JP2824071B2 - Optimal design support method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Industrial Application Field] The present invention relates to an optimal design support method using a computer, and particularly to a case where input parameters are optimized by trial and error so that an output of a simulator becomes a target output. The present invention relates to a suitable optimal design support method.

[Conventional technology]

Conventionally, when a computer is used to support optimal design in general, various mathematical programming methods have been used. In addition, a method of supporting optimization by introducing the know-how of a designer into a design calculation by trial and error is described in, for example, the Journal of Japan System Engineering,
Volume 8, Number 2, (1984), pp. 3-12. In this method, a designer preliminarily specifies know-how such as selection of an input parameter to be changed, a priority order of a change, and a range of a change, and a computer performs optimization based on the know-how.

[Problems to be solved by the invention]

In the optimization using the mathematical programming described above, although the main body of the optimal design is a computer and labor saving can be achieved, since the method itself is versatile and it is difficult to incorporate information according to the problem, veteran designers generally use know-how. This is particularly problematic when the efficiency is low compared to performing optimization and the burden per simulator is large. In addition, since it is generally difficult to sufficiently reflect the intention of the designer, there is also a problem that the optimal solution finally obtained does not easily match the one expected by the designer. The latter problem can be solved to some extent by a known example using know-how reported in the above-mentioned paper, but it is hard to say that the intention of the designer is sufficiently reflected. In particular, it is difficult to extract in advance the designer's judgment at each stage of the optimization and reflect it in the computer system, which has been a problem in performing efficient optimization.

An object of the present invention is to solve the above problems and to provide an optimal design support method capable of efficiently performing an optimal design by utilizing the know-how of a veteran designer.

[Means for solving the problem]

The above-described object is to cause a computer to store the contents of the correction and the result of the correction and the history of the content of the correction leading to the optimal design each time the correction is made. Either let the computer automatically create the amendment from the similar situation, or the designer creates the amendment, and the computer evaluates the amendment with reference to the above-mentioned memory to make a reasonable amendment. This is achieved by simulating a modification plan created by such a method.

[Action]

In the present invention, the designer assists the process of optimizing interactively. The designer determines the initial values of the input parameters and obtains the output by the simulator. Next, the designer compares the output with the target output, corrects the input parameters so as to obtain the target output, obtains the output by the simulator, and repeats the above procedure until the output satisfies the target output. In the flow of the optimal design, the designer stores the contents of the correction in a predetermined format using the graphic terminal every time the correction is performed. On the other hand, at the stage of performing the correction, for example, attention is paid to an output that is desired to be close to the target output, and it is searched whether the correction performed under a similar situation is already stored by the above operation. Here, two types of correction support are performed. One is that the computer makes a correction based on the current correction status based on the content of the correction that has been successfully completed among the searched corrections, and the computer proposes a correction plan for the input parameters. The other is that the designer proposes a modification and compares it with the contents of the retrieved modification, so that the computer evaluates the pros and cons of the modification. For example, if it is determined that the proposed amendment is similar to the content of the successfully completed amendment according to a predetermined criterion, a positive evaluation is made based on the similarity. Further, if it is determined that the proposed amendment is similar to the contents of the failed amendment by a predetermined criterion, a negative evaluation is performed based on the similarity. If it is determined that the proposed amendment is not similar to the content of the amended revision, a negative evaluation will be made. If it is determined that the proposed amendment is not similar to the content of the failed amendment, a positive evaluation is performed. In this way, not only the history of the modification process of the optimal design is managed, but also computer support for the modification can be provided interactively.

In addition, the content of the modification to be searched may be not only the modification content stored in the ongoing optimal design to be searched but also a modification content extracted in another optimal design or a modification content input separately. .

〔Example〕

 Hereinafter, embodiments of the present invention will be described.

FIG. 2 is a functional block diagram showing one configuration example of a device for realizing the support of the optimal design. The first data storage means 1 stores target outputs, input parameters and simulation outputs.
1. Second to store the parameter correction contents and the correction process history
A data storage unit 12, a simulation execution unit 21, an extraction unit 22 for extracting correction details and a correction process history, a search unit 23 for searching for similar corrections, a generation unit 24 for generating a correction plan,
It consists of an evaluation means 25 for evaluating the proposed amendment, an editing means 26 for adding or changing the contents of the correction, an input / output means 31 for inputting / outputting data and correction contents, and a control means 41 for controlling the storage means and the information processing means. ing.

FIG. 1 is a flowchart (Flow Cha) for explaining the flow of processing when supporting the optimal design by the apparatus shown in FIG.
rt). First, a target output is input to the first data storage means 11 by the input / output means 31 shown in FIG. 2 in step 101 of FIG. Next, in step 102, the initial values of the input parameters are input to the first data storage unit 11 by the input / output unit 31. Next, in step 103, the simulation executing means 2
The simulation is performed by 1 and the simulation output is stored in the first data storage unit 11. Next, at step 104, the extracting means 22 extracts the contents of the correction accompanying the current correction of the input parameters and the history of the correction process, and stores them in the second data storage means 12. Next, at step 105, the simulation output stored in the first data storage means 11 is compared with the target output, and it is determined whether or not all outputs satisfy the target output. To end. Step 10 if you are not satisfied with the output
Move to 6. In step 106, the designer inputs a determination as to whether or not the designer himself / herself inputs the correction of the next input parameter through the input / output means 31. The process moves to step 108 when inputting by yourself, and moves to step 107 when automatically generating by a computer. In step 107, the search means 23 determines the current situation from the data stored in the first data storage means 11, and selects similar correction contents among the correction contents stored in the second data storage means 12. Search for. The generating means 24 generates a correction plan based on the similar correction contents and the current situation. If the designer is unsatisfied with the proposed amendment, the editing means 26 adds or changes the contents of the amendment stored in the second data storage means 12. Step 107
Is completed, the process returns to step 103, and the steps from step 103 onward are repeated. In step 108, the designer inputs a correction plan through the input / output means 31. The search means 23 determines the current situation from the data stored in the first data storage means 11 and searches for similar correction contents among the correction contents stored in the second data storage means 12. . Based on the similar correction contents and the input correction plan, the evaluation means 25 performs evaluation and advice on the correction plan. If the designer is dissatisfied with this evaluation, the editing means 26 sends the data to the second data storage means.
Make additions or changes to the corrections stored in 12. Next, in step 109, based on the evaluation / advice, the designer changes the correction plan by the input / output means 31 if necessary. When step 109 is completed, the process returns to step 103, and the steps from step 103 onward are repeated.

FIG. 3 is a diagram showing examples of input parameters and outputs stored in the first data storage means 11. When the optimal design of the room air conditioner is performed, for example, the simulator is the performance calculation program P, and the input parameters are the shape (height × width) A of the heat exchanger, the air volume B, and the rotation speed C of the compressor.
And outputs are cooling capacity X, power consumption Y, and noise Z. In this case, the optimum design is that the output cooling capacity X, power consumption Y, and noise Z are the target output cooling capacity XX, power consumption YY, and noise ZZ.
The input parameters are to determine the shape (height x width) A of the heat exchanger, the air volume B, and the number of revolutions C of the compressor, which are input parameters. FIG. 3 shows an example in which a graphic is displayed on the screen by the input / output means. The input parameters and the output which are the input / output data of the simulator are enclosed by an ellipse, the simulation which is the processing is enclosed by a rectangle, and both are input as nodes. It is expressed as a directed graph by connecting output relations with directed arcs.

FIG. 4 shows an example of a method of storing input parameters, target outputs, and outputs in the first data storage means 11. In FIG. 4, the process of changing the value of each data by optimization is stored as a set of data name, change order (#), and value. The first (# 1) of the change order represents the initial value, and the last of the change order represents the last situation in the optimization process or the final result of the optimization.

FIG. 5 shows an embodiment of the input method of the input parameter correction contents. FIG. 5A shows a screen in the middle of optimization, and FIG.
In addition to the screen shown in the figure, the correction processing up to that time is represented in a graph, and the current input parameter value and output value are separately displayed. FIG. 5 (b) is a graph in which data to be corrected and data to be corrected are selected and input, and the result is displayed as a graph.
In the present embodiment, focusing on the target value ZZ of the noise and the current output Z, the rotational speed C of the compressor is corrected for the purpose of reducing the noise Z. Next, the contents of correction are input in FIG. 5 (c). The content of the correction relates to the content of the judgment condition of the correction and the amount of the correction. In this example, it is a condition that the target value ZZ of the noise is smaller than the current noise Z, and the correction is made by reducing the rotation speed C of the compressor, specifically, the rotation speed C− (Z−ZZ). ×
25 is a new rotation speed C. At this stage, the expected effect of the correction is also input. Thereby, the intention of the correction can be input. Next, the input parameters are corrected based on the input correction contents, and the simulation is repeated. FIG. 5D shows the result. Enter an assessment of the success or failure of the appropriate modifications at this stage. This evaluation compares the expected effect with the new output,
It can be done automatically.

FIG. 6 shows an example of one correction content storage method by the second data storage means 12. The correction content includes a data name serving as a judgment material, a judgment content, a correction content, a data name to be corrected, an expected effect, an actual output, and an evaluation of the correction. FIG. 6 shows the storage contents of the correction shown in FIG. The subscripts in parentheses in the figure indicate the data change order.

FIG. 7 shows an example of a method of storing the history of the correction process by the second data storage means 12. FIG. 7 (a) shows an example of realizing the storage of the history in the time series of the name of the correction processing performed, and FIG. 7 (b) shows the storage of the history in the time series of the name of the correction processing performed. This is an example in which the input (judgment material) and the output (correction parameter) of the correction process are attached to each other.

FIG. 8 shows a PAD (Problem A) which is an example of a method of searching for similar correction contents from the correction contents stored in the second data storage means based on the situation at the stage of making a correction.
nalysis Diagram). First, step 201 in FIG.
The situation 1 in which correction is to be performed is extracted from the first data storage means 11. This situation is represented by, for example, a difference between the current output and the target output. Next, the process proceeds to step 202, and steps 203 to 205 are repeated for all the correction contents stored in the second data storage means. Step 20
In step 3, the situation 2 at the time of the modification is extracted from the modification contents already stored. This situation is, for example, the difference between the output before the execution of the correction and the target output, and the difference between the output before the execution of the correction and the output after the execution of the correction. Next, step 204
Then, the similarity between the situation 1 and the situation 2 is compared in a fixed format. If they are similar, the process proceeds to step 205, and if not, the process returns to step 202. In this similarity comparison, for example, if the directions (signs) of the differences between all the outputs and the target outputs match, it is determined that they are similar. In step 205, the correction content is registered as a similar correction, and the process returns to step 203.

Fig. 9 shows a PAD (Problem Ana
lysis diagram). In step 301 of FIG. 9, the situation 1 in which correction is to be performed is extracted from the first data storage unit 11. Next, the process proceeds to step 302, where the step 303 is repeated for the correction that has been successfully completed among all the similar corrections searched by the similar correction search unit 23. In step 303, the degree of similarity between the situation 2 and the situation 1 of each similarity modification is quantitatively calculated in a fixed format. This degree is obtained by normalizing the difference between the current output and the target output and the difference between the similar correction output and the target output for each item and adding them. Next, in step 304, the most similar correction is determined based on the degree of similarity. Next, step 305
Then, the correction amount of the similar correction is corrected based on the difference between the situation 1 and the situation 2. For example, if the difference between the current output and the target output is about half of the difference at the time of similar correction, the correction amount is halved.

FIG. 10 is an example of a display method of a correction plan. FIG. 10 shows a drawing. In the graph in the figure, hatched lines indicate the judgment material of the correction plan and the correction target, and the contents of the correction plan are separately displayed to facilitate the designer's understanding.

Fig. 11 shows an example of the PAD (Probl
em Analysis Diagram). In step 401 of FIG. 11, steps 402 to 408 are repeated for all similar corrections. In step 402, it is determined whether the similar correction has been successfully completed. If the operation has been successfully completed, the process proceeds to step 403. If the operation has failed, the process proceeds to step 406. In step 403, the similarity between the correction of the correction plan and the correction of the similar correction is compared in a fixed format. Step 40 if similar
If no, go to step 405. In the similarity comparison, for example, if the signs (directions) of the corrections match, it is determined that they are similar, and if they do not match, it is determined that they are not similar. In step 404, the amendment is evaluated positively. In step 405, the revised proposal is evaluated negatively for the time being, and the basis thereof is shown. In step 406, the similarity between the correction of the correction plan and the correction of the similar correction is compared in a fixed format. If they are similar, go to step 407, otherwise go to step 408. In the similarity comparison, for example, if the signs (directions) of the corrections match, it is determined that they are similar, and if they do not match, it is determined that they are not similar. In step 407, the amendment is negatively evaluated. In step 408, the proposed amendment is evaluated positively for the time being, and the basis thereof is shown.

FIG. 12 is an example of a method of displaying an evaluation for a correction plan. FIG. 12 (a) is a display example of a positive evaluation in step 404 of FIG. 11, and FIG. 12 (b) is a display example of a negative evaluation of step 405 in FIG. (C) is the eleventh
FIG. 12D is a display example of a negative evaluation in step 407 in FIG. 11, and FIG. 12D is a display example of a positive evaluation in step 408 in FIG.

FIG. 13 shows an example of a modification content changing method. FIG. 13 (a)
Represents the state before the change, and the case where the modification process 3 is changed is taken as an example. FIG. 13 (b) is an example of a case of adding judgment material, which is realized by changing the graph on the screen and changing the content. It should be noted that the judgment material is deleted in the same manner. FIG. 13 (c) shows an example of adding a correction parameter, which is realized by changing the graph on the screen and the content. Note that the correction parameters are deleted in the same manner. Fig. 13 (d)
Is an example in which only the correction content is changed, and is realized only by changing the content on the screen.

FIG. 14 shows an embodiment in which the present invention is used for learning an optimal design method. FIG. 14 is to improve the learner's optimal design technique by storing the history of the correction process of the optimal design as a model in advance and displaying the correction process of the model in the learner's correction process. is there.

Although not shown in the present embodiment, it is possible to support the optimum design in the case where there are a plurality of simulators and their input parameters and outputs are related, similarly to the present embodiment.

〔The invention's effect〕

As described above, according to the present embodiment, the correction process for performing the actual optimum design is stored by the graphic display means that can be easily changed or added, and at the same time, the interactive support for the correction is made in the form of using the graphic display means. Can be realized. Therefore, there is an effect that the design support of design history management and the design support of correction advice can be performed at the same time. In addition, there is also an effect of extracting, accumulating, utilizing, or transmitting knowledge about optimization, which has been difficult to transmit as conventional know-how.

[Brief description of the drawings]

FIG. 1 is a flowchart of one embodiment of a processing procedure which is a feature of the method of the present embodiment, FIG. 2 is a functional block diagram showing an example of the configuration of an apparatus for realizing the method of the present invention, and FIG. FIG. 4 is an explanatory view of an example of a simulation of an optimal design to be supported, FIG. 4 is an explanatory view of a storage method of input parameters, a target output and an output, FIG. 5 is an explanatory view of an input method of input parameter correction contents, FIG. FIG. 7 is an explanatory view of a method of storing input parameter correction contents, FIG. 7 is an explanatory view of a storage method of a history of a correction process, FIG. 8 is an explanatory view of a search method of similar correction contents, and FIG. PAD explaining the generation method, FIG. 10 is an explanatory view of the display method of the correction plan, FIG. 11 is a PAD explaining the evaluation method of the correction plan, FIG. 12 is an explanatory view of the display method of the evaluation of the correction plan, FIG. Fig. 13 is an explanatory diagram of the method of changing the correction contents, and Fig. 14 is a diagram of the optimization technique. Illustration of the method used for learning,
It is. 11 first data storage means, 12 second data storage means, 21 simulation execution means, 22 extraction means, 23 search means, 24 generation means, 25 evaluation means , 26... Editing means, 31... Input / output means, 41.

──────────────────────────────────────────────────の Continued on the front page (72) Inventor Shunsuke Minami 4026 Kuji-cho, Hitachi City, Ibaraki Prefecture Inside Hitachi, Ltd.Hitachi Research Laboratories Co., Ltd. In-house (56) References JP-A-63-118878 (JP, A) JP-A-62-226279 (JP, A) 8th Knowledge Engineering Symposium Documents pp. 57-62 Kotaro Nakamura et al. (58) Fields surveyed (Int. Cl. ⁶ , DB name) G06F 17/50 G06F 9/44 550 JICST file (JOIS)

Claims (4)

(57) [Claims] 1. An optimal design support method for supporting an optimal design by repeating an operation of correcting an input parameter to a simulator so that an output of the simulator satisfies a target output. When an input parameter is corrected based on one simulation result, the output of the simulation result and the target A record of a correction process having a similar relationship with the output is retrieved and retrieved from the extraction storage means, and the input parameter at that time is corrected with reference to the retrieved record. Optimal design support method characterized in that the contents of correction in the above are graphically represented in a predetermined format and displayed on a display means. 2. The method according to claim 1, wherein the predetermined format is a graph in which at least data for determining input parameter correction, correction processing, and correction parameters are nodes, and an input / output relationship between the nodes is an arc. Claim 1.
The optimal design support method described. 3. The optimum design support method according to claim 1, wherein the contents of correction in the process of correcting the input parameters are graphically represented in a predetermined format and input to a simulator. 4. The method according to claim 1, wherein the predetermined format is a graph in which at least data for determining input parameter correction, correction processing, and correction parameters are nodes, and an input / output relationship between the nodes is an arc. Claim 3
The optimal design support method described.