JP2021135611A - Diversion design support system and diversion design support method - Google Patents

Abstract

To provide efficient diversion design.SOLUTION: A diversion design support system comprises: a search condition input part 110 to which a search condition is input, the search condition being information of a design item which will be designed; a random number generating part 121 for generating a random number on the basis of a value included in the search condition; a similarity determining part 122 for setting the random number to a threshold, comparing a value in a previous design item with the threshold, for narrowing previous design items stored in a previous design item database 201 as first similar items; a similarity score calculating part 130 for performing N times (N>1), processing performed by the random number generating part 121 and the similarity determining part 122, for calculating a similarity score obtained by digitizing, how many times each previous design item is narrowed as the first similar item, out of N times; and a similar item output part 140 for, on the basis of the similarity score, outputting prescribed previous design items as second similar items, to an output part 302.SELECTED DRAWING: Figure 1

Description

The present invention relates to a technique of a diversion design support system and a diversion design support method for diverting a project designed in the past.

When designing a manufacturing industry such as a plant, a similar project may be selected from past projects with a proven track record, and a diversion design may be performed to change the design based on the specifications of the project. By performing such a diversion design, it is possible to shorten the development lead time as compared with designing from scratch. In such a diversion design, a skilled designer selects the optimum similar project based on past experience and in consideration of many design specification values, constraints, and the like. At that time, there is a problem that the projects to be selected differ depending on the skill level of the designer who makes the selection and the difference in the priority of the items to be considered.

The techniques described in Patent Documents 1 to 3 are disclosed for such problems.
In Patent Document 1, "a primary search is performed to search for design actual data corresponding to the required specification actual data satisfying the specifications based on the required specification input data from the design case database (S4), and the above is described in accordance with a predetermined selection rule. A representative value is selected as a design proposal from the contents of the design record data obtained by the primary search (S6), and corresponds to other design record data having the same value as the selected representative value from the design case database. The required specification actual data is searched (S11), and each search result and the representative value are output together (S14). ”A design case search device and a design case search program are disclosed (see summary).

Further, Patent Document 2 describes "a requirement specification input means for inputting a requirement specification for a design object, a knowledge base storing knowledge for determining a design specification to be applied from the requirement specification, and a knowledge base based on the requirement specification. A design specification determination means equipped with an inference engine for determining a design specification to be applied using, a design case storage means for storing a design case base storing past design cases, and a design specification determination means for determination. A similar case search means for searching past design cases having substantially the same design specifications as the designed design specifications from the design case base, and a search result output means for outputting the contents of the past design cases searched by the similar case search means. A similar design case search device is disclosed (see summary).

Then, in Patent Document 3, "a design case storage unit that manages information such as specification values of past product design cases and costs and periods for producing the product, and recommended weights for each required specification item are stored. The recommended weight storage unit, the requirement specification reception unit that accepts the requirement specification value, and the requirement specification that the specification value of each design example stored in the design information storage unit is satisfied with the requirement specification value. The required specification satisfaction calculation unit that calculates the satisfaction level, the recommendation level calculation unit that calculates the recommended value of the searched design example from the recommended weight stored in the recommended weight storage unit, and the searched product as the similarity. It is provided with a search result display unit that displays together with the recommendation level and can be sorted according to the similarity level and the recommendation level. ”A similar design case search device is disclosed (see summary).

Japanese Unexamined Patent Publication No. 2010-02617 Japanese Unexamined Patent Publication No. 7-219989 Japanese Unexamined Patent Publication No. 2012-185755

In the methods described in Patent Document 1 and Patent Document 2, the empirical extraction rule of the designer who selects similar cases is reproduced by the program. At that time, the importance of priority items is weighted, and similar projects that take into consideration the necessary constraints and interference conditions are searched. However, it is difficult to judge whether the extraction rule such as which item is emphasized among the empirical extraction rules that actually differ depending on the designer is appropriate. For example, it is often unclear whether the extraction rules used by a person are really valid. In addition, it is difficult to quantitatively grasp how similar the cases judged to be similar are. Further, the technique described in Patent Document 3 also needs further improvement.

The present invention has been made in view of such a background, and an object of the present invention is to realize an efficient diversion design.

In order to solve the above-mentioned problems, the present invention is a search condition for searching the past design project similar to the design project to be designed from now on among the past design projects which are a plurality of projects designed in the past. As a search condition input unit in which the value of the design project is input as information, a random number generation unit that generates a random number based on the value included in the search condition, and the random number generated by the random number generation unit. By setting an evaluation value in, comparing the values in the plurality of past design projects with the evaluation values, and narrowing down the past design projects, the first design project that is similar to the search condition. The process performed by the first similar project extraction unit that narrows down similar projects, the random number generation unit, and the first similar project extraction unit is performed N times (N> 1), and among the plurality of past design projects, each of the past designs A similarity score calculation unit that calculates a similarity score that quantifies how many times the design project is narrowed down as the first similar project out of the N times, and a predetermined past design project based on the similarity score. It is characterized by having a second similar project output unit that outputs the second similar project to the output unit.
Other solutions will be described as appropriate in the embodiments.

According to the present invention, an efficient diversion design can be realized.

It is a figure which shows the structure of the similar matter extraction system which concerns on 1st Embodiment. It is a figure which shows the example of the past design matter database. It is a flowchart which shows the procedure of the similar design matter extraction method in 1st Embodiment. It is a figure which shows an example of the input screen in 1st Embodiment. It is a figure which shows the random number generated in 1st Embodiment. It is a flowchart which shows the detailed procedure of the similarity determination process in 1st Embodiment. It is a figure which shows the example of the output screen in 1st Embodiment. It is a functional block diagram of the similar matter extraction system which concerns on 2nd Embodiment. It is a figure which shows the structure of the design specification draft estimation part in 2nd Embodiment. It is a flowchart which shows the procedure of the similar design project extraction method in 2nd Embodiment. It is a flowchart which shows the detailed procedure of the design specification draft estimation process in 2nd Embodiment. It is a functional block diagram of the similar matter extraction system which concerns on 2nd Embodiment. It is a flowchart which shows the procedure of the similar design project extraction method in 3rd Embodiment. It is a flowchart which shows the detailed procedure of the similarity determination process in 3rd Embodiment. It is a functional block diagram of the similar matter extraction system which concerns on 2nd Embodiment. It is a flowchart which shows the procedure of the similar design project extraction method in 4th Embodiment. It is a figure which shows the example of the design specification check output screen in 4th Embodiment. It is a figure which shows the hardware configuration of the similar matter extraction apparatus.

Next, an embodiment (referred to as “embodiment”) for carrying out the present invention will be described in detail with reference to the drawings as appropriate.

[First Embodiment]
(system)
FIG. 1 is a diagram showing a configuration of a similar project extraction system 1 according to the first embodiment.
The similar matter extraction system 1 shown in FIG. 1 has a similar matter extraction device 100, a past design matter database 201, an extraction rule database 202, and an input / output terminal 300. The similar project extraction device 100, the past design project database 201, the extraction rule database 202, and the input / output terminal 300 are connected to each other via a network such as the Internet or an intranet. In this embodiment, it is assumed that the similar project extraction device 100, the past design project database 201, the extraction rule database 202, and the input / output terminal 300 are installed as separate devices, but the present invention is not limited to this. For example, at least two of the similar matter extraction device 100, the past design matter database 201, the extraction rule database 202, and the input / output terminal 300 may be integrated.

The similar case extraction device 100 has a search condition input unit 110, a Monte Carlo simulation execution unit 120, a similarity score calculation unit 130, and a similar case output unit 140.
In the search condition input unit 110, the search conditions and the like to be the target of the design this time, which are input by the user via the input unit 301, are input.
The Monte Carlo simulation execution unit 120 extracts the first similar matter based on the past design matter database 201, the extraction rule database 202, and the input search conditions. The first similar case will be described later. Specifically, the Monte Carlo simulation execution unit 120 repeats a process of extracting past design projects similar to the input search conditions many times according to the extraction rules made different by the Monte Carlo method. Here, the variation of random numbers follows, for example, a normal distribution.

Further, the Monte Carlo simulation execution unit 120 has a random number generation unit 121 and a similarity determination unit 122.
The random number generation unit 121 generates a random number used in the Monte Carlo method.
The similarity determination unit 122 determines whether or not the past design project is similar to the search condition by using the random number generated by the random number generation unit 121.

The similarity score calculation unit 130 totals the number of times extracted by the Monte Carlo simulation and calculates the similarity score.
The similar matter output unit 140 outputs the calculated matter with a high similarity score. Similar scores will be described later.
The processing performed by each part 110-140, 121-122 will be described later.

The past design matter database 201 is a database in which the specification information of the matter (past design matter) designed in the past is stored. The past design project database 201 will be described later.
The extraction rule database 202 is a database in which a rule (extraction rule) that enables a computer to use an empirical first similar matter extraction rule of an expert is stored. The extraction rules will be described later. The first similar case will be described later.
In this embodiment, the past design project database 201 and the extraction rule database 202 constitute the similar project extraction system 1, but the present embodiment is not limited to this. For example, the past design matter database 201 and the extraction rule database 202 may be installed outside the similar matter extraction system 1 such as being installed on a cloud (not shown). In this case, the similar matter extraction device 100 appropriately acquires data from the past design matter database 201 and the extraction rule database 202 installed on the cloud.

The input / output terminal 300 has an input unit 301 and an output unit 302.
The input unit 301 is various input devices such as a keyboard and a mouse, and is used when a user inputs information to the similar matter extraction system 1.
The output unit 302 is an output device such as a display device, and displays a screen for a process of processing by the similar matter extraction device 100 and, as a result, a screen for interactive processing for the user by the similar matter extraction device 100.

In this embodiment, an example in which the similar project extraction device 100 is applied to processing in a power plant (hereinafter, simply referred to as a plant) is shown. Specifically, when designing a new plant (referred to as "plant α"), refer to the design of any of the plants designed in the past (referred to as "plant A" to "plant Z"). Here is an example of searching for what to do.

(Past Design Project Database 201)
FIG. 2 is a diagram showing an example of the past design project database 201.
The past design matter database 201 stores information about "plant A" to "plant Z" which are plants designed in the past.
Then, in the past design project database 201, numerical data such as "output", "fluid pressure", "fluid temperature", "pump head", "pipe flow velocity (pipe A)", and "pipe flow velocity (pipe B)" are stored. It is stored. Further, the past design project database 201 stores category data such as "boiler model" and "cooling method". As shown in FIG. 2, each data (numerical data, category data) is stored in association with each plant. In addition, items such as "output", "fluid pressure", "fluid temperature", "pump head", "piping flow velocity (piping A)", "piping flow velocity (piping B)", "boiler model", "cooling method", etc. It will be referred to as (past design project item). Here, eight items are shown, but in reality, the past design matter database 201 stores thousands and tens of thousands of items.

Here, the outline of the processing performed in this embodiment will be described.
In the present embodiment, the values of "output", "fluid pressure", "fluid temperature", "boiler model", and "cooling method" of FIG. 2 are input as search conditions (search item). Then, in the example shown in the first embodiment, the similar project extraction device 100 has similar "output", "fluid pressure", and "fluid temperature", and has the same "boiler model" and "cooling method". The past design projects are extracted as a reference (second similar project) for "plant α". Incidentally, the "pump head", "piping flow velocity (piping A), and" piping flow velocity (piping B)", which are not adopted as the search conditions, are the specification values (design specification values) set at the design stage.

Then, the Monte Carlo simulation execution unit 120 sets threshold values on the plus side and the minus side of the input "output", "fluid pressure", and "fluid temperature". Then, the similar project extraction device 100 extracts the past design project in which the "output", "fluid pressure", and "fluid temperature" are within the range of the provided threshold values as the first similar project. Details will be described later.

(Overall processing)
Next, the processing of the similar design project extraction method in the first embodiment will be described.
FIG. 3 is a flowchart showing the procedure of the similar design project extraction method in the first embodiment.
First, a search condition input process for a user to input a search condition is performed via the search condition input unit 110 (S101). As described above, among the items stored in the past design project database 201, the values of "output", "fluid pressure", "fluid temperature", and "boiler model" and "cooling method" are used here. An example of input as a search condition is shown.

FIG. 4 is a diagram showing an example of the input screen 400 according to the first embodiment.
The input screen 400 shown in FIG. 4 is a screen displayed on the output unit 302 in step S101 of FIG.
In the first embodiment, the values of "output", "fluid pressure", and "fluid temperature" of the plant are input as search conditions. In addition, the category data "boiler model" and "cooling method" are also input as search conditions.

Then, as shown in the figure, "output", "fluid pressure", and "fluid temperature" are input by the numerical input window 401, and "boiler model" and "cooling method" are selected and input by the pull-down menu 402.

Then, when the user selects and inputs the search execution button 403, step S102 in FIG. 3 is executed. When the user selects and inputs the end button 404, the similar matter extraction device 100 ends the process of FIG.

Return to the description of FIG.
When the search condition is input in step S101, the random number generation unit 321 performs a random number generation process for generating a random number to be used in the Monte Carlo method (S102).

Here, step S102 of FIG. 3 will be described. The random numbers generated by the Monte Carlo simulation execution unit 120 are threshold values (evaluation values) for each of the “output”, “fluid pressure”, and “fluid temperature” that are the targets of the search conditions.
FIG. 5 is a diagram showing a random number generated in step S102. In FIG. 5, the horizontal axis represents the output and the vertical axis represents the frequency.
Here, it is assumed that past design projects within a range of ± m1 with respect to m of “output” are extracted. Here, m is a value input as a search condition in step S101, and the value of m1 with respect to m is a preset value.

As an example, the random number generation unit 121 generates a random number based on the normal distribution represented by the following equation (1).

Here, x is the value of the generated random number, f (x) is the generation frequency, μ is the average value of the generated random numbers, and σ is the standard deviation. The average value and standard deviation are set in advance based on the experience of an expert.

That is, as shown in FIG. 5, the random number generation unit 121 generates a random number with a threshold value on the + side within the range of the normal distribution 411 with + m1 as the average value. Similarly, the random number generation unit 121 generates a random number with a threshold value on the − side within the range of the normal distribution 412 having −m1 as the average value. When a certain random number r is generated, the random number generation unit 121 sets + r as the threshold value on the + side and −r as the threshold value on the − side. Alternatively, the random number generation unit 121 generates a random number r1 as the threshold value on the + side, generates a random number r2 as the threshold value on the − side, and so on, and generates a random number of the threshold value on the + side and a random number of the threshold value on the − side individually. You may.

In the present embodiment, a random number based on a normal distribution is generated, but a random number based on a probability distribution other than the normal distribution such as a uniform distribution, a gamma distribution, and a beta distribution may be generated. In addition, algorithms such as linear congruential method, Mersenne Twister, and chaotic random numbers are used to generate random numbers.

Returning to the description of FIG.
When step S102 is completed, the similarity determination unit 122 performs the similarity determination process (S103).
The process of step S103 will be described with reference to FIG.
FIG. 6 is a flowchart showing a detailed procedure of the similarity determination process in the first embodiment.
First, the similarity determination unit 122 narrows down the past design projects whose "output" is within ± m1 from the past design projects (S201).
Next, the similarity determination unit 122 narrows down the past design projects whose "fluid pressure" is within ± g1 from the past design projects (S202).
Then, the similarity determination unit 122 narrows down the past design projects whose "fluid temperature" is within ± h1 from the past design projects (S203).
Here, each value of g1 and h1 is the same as m1, and is an average value of random numbers generated by the random number generation unit 121 in step S102. Incidentally, it is assumed that g of "fluid pressure" and h of "fluid temperature" corresponding to m of "output" are input as search conditions.

Subsequently, the similarity determination unit 122 narrows down the past design projects having the same "boiler type" as the "boiler model" in the search condition input in step S101 of FIG. 3 among the past design projects (S204).
Finally, the similarity determination unit 122 narrows down the past design projects in which the “cooling method” is the same as the “cooling method” in the search condition input in step S101 of FIG. 3 (S205).
Here, in steps S204 and S205, if there is no plant having the same boiler type and cooling method, the similarity determination unit 122 does not execute the narrowing down process in steps S204 and S205.

Then, the similarity determination unit 122 outputs the finally remaining past design matter as the first similar matter as a result of the processing of steps S201 to S205 (S206).
Then, the Monte Carlo simulation execution unit 120 returns to step S104 of FIG.
The extraction rule database 202 stores a collection of extraction rules for the empirical first similar matter of such an expert.

Returning to the description of FIG.
When the processing of step S103 is completed, the Monte Carlo simulation execution unit 120 determines whether or not the processing of steps S103 and S104 has been performed N times (S104). N is a preset value, for example, N = 1000, N = 10000, and the like.

When the processing of steps S103 and S104 has not been performed N times (S104 → No), the Monte Carlo simulation execution unit 120 returns the processing to step S102. That is, the random number generation unit 121 newly generates random numbers for each of the threshold values of "output", "fluid pressure", and "fluid temperature". Then, the similarity determination unit 122 executes the random number generation process in step S102, using the generated random numbers as threshold values for each of the “output”, “fluid pressure”, and “fluid temperature”. That is, a Monte Carlo simulation is performed.

When the processes of steps S103 and S104 are performed N times (S104 → Yes), the similarity score calculation unit 130 performs the similarity calculation process for calculating the similarity score for each past design project (S105). Specifically, the similarity score calculation unit 130 calculates the similarity score by, for example, calculating the equation (2).

S = n / N ... (2)

Here, S is a similar score. Further, n is the number of times that a certain past design project is determined to be similar by the similarity determination unit 122. Taking "Plant A" as an example, when N = 1000 in step S104, it is the number of times that "Plant A" remains as the first similar project among 1000 times of steps S102 and S103. As described above, N is the number of repetitions of the Monte Carlo simulation. That is, N is N in step S104. For example, if the number of times "Plant A" remains as the first similar project is 541 out of 1000 steps S102 and S103, the similarity score S is 0.541. The similarity score calculation unit 130 calculates the similarity score for each of the past design projects of "Plant B" to "Plant Z" by the same method.

Finally, the similar matter output unit 140 performs the second similar matter output processing of outputting the past design matter having the higher similarity score to the output unit 302 as the second similar matter (S106).

(Output screen 500)
FIG. 7 is a diagram showing an example of the output screen 500 output in step S106 of FIG.
In the example of the output screen 400 shown in FIG. 7, the project names of the past design projects (second similar projects) having the top 5 similar scores and the bar graph showing the similarity scores are output in association with each other.

The output screen 500 shown in FIG. 7 is an example, and may have a configuration different from that of the output screen 500 shown in FIG. 7. For example, the project name of the past design project and the similarity score may be displayed in a list format. Further, in FIG. 7, the past design projects having the highest similarity score are displayed, but the past design projects having the similarity score equal to or higher than a certain threshold value may be displayed as the second similar project. In addition to the similarity score, in order to show how similar the values of each item of the search condition are, for example, a radar chart may show the difference in the values of each item. When a radar chart is used, each item of the displayed search condition is numerical data such as "output", "fluid pressure", and "fluid temperature".

The user can consider the past design project (second similar project) having a large similarity score as a candidate for the diversion source of the diversion design. That is, in the example of FIG. 7, "Plant D" is considered to be the most suitable reference for the "Plant α" to be designed from now on.

Further, in the first embodiment, the similarity determination unit 122 generates a random number for the threshold value when narrowing down the past design projects, but even if the random number is generated for the parameter of the extraction rule other than the threshold value. good. Further, in the first embodiment, the value of m in FIG. 5 is input as a search condition, and the average value of random numbers m1 is set in advance, but the present invention is not limited to this. By inputting the value of m1 as a search condition, m1 may be changed as appropriate (the same applies to g, h, g1, and h1).

Further, the process shown in FIG. 6 is an example of the similarity determination process, but the order of narrowing down may be different. In the example of FIG. 6, the order of narrowing down is different. In the example of FIG. 6, the narrowing down is performed in the order of "output"-> "fluid pressure"-> "fluid temperature"-> "boiler model"-> "cooling method", but it is not always in this order. It means that it is not necessary to narrow down.

In the first embodiment, the processes of steps S201 to S205 of FIG. 6 are executed N times throughout, but the present invention is not limited to this. For example, the similarity determination unit 122 may execute step S201 N times, then step S202 N times, and then step S203 N times. That is, each of steps S201, S202, and S203 may be executed N times. In this case, a random number is generated every time each of steps S201, S202, and S203 is executed N times. When this method is executed, the similarity score is expressed by the following equation (3).

S = (n1 + n2 + n3) / 3N ... (3)

Here, S is a similarity score, and n1 is the result of executing step S201 N times, and as a result, the “output” of a certain past design project (for example, “plant A”) is the threshold on the + side and the threshold on the − side. The number of times it existed in between. Further, n2 is the number of times that the “fluid pressure” of a certain past design project exists between the threshold value on the + side and the threshold value on the − side as a result of executing step S202 N times. Similarly, n3 is the number of times that the “fluid temperature” of a certain past design project exists between the threshold value on the + side and the threshold value on the − side as a result of executing step S203 N times.
The extraction rule database 201 stores a rule (extraction rule) for extracting a past design matter by a search. As the extraction rule, the order of narrowing down in the processing shown in FIG. 6, the average value of threshold values such as m1, g1, h1 and the like in FIG. 6, the standard deviation σ of the normal distribution when generating random numbers, and the like are stored. Each value stored in the extraction rule database 202 is set based on experience.

In Monte Carlo simulation, the phenomenon to be simulated is stochastically solved by giving a large number of random numbers to the input and observing the output value. In the first embodiment, the probabilistic evaluation index (similarity score S) by Monte Carlo simulation is applied to the determination of the second similar project. By doing so, it is possible to obtain a more robust solution to the similarity criterion whose validity has not been clarified so far. In addition, the same project can be extracted even by an unskilled person while making the best use of the experience of a skilled person. In addition, the degree of similarity can be quantitatively grasped by calculating the similarity score. By doing so, an efficient diversion design can be realized.

In the first embodiment, it is natural that the items in the search condition are not limited to "output", "fluid pressure", "fluid temperature", "boiler model", and "cooling method".

[Second Embodiment]
In this embodiment, a similar design project extraction method and a system for estimating a design specification value using the design data of the extracted project after the extraction of the second similar project will be described. Here, as described above, the design specification value is an item other than the item set as the search condition in the past design matter database 201, and is a specification value set at the design stage. In the example of FIG. 2, "pump head", "piping flow velocity (piping A)", and "piping flow velocity (piping B)" are design specification values. The similar matter extraction system 1a of the second embodiment presents to the user what values should be set for these design specification values.

(system)
FIG. 8 is a functional block diagram of the similar project extraction system 1a according to the second embodiment. In FIG. 8, the same components as those in FIG. 1 are designated by the same reference numerals, and the description thereof will be omitted.
The similar case extraction device 100a in the similar case extraction system 1a is different from the similar case extraction device 100 shown in FIG. 1 in that the similar case extraction device 100a has a design specification proposal estimation unit 150 and a design specification proposal output unit 160.
The design specification draft estimation unit 150 estimates the design specification value of the project (design project) to be designed from now on by using learning or the like.
The design specification draft output unit 160 outputs the design specification value estimated by the design specification draft estimation unit 150 to the output unit 302.

FIG. 9 is a diagram showing the configuration of the design specification draft estimation unit 150 in the second embodiment.
The design specification draft estimation unit 150 includes a learning data input unit 151, a learning model generation unit 152, and a design specification estimation processing unit 153.
The learning data input unit 151 inputs data used for learning used when estimating the design specification value. Then, the learning data input unit 151 sets explanatory variables and objective variables for learning.
The learning model generation unit 152 performs machine learning using the set explanatory variables and objective variables, and generates a learning model.
The design specification estimation processing unit 153 estimates the design specification value of the design project to be designed from now on from the search condition input from the search condition input unit 110 and the learning model.

(Overall processing)
FIG. 10 is a flowchart showing the procedure of the similar design project extraction method in the second embodiment. In FIG. 10, the same processing as in FIG. 3 is assigned the same step number and the description thereof will be omitted.
A design specification draft estimation process (S111) and a design specification draft output process (S112) are added to the process of FIG. 10 with respect to FIG. 3 of the first embodiment.

When the second similar matter is output in step S106, the design specification draft estimation unit 150 acquires each value in the second similar matter from the past design matter database 201.

Next, the design specification estimation processing unit 153 performs a design specification estimation process for estimating the design specification value of the design project (here, “plant α”) to be designed by using a method such as machine learning (S111).

FIG. 11 is a flowchart showing a detailed procedure of the design specification draft estimation process in step S111 of FIG.
First, learning data is input to the learning data input unit 151, and explanatory variables and objective variables for learning are set (S301). Here, as learning data, among the items recorded in the past design project database 201 for the second similar project, the value of the search condition item (referred to as the search item) is used as the explanatory variable, and the other items (design specifications). The value of (referred to as an item) is used as the objective variable. For example, the learning data input unit 151 uses "output", "fluid pressure", and "fluid temperature" constituting the search conditions as explanatory variables in the items of the past design project database 201 of FIG. 2 for the second similar project. do. Then, the learning data input unit 151 sets the design specification item, which is an item other than the configuration of the search condition, as the objective variable among the items of the past design matter database 201. In the example of FIG. 2, the design specification items are "pump head", "piping flow velocity (piping A)", "piping flow velocity (piping B)", and the like.

Next, the learning model generation unit 152 generates a learning model from the relationship between the explanatory variable and the objective variable using the learning data (S302). As a learning model, for example, a model such as a support vector machine or a random forest is used. If the objective variable is a categorical type, a classification model is used, and if it is a numerical type, a regression model is used. That is, it is preferable that the numerical data such as "output", "fluid pressure", and "fluid temperature" and the category data such as "boiler model" and "cooling method" are learned separately.

Next, the value (search condition) of the search condition input by the search condition input unit 110 is input to the design specification draft estimation unit 150 (S303). Here, the search condition input is the same as the value of the search condition input in step S101 of FIG. Further, as the search condition input in step S303, the same items as the items input as explanatory variables in step S301 are input.

Finally, the design specification estimation processing unit 153 estimates the design specification value using the learning model generated by the learning model generation unit 152 (S304). Specifically, the design specification draft estimation processing unit 153 applies the value of the search condition input in step S303 to the learning model generated in step S302. Applying to the learning model generated in step S302 means, for example, determining which group the value of the search condition input in step S303 belongs to among the groups classified by the learning model. be.

The design specification draft output unit 160 performs a design specification draft output process for outputting the estimated design specification value to the output unit 302 as a design specification draft (step S112 in FIG. 14). Specifically, the design specification estimation processing unit 153 calculates the average value of each design specification value in the learning model group to which the value of the search condition input in step S303 belongs. Then, in step S112, the design specification draft output unit 160 outputs the average value calculated by the design specification estimation processing unit 153 to the output unit 302 as a design specification draft. In step S112, the confidence interval of the draft design specification may be calculated based on the average value and variance of the group of learning models to which the value of the search condition input in step S303 belongs. In this case, the calculated confidence interval may be output to the output unit 302 together with the draft design specifications.

According to the second embodiment, in addition to the effect of the first embodiment, an estimated value (design specification draft) of the design specification value referring to the past design project (second similar project) having a high similarity score is given to the user. Presented. As a result, the user can efficiently proceed with the diversion design.

In the second embodiment, the output of the second similar project (S106) and the output of the design specification draft (S112) are performed separately in FIG. 10, but they may be output at the same time or on the same screen. However, in this case, the determination of the second similar matter is made before step S111.

[Third Embodiment]
Next, a third embodiment of the present invention will be described with reference to FIGS. 12 to 14. In the third embodiment, a method of applying the method using the distance of each item in the multidimensional space as the extraction rule will be described.

(system)
FIG. 12 is a functional block diagram of the similar project extraction system 1b according to the second embodiment. In FIG. 12, the same components as those in FIG. 1 are designated by the same reference numerals, and the description thereof will be omitted.
The similar case extraction device 100b in the similar case extraction system 1b shown in FIG. 16 is different from the similar case extraction system 1 shown in FIG. 1 in that the Monte Carlo simulation execution unit 120b has a random number generation unit 121b and a similarity determination unit 122b. ..
The random number generation unit 121b generates a random number used by the similarity determination unit 122b.
The similarity determination unit 122b calculates the distance between each item using the random number generated by the random number generation unit 121b as a weighting coefficient. The processing of the similarity determination unit 122b will be described later.

(Overall processing)
FIG. 13 is a flowchart showing the procedure of the similar design project extraction method in the third embodiment. In FIG. 13, the same process as in FIG. 3 is assigned the same step number and the description thereof will be omitted.
After step S101, the random number generation unit 121b performs a random number generation process of generating the weighting coefficient used in step S212 of FIG. 14 with a random number (S102b). The generated random numbers are based on the normalized distribution and the like as in the first embodiment.
Then, the similarity determination unit 122b uses the random numbers generated by the random number generation unit 121b to perform a similarity determination process for determining a past design project that is close to the value of the search condition from the past design project (S103b). The process of step S103b will be described later.

(Similarity judgment processing)
FIG. 14 is a flowchart showing a detailed procedure of the similarity determination process in step S103b of FIG.
First, the similarity determination unit 122 performs standardization processing for standardizing the numerical data of all the items in the past design matter database 201 (S211). Standardization (normalization) is performed by Eq. (4).

Zi = (xi−μ) / σ ・ ・ ・ (4)

Here, z _i is data after normalization, x _i is pre-standard data, the average of μ is generate, sigma is the standard deviation. By standardizing, the importance of each item in the search conditions becomes uniform once.
Next, the similarity determination unit 122 performs a weighting process that weights the items according to their importance in the similarity determination (S212). As the weighting coefficient at the time of weighting, the random number generated in step S102b of FIG. 13 is used. The parameters (mean value, standard deviation, etc.) of the probability distribution for generating this random number are values preset by the experience of an expert. Weighting will be described later.

Next, the similarity determination unit 122 performs a distance calculation process for calculating the distance between data in the multidimensional coordinate space with the item in the search condition as the coordinate axis (S213). As the distance, for example, the Euclidean distance as shown in the equation (4) is used.

Here, d is the distance, n represents the number of items (the number of axes; _{dimensionality),} q _{i, p} i denotes the data of the two points of interest. Of the two points of data, one is the value of the past design matter database 201 and the other is the value of the search condition. However, these data belong to the same item. Well, the target past design matter is all the past design matter stored in the past design matter database 201. Although the Euclidean distance is used here, other distances such as the Mahalanobis distance may be used.
Here, the weighting is to extend one scale of the coordinate axes corresponding to the weighted items according to the weighting value. That is, the distance of the weighted items increases according to the weighted value.

Next, the similarity determination unit 122 sets the past design project in which the average value of the distances calculated in step S213 is smaller than the preset threshold value, or the top several past design projects in which the distance is small, as the first similar project. The first similar matter output process to be output as is performed (S214). The threshold value used in step S214 is not a random number generated in step S102b, but is set in advance, but is not limited to this. That is, the threshold value used in step S214 may also be a random number generated in step S102b.

Subsequent processing is the same as that of the first embodiment. That is, n in the equation (2) is the number of times that is output as the first similar matter in step S214.

That is, in the third embodiment, for the weighting coefficient, for example, a random number following a normal distribution is generated N times as in the first embodiment. Then, the similarity determination unit 122 performs the similarity determination process shown in FIG. 14 using the generated random number as a weighting coefficient.
By doing so, in the third embodiment, the Monte Carlo simulation regarding the weighting coefficient is executed.

According to the third embodiment, the same effect as that of the first embodiment can be obtained.
A similarity determination in which the similarity determination of the first embodiment and the similarity determination of the third embodiment are combined may be performed. If there are different similarity determination methods for the similarity determination of the first embodiment and the similarity determination of the third embodiment, these three or more similarity determination methods may be combined.

[Fourth Embodiment]
Next, a fourth embodiment of the present invention will be described with reference to FIGS. 15 to 17. In the fourth embodiment, after designing the plant, a design specification check screen 600 (see FIG. 17) for checking the design specification value of the designed plant is output using the extracted second similar project.

(System configuration)
FIG. 15 is a functional block diagram of the similar project extraction system 1c according to the second embodiment. In FIG. 15, the same components as those in FIG. 1 are designated by the same reference numerals, and the description thereof will be omitted.
It differs from the similar case extraction system 1 shown in FIG. 1 in that the similar case extraction device 100c of the similar case extraction system 1c has the design specification check unit 170.
The design specification check unit 170 compares and outputs to check how much the design specification value of the design project designed with reference to the second similar project differs from the past design project.

(flowchart)
FIG. 16 is a flowchart showing the procedure of the similar design project extraction method in the fourth embodiment. In FIG. 16, the same process as in FIG. 3 is assigned the same step number and the description thereof will be omitted.
In the process shown in FIG. 16, the user designs a design matter (here, “plant α”) based on the second similar matter output in step S106 (S121). Since the process of step S121 is a work by the user, it is indicated by a broken line box.
Then, the design specification check unit 170 performs a check screen generation process for generating the design specification check screen 600 (see FIG. 17) (S122). On the design specification check screen 600, how much the design specification value of the second similar project in which the similarity score calculated by the similarity score calculation unit 130 is higher and the design specification value of the designed design project are different. It is a screen for comparing.
After that, the design specification check unit 170 performs comparison processing and output processing to output the comparison result of the generated step S502 to the output unit 302 (S123).

(Design specification check screen 600)
FIG. 17 is a diagram showing an example of the design specification check output screen 600 output in step S123 of FIG.
FIG. 17 shows a boxplot as an example of the visual output of the design specification check.
Here, the error of the design specification value is checked by comparing the pipe flow velocity between "Piping A" and "Piping B" with the design specification value (broken line 601) of the designed project and the actual value of the past project. It shall be. The designed project here is the design project (that is, “plant α”) in the first to third embodiments.

Boxplot 610 shown in FIG. 17 shows the distribution of data of past design projects. In the boxplot 610, reference numeral 611 indicates the minimum value of the corresponding item in all past design projects, and reference numeral 612 indicates the maximum value of the corresponding item in all past design projects. Further, reference numeral 613 indicates a value of 25% from the minimum value, and reference numeral 614 indicates a value of 25% from the maximum value. In addition, line 615 shows the average value in the corresponding item. In the example of FIG. 17, the corresponding items are "pipe flow velocity (pipe A)" and "pipe flow velocity (pipe B)".

In FIG. 17, the values of the second similar projects (here, the past design projects with the top five similar scores) are shown in a circle plot 620 in the boxplot 610. Further, as described above, the design specification value of the designed project is indicated by the broken line 601. In the example of FIG. 17, in the pipe flow velocity of the “pipe A”, the design specification value (broken line 601A) of the designed project is substantially the same as the design specification value (round plot 620) of the second similar project. On the other hand, regarding the pipe flow velocity of "Piping B", the design specification value (broken line 601B) of the designed project is significantly different from the design specification value (round plot 620) of the second similar project. Therefore, the user can determine that "Piping B" needs to be rechecked.

According to the fourth embodiment, it is possible to check whether or not the design specification value of the plant designed based on the second similar project is appropriate. This makes it possible to check the validity of the designed project designed by the diversion design.

In the fourth embodiment, the output of the second similar project (S106) and the output of the design specification check screen 600 (S123) are performed separately in FIG. 16, but they may be output at the same time or on the same screen. good.

[Hardware configuration diagram]
FIG. 18 is a diagram showing a hardware configuration of similar project extraction devices 100, 100a to 100c.
Similar project extraction devices 100, 100a to 100c include a memory 181 and a CPU (Central Processing Unit) 182, a storage device 183 such as an HD (Hard Disk), and a communication device 184 such as a NIC (Network Interface Card).
The program stored in the storage device 183 is loaded into the memory 181. Then, by executing the loaded program by the CPU 182, each part 110-170, 121, 121b, 122, 122b, 151-153 is embodied.

In this embodiment, a plant such as a power plant is applied, but the present invention is not limited to this.
Further, in the present embodiment, the larger the similarity score is, the closer the past design project is to the search condition, but the present invention is not limited to this. If the similarity score is small, the similarity score may be set so that the past design project is close to the search condition.

The present invention is not limited to the above-described embodiment, and includes various modifications. For example, the above-described embodiment has been described in detail in order to explain the present invention in an easy-to-understand manner, and is not necessarily limited to those having all the described configurations. Further, it is possible to replace a part of the configuration of one embodiment with the configuration of another embodiment, and it is also possible to add the configuration of another embodiment to the configuration of one embodiment. Further, it is possible to add / delete / replace other configurations with respect to a part of the configurations of each embodiment.

Further, each configuration, function, each part 110-170, 121, 122, 151-153, database 201, 202, etc. described above are realized by hardware by designing a part or all of them, for example, by an integrated circuit. You may. Further, as shown in FIG. 18, each of the above-described configurations, functions, and the like may be realized by software by interpreting and executing a program in which a processor such as a CPU 182 realizes each function. In addition to storing information such as programs, tables, and files that realize each function in HD (Hard Disk), memory, recording devices such as SSD (Solid State Drive), IC (Integrated Circuit) cards, etc. , SD (Secure Digital) card, DVD (Digital Versatile Disc) and other recording media.
Further, in each embodiment, the control lines and information lines are shown as necessary for explanation, and not all the control lines and information lines are necessarily shown in the product. In practice, almost all configurations can be considered interconnected.

1,1a-1c Similar project extraction system (diverted design support system)
100,100a-100c Similar matter extraction device 110 Search condition input unit (search condition is input)
120,120b Monte Carlo simulation execution unit 121,121b Random number generation unit 122,122b Similarity judgment unit (first similar matter extraction unit)
130 Similarity score calculation unit 140 Similar matter output unit (2nd similar matter output unit)
152 Learning model generation unit (learning unit)
153 Design specification estimation processing unit (design project specification estimation processing unit)
170 Design specification check section (comparison display section)
201 Past design matter database (past design matter information, past design matter items, search condition items)
302 Output unit S101 Search condition input process (search condition input step)
S102, S102b Random number generation process (random number generation step)
S103, S103b Similarity determination process (first similar matter extraction step)
S105 Similarity score calculation process (similarity score calculation step)
S106 Second similar matter output processing (second similar matter output step)

Claims (6)

Among the past design projects that are a plurality of projects designed in the past, the value of the design project is input as information as a search condition for searching the past design project that is similar to the design project that is the project to be designed from now on. Search condition input section and
A random number generator that generates random numbers based on the values included in the search conditions,
An evaluation value is set based on the random number generated by the random number generation unit, the value in the plurality of past design projects is compared with the evaluation value, and the past design project is narrowed down to obtain the search condition. The first similar project extraction unit that narrows down the first similar project that is similar to the past design project,
The process performed by the random number generation unit and the first similar project extraction unit is performed N times (N> 1), and each of the past design projects is performed N times out of the N times. , The similarity score calculation unit that calculates the similarity score that quantifies whether it was narrowed down as the first similar project,
A second similar matter output unit that outputs a predetermined past design matter to the output unit as a second similar matter based on the similarity score, and
A diversion design support system characterized by having. The diversion design support system according to claim 1, wherein the evaluation value is a threshold value for narrowing down the past design projects. The past design matter information includes a plurality of past design matter items in which values are stored.
The search condition includes at least a part of the past design matter items in the information of the past design matter as search items, and the value of the design matter is stored in each of the search items.
In the second similar matter, the value of the past design matter item corresponding to the search item is used as an explanatory variable, and in the second similar matter, the past design matter item other than the past design matter item which is the explanatory variable. A learning unit that uses the design specification value, which is the value of, as the objective variable, and generates a learning model based on the explanatory variables and the objective variables.
A design project specification estimation processing unit that estimates the design specification value in the design project by applying the value included in the search condition to the learning model.
The diversion design support system according to claim 1. The past design matter information includes a plurality of past design matter items in which values are stored.
The search condition includes at least a part of the past design matter items in the information of the past design matter as search items, and the value of the design matter is stored in each of the search items.
The first similar matter extraction unit
Using the search item as a coordinate axis
In each of the plurality of past design projects, the values of the past design project items corresponding to the search items are plotted on the coordinates according to the coordinate axes.
The values in the search condition are plotted on the coordinates.
The distance in the coordinates between the value in the search condition plotted in the coordinates and the value in the information of the past design project plotted in the coordinates is calculated as the evaluation value.
Based on the distance, narrow down the past design projects,
The diversion design support system according to claim 1, wherein the random number generated by the random number generation unit is a weighted value for increasing the distance in a predetermined coordinate axis. A claim characterized by having a comparison display unit that displays a design value designed based on the narrowed-down past design project and a value corresponding to the design value in the second similar project in a comparable manner. The diversion design support system according to item 1. Among the past design projects that are a plurality of projects designed in the past, the value of the design project is input as information as a search condition for searching the past design project that is similar to the design project that is the project to be designed from now on. Search condition input step and
A random number generation step that generates a random number based on the value included in the search condition, and
An evaluation value is set based on the random number generated by the random number generation step, the value in the plurality of past design projects is compared with the evaluation value, and the past design project is narrowed down to obtain the search condition. The first similar project extraction step that narrows down the first similar project that is the similar past design project, and
The random number generation step and the first similar matter extraction step are performed N times (N> 1), and each of the past design projects among the plurality of past design projects is the number of times of the N times. 1 Similar score calculation step to calculate the similar score that quantifies whether it was narrowed down as a similar project,
A second similar matter output step that outputs a predetermined past design matter to the output unit as a second similar matter based on the similarity score, and
A diversion design support method characterized by being executed.

Images