JP7144181B2 - Design support device, structure production method and program - Google Patents

Description

The present invention is a design support device, a structure production method, and a program.

2. Description of the Related Art A technology related to a design support device that supports a design work that has conventionally been performed manually by a computer has been disclosed. For example, Patent Literature 1 discloses a system capable of setting constraints and specifications for each element that constitutes a design object, satisfying the constraints, and optimizing the specifications.
Further, for example, Patent Literature 2 discloses a design support apparatus that uses past analysis results for CAD data to enable shortening the analysis time for newly input CAD data.

U.S. Patent Application Publication No. 2008/0234991 JP 2017-111658 A

However, according to the prior art disclosed in Patent Document 1, a human must define the basic relationships and constraints between elements, and as the number of elements increases, the combination pattern changes according to the number of elements. , and it becomes difficult to describe a large number of rules or constraints without contradiction, which complicates the work.
Further, according to the prior art disclosed in Patent Document 2, the purpose is to omit the act of analyzing the range to be analyzed. Therefore, the effect of labor saving in design work is limited for complex design objects with information hierarchies.

The present invention has been made in view of the above-described problems, and can reduce the complexity of the work compared to manually setting the layout relationships and layout constraints between design elements. An object of the present invention is to provide a design support device, a structure production method, and a program.

An embodiment of the present invention includes an acquisition unit that acquires design element placement conditions in a design target area, and a generation unit that generates placement results of the design elements based on the placement conditions acquired by the acquisition unit using a neural network model. and an output unit that outputs the placement result generated by the generation unit.
According to an embodiment of the present invention, an acquisition unit acquires a placement condition of design elements in a design target area, the placement condition including a constraint condition due to a structure outside the design target area, and the acquisition unit acquires the placement condition. The design support apparatus includes a generation unit that generates a layout result of the design elements based on the layout condition using a neural network model, and an output unit that outputs the layout result generated by the generation unit.

In one embodiment of the present invention, in the above-described design support apparatus, the acquisition unit obtains data indicating arrangement of design elements in the design area, the coordinate positions at which the design elements are arranged in the design area. and the number of types of design elements as the placement condition, and the generation unit has an input dimension corresponding to the dimension of the placement condition acquired by the acquisition unit. A neural network model including an input layer, a hidden layer having a dimension smaller than the input dimension, and an output layer having an output dimension corresponding to the input dimension generates layout data of the design elements corresponding to the layout condition. The layout result is generated for each design element.

In one embodiment of the present invention, the design elements include a generation target element that is a design element of a type not included in the placement condition, and a default element that is a design element of a type included in the placement condition, The acquisition unit acquires, as the placement condition, layout data in which some types of the design elements are missing as the generation target elements, and the generation unit selects at least one of the default element and the generation target element. Placement data for the generation target element is generated as the placement result.

In one embodiment of the present invention, in the design support apparatus described above, the acquisition unit acquires the placement result output by the output unit as the placement condition, and the generation unit arranges Placement data corresponding to the placement result output by the output unit acquired as a new placement result is generated.

In one embodiment of the present invention, in the design support apparatus described above, the design elements include generation target elements that are design elements of a type not included in the placement conditions, and design elements that are included in the placement conditions. In the case where there is a certain default element, the acquisition unit causes the output unit to output when the output unit outputs the placement data of the default element and the placement data of the generation target element as the placement result. The layout data of the element to be generated output by the output unit is obtained as the layout condition without obtaining the layout data of the default element.

In one embodiment of the present invention, in the design support apparatus described above, the arrangement condition acquired by the acquisition unit includes the type of the design element that matches the type of the design element included in the arrangement result output by the output unit. Contains placement data for design elements.

In one embodiment of the present invention, in the design support apparatus described above, the arrangement condition acquired by the acquisition unit includes the type of design elements included in the arrangement result output by the output unit. and the layout data of the design elements whose layout in the design target area is different from the layout of the design elements included in the layout result is included.

In one embodiment of the present invention, in the design support apparatus described above, the generating unit generates the placement result for each neural network model based on a plurality of neural network models with mutually different weightings for learning results.

In one embodiment of the present invention, in the design support apparatus described above, the placement condition and the placement result indicate the placement of the design elements when the design target area is viewed from above.

In one embodiment of the present invention, in the design support apparatus described above, the placement condition includes a constraint condition based on the maximum range in which the design element can be placed in the design target area.

In one embodiment of the present invention, in the design support apparatus described above, the arrangement conditions include constraints due to structures outside the design target area.

An embodiment of the present invention is an acquisition procedure for acquiring arrangement conditions based on a design design result of a structure, the arrangement conditions of design elements in a design target area of a structure; A generation procedure for generating a layout result of the design elements based on layout conditions by a neural network model ; and a structure production method for producing a structure based on the arrangement result obtained by the output procedure.
An embodiment of the present invention is an acquisition procedure for acquiring arrangement conditions of design elements in a design target area of a structure, the arrangement conditions including constraints imposed by structures outside the design target area, and the acquisition procedure. The placement result obtained by a generation procedure for generating the placement result of the design elements based on the placement condition acquired in the neural network model by a neural network model, and an output procedure for outputting the placement result generated in the generation procedure. A structure production method for producing a structure based on

An embodiment of the present invention is an acquisition step of acquiring, in a computer provided in a design support device , a layout condition based on a structural design result, which is a layout condition of design elements in a design target area of a structure; a generation step of generating, by a neural network model, the layout result of the design elements based on the layout conditions obtained in the obtaining step; and an output step of outputting the placement result of the structural members for the execution of the program.
In one embodiment of the present invention, a computer provided in a design support device acquires arrangement conditions for design elements in a design target area of a structure, which include constraints imposed by structures outside the design target area. a generating step of using a neural network model to generate the layout result of the design elements based on the layout conditions obtained in the obtaining step; and an output step of outputting the layout result generated in the generating step. This is a program for executing

According to the present invention, design support that can reduce the complexity of the work or reduce the number of design man-hours compared to the case of manually setting the arrangement relationships and arrangement constraints between design elements. An apparatus, a method for producing a structure, and a program are provided.

It is a figure which shows an example of the design element of the design support apparatus of this embodiment. It is a figure showing an example of functional composition of a design support device of this embodiment. It is a figure which shows an example of the arrangement|positioning data of the column of this embodiment. It is a figure which shows another example of the arrangement|positioning data of the column of this embodiment. It is a figure which shows an example of the functional structure of the production|generation part of this embodiment. It is a figure which shows an example of the design flow of a building. It is a figure which shows an example of operation|movement of the design support apparatus of this embodiment. It is a figure which shows the 1st modification of the design support apparatus of this embodiment. It is a figure which shows the 2nd modification of the design support apparatus of this embodiment.

[Embodiment]
BEST MODE FOR CARRYING OUT THE INVENTION Hereinafter, embodiments of the present invention will be described with reference to the drawings. The design support device 1 can be applied to design support for layout of structures or functions in buildings, machine tools, automobiles, ships, aircraft, integrated circuits, computer software, and the like. In this embodiment, a case where the design support device 1 supports the design of a building will be described as an example. In addition, the structure of the building which the design support apparatus 1 makes the target of design support is not limited. The structures of buildings to which the design support device 1 provides design support include various structures such as steel structures, wooden structures, and RC (Reinforced Concrete) structures.

FIG. 1 is a diagram showing an example of the design element DE of the design support device 1 of this embodiment. The design elements DE include walls WL, pillars PL, floor surface perimeter FL, and windows WD. The walls WL include an outer wall OWL and an inner wall IWL. When the outer wall OWL and the inner wall IWL are not distinguished, they are collectively referred to as the wall WL. The floor surface perimeter FL includes the floor surface perimeter FL of each floor. For example, in the case of a two-story building, the floor surface perimeter FL includes the floor surface perimeter FL1 that is the perimeter of the floor surface of the first floor, the floor surface perimeter FL2 that is the perimeter of the floor surface of the second floor, and the floor surface of the roof. A floor perimeter FL3, which is the perimeter of the face, is included. When the floor surface perimeter FL1 to the floor surface perimeter FL3 are not distinguished, they are collectively referred to as the floor surface perimeter FL. Further, for example, in the case of a two-story building, the pillars PL include a first-floor pillar PL1 and a second-floor pillar PL2.
The design support device 1 supports the design of a building by arranging these various design elements DE at any position within the design target area AR. In one example of the present embodiment, the design target area AR is the range of the layout targets of the design elements DE when the building is viewed from above for each floor.

[Functional Configuration of Design Support Device 1]
FIG. 2 is a diagram showing an example of the functional configuration of the design support device 1 of this embodiment. The design support device 1 includes an acquisition unit 110 , a generation unit 120 and an output unit 130 .
Acquisition unit 110 acquires arrangement data PD. Here, the placement data PD is data indicating the placement of the design elements DE in the design target area AR. In this example, the layout data PD includes data such as walls WL, columns PL, windows WD, floor surface perimeter FL, fittings DR, earthquake-resistant frames BR, electrical equipment EL, and sanitary equipment SA. An example of the placement data PD will be described with reference to FIGS. 3 and 4. FIG.

FIG. 3 is a diagram showing an example of the arrangement data PD of the pillar PL1 of this embodiment. The pillar PL1 placement data PD indicates the coordinate position where the pillar PL1 is placed when the floor surface of the first floor of the building is viewed from above. In this example, the arrangement data PD indicates the coordinate position by two orthogonal axes (axis x1 and axis y1) with the coordinate plane being the floor surface of the first floor of the building.

FIG. 4 is a diagram showing an example of the arrangement data PD of the pillar PL2 of this embodiment. The arrangement data PD of the pillar PL2 indicates the coordinate position where the pillar PL2 is arranged when the floor surface of the second floor of the building is viewed from above. In this example, the arrangement data PD indicates the coordinate position by two orthogonal axes (axis x2 and axis y2) with the coordinate plane being the floor surface of the second floor of the building.
The axis x1 and the axis x2 are collectively referred to as the axis x. Also, the axis y1 and the axis y2 are collectively referred to as the axis y.
Also, the arrangement data PD is not limited to a planar two-dimensional array, and may be a three-dimensional or higher-dimensional array.

In other words, the placement data PD indicates the placement of the design elements DE when the design target area AR is viewed from above. In the example shown in FIGS. 3 and 4, the placement data PD has 64 coordinate positions along the x-axis direction and 64 coordinate positions along the y-axis direction, that is, 4096 coordinate positions. "1" shown at the coordinate position of the arrangement data PD in the figure indicates that the pillar PL is arranged at that coordinate position. Also, "0" shown at the coordinate position of the arrangement data PD in the figure indicates that the pillar PL is not arranged at that coordinate position. That is, the placement data PD indicates, for example, the coordinate position where the design element DE such as the pillar PL is placed or not by binary information of "1" or "0".
Regardless of the type of design element DE, the placement data PD is normalized by expressing the presence of each design element with "1" and the non-existence of each design element with "0". By normalizing the placement data PD, the learning of the placement data PD by the machine can proceed appropriately. Also, on the output side, the probability that the design element exists at that coordinate is indicated by being represented by a value between "0" and "1".
Numerical values such as the number of coordinate positions (for example, 4096) and the number of coordinate positions for each axis (for example, 64) in the arrangement data PD are merely examples and may be changed as appropriate. be.

Here, the number of dimensions of the layout data PD is indicated by the product of the number nDE of types of design elements DE and the number nPX of coordinate positions. In this example, the layout data PD includes data of 13 types of design elements DE such as walls WL, columns PL, windows WD, floor surface perimeter FL, fittings DR, earthquake-resistant frames BR, electrical equipment EL, sanitary equipment SA, and so on. included. In the example shown in FIG. 2, only some of these 13 types of design elements DE are shown.
In this case, the number nDE of types of design elements DE is thirteen. Also, the number nPX of coordinate positions of the arrangement data PD is 4096 as described above. In this example, the number of dimensions of the layout data PD is 53248, which is the product of the number nDE of types of design elements DE and the number nPX of coordinate positions. In other words, the placement data PD has dimensions based on the number nPX of coordinate positions at which the design elements DE are placed in the design target area AR and the number nDE of the types of the design elements DE.
Note that the numerical value of the number nDE (for example, 13) of the types of design elements DE is an example and is not limited to this, and can be changed as appropriate.

Returning to FIG. 2, the acquisition unit 110 acquires the placement data PD as the placement condition CP of the design element DE. The acquisition unit 110 supplies the acquired arrangement data PD, that is, the arrangement condition CP to the generation unit 120 .

The generation unit 120 generates the placement result RP based on the placement condition CP supplied from the acquisition unit 110 . A specific example of the configuration of the generator 120 will be described with reference to FIG.

FIG. 5 is a diagram showing an example of the functional configuration of the generation unit 120 of this embodiment. The generator 120 comprises a neural network model NNM. This neural network model NNM comprises an input layer IL, a hidden layer HL and an output layer OL.
The input layer IL has an input dimension number nID corresponding to the dimension of the arrangement condition CP acquired by the acquisition unit 110 . In this example, the number of dimensions of the input layer IL (that is, the number of input dimensions nID) is 53248 dimensions.
The hidden layer HL has dimensions less than the input number of dimensions nID. In this example, the number of dimensions of the hidden layer HL (ie, the number of hidden layer dimensions nHD) is 832 dimensions.
The output layer OL has an output dimensionality nOD corresponding to the input dimensionality nID. In this example, the number of dimensions of the output layer OL (that is, the number of output dimensions nOD) is 53248 dimensions.
That is, in this example, the neural network model NNM is a model in which the features of the arrangement data PD are learned by dimensional compression based on the autoencoder algorithm.

In the learning stage of the neural network model NNM, a placement result RP output from the output layer OL when the placement condition CP is supplied to the input layer IL is compared with the supplied placement condition CP. In this case, the neural network model NNM is made to learn when the arrangement condition CP supplied to the neural network model NNM and the arrangement result RP output from the neural network model NNM match with each other as a correct output result.
In this example, a three-layer neural network model NNM will be described, but the present invention is not limited to this. The neural network model NNM may have multiple layers of intermediate layers (hidden layers HL).

Using this neural network model NNM, the generation unit 120 generates the placement data PD of the design elements DE corresponding to the placement condition CP as the placement result RP for each design element DE.

Returning to FIG. 2 , the output unit 130 outputs the placement result RP generated by the generation unit 120 . For example, the output unit 130 is connected to a display device (not shown) and outputs the arrangement result RP as an image.

[Regarding the default element EE and generated element GTE]
Here, the design elements DE include default elements EE and generation target elements GTE. A default element EE is a design element DE of a type included in the arrangement condition CP among the design elements DE. A generation target element GTE is a type of design element DE that is not included in the arrangement condition CP among the design elements DE.
As described above, the neural network model NNM of the generation unit 120 has learned the features of the arrangement data PD by dimensional compression based on the autoencoder algorithm. Through this learning, when the neural network model NNM is supplied with a placement condition CP in which some of the types of design elements DE supplied to the input layer IL at the time of learning are missing some of the design elements DE, A placement result RP can be generated for this missing design element DE.

In one example of the present embodiment, the neural network model NNM includes 13 types of walls WL, columns PL, windows WD, floor outer perimeter FL, fittings DR, earthquake-resistant frames BR, electrical equipment EL, sanitary equipment SA, . . . is learning about the design element DE of
In the example shown in FIG. 2, the layout condition CP in which the layout data PD of the window WD is missing among the layout data PD of the plurality of types of design elements DE is supplied to the design support device 1 . That is, in this example, the window WD among the design elements DE is the generation target element GTE. In addition, among the design elements DE, elements other than the window WD, that is, 12 types of design elements DE such as walls WL, columns PL, floor surface perimeter FL, fittings DR, earthquake-resistant frames BR, electrical equipment EL, sanitary equipment SA, . . . , the default element EE.

The acquisition unit 110 acquires the arrangement data PD in which the window WD is missing as the generation target element GTE as the arrangement condition CP. The generation unit 120 generates the placement data PD of the generation target element GTE (the window WD in this example) as the placement result RP based on the placement condition CP acquired by the acquisition unit 110 . Note that the generation unit 120 generates arrangement data of the default elements EE (in this example, 12 types of design elements DE such as walls WL, columns PL, floor surface perimeter FL, earthquake-resistant frame BR, electrical equipment EL, sanitary equipment SA, . . . ). PD may be generated together with the placement data PD of the generation target element GTE. That is, the generation unit 120 generates the placement data PD for at least the generation target element GTE among the default element EE and the generation target element GTE as the placement result RP.
That is, the generation unit 120 generates the placement data PD for the design element DE (generation target element GTE) missing in the placement condition CP based on the learning result and the default element EE.

[Operation of design support device 1]
Next, an example of the operation of the design support device 1 will be described. First, with reference to FIG. 6, an example of the design flow of a general building will be described.

FIG. 6 is a diagram showing an example of a building design flow.
(Step S10) The designer designs the architectural design of the building.
(Step S20) The designer designs the arrangement of structural members to realize the design designed in step S10.
(Step S30) The designer performs structural calculations based on the layout of the structural members designed in step S20.
(Step S40) When the structural calculation result in step S30 satisfies a predetermined requirement (step S40; YES), the designer ends the design. If the structural calculation result in step S30 does not satisfy the predetermined requirements (step S40; NO), the designer returns to step S20 and continues the design.
As an example, the design support device 1 of the present embodiment supports the layout design of the structural members in step S20 described above.

FIG. 7 is a diagram showing an example of the operation of the design support device 1 of this embodiment.
(Step S110) The design support device 1 acquires design constraints. Here, the design constraints include constraints on the maximum range in which design elements DE can be arranged in the design area AR and constraints on structures outside the design area AR.
Here, the constraints based on the maximum range in which the design elements DE can be arranged in the design target area AR include, for example, the shape of the site, the oblique road restriction line, the oblique line on the north side, the building coverage ratio, the floor area ratio, the sky ratio, the type of use zone, and the like. is included. Constraints due to structures outside the design target area AR include borders of neighboring land, diagonal lines of neighboring land, connecting roads, positions of windows, gardens and balconies of neighboring houses, positions of trees and utility poles (distribution equipment), the site and neighboring land and the height difference between the site and the connecting road.

(Step S120) Based on the constraint conditions acquired in step S110, the design support device 1 generates layout data PD of the possible build area within the design target area AR.
Thereafter, in steps S130 to S180, the design support device 1 repeats the process of generating the placement result RP based on the placement condition CP. The operation of the design support apparatus 1 in steps S130 to S180 will be specifically described.

(Step S130) The design support device 1 generates a floor plan based on the layout data PD of the buildable area generated in step S120. The floor plan referred to here is the form of arrangement of the floor outer periphery FL and the walls WL within the design target area AR. In this example, in the learning stage, the neural network model NNM is made to learn in advance a combination of layout data PD of the buildable area, layout data PD of the floor outer periphery FL, and layout data PD of the walls WL. When the layout data PD of the buildable area is input, the neural network model NNM learned in this way generates the layout data PD of the floor outer periphery FL and the layout data PD of the walls WL suitable for this buildable area. can do.
Specifically, in step S130, the acquisition unit 110 sets the layout data PD of the buildable area to the default element EE, and sets the layout data PD of the floor outer circumference FL and the layout data PD of the wall WL to the generation target element GTE. Acquire the arrangement condition CP. In other words, the acquisition unit 110 acquires the arrangement condition CP in which the arrangement data PD of the floor outer circumference FL and the arrangement data PD of the walls WL are missing. Based on this layout condition CP, the generation unit 120 generates the layout data PD of the floor surface perimeter FL and the layout data PD of the walls WL, which are the elements GTE to be generated, that is, the floor plan layout data PD as the layout result RP.
The placement condition CP acquired in step S130 is also referred to as "placement condition CP1", and the placement result RP generated in step S130 is also referred to as "placement result RP1".

(Step S140) Based on the layout data PD of the floor plan generated in step S130, the design support device 1 generates layout data PD of windows WD and layout data PD of fittings DR (for example, hinged doors and sliding doors). In this example, the neural network model NNM is made to learn in advance a combination of layout data PD, window WD layout data PD, and fixture DR layout data PD. When the layout data PD of the floor plan is input, the neural network model NNM learned in this way can generate the layout data PD of the windows WD and the layout data PD of the fittings DR suitable for the floor plan.
Specifically, in step S140, the acquiring unit 110 acquires the layout condition CP in which the layout data PD of the floor plan is set to the default element EE, and the layout data PD of the window WD and the layout data PD of the fittings DR are set to the generation target element GTE. to get That is, the acquisition unit 110 acquires the arrangement condition CP in which the arrangement data PD of the window WD and the arrangement data PD of the fittings DR are missing. Based on this arrangement condition CP, the generating unit 120 generates the arrangement data PD of the window WD and the arrangement data PD of the fittings DR, which are the generation target elements GTE, as the arrangement result RP.

The placement condition CP acquired in step S140 is also referred to as "placement condition CP2", and the placement result RP generated in step S140 is also referred to as "placement result RP2". In this example, the placement result RP1 generated in step S130 and the placement condition CP2 generated in step S140 match.
That is, the placement condition CPn (n is a natural number. In this example, placement condition CP2) acquired by the acquisition unit 110 includes the placement result RP(n−1) output by the output unit 130 (in this example, placement result RP(n−1)). RP1) contains placement data PD for design element DEs of a type that matches the type of design elements DE contained in RP1).

(Step S150) Based on the layout data PD of the floor plan generated up to step S140, the layout data PD of the windows WD, and the layout data PD of the fittings DR, the design support device 1 generates the layout data PD of the pillars PL and the layout data PD of the fittings DR. Placement data PD of the earthquake-resistant frame BR is generated. In this example, the neural network model NNM includes layout data PD, window WD layout data PD, fittings DR layout data PD, column PL layout data PD, and seismic frame BR layout data PD. The combination is learned in advance. The neural network model NNM learned in this way, when the layout data PD of the floor plan, the layout data PD of the windows WD, and the layout data PD of the fittings DR are inputted, the layout, the layout of the windows WD, and the layout of the fittings DR. It is possible to generate the arrangement data PD of the pillar PL and the arrangement data PD of the seismic frame BR that are suitable for .
Specifically, in step S150, the acquisition unit 110 sets the layout data PD, the window WD layout data PD, and the fitting DR layout data PD to the default elements EE, and sets the column PL layout data PD and the earthquake-resistant frame BR to the default elements EE. is set to the generation target element GTE. That is, the acquisition unit 110 acquires the arrangement condition CP in which the arrangement data PD of the pillar PL and the arrangement data PD of the earthquake-resistant frame BR are missing. The generation unit 120 generates, as the placement result RP, the placement data PD of the column PL, which is the generation target element GTE, and the placement data PD of the earthquake-resistant frame BR based on the placement condition CP.
The placement condition CP acquired in step S150 is also referred to as "placement condition CP3", and the placement result RP generated in step S150 is also referred to as "placement result RP3". In this example, the placement result RP2 generated in step S140 and the placement condition CP3 acquired in step S150 match.

(Steps S160 to S180) In the same manner as in steps S140 and S150, the design support device 1 sequentially converts the placement data PD of the beams BM, the sanitary facilities SA, the hot water supply facilities HW, and the electric facilities EL as the placement results RPn. Generate and end a series of actions.

That is, the acquisition unit 110 acquires the placement result RP(n−1) output by the output unit 130 in the previous stage of the generation procedure as the placement condition CPn. Further, the generation unit 120 generates the placement data PD corresponding to the placement condition CPn (that is, the placement result RP(n−1)) acquired by the acquisition unit 110 as a new placement result RPn.

Note that the design support apparatus 1 may have a function of generating a component layout diagram based on each arrangement data PD generated in steps S130 to S180 (step S210). The design support device 1 may also have a function of generating a parts order list based on the parts allocation diagram generated in step S210 (step S220).
The design support device 1 also has a function of supplying each arrangement data PD to other devices having a function of generating these component layout drawings and a function of generating a component order list. good too. In other words, the design support device 1 can generate data for producing a building by generating (or causing to generate) a component layout diagram and a component order list based on the layout data PD.

The design support device 1 may also have a conversion function to convert the layout data PD into another CAD (computer-aided design) data format. The design support device 1 can generate data for producing a building by converting the layout data PD into a CAD data format.

[Summary of embodiment]
As described above, the design support device 1 of this embodiment includes the generator 120 that generates the placement result RP based on the given placement condition CP. This generation unit 120 generates the arrangement result RP by the trained neural network model NNM. The design support device 1 can support the design of the arrangement of structures by automating part of the design work that has conventionally been done manually. Therefore, the design support apparatus 1 can reduce the complexity of the work as compared with the case of manually setting the arrangement relationships and arrangement constraints between the design elements DE.

In addition, in the design support device 1 of the present embodiment, the generation unit 120 generates the layout data PD by a neural network model NNM having a hidden layer HL having dimensions smaller than the number of input dimensions nID. In the neural network model NNM configured in this way, since the feature amount of the arrangement data PD is dimensionally compressed, learning is performed by grasping the features of the arrangement data PD more than when the dimensionality is not compressed. Therefore, according to the design support apparatus 1 of the present embodiment, it is possible to obtain a more preferable design result (for example, comparable to or better than manual design). can.

In addition, in the design support device 1 of the present embodiment, the generation unit 120 generates the generation target element GTE based on the placement data PD in which some types of the design element DE are missing. As described above, the neural network model NNM of this embodiment learns combinations of placement conditions CP and placement results RP in the learning stage. Even if some of the types of design elements DE included in the placement conditions CP are missing in the learning stage, the neural network model NNM will obtain the placement data of the missing design elements DE based on the learning results. PD can be generated.

The acquisition unit 110 of this embodiment acquires the placement result RP(n−1) output by the output unit 130 as the placement condition CPn. Further, in the design support apparatus 1 of the present embodiment, the arrangement condition CP acquired by the acquisition unit 110 includes design element DEs of the same type as the design element DEs included in the placement result RP output by the output unit 130. is included.
In the design of buildings, machine tools, automobiles, ships, integrated circuits, computer software, etc., a so-called waterfall model is used, in which design targets are designed in stages, from those with a high degree of abstraction to those with a low degree of abstraction. design procedures may be adopted.
The design support apparatus 1 of the present embodiment can be applied to the above-described waterfall type design procedure by acquiring the placement result RP output by the output unit 130 as the placement condition CP.
Here, it is generally difficult to create a trained model capable of generating an appropriate final placement result (design result) in one step based on certain placement conditions (initial conditions). The design support device 1 of this embodiment generates the placement result RP step by step. In this case, if the generated placement result RP is unfavorable, only that design stage should be retried. In other words, the design support device 1 of the present embodiment can generate a more preferable placement result RP by generating the placement result RP in stages.
Also, based on a certain placement condition A (design element a ₀ , design element a ₁ , design element a ₂ , . . . , design element a _n ,), a certain placement result A′ (design element a ₀ , _{design element a 1} , design element a ₂ , . In this case, a layout condition _A '' ₍ design element _{a 0} , design element a ₂ , . An arrangement result B (design element a ₀ , design element b ₁ , design element a ₂ , . . . , design element an, design element an ₊₁ ) can be obtained using the network model _NNM2 . Obtaining the placement result B after obtaining the placement result A' in this way means going back through the waterfall type design development process. Here, in the autoencoder algorithm, the number of dimensions of the input layer and the number of dimensions of the output layer of the neural network model NNM are the same, that is, the input and the output are symmetrical. For this reason, according to the neural network model NNM based on the autoencoder algorithm, the waterfall type design development process can be traced back, so it is possible to respond flexibly to various design requirements and changes in requirements.
As described above, when the layout data PD of a certain design element DE is input as the default element EE, the neural network model NNM outputs the layout data PD of the generation target element GTE, and also outputs the layout data PD of the input design element DE. is output as the default element EE. For example, when the layout data PD input to the neural network model NNM includes "wall WL", the neural network model NNM outputs the layout data PD of "wall WL" as the default element EE. Here, in the case of a neural network model NNM based on an autoencoder algorithm, the layout data PD of the design elements DE input to the neural network model NNM and the layout data PD output from the neural network model NNM for the design elements DE are: , may not match exactly. For example, if the design element DE input to the neural network model NNM is "wall WL", the placement data PD of the wall WL output from the neural network model NNM is the placement of the wall WL input to the neural network model NNM. The data PD may not match completely. In other words, there is a case where the layout data PD to be output deteriorates with respect to the layout data PD to be input to the neural network model NNM. Therefore, if the output placement data PD is degraded, the default element EE placement data PD output by the neural network model NNM can be used as it is in the next design stage of the waterfall type design development process. The input to the neural network model NNM may be omitted. In this case, in the waterfall type design development process, instead of the placement data PD output from the neural network model NNM in the previous design stage, the placement data PD input to the neural network model NNM in the previous design stage is input to the neural network model NNM at the next stage.
In addition, the placement data PD output from the neural network model NNM may be incomplete or have low accuracy for supporting design. In this case, the placement data PD output from the neural network model NNM may be transformed into complete or highly accurate data by shaping the placement data PD using a procedural computer program.

[Modification (1)]
FIG. 8 is a diagram showing a first modification of the design support device 1 of this embodiment. In this modification, the output unit 130 outputs the placement data PD for the default element EE and the placement data PD for the generation target element GTE as the placement result RP.
In the case of the specific example shown in FIG. 1, the design support device 1 sets the wall WL and the floor surface outer periphery FL as the default elements EE, sets the window WD as the generation target element GTE, and sets the placement result RP(n −1). In this case, the design support device 1 generates the layout data PD of the window WD, which is the generation target element GTE, and also generates the layout data PD of the wall WL and the floor circumference FL, which are the default elements EE. Here, among the generated layout results RP(n−1), layout data PD may be generated that is undesirable from a design standpoint or a request of the orderer with respect to the floor surface perimeter FL. In this case, at the n stage, the design support device 1 generates the placement result RPn without using the layout data PD of the floor surface perimeter FL generated at the (n−1) stage. That is, the acquisition unit 110 does not acquire the placement data PD for the default element EE output by the output unit 130, but acquires the placement data PD for the generation target element GTE output by the output unit 130 as the placement condition CPn.

In this case, the layout data PD in which the layout generated in the (n−1) stage is different from the layout generated in the (n−1) stage is used as the n-stage layout condition CP for the floor surface perimeter FL.
That is, the layout condition CPn acquired by the acquisition unit 110 includes the same type of design element DE as the type of the design element DE included in the layout result RP(n−1) output by the output unit 130. The placement data PD of the design elements DE whose placement in the target area AR is different from the placement of the design elements DE included in the placement result RP(n-1) is included.

According to the design support apparatus 1 configured in this way, a more preferable placement result RP can be generated by acquiring a more preferable placement condition CP at each stage of design. In other words, according to the design support device 1, more preferable design results can be handed over to subsequent designs.

[Modification (2)]
In this modification, the generation unit 120 generates the placement result RP for each neural network model NNM based on a plurality of neural network models NNM with different weightings for learning results.

FIG. 9 is a diagram showing a second modification of the design support device 1 of this embodiment. In this modification, the generator 120 has multiple types of neural network models NNM. In this example, the generator 120 has a neural network model NNM1 and a neural network model NNM2. The generation unit 120 generates a placement result RP by the neural network model NNM1 and a placement result RP by the neural network model NNM2 for one placement condition CPn. The neural network model NNM1 and the neural network model NNM2 have different weightings for learning results. Therefore, the arrangement result RP by the neural network model NNM1 and the arrangement result RP by the neural network model NNM2 are different in the arrangement of the design elements DE.

According to the design support apparatus 1 configured in this way, it is possible to generate a plurality of placement results RP in which the placement of the design elements DE is different from each other for one placement condition CP, thereby increasing the variations of the placement results RP. , and the options for arranging the design elements DE can be increased.
Furthermore, according to the design support apparatus 1 configured in this way, it is possible to deal with various specifications regarding the arrangement of the design elements DE.

Although embodiments of the present invention and variations thereof have been described above, these embodiments and variations thereof are presented as examples and are not intended to limit the scope of the invention. These embodiments and modifications thereof can be implemented in various other forms, and various omissions, replacements, and modifications can be made without departing from the scope of the invention. These embodiments and their modifications are included in the scope and gist of the invention, and are included in the scope of the invention described in the claims and equivalents thereof.

Each of the devices described above has a computer inside. The process of each process of each device described above is stored in a computer-readable recording medium in the form of a program, and the above process is performed by reading and executing this program by a computer. Here, the computer-readable recording medium refers to magnetic disks, magneto-optical disks, CD-ROMs, DVD-ROMs, semiconductor memories, and the like. Alternatively, the computer program may be distributed to a computer via a communication line, and the computer receiving the distribution may execute the program.

Further, the program may be for realizing part of the functions described above.
Further, it may be a so-called difference file (difference program) that can realize the above-described functions in combination with a program already recorded in the computer system.

DESCRIPTION OF SYMBOLS 1... Design support apparatus 110... Acquisition part 120... Generation part 130... Output part NNM... Neural network model DE... Design element EE... Predefined element GTE... Generation object element

Claims (15)

an acquisition unit that acquires arrangement conditions of design elements in a design target area, which are arrangement conditions based on a result of architectural design of a structure ;
a generation unit that uses a neural network model to generate the arrangement result of the design elements based on the arrangement conditions acquired by the acquisition unit;
an output unit that outputs the layout result generated by the generation unit as a layout result of structural members for realizing a design based on the design design ;
A design support device comprising: an acquisition unit that acquires arrangement conditions for design elements in a design target area, the arrangement conditions including constraints imposed by structures outside the design target area ;
a generation unit that uses a neural network model to generate the arrangement result of the design elements based on the arrangement conditions acquired by the acquisition unit;
an output unit that outputs the arrangement result generated by the generation unit;
A design support device comprising: The acquisition unit
data indicating arrangement of design elements in the design area, the arrangement data having dimensions based on the number of coordinate positions at which the design elements are arranged in the design area and the number of types of the design elements; obtained as the arrangement condition,
The generating unit
An input layer having an input dimension corresponding to the dimension of the arrangement condition acquired by the acquisition unit, a hidden layer having a dimension smaller than the input dimension, and an output layer having an output dimension corresponding to the input dimension. 3. The design support apparatus according to claim 1, wherein the placement data of the design elements corresponding to the placement condition is generated as the placement result for each design element using a neural network model. The design elements include generation target elements, which are design elements of a type not included in the placement conditions, and default elements, which are design elements of a type included in the placement conditions,
The acquisition unit
obtaining, as the placement conditions, placement data in which some types of the design elements are missing as the generation target elements;
The generating unit
4. The design support apparatus according to any one of claims 1 to 3, wherein layout data for at least the generation target element among the default element and the generation target element is generated as the placement result. The acquisition unit
obtaining the placement result output by the output unit as the placement condition;
The generating unit
5. The design according to any one of claims 1 to 4 , wherein placement data corresponding to the placement result output by the output unit acquired by the acquisition unit as the placement condition is generated as the new placement result. support equipment. When the design elements include a generation target element that is a design element of a type not included in the placement condition and a default element that is a design element of a type included in the placement condition,
The acquisition unit
When the output unit outputs the layout data about the default element and the layout data about the generation target element as the layout result, without acquiring the layout data about the default element output by the output unit, 6. The design support device according to claim 5 , wherein layout data of said generation target element output by said output unit is obtained as said layout condition. 5. The placement condition acquired by the acquisition unit includes placement data for the design element of a type that matches the type of the design element included in the placement result output by the output unit. 7. The design support device according to 6 . The arrangement conditions acquired by the acquisition unit include that the type of design element is the same as the type of the design element included in the arrangement result output by the output unit, and that the arrangement in the design target area is the same as the arrangement condition. 8. The design support apparatus according to any one of claims 5 to 7 , wherein layout data of said design elements different from the layout of said design elements included in the result is included. 9. The design support according to any one of claims 1 to 8 , wherein the generation unit generates the arrangement result for each neural network model based on a plurality of neural network models with mutually different weightings for learning results. Device. 10. The design support apparatus according to any one of claims 1 to 9 , wherein the arrangement condition and the arrangement result indicate the arrangement of the design elements when the design target area is viewed from above. 11. The design support apparatus according to any one of claims 1 to 10 , wherein said placement conditions include constraints on a maximum range in which said design elements can be placed in said design target area. an acquisition procedure for acquiring arrangement conditions of design elements in a design target area of the structure, which are arrangement conditions based on the result of the architectural design of the structure;
a generating step for generating, by means of a neural network model, placement results of the design elements based on the placement conditions acquired in the acquiring step;
an output step of outputting the layout result generated in the generation step as a layout result of structural members for realizing a design based on the design design ;
A structure production method for producing a structure based on the arrangement result obtained by. The computer provided in the design support device,
an acquisition step of acquiring arrangement conditions of design elements in a design target area of the structure, which are arrangement conditions based on the result of architectural design of the structure;
a generation step of generating a layout result of the design elements based on the layout conditions obtained in the obtaining step, using a neural network model;
an output step of outputting the placement result generated in the generating step as a placement result of structural members for realizing a design based on the design design ;
program to run the an acquisition procedure for acquiring arrangement conditions for design elements in a design target area of a structure, the arrangement conditions including constraints imposed by structures outside the design target area;
a generating step for generating, by means of a neural network model, placement results of the design elements based on the placement conditions acquired in the acquiring step;
an output procedure for outputting the arrangement result generated in the generation procedure;
A structure production method for producing a structure based on the arrangement result obtained by. The computer provided in the design support device,
an acquisition step of acquiring arrangement conditions for design elements in a design target area of a structure, the arrangement conditions including constraints imposed by structures outside the design target area;
a generation step of generating a layout result of the design elements based on the layout conditions obtained in the obtaining step, using a neural network model;
an output step of outputting the placement result generated in the generating step;
program to run the

Images