification. It is a figure which shows an example of the content of the measurement data. It is a figure which shows an example of the content of the data about the structure which is the target of maintenance support. It is a figure which shows an example of the display mode of a display. It is a flowchart which shows the probability information output processing. It is a block diagram which shows an example of a prediction model. It is sectional drawing of the structure for demonstrating the predicted value of the margin wall thickness. It is a figure for demonstrating the acquisition process of probability information. It is a flowchart which shows the numerical value range setting process. It is sectional drawing of the structure for demonstrating the predicted margin wall thickness. It is sectional drawing of the structure for demonstrating the actual margin wall thickness. It is a figure for demonstrating the setting method of the 1st to 3rd boundary values, (a) is a figure for demonstrating the setting method of the 1st boundary value, and (b) is the figure for demonstrating the 2nd boundary value. (C) is a diagram for explaining a method of setting a third boundary value, and (d) is a diagram for explaining the first to fourth numerical ranges and the first to first ones. It is a figure for demonstrating the relationship with the boundary value of 3. It is a figure for demonstrating the setting method of the 1st to 3rd boundary values. It is a flowchart which shows the operation sequence of the maintenance system. It is a figure for demonstrating the acquisition procedure of probability information, (a) is the figure which shows the case where the control wall thickness is "0 mm", and (b) is the figure which shows the case where the control wall thickness is "1 mm". is there. It is a flowchart which shows the operation sequence of the maintenance system.

Hereinafter, the maintenance support device, the maintenance support program, and the maintenance support method according to the present disclosure will be described with reference to the drawings. The maintenance support device according to the present disclosure constitutes a maintenance support system for implementing maintenance support services that support determination of maintenance methods, inspection specifications, etc. of plant structures. The structure is, for example, a pipe, a heat exchanger, a tower, or a tank covered with a heat insulating material. As will be described later, the maintenance support system according to the present disclosure includes a maintenance support device and a service use terminal. The maintenance support device acquires data from the service use terminal, and based on the acquired data, the structure becomes unusable due to corrosion under the heat insulating material as an index indicating the degree of necessity for maintenance and repair of the structure. The probability information indicating the probability is acquired, and the acquired probability information is transmitted to the server user terminal of the data transmission source.

In addition, the maintenance support device according to the present disclosure predicts the degree of corrosion of the structure to be maintained based on the measurement data including the information on the attributes of the structure collected in advance and the information on the degree of corrosion. Generate a model.

It should be noted that the embodiments shown below exemplify methods and devices for embodying the technical idea of the present invention, and the technical idea of the present invention describes the structure, arrangement, etc. of the constituent elements as follows. It is not specific to the thing. The technical idea of the present invention may be modified in various ways within the technical scope specified by the claims.

(Constitution)
As shown in FIGS. 1 and 2, the maintenance support system 1 includes a maintenance support device 2 and a service use terminal 3. The maintenance support device 2 and the service use terminal 3 are configured to be able to send and receive information to and from each other via a communication path 4 such as the Internet.

(Maintenance support device)
The maintenance support device 2 is, for example, a server device owned by a service provider of the maintenance support service. As the maintenance support device 2, for example, a Web server (computer) can be adopted.

As shown in FIG. 1, the maintenance support device 2 provides information with another device (for example, a service use terminal 3 described later) via a storage device 5, an arithmetic unit 6 including a processor, and a communication path 4 (network). It is provided with hardware resources such as a communication interface (hereinafter, referred to as "communication I / F") 30 for transmitting and receiving. The maintenance support device 2 is configured such that the arithmetic unit 6 estimates at least the degree of corrosion of the maintenance object by executing information processing based on a predetermined program and data stored in the storage device 5. In the present disclosure, the wall thinning depth is used as the degree of corrosion of the object to be maintained.
The arithmetic unit 6 includes a CPU (Central Processing Unit) 6a as a hardware processor, a RAM (Random Access Memory) 6b, and a ROM (Read Only Memory) 6c. The arithmetic unit 6 executes specific information processing using the predetermined data stored in the storage device 5 and the predetermined program. The hardware processor is not limited to the CPU, and various processors such as GPU, ASIC, and FPGA can be used.

The storage device 5 is an auxiliary storage device composed of memories such as a hard disk drive and a solid state drive. The storage device 5 stores a predetermined program and predetermined data required for specific information processing by the arithmetic unit 6. In the present disclosure, the storage device 5 includes a program 9 for the arithmetic device 6 to execute predetermined information processing such as maintenance support processing, probability information generation processing, prediction model generation processing, and probability information 7, and measurement. Data 10 and learning result data 35 are stored. Even if each of the probability information 7, the measurement data 10, the program 9, and the learning result data 35 is separately stored in a predetermined area of the storage device 5, for example, a probability information storage area, a measurement data storage area, and a program storage area. Good. Further, the probability information 7, the measurement data 10, the program 9, and the learning result data 35 are each stored in the separately configured probability information storage device, measurement data storage device, and program storage device. It may be configured.

The predetermined data and the predetermined program may not be stored in the constant storage device 5. For example, when the arithmetic unit 6 executes a specific information processing, a part or the whole thereof may be acquired from another device via the communication path 4. Further, the predetermined data and the predetermined program may be read from the storage medium 32 via the drive 31 described later when the arithmetic unit 6 executes specific information processing.

The communication I / F 30 is, for example, a wired LAN module, a wireless LAN module, or the like, and is an interface for performing wired or wireless communication with another device via the communication path 4. In the present disclosure, the communication I / F is configured to receive predetermined data and the like from a service-using terminal 3 and other devices (not shown) described later via the communication path 4.

The maintenance support device 2 may further include an input device 33 and a display device (output device) 34. As the input device 33, a mouse, a keyboard, a touch panel, or the like can be used. Further, as the display device (output device) 34, a display, a speaker, or the like can be used. Further, the maintenance support device 2 may further include a drive 31 for reading data and programs stored in the storage medium 32 such as a CD drive and a DVD drive.

In the probability information storage area, probability information 7 used for maintenance support processing described later is stored. In the present disclosure, the probability information 7 is a candidate value of a reference wall thickness, a numerical range in which a predicted value of the wall thickness when corrosion occurs can be taken, and a probability as an index indicating the degree of necessity of maintenance or repair of a structure. ,including.

The standard wall thickness is the wall thickness (plate thickness) that the structure to be maintained should satisfy, and is a parameter determined in advance before the calculation of the risk in the maintenance support process. In the present disclosure, the reference wall thickness is set as one or more candidate values. A plurality of candidate values are set so as to be different from each other. The candidate value of the reference wall thickness can be determined based on, for example, the required wall thickness or the controlled wall thickness described later.

The required wall thickness is, for example, a value of the wall thickness defined by laws and regulations and standards to be complied with. Examples of laws and regulations and standards to be complied with include laws and regulations and standards in which the wall thickness required for a structure is determined according to the pressure inside the structure during operation. The control wall thickness is a wall thickness arbitrarily set by the maintenance entity that manages the structure of the plant. The required wall thickness may be used as the control wall thickness, or may be set to a value larger than the required wall thickness by adding a certain value to the required wall thickness, multiplying by a safety factor, or the like. The controlled wall thickness is, for example, the required wall thickness that causes ductile fracture, or the wall thickness specified by the plant manager on the safe side (thick) than the required wall thickness in order to determine whether or not to repair the structure. Thickness can be adopted.

As the probability information 7, the probability information storage area has a plurality of predetermined numerical ranges b1, b2, b3, b4, b5, b6, b7, and b8 as the probability information 7, as shown in FIG. , A2 and memorized in association with it. In the present disclosure, the plurality of numerical ranges b1, b2, b3, b4, b5, b6, b7, and b8 are predicted values (predicted margin wall thickness) of the wall thickness (margin wall thickness) remaining even after the progress of corrosion under the heat insulating material. ), But not limited to this. The prediction margin wall thickness is calculated using a prediction model generated from the measurement data collected in advance. Here, in the present disclosure, each of the plurality of numerical ranges b1, b2, b3, b4, b5, b6, b7, and b8 are values set using the same prediction model. The method of calculating the predicted margin wall thickness will be described later with reference to FIG. 10 and the like.

In FIG. 3, “0 mm” and “1 mm” are used as the candidate values a1 and a2 for the reference wall thickness. Further, the numerical ranges b1 to b8 are numerical ranges that divide the range in which the predicted value of the margin thickness shown in FIG. 10 can be taken.

Further, in the probability information storage area, an index indicating the degree of necessity of maintenance and repair of the structure is stored. In the present disclosure, the probability information storage area maintains and repairs a plurality of probability indexes c1, c2, c3, c4, c5, c6, c7, and c8 indicating the probability that the structure becomes unusable due to corrosion under the heat insulating material. It is memorized as an index showing the degree of necessity of. The probability indexes c1, c2, c3, c4, c5, c6, c7, and c8 are associated with the reference wall thickness and a plurality of numerical ranges and are stored.

As the probability indexes c1 to c8, for example, ranks (for example, "A", "B", "C", "D") indicating the probability that the structure becomes unusable due to corrosion under the heat insulating material can be adopted. Rank "A" indicates a probability of "less than 0.1%", rank "B" indicates a probability of "0.1% or more and less than 1%", and rank "C" indicates a probability of "1% or more and less than 10%". It is assumed that the rank "D" indicates the probability "10% or more". That is, it is shown that the probability that the structure becomes unusable increases in order from the rank "A" to the rank "D" among the ranks "A", "B", "C", and "D".

Further, the plurality of numerical values ranges b1 to b8 and the plurality of probability indexes c1 to c8 are associated with each other. In FIG. 3, the numerical range b1 is associated with the probability index c1, the numerical range b2 is associated with the probability index c2, the numerical range b3 is associated with the probability index c3, and the numerical range b4 is associated with the probability index. c4 is associated, the numerical range b5 is associated with the probability index c5, the numerical range b6 is associated with the probability index c6, the numerical range b7 is associated with the probability index c7, and the numerical range b8 is associated with the probability index c8. It is associated.

In the maintenance support device 2 according to the present disclosure, as shown in FIG. 3, ranks (“A”, “B”, “C”, “D”) indicating the probability of becoming unusable are assigned as probability indexes c1 to c8. An example of use is shown, but other configurations can also be adopted. For example, as shown in FIG. 4, as the probability indexes c1 to c8, numerical values indicating the probability of becoming unusable (for example, "less than 0.1%", "0.1% or more and less than 1%", "1% or more 10""Lessthan%" and "10% or more") may be used.
Further, the numerical ranges b1 to b8 are set by the numerical range setting process described later.

According to this, the large value of the candidate value of the standard wall thickness means that the conditions requiring maintenance and repair are strictly set for the structure to be maintained, in other words, it is set on the safer side. Show that you are doing. As a result, even when the predicted value of the margin wall thickness (predicted margin wall thickness) is relatively large, the probability of becoming unusable is set to be equal to or higher than that. For example, as shown in FIGS. 3 and 4, the candidate value a1 (“0”) of the reference wall thickness and the candidate value a2 (“1”) of the reference wall thickness are under the condition that the candidate value a2 is stricter. is there. Here, the numerical range b4 of the predicted margin wall thickness of the candidate value a1 is "0 to 2 mm", and the numerical range b8 of the predicted margin wall thickness of the candidate value a2 is "0 to 4 mm", and any numerical value. Rank D (10% or more) is associated with the range as the probability index. That is, the candidate value a2 of the reference wall thickness sets the conditions requiring maintenance and repair relatively stricter than the candidate value a1, and thus the numerical value of the predicted margin wall thickness corresponding to the candidate value a1. Rank D (10% or more), which is the same probability index, is associated with the numerical range b8 containing a value larger than the range b4.

In the measurement data storage area, as shown in FIG. 5, the measured value d of the wall thinning depth due to corrosion under the heat insulating material of the structure of the plant, the attributes e1, e2, ... Of the structure, and the initial stage of the structure. A large number of measurement data 10 including the plate thickness f are stored. The attribute of the structure is some information indicating the characteristics of the structure of the plant, and is information that is a factor that changes the degree of corrosion (probability of occurrence, speed of progress, depth of corrosion, etc.). The attributes e1, e2, ... Of the structure include, for example, numerical attributes such as the period of use and temperature of the structure, the type of the structure, the environment (distance from the sea, distance from the cooling tower, annual rainfall, etc.). At least one of qualitative attributes such as the presence or absence of coating, the type of coating, and the type of heat insulating material can be used. Further, as the initial plate thickness f, for example, the measured wall thickness of the structure before being corroded under the heat insulating material can be used. Further, for example, the nominal wall thickness of the structure described in the design drawing can be adopted.

The measurement data 10 is data acquired at the time of inspection of various plants. After the start of information processing using the measurement data 10 described later, it is possible to perform inspections at various plants in order to acquire the measurement data 10. However, it is often difficult to collect a large amount of data in a short period of time by inspection at various plants. Therefore, it is preferable to collect the data acquired at the time of the past inspection in various plants in advance and store it in the storage device 5 before starting the information processing described later.
Further, the measurement data storage area is configured to collect measurement data 10 from the service use terminal 3 or the like at any time and continuously increase the total number of the stored measurement data 10.
The maintenance support program that can be executed by the processor is stored in the program storage area. Further, as described above, the storage device 5 stores various data necessary for executing the maintenance support program.

The arithmetic unit 6 uses the maintenance support program for outputting the probability information stored in the program storage area and the data 16 of the structure transmitted from the service use terminal 3 and being the object of the maintenance support. Execute the process. As a result, the acquisition unit 11, the wall thinning depth prediction unit 12, the margin thickness calculation unit 13, the probability index acquisition unit 14, the output unit 15, and the like are realized. As described above, the data 16 is not only received from another device such as the service use terminal 3 via the communication path 4, but is also stored in the maintenance support device 2 by using, for example, a drive 31. It may be read from, or may be directly input using, for example, an input device 33.

The data 16 of the structure transmitted from the service use terminal 3 and being the object of maintenance support and for determining whether or not the structure becomes unusable is the data related to the attributes of the maintenance object. including. Therefore, unlike the measurement data 10, the data 16 is data of a structure whose wall thinning depth is unknown. For the data 16, for example, as shown in FIG. 6, an information group including the attributes g1, g2, ... Of the structure of the plant, the initial plate thickness h, and the control wall thickness i as the reference wall thickness can be adopted. As the control wall thickness i, for example, “0 mm” can be used when there is almost no internal pressure in the structure of the plant. Further, for example, when there is an internal pressure of the structure, "1 mm" can be used as the control wall thickness.

The arithmetic unit 6 executes the probability information output process in cooperation with other hardware resources such as the storage device 5 and the communication I / F30, so that the acquisition unit 11, the wall thinning depth prediction unit 12, and the margin thickness The calculation unit 13, the probability index acquisition unit 14, the output unit 15, and the like are realized. The details of the probability information output process will be described later.

Further, the arithmetic unit 6 executes the numerical range setting process by using the instruction for setting / updating the numerical ranges b1 to b8 shown in FIG. 3 and the maintenance support program for setting the numerical ranges b1 to b8. As a result, the machine learning unit 17, the predicted wall thinning depth calculation unit 18, the predicted value calculation unit 19, the actual margin thickness calculation unit 20, the boundary value setting unit 21, and the like are realized. The service provider or the service user operates the input device 33 of the maintenance support device 2 or is connected to the maintenance support device 2 via the communication path 4 to instruct the setting / updating of the numerical ranges b1 to b8. It is done by operating other devices.

The arithmetic unit 6 executes the numerical range setting process in cooperation with other hardware resources such as the storage device 5 and the communication I / F30, so that the predicted wall thinning depth calculation unit 18 and the predicted value calculation unit 19 The actual margin wall thickness calculation unit 20, the boundary value setting unit 21, and the like are realized. The details of the numerical range setting process will be described later.

(Service terminal)
The service use terminal 3 is a terminal device used by a service user of the maintenance support service. As the service use terminal 3, for example, a personal computer, a smartphone, a tablet, or the like can be used. The service use terminal 3 has an input device 22 such as a keyboard, a mouse, and a touch panel, a display (display device) 23 for displaying various information, a storage device 24, and a processor (not shown) as an example of an output device for various information. A communication interface (hereinafter referred to as "communication I / F") 42 or the like that transmits / receives information to / from another device (for example, a service-using terminal 3 described later) via a computing device 25 and a communication path 4 (network). It has hardware resources.

The storage device 24 stores one or more programs that can be executed by the arithmetic unit 25. Further, the storage device 24 stores various data necessary for executing the program. The arithmetic unit 25 may have the same configuration as the arithmetic unit 6 provided in the maintenance support device 2. Further, the arithmetic unit 6 may be configured by using various processors according to the type of arithmetic processing and the like. Further, the storage device 24 may have the same configuration as the storage device 5 provided in the maintenance support device 2. Further, the communication I / F 42 may have the same configuration as the communication I / F provided in the maintenance support device 2. The service use terminal 3 may further include a drive 40 for reading data and programs stored in the storage medium 41 such as a CD drive and a DVD drive.

The arithmetic unit 25, in cooperation with other hardware resources, detects an operation requesting the service use terminal 3 to use the maintenance support service, reads a program from the storage device 24, and executes the program. As a result, the data transmission unit 26, the probability information display unit 27, the data reception unit 28, the acquisition unit 29, and the like are realized.

The acquisition unit 29 accepts the input of data 16 including the attributes g1, g2, ... Of the structure of the plant for which the probability indexes c1 to c8 are to be obtained, the initial plate thickness h, and the control wall thickness i as the reference wall thickness. The data 16 is input by the service user using the input device 22. The information on the reference wall thickness included in the data 16 can be obtained by the maintenance entity of the structure of the plant for which the probability indexes c1 to c8 are to be obtained. Information about (referred to as "thickness").

The data transmission unit 26 transmits the input data 16 to the maintenance support device 2 via the communication I / F 42. The data 16 may be read from the storage medium 41 via the drive 40, for example.

The data receiving unit 28 receives the probability indexes c1 to c8 transmitted from the maintenance support device 2 via the communication I / F42.

As shown in FIG. 7, the probability information display unit 27 displays the received probability indexes c1 to c8 on the display 23. Thereby, the probability indexes c1 to c8 can be presented to the service user.

(Probability information output processing)
Next, the probability information output process executed by the acquisition unit 11, the wall thinning depth prediction unit 12, the margin thickness calculation unit 13, the probability index acquisition unit 14, the output unit 15, and the like will be described. As described above, in the process of outputting the probability information, the arithmetic unit 6 of the maintenance support device 2 executes specific information processing based on the predetermined data and the predetermined program, so that the acquisition unit 11 and the wall thinning depth are reduced. This is achieved by realizing the prediction unit 12, the margin thickness calculation unit 13, the probability index acquisition unit 14, the output unit 15, and the like.

As shown in FIG. 8, the process of acquiring data is performed in step S101. In step S101, the acquisition unit 11 acquires the data 16 (structure attributes g1, g2, ..., Initial plate thickness h, management wall thickness i) transmitted from the service use terminal 3.

Subsequently, the process proceeds to step S102 to calculate the predicted value of the wall thinning depth. In step S102, the wall thinning depth prediction unit 12 calculates a predicted value of the wall thinning depth due to corrosion under the heat insulating material of the structure based on the attributes g1, g2, ... Included in the data 16 acquired in step S101. As a method of calculating the predicted value of the wall thinning depth, for example, as shown in FIG. 9, the predicted value of the wall thinning depth due to corrosion under the heat insulating material is calculated based on the attributes g1, g2, ... Of the structure. A method using the model 50 can be used. As the prediction model 50, for example, a model derived by machine learning based on the measurement data 10 stored in the storage device 5 can be adopted.

Subsequently, the process proceeds to step S103 to calculate the predicted value of the margin thickness. In step S103, as shown in FIG. 10, the margin thickness calculation unit 13 subtracts the predicted value of the wall thinning depth calculated in step S102 from the initial plate thickness h included in the data 16 acquired in step S101. The result (hereinafter referred to as "predicted value of margin thickness") is calculated. FIG. 10 shows "initial plate thickness", "predicted value of wall thinning depth", and "predicted value of margin wall thickness" when the structure is a pipe.

Subsequently, the process proceeds to step S104 to acquire probability information. In step S104, the probability index acquisition unit 14 refers to the storage device 5 and specifies a numerical range including the predicted value of the margin thickness calculated in step S103 from the plurality of numerical ranges b1 to b8. Specifically, first, as shown in FIG. 11, the candidate value a1 (or a2) of the same reference wall thickness as the control wall thickness i as the reference wall thickness included in the data 16 acquired in step S101 from the storage device 5). A plurality of numerical ranges b1 to b4 (or b5 to b8) and a plurality of probability indexes c1 to c4 (or c5 to c8) corresponding to the above are read out (step S201). Subsequently, the plurality of read numerical ranges b1 to b4 (or b5 to b8) and the plurality of probability indexes c1 to c4 (or c5 to c8) are referred to, and the plurality of numerical ranges b1 to b4 (or b5 to b8) are referred to. ), The numerical range b1 (or any of b2 to b8) including the predicted value of the margin thickness calculated in step S103 is specified (step S202). Subsequently, the probability index c1 (or any of c2 to c8) associated with the specified numerical range b1 (or any of b2 to b8) is acquired (step S203). FIG. 11 shows a case where the controlled wall thickness i as the reference wall thickness included in the data 16 is “0 mm” and the predicted value of the margin wall thickness is “3 mm”.

Subsequently, the process proceeds to step S105 to output probability information. In step S105, the output unit 15 outputs (transmits) the probability index c1 (or any of c2 to c8) acquired in step S104 to the service-using terminal 3 that is the source of the data 16, and then outputs the probability information. End the process. As described above, the output of the probability information may be performed on the output device 34 of the maintenance support device 2.

(Numerical range setting process)
Next, the numerical range setting process executed by the machine learning unit 17, the predicted wall thinning depth calculation unit 18, the predicted value calculation unit 19, the actual margin thickness calculation unit 20, the boundary value setting unit 21, and the like will be described. As described above, in the process of setting the numerical range, the arithmetic unit 6 of the maintenance support device 2 executes specific information processing based on the predetermined data and the predetermined program, so that the machine learning unit 17 and the predicted wall thinning are reduced. This is achieved by realizing the depth calculation unit 18, the predicted value calculation unit 19, the actual margin thickness calculation unit 20, the boundary value setting unit 21, and the like.
As shown in FIG. 12, a process of generating the prediction model 50 is performed in step S301. In the present disclosure, in step S301, the machine learning unit 17 generates a prediction model 50 based on a large number of measurement data 10 stored in the storage device 5. As a method of generating the prediction model 50, for example, a method based on a machine learning algorithm can be adopted.

Specifically, first, a machine learning algorithm is used with the attributes e1, e2, ... Of the structure included in the measurement data 10 as explanatory variables and the measured value d of the wall thinning depth included in the measurement data 10 as the objective variable. By doing so, the prediction model 50 that learned the relationship between the explanatory variable and the objective variable is derived. That is, the machine learning unit 17 is a set of the attributes e1, e2, ... Of the specific structure and the measured value d of the wall thinning depth of the specific structure associated with the attributes e1, e2, ... The prediction model 50 is generated by machine learning using training data (training data) including a plurality of. The training data can be generated based on the measurement data 10.

As the prediction model 50 (learner), it is preferable to perform machine learning using a model whose prediction accuracy is high and whose source code is open to the public. For example, a neural network can be used as the prediction model 50. A neural network, which is an example of a learner, includes an input layer, an intermediate layer (hidden layer), and an output layer in this order from the input side. The intermediate layer is not limited to one layer, and may include two or more intermediate layers.

Each layer comprises one or more neurons, respectively. For example, the number of neurons in the input layer can be appropriately set according to the number of types of information indicating the attributes e1 and e2 of the structure input to the learner. The values input to one or more neurons in the input layer are appropriately set according to the types of attributes e1, e2, .... For example, when the attributes e1, e2, ... Are attributes expressed numerically such as the period of use of the structure and the operating temperature, the numerical values can be input to the neurons of the input layer. If the attributes e1, e2, ... Are attributes that are not expressed numerically, such as the type of structure and environment, they are quantified according to a predetermined rule and the numerical values are input to the neurons in the input layer. be able to. For example, attributes e1, e2, ... Indicates the type of structure, and when the type of structure includes piping, heat exchanger, tower, and tank, different numerical values are assigned to each of piping, heat exchanger, tower, and tank. Correspondingly, the numerical value is input to the neurons in the input layer. For example, the number of pipes is 1, the number of heat exchangers is 2, the number of towers is 3, and the number of tanks is 4. The number of neurons in the intermediate layer can be appropriately set according to the prediction accuracy of the predicted value of the wall thinning depth to be output by the prediction model 50, the processing speed related to the prediction, the time required for learning the prediction model 50, and the like. .. The output layer may be composed of, for example, a single neuron that outputs a numerical value of the predicted value of the thinning depth. Further, a plurality of neurons may be provided in the output layer. For example, each neuron may be associated with a numerical range of different wall thicknesses, and each neuron may output a probability of being the associated numerical range. In this case, the prediction model 50 may be configured to output an average value calculated from the numerical range and probability of the wall thinning depth associated with each neuron.

A threshold is set for each neuron, and the output of each neuron is basically determined by whether or not the sum of the products of each input and each weight exceeds the threshold. The machine learning unit 17 reads out a set of the attributes e1, e2, ... And the measured value d of the wall thinning depth included in the training data. When the attribute is input, the machine learning unit 17 trains the learner to output an output value corresponding to the measured value d of the wall thinning depth, and generates a prediction model 50.

Information indicating the configuration of such a neural network (for example, the number of layers of the neural network, the number of neurons in each layer, the connection relationship between neurons, the transmission function of each neuron), the weight of the connection between each neuron, and the threshold value of each neuron. Is stored in the storage device 5 as learning result data.

Subsequently, the process proceeds to step S302 to calculate the predicted wall thinning depth. In the present disclosure, in step S302, the predicted wall thinning depth calculation unit 18 uses the prediction model 50 for each measurement data 10 stored in the storage device 5, and the attributes e1, e2, which are included in the measurement data 10. The predicted value of the wall thinning depth (hereinafter, referred to as “predicted wall thinning depth”) is calculated based on the above. Specifically, first, the predicted wall thinning depth calculation unit 18 reads out one measurement data 10 stored in the measurement data storage area. Subsequently, the predicted value (predicted wall thinning depth) of the wall thinning depth is calculated using the prediction model 50 based on the attributes e1, e2, ... Included in the read measurement data 10. Then, the flow of “reading the measurement data 10” → “calculating the predicted wall thinning depth” is repeatedly executed for all the remaining measurement data 10.

Subsequently, the process proceeds to step S303 to calculate the predicted value of the margin wall thickness. In the present disclosure, in step S303, the predicted value calculation unit 19 starts from the initial plate thickness f included in the measurement data 10 for each measurement data 10 stored in the measurement data storage area, as shown in FIG. The predicted value of the margin thickness (hereinafter referred to as "predicted margin thickness"), which is the subtraction result obtained by subtracting the predicted wall thickness reduction corresponding to the measurement data 10, is calculated. Specifically, first, one measurement data 10 stored in the measurement data storage area is read out. Subsequently, the predicted wall thinning depth calculated in step S302 based on the read data (measurement data 10) from the initial plate thickness f included in the read measurement data 10 (hereinafter, referred to as “read data”). Is subtracted to calculate the predicted margin wall thickness. Then, the flow of “reading the measurement data 10” → “calculating the predicted margin wall thickness” is repeatedly executed for all the remaining measurement data 10. FIG. 13 shows "predicted wall thinning depth", "predicted margin wall thickness", and the like when the structure is a pipe.

Subsequently, the process proceeds to step S304 to calculate the actual value of the margin thickness calculated based on the measured value. In the present disclosure, in step S304, the actual margin wall thickness calculation unit 20 starts from the initial plate thickness f included in the measurement data 10 for each measurement data 10 stored in the measurement data storage area, as shown in FIG. , The actual value of the margin wall thickness (hereinafter, referred to as "actual margin wall thickness"), which is the subtraction result of subtracting the measured value d of the wall thinning depth included in the measurement data 10, is calculated. Specifically, first, the actual margin wall thickness calculation unit 20 reads out one measurement data 10 stored in the measurement data storage area. Subsequently, the actual margin wall thickness is calculated by subtracting the measured value d of the wall thinning depth included in the measurement data 10 from the initial plate thickness f included in the read measurement data 10. Then, the flow of “reading the measurement data 10” → “calculating the actual margin wall thickness” is repeatedly executed for all the remaining measurement data 10. FIG. 14 shows "measured value of wall thinning depth", "actual margin wall thickness", etc. when the structure is a pipe.

Subsequently, the process proceeds to step S305 to set a plurality of numerical ranges and probability information. In the present disclosure, in step S305, the boundary value setting unit 21 sets a plurality of numerical values b1 to b8 stored in the probability information storage area. As described above, in the present disclosure, the plurality of numerical ranges b1 to b8 indicate the numerical range of the predicted margin wall thickness. Specifically, first, for each of the candidate values a1 and a2 of the reference wall thickness, the predicted margin meat included in the numerical range b1 (or any of b2 to b8) for each of the plurality of numerical ranges b1 to b8. The ratio of the actual margin wall thickness corresponding to the thickness to include the actual margin wall thickness of the candidate value a1 (or a2) or less is associated with the numerical range b1 (or either b2 to b8). Same as the probability indicated by the probability index c1 (or any of c2 to c8) ("less than 0.1%", "0.1% or more and less than 1%", "1% or more and less than 10%", "10% or more", etc.) A boundary value for dividing a plurality of numerical values ranges b1 to b8 is set so as to have a numerical value of.

Here, as shown in FIG. 15 (d), a numerical range b1 (hereinafter referred to as "first numerical range b1") and a numerical range b2 (hereinafter referred to as "second numerical range b2"). First boundary value j between and, second boundary value k between the second numerical range b2 and the numerical range b3 (hereinafter, referred to as "third numerical range b3"), third The procedure for setting the third boundary value l between the numerical range b3 and the numerical range b4 (hereinafter, referred to as “fourth numerical range b4”) will be described. First, a tentative numerical value is set as the first boundary value j. Subsequently, as shown in FIG. 15A, the prediction margin included in the first numerical range b1 determined based on the temporarily set first boundary value j from the prediction margin wall thickness calculated in step S303. Identify the wall thickness. Subsequently, from the actual margin wall thickness calculated in step S304, the actual margin wall thickness corresponding to the specified predicted margin wall thickness is specified.

As a method for specifying the actual margin wall thickness, for example, the actual margin wall thickness is calculated from the CUI thinning depth under the condition in which the predicted margin wall thickness is calculated. Subsequently, the ratio of the specified actual margin wall thickness including the actual margin wall thickness of the selected control wall thickness (0 mm) or less is calculated. Subsequently, in consideration of the calculation result, the first boundary value j is changed so that the calculated ratio approaches "0.1%". In other words, as shown in FIG. 16, when the cumulative probability of the specified actual margin wall thickness is represented by the approximate straight line 51, the actual margin wall thickness on the approximate straight line 51 is the same value as the selected control wall thickness (0 mm). The first boundary value j is changed so that the value of the cumulative probability when is close to "0.1%". Then, the flow of "specifying the predicted margin wall thickness"-> "specifying the actual margin wall thickness"-> "changing the boundary value" is repeatedly executed, and the calculated ratio becomes "0.1%". Find the boundary value j of 1.

Subsequently, a tentative numerical value is set as the second boundary value k. Subsequently, as shown in FIG. 15B, the second boundary value j determined from the predicted margin wall thickness calculated in step S303 and the second boundary value k tentatively set are determined. The predicted margin wall thickness included in the numerical range b2 of is specified. Subsequently, from the actual margin wall thickness calculated in step S304, the actual margin wall thickness corresponding to the specified predicted margin wall thickness is specified. Subsequently, the ratio of the specified actual margin wall thickness including the actual margin wall thickness of the selected control wall thickness (0 mm) or less is calculated. Subsequently, the second boundary value k is changed so that the calculated ratio approaches "1%" in consideration of the calculation result. In other words, as shown in FIG. 16, when the cumulative probability of the specified actual margin wall thickness is represented by the approximate straight line 52, the actual margin wall thickness on the approximate straight line 52 is the same value as the selected control wall thickness (0 mm). The second boundary value k is changed so that the value of the cumulative probability when is close to "1%". Then, the flow of "specifying the predicted margin wall thickness"->"specifying the actual margin wall thickness"->"changing the boundary value" is repeatedly executed, and the calculated ratio becomes "1%". Find the boundary value k.
In addition, in FIG. 16, the reason why the actual margin wall thickness includes a negative value is that the nominal wall thickness is input as the initial plate thickness f, but the wall thickness of the actual structure is thicker than the nominal wall thickness. it is conceivable that.

Subsequently, a tentative numerical value is set as the third boundary value l. Subsequently, as shown in FIG. 15C, a third boundary value k determined based on the second boundary value k and the tentatively set third boundary value l from the predicted margin wall thickness calculated in step S303. The predicted margin wall thickness included in the numerical range b3 of is specified. Subsequently, from the actual margin wall thickness calculated in step S304, the actual margin wall thickness corresponding to the specified predicted margin wall thickness is specified. Subsequently, the ratio of the specified actual margin wall thickness including the actual margin wall thickness of the selected control wall thickness (0 mm) or less is calculated. Subsequently, in consideration of the calculation result, the third boundary value l is changed so that the calculated ratio approaches "10%". In other words, as shown in FIG. 16, when the cumulative probability of the specified actual margin wall thickness is represented by the approximate straight line 53, the actual margin wall thickness on the approximate straight line 53 is the same value as the selected control wall thickness (0 mm). The third boundary value l is changed so that the value of the cumulative probability when is close to "10%". Then, the flow of "specifying the predicted margin wall thickness"-> "specifying the actual margin wall thickness"-> "changing the boundary value" is repeatedly executed, and the calculated ratio is "10%". Find the value l.

By such a procedure, as shown in FIG. 15D, the first boundary value j, the second boundary value k, and the third boundary value l that divide the numerical ranges b1 to b4 can be set. ..
The boundary values in the numerical ranges b5 to b8 can also be set by using the same procedure.
Subsequently, the boundary value setting unit 21 sets the numerical values ranges b1 to b8 based on the set boundary values and stores them in the storage device 5, and then ends the numerical range setting process. When the numerical range b1 to b8 is already stored in the storage device 5, the newly obtained numerical range b1 to b8 is overwritten with the numerical range b1 to b8 stored in the storage device 5.

In other words, the process of setting the numerical range and the probability index shown in step 305 can be expressed as follows.

Before the process of setting the numerical range and the probability information, as shown in steps S301 to S304, the actual margin wall thickness and the predicted margin wall thickness are calculated for the measurement data 10. It should be noted that the calculation of the actual margin wall thickness and the predicted margin wall thickness does not have to be performed for all the measurement data 10, and the actual margin wall thickness is obtained for the measurement data 10 in an amount sufficient to generate the probability information 7. The thickness and the predicted margin wall thickness may be calculated. For example, the actual margin wall thickness and the predicted margin wall thickness can be calculated for 1000 measurement data 10, and the probability information 7 can be generated based on the calculated actual margin wall thickness and the predicted margin wall thickness.

A data group including a plurality of sets of the calculated actual margin wall thickness and the predicted margin wall thickness is divided into a plurality of data groups according to the numerical range of the predicted margin wall thickness so that a predetermined condition is satisfied. The plurality of numerical ranges that divide the data group are set as a plurality of numerical ranges of the probability information 7. Here, the predetermined condition is a predetermined predetermined condition defined in the probability index as the ratio of the data in which the actual margin wall thickness is equal to or less than the candidate value of the standard wall thickness in the data group divided in a certain numerical range. To meet the criteria of. As described above, quantitative or qualitative criteria such as probability and rank can be appropriately adopted as predetermined criteria.

For example, in the first numerical range b1 for the predicted margin wall thickness, the actual margin wall thickness occupies the first data group classified by the first numerical range b1 and the actual margin wall thickness is a candidate value of the reference wall thickness (for example, 0 mm). ) The percentage of data below is set in a numerical range that serves as the first criterion (eg, less than 0.1%). As shown in FIG. 15, the first numerical range b1 is, for example, 8 mm or more.

Further, in the second numerical range b2 for the predicted margin wall thickness, the actual margin wall thickness occupies the second data group classified by the second numerical range b2, and the actual margin wall thickness is a candidate value of the reference wall thickness (for example, 0 mm). ) The ratio of the data below is set in a numerical range that serves as a second criterion (for example, 0.1% or more and less than 1%). As shown in FIG. 15, the second numerical range b2 is, for example, 5 mm or more and less than 8 mm.

Similarly, the third numerical range b3 and the fourth numerical range b4 are set. That is, the third numerical range b3 is the data in which the actual margin wall thickness is equal to or less than the candidate value (for example, 0 mm) of the reference wall thickness in the third data group classified by the third numerical range b3. The ratio is set in a numerical range that serves as a third criterion (for example, 1% or more and less than 10%). Further, the fourth numerical range d4 is the data in which the actual margin wall thickness is equal to or less than the candidate value (for example, 0 mm) of the reference wall thickness in the fourth data group classified by the fourth numerical range d4. The ratio is set in a numerical range that serves as a fourth criterion (for example, 10% or more). According to FIG. 15, the third numerical range d3 is 2 mm or more and less than 5 mm, and the fourth numerical range d4 is 0 mm or more and less than 2 mm.

As described above, a plurality of numerical ranges for the predicted margin wall thickness and a boundary value that defines the numerical range are set. This is done for one or more reference wall thickness candidate values.

That is, in the present disclosure, as an index indicating the high possibility that the plant structure becomes unusable, the ratio of data in which the actual margin wall thickness is equal to or less than the candidate value of the reference wall thickness is used as a probability index. ..

Here, it is preferable that the plurality of numerical ranges are set so that the index indicating the high possibility that the plant structure becomes unusable increases (or decreases) in order. As described above, the predetermined standard defined in the probability index is not particularly limited in the type of the numerical value and the display format. For example, as shown in FIG. 3, a predetermined criterion may be displayed as a qualitative rank, or as shown in FIG. 4, it may be displayed as a probability. Further, in the present disclosure, the data group is divided into four numerical ranges, but the number thereof is not particularly limited. Further, the values of adjacent numerical ranges and probability indexes may partially overlap.

(Operation and others)
Next, the operation at the time of providing the maintenance support service in the maintenance support system 1 according to the present disclosure will be described. Hereinafter, an example in which the controlled wall thickness i is used as the reference wall thickness is shown.
First, the service user operates the service-using terminal 3, and as shown in FIG. 17, the service-using terminal 3 passes through the communication path 4 and the data 16 (attributes g1, g2, ..., It is assumed that the initial plate thickness h and the control wall thickness i) are transmitted to the maintenance support device 2 (step S401). Then, the maintenance support device 2 receives the transmitted data 16 (step S402), and based on the attributes g1, g2, ... Included in the acquired data 16, the prediction model 50 is used under the heat insulating material of the structure. The predicted value of the wall thinning depth due to corrosion is calculated (step S403). Subsequently, as shown in FIG. 10, the maintenance support device 2 subtracts the calculated predicted value of the wall thinning depth from the initial plate thickness h included in the acquired data 16 to calculate the predicted value of the margin wall thickness. (Step S404).

Subsequently, the maintenance support device 2 refers to the probability information storage area and specifies a numerical range including the calculated marginal wall thickness prediction value from the plurality of numerical ranges b1 to b8 shown in FIG. 3 (step). S405), the probability index associated with the specified numerical range is acquired (step S406). Here, as shown in FIG. 18A, when the management wall thickness i included in the data 16 is “0 mm” and the predicted value of the margin wall thickness is “3 mm”, the predicted value of the margin wall thickness “ The numerical range b3 is specified as the numerical range including "3 mm", and the rank "C" which is the probability index c3 associated with the numerical range b3 is acquired. On the other hand, as shown in FIG. 18B, when the management wall thickness i included in the data 16 is "1 mm" and the predicted value of the margin wall thickness is "3 mm", the predicted value of the margin wall thickness is "3 mm". The numerical range b8 is specified as the numerical range including the above, and the rank “D” which is the probability index c8 associated with the numerical range b8 is acquired. That is, different ranks (ranks "C", "D", etc.) are acquired according to the management wall thickness i included in the data 16. Subsequently, the maintenance support device 2 outputs (transmits) the acquired probability index to the service use terminal 3 that is the source of the data 16 via the communication path 4 (step S407).

Subsequently, the service use terminal 3 receives the transmitted probability index (step S408), and displays the received probability index on the display 23 (step S409). As a result, the service user can know the probability that the structure becomes unusable due to the corrosion under the heat insulating material, and can more appropriately determine the maintenance method, inspection specifications, etc. of the structure of the plant.

Next, the operation at the time of setting the prediction model 50 and the numerical ranges b1 to b8 in the maintenance support system 1 according to the present disclosure will be described.
First, the service provider operates the maintenance support device 2, the maintenance support device 2 executes the numerical range setting process, and as shown in FIG. 19, a large number of measurement data 10 stored in the measurement data storage area. Based on the above, a prediction model 50 is derived by learning the relationship between the attributes e1, e2, ... And the measured value d of the wall thinning depth using a machine learning algorithm (step S501). Subsequently, the maintenance support device 2 uses the derived prediction model 50 to store the attributes e1, e2, ... The predicted value of the wall thinning depth (predicted wall thinning depth) is calculated based on (step S502). Subsequently, the maintenance support device 2 corresponds to the measurement data 10 from the initial plate thickness f included in the measurement data 10 for each measurement data 10 stored in the measurement data storage area, as shown in FIG. The subtraction result (predicted margin wall thickness) obtained by subtracting the predicted wall thinning depth is calculated (step S503).

Subsequently, for each measurement data 10 stored in the measurement data storage area, the maintenance support device 2 starts with the initial plate thickness f included in the measurement data 10 and the measured value d of the wall thinning depth included in the measurement data 10. The subtraction result (actual margin wall thickness) obtained by subtracting is calculated (step S504). Subsequently, the maintenance support device 2 sets a boundary value for dividing the plurality of numerical values ranges b1 to b8 for each of the management wall thickness candidate values a1 and a2 (step S505), and a numerical value is based on the set boundary value. The ranges b1 to b8 are set and stored in the probability information storage area (step S506). As a result, the prediction model 50 and the numerical ranges b1 to b8 can be updated as the measurement data 10 increases, and the service user can be notified of more accurate probability information.

Although an example of the operation when the prediction model 50 and the numerical range b1 to b8 are set is shown, for example, the same operation may be performed when the prediction model 50 and the numerical range b1 to b8 are updated. The prediction model 50 and the numerical ranges b1 to b8 are updated, for example, when the number of stored measurement data 10 increases by a predetermined number (for example, 100) from the time of the previous update.

As described above, in the maintenance support device 2 according to the present disclosure, a plurality of numerical ranges b1 to b8 and a plurality of probability indexes c1 to c8 indicating the probability that the structure becomes unusable due to corrosion under the heat insulating material are provided. With reference to the probability information storage area stored in association with each other, a numerical range including the calculated estimated value of the margin wall thickness is specified from a plurality of numerical ranges b1 to b8, and the numerical range is associated with the specified numerical range. The probability index acquisition unit 14 for acquiring the acquired probability index and the output unit 15 for outputting the acquired probability index are provided. Therefore, it is possible to provide the maintenance support device 2 capable of presenting the probability that the structure of the plant becomes unusable. Then, by using the maintenance support device 2, it is possible to know the probability that the structure becomes unusable due to corrosion under the heat insulating material, and it is possible to more appropriately determine the maintenance method, inspection specifications, etc. of the structure of the plant.

Similarly, it is possible to provide a maintenance support program that can present the probability that the structure of the plant becomes unusable and can more appropriately determine the maintenance method and inspection specifications of the structure of the plant.
Further, in the maintenance support device 2 according to the present disclosure, the data 16 includes, for example, the control wall thickness i as the reference wall thickness. Further, in the probability information storage area, a plurality of numerical values ranges b1 to b8 and a plurality of probability indexes c1 to c8 are stored for each of the candidate values a1 and a2 of the plurality of management thicknesses. Further, the probability index acquisition unit 14 corresponds to the candidate value a1 (or a2) of the management wall thickness same as the management wall thickness i included in the data 16 acquired by the acquisition unit 11 stored in the probability information storage area. A numerical value including a predicted value of the margin wall thickness calculated by the margin wall thickness calculation unit 13 from the plurality of numerical value ranges b1 to b8 with reference to a plurality of numerical value ranges b1 to b8 and a plurality of probability indexes c1 to c8. Specify the range b1 (or any of b2 to b8) and acquire the probability index c1 (or any of c2 to c8) associated with the specified numerical range b1 (or either b2 to b8). I did. Therefore, it is possible to output a probability index according to the management wall thickness i of the structure.

Therefore, for example, when the internal pressure of the structure is low and the data 16 includes "0 mm" as the control wall thickness i, a probability index that the wall thickness of the structure becomes "0 mm" can be presented. Further, for example, when the internal pressure of the structure is high and the data 16 includes, for example, "1 mm" as the control wall thickness i, a probability index that the wall thickness of the structure becomes "1 mm" can be presented. Therefore, it is possible to know the probability that the structure becomes unusable due to corrosion under the heat insulating material with higher accuracy, and it is possible to more appropriately determine the maintenance method, inspection specifications, etc. of the structure of the plant.

Further, in the maintenance support device 2 according to the present disclosure, the wall thinning depth prediction unit 12 uses the prediction model 50 that calculates the predicted value of the wall thinning depth based on the attributes g1, g2, ... The predicted value of the wall thinning depth is calculated based on the attributes g1, g2, ... Included in the data 16 acquired in. In addition, a large number of measurement data 10 including the measured value d of the wall thinning depth due to corrosion under the heat insulating material of the structure of the plant, the attributes e1, e2, ... Of the structure and the initial plate thickness f of the structure are stored. A data storage area is provided. Further, using the prediction model 50, the predicted wall thinning depth (predicted wall thinning depth) for calculating the predicted value of the wall thinning depth (predicted wall thinning depth) based on the attributes e1, e2, ... Included in the measurement data 10 for each measurement data 10. The calculation unit 18 is provided. Further, for each measurement data 10, a prediction value calculation unit that calculates a subtraction result (prediction margin wall thickness) obtained by subtracting the predicted wall thinning depth corresponding to the measurement data 10 from the initial plate thickness f included in the measurement data 10. 19 is provided. Further, for each measurement data 10, the actual margin for calculating the subtraction result (actual margin wall thickness) obtained by subtracting the measured value d of the wall thinning depth included in the measurement data 10 from the initial plate thickness f included in the measurement data 10. A wall thickness calculation unit 20 is provided.

Further, for each of the candidate values a1 and a2 of the management wall thickness, for each of the plurality of numerical values ranges b1 to b8, the predicted margin wall thickness included in the numerical range b1 (or any of b2 to b8) is supported. Probability index c1 (or c2-) in which the ratio of the actual margin wall thickness that includes the actual margin wall thickness equal to or less than the control wall thickness is associated with the numerical range b1 (or either b2 to b8). A boundary value setting unit 21 for setting a boundary value for dividing a plurality of numerical values ranges b1 to b8 is provided so that the probability is the same as the probability indicated by any of c8). Therefore, for example, the numerical ranges b1 to b8 can be appropriately set, and a highly accurate probability index can be presented to the service user.

Further, in the maintenance support device 2 according to the present disclosure, the attributes e1, e2, ... Of the structure included in the measurement data 10 stored in the storage device 5 are used as explanatory variables, and the wall thinning depth included in the measurement data 10 is measured. A machine learning unit 17 for deriving a prediction model 50 that has learned the relationship between the explanatory variable and the objective variable is provided by using a machine learning algorithm with the value d as the objective variable. Therefore, the maintenance support device 2 (computer) can derive the prediction model 50, and the prediction model 50 with higher accuracy can be obtained with relatively little effort.
Further, in the maintenance support device 2 according to the present disclosure, the rank indicating the probability that the structure becomes unusable and the numerical value indicating the probability are used as the probability indexes c1 to c8. For example, by using the ranks "A", "B", "C", and "D" indicating the probability, the probability can be grasped more intuitively. Further, for example, the probability can be grasped more accurately by using the numerical values "less than 0.1%", "0.1% or more and less than 1%", "1% or more and less than 10%", and "10% or more" indicating the probability. ..

(Modification example)
(1) In the above embodiment, an example is shown in which the maintenance support system 1 including the maintenance support device 2 and the service use terminal 3 is configured and the probability indexes c1 to c8 are presented to the service users. The configuration can also be adopted. For example, it may be configured to perform various processes in one maintenance support device 2 and present the probability indexes c1 to c8 to the service user.
(2) Further, in the above embodiment, an example of using a model derived by machine learning from the attributes e1, e2, ... And the measured value d of the wall thinning depth is shown as the prediction model 50, but other configurations are shown. Can also be adopted. For example, a model derived by a statistical method may be used.

1 ... maintenance support system, 2 ... maintenance support device, 3 ... service use terminal, 4 ... communication path, 5 ... storage device, 6 ... arithmetic device, 7 ... probability information, 9 ... program, 10 ... measurement data, 11 ... acquisition Unit, 12 ... Wall thinning depth prediction unit, 13 ... Margin wall thickness calculation unit, 14 ... Probability index acquisition unit, 15 ... Output unit, 16 ... Data, 17 ... Machine learning unit, 18 ... Predicted wall thickness reduction depth calculation unit , 19 ... Predicted value calculation unit, 20 ... Actual margin thickness calculation unit, 21 ... Boundary value setting unit, 22 ... Input device, 23 ... Display (output device), 24 ... Storage device, 25 ... Calculation device, 26 ... Data Transmission unit, 27 ... Probability information display unit, 28 ... Data reception unit, 29 ... Acquisition unit, 30 ... Communication I / F, 31 ... Drive, 32 ... Storage medium, 33 ... Input device, 34 ... Output device, 35 ... Learning Result data, 40 ... drive, 41 ... storage medium, 42 ... communication I / F, 50 ... prediction model, 51 to 53 ... approximate straight lines, a1, a2 ... candidate values for management wall thickness, b1 to b8 ... numerical range, c1 ~ C8 ... Probability information, d ... Measured value of wall thinning depth, e1, e2 ... Attribute, f ... Initial plate thickness, g1, g2 ... Attribute, h ... Initial plate thickness, i ... Control wall thickness, j ... First Boundary value, k ... second boundary value, l ... third boundary value

Claims (7)

An acquisition unit that acquires data including the attributes and initial plate thickness of the plant structure,
Based on the attributes included in the data acquired by the acquisition unit, the wall thickness prediction unit that calculates the predicted value of the wall thickness reduction due to corrosion under the heat insulating material of the structure, and the wall thickness prediction unit.
The margin wall thickness for calculating the margin wall thickness prediction value, which is the subtraction result of subtracting the wall thickness reduction depth prediction value calculated by the wall reduction depth prediction unit from the initial plate thickness included in the data acquired by the acquisition unit. Calculation unit and
A probability information storage device that stores a plurality of predetermined numerical values in association with a plurality of probability indexes indicating the probability that the structure becomes unusable due to corrosion under a heat insulating material.
With reference to the probability information storage device, a numerical range including the predicted value of the margin thickness calculated by the margin thickness calculation unit is specified from the plurality of numerical ranges, and is associated with the specified numerical range. The probability index acquisition unit that acquires the probability index that exists,
A maintenance support device including an output unit that outputs a probability index acquired by the probability index acquisition unit. The data includes the required wall thickness or a wall thickness set to be larger than the required wall thickness as a reference wall thickness.
The probability information storage device stores the plurality of numerical values and the plurality of probability indexes for each of a plurality of predetermined reference wall thickness candidate values.
The probability index acquisition unit has a plurality of numerical values corresponding to the candidate values of the reference wall thickness which is the same as or closest to the reference wall thickness included in the data acquired by the probability information storage device. With reference to the range and a plurality of probability indexes, a numerical range including the predicted value of the margin thickness calculated by the margin thickness calculation unit is specified from the plurality of numerical ranges, and the numerical range is associated with the specified numerical range. The maintenance support device according to claim 1, wherein the probability index is acquired. The wall-thinning depth prediction unit uses a prediction model that calculates a predicted value of the wall-thinning depth based on the attributes, and uses the prediction model of the wall-thinning depth based on the attributes included in the data acquired by the acquisition unit. Calculate the predicted value,
A measurement data storage device that stores a large number of measurement data including the measured value of the wall thinning depth due to corrosion under the heat insulating material of the structure of the plant, the attributes of the structure, and the initial plate thickness of the structure.
Using the prediction model, for each measurement data, a predicted wall thinning depth calculation unit that calculates the predicted wall thinning depth, which is a predicted value of the wall thinning depth based on the attributes included in the measurement data,
For each measurement data, a prediction value calculation unit for calculating a prediction margin wall thickness, which is a subtraction result obtained by subtracting the predicted wall thinning depth corresponding to the measurement data from the initial plate thickness included in the measurement data.
For each measurement data, an actual margin wall thickness calculation unit that calculates the actual margin wall thickness, which is a subtraction result obtained by subtracting the measured value of the wall thinning depth included in the measurement data from the initial plate thickness included in the measurement data. ,
For each candidate value of the reference wall thickness, for each of the plurality of numerical values, the actual margin wall thickness corresponding to the predicted margin wall thickness included in the numerical range is equal to or less than the standard wall thickness. Boundary value setting that sets the boundary value that divides the plurality of numerical ranges so that the ratio including the actual margin wall thickness is the same as the probability indicated by the probability index associated with the numerical range. The maintenance support device according to claim 2, further comprising a unit. Using a machine learning algorithm, the above description will be made using the attributes of the structure included in the measurement data stored in the measurement data storage device as explanatory variables and the measured value of the wall thinning depth included in the measurement data as the objective variable. The maintenance support device according to claim 3, further comprising a machine learning unit for deriving the prediction model that has learned the relationship between the variable and the objective variable. The maintenance support device according to any one of claims 1 to 4, wherein the probability index is a rank indicating the probability or a numerical value indicating the probability. Computer,
Acquisition unit that acquires data including the attributes and initial plate thickness of the plant structure,
The wall thinning depth prediction unit, which calculates the predicted value of the wall thinning depth due to corrosion under the heat insulating material in the portion of the structure covered with the heat insulating material, based on the attributes included in the data acquired by the acquisition unit.
The margin wall thickness for calculating the margin wall thickness prediction value, which is the subtraction result of subtracting the wall thickness reduction depth prediction value calculated by the wall reduction depth prediction unit from the initial plate thickness included in the data acquired by the acquisition unit. Calculation unit,
Referencing the probability information storage device that stores a plurality of predetermined numerical values in association with a plurality of probability indexes indicating the probability that the structure becomes unusable due to corrosion under the heat insulating material. From the numerical range of, the probability index acquisition unit that specifies the numerical range including the predicted value of the margin thickness calculated by the margin thickness calculation unit and acquires the probability index associated with the specified numerical range.
And a maintenance support program for functioning as an output unit that outputs the probability index acquired by the probability index acquisition unit. Acquiring data including the attributes and initial plate thickness of the plant structure,
Based on the attributes included in the acquired data, the predicted value of the wall thinning depth due to corrosion under the heat insulating material of the structure is calculated, and
To calculate the predicted value of the margin wall thickness, which is the subtraction result of subtracting the predicted value of the wall thinning depth from the initial plate thickness included in the acquired data.
Referencing the probability information stored in association with a plurality of predetermined numerical ranges and a plurality of probability indexes indicating the probability that the structure becomes unusable due to corrosion under the heat insulating material.
From the plurality of numerical ranges, the numerical range including the predicted value of the margin thickness is specified, and the probability index associated with the specified numerical range is acquired.
To output the probability index, including
Maintenance support method.

Images