Steel frame construction support system

Abstract

We provide a front-loading steel frame construction support system that reduces the burden of design. [Solution] The steel frame construction support system 1 includes a case database 11 that stores case data consisting of drawing data (D2 / D3 / D4) and know-how information D5, corresponding to a specific premise state D1; an AI inference unit 12 that takes design data D6 as input and outputs a specific premise state inferred using a first learning model that has been previously learned from the design data and the specific premise state corresponding to the design data to the case database; a generation AI cooperation unit 13 that outputs know-how instruction information D7, which includes know-how information D5 and instruction information, to a text generation AI system G that generates and outputs text according to the input information, causing the text generation AI system to generate proposal information D8; and a display unit 14 that displays the proposal information obtained from the text generation AI system and the case data (D2 / D3 / D4) obtained from the case database.

Description

Technical Field

[0001] The present invention relates to a steel frame construction support system for supporting the design of steel frame construction and ordering to fabricators.

Background Art

[0002] In the conventional domestic construction industry, there has been a problem that productivity is extremely low due to factors in the design process. In order to improve the design process, the use of a BIM (Building Information Modeling) system is expected to replace the conventional CAD (Computer Aided Design) system that handles two-dimensional data and three-dimensional data. BIM is a mechanism that gives a three-dimensional digital model of a building all kinds of data from design to construction management, cost management, and building maintenance management. Specifically, as its utilization, there has been a technology for efficiently incorporating manufacturer products into a BIM model (see, for example, Patent Documents 1 and 2). Also, there has been a technology for efficiently creating a BIM structural model composed of BIM objects having correct object information (see, for example, Patent Document 3). However, regarding these technologies, due to the unique contract system and production system in Japan, important information flows to the downstream process, that is, the construction process rather than the design process, and thus there has been a problem that their utilization has not been sufficient.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Patent Document 2

Patent Document 3

Summary of the Invention

Problems to be Solved by the Invention

[0004] In view of these problems, this invention proposes front-loading, that is, allocating resources from downstream processes to upstream processes such as design, and using AI (Artificial Intelligence) to reduce the design burden resulting from front-loading. This will improve efficiency and enable work to proceed ahead of schedule. As a result, the aim is to provide a steel frame construction support system that achieves overall optimization. Furthermore, the aim is to provide a steel frame construction support system that achieves further overall optimization by utilizing streamlined upstream design information and AI to enable efficient ordering to fabricators. [Means for solving the problem]

[0005] The steel frame construction support system according to the present invention is a steel frame construction support system that supports the design of steel frame construction and ordering from fabricators, and comprises: a case database that stores case data consisting of drawing data, image data, numerical data, and know-how information, which corresponds to specific presuppositions and is classified into appropriate implementation examples or appropriate estimation examples; an AI inference unit that takes design data as input and outputs specific presuppositions inferred using a first learning model that has been previously learned from the design data and specific presuppositions corresponding to the design data, to the case database; a generation AI cooperation unit that outputs know-how instruction information consisting of know-how information obtained from the case database and instruction information indicating proposal information which is information that shows proposed design changes based on the know-how information to a text generation AI system that generates and outputs text according to the input information, causing the text generation AI system to generate proposal information; and a display unit that displays the proposal information obtained from the text generation AI system and the drawing data, image data, and numerical data obtained from the case database, and the steel frame construction support system is a system added as an add-on to a BIM system, and the display unit utilizes the display function installed in the BIM system. [Effects of the Invention]

[0006] The steel frame construction support system of the present invention utilizes an AI inference unit to effectively make use of a database containing information such as know-how from downstream processes, thereby reducing the burden of design and enabling work to proceed ahead of schedule. This improves efficiency and, as a result, achieves the effect of realizing an overall optimized steel frame construction support system. The steel frame construction support system of the present invention enables efficient ordering to fabricators based on streamlined upstream design information and by utilizing an AI inference unit. As a result, it achieves the effect of realizing a steel frame construction support system that is further optimized overall. [Brief explanation of the drawing]

[0007] [Figure 1] This is a schematic diagram showing the steel frame construction support system in the first embodiment. [Figure 2] This figure shows an example of the display screen (BIM screen) of the display unit in the first embodiment, and displays the results of the estimated cost, proposal information regarding value engineering, and image data of examples of changes to the proposal. [Figure 3] This figure shows an example of the display screen (BIM screen) of the display unit in the first embodiment, and shows proposal information for ordering a quote from a fabricator and image data corresponding to the proposal. [Figure 4] This figure shows an example of the display screen (BIM screen) of the display unit in the first embodiment, and illustrates the selection of member attribute data and the search results for standardized member data. [Modes for carrying out the invention]

[0008] Hereinafter, an embodiment of the steel frame construction support system 1 according to this embodiment will be described with reference to the drawings. [Definition of Terms] In this invention, "specific prerequisite state" is defined as the state that serves as the premise for case data (especially know-how information), that is, a specific state in which case data can be adopted. In this invention, "prerequisite" refers to a condition that must be met beforehand for something to be true. [Examples]

[0009] This system is a steel frame construction support system 1 that assists with the design of steel frame construction and ordering from fabricators.

[0010] [Configuration of Steel Frame Construction Support System 1] The configuration of the steel frame construction support system 1 according to this embodiment is shown below. Figure 1 shows a schematic diagram of the steel frame construction support system 1 according to this embodiment.

[0011] [Component 1 of Steel Frame Construction Support System 1]

[0012] <Configuration of Component 1> As shown in Figure 1, the steel frame construction support system 1 of this embodiment comprises a case database 11, an AI inference unit 12, a generation AI collaboration unit 13, and a display unit 14. The case database 11 stores case data that corresponds to a specific premise state D1 and is classified as either an appropriate implementation case or an appropriate estimation case. The case data consists of drawing data D2, image data D3, numerical data D4, and know-how information D5. In other words, as shown in Figure 1, the case database 11 takes a specific premise state D1 as input and outputs case data (D2 / D3 / D4 / D5) that corresponds to that, i.e., stored in relation to the specific premise state D1. As shown in Figure 1, the AI ​​inference unit 12 takes design data D6 as input and outputs a specific premise state D1 inferred using the first learning model. The first learning model is a model that has been previously trained using the design data D6 and the specific premise state D1 corresponding to the design data D6. The generation AI cooperation unit 13 outputs know-how instruction information D7 to the text generation AI system G that generates and outputs a text according to the input information. The know-how instruction information D7 consists of know-how information D5 obtained from the case database 11 and instruction information for outputting proposal information D8 indicating a proposal for a design change plan based on the know-how information D5. By outputting the know-how instruction information D7 to the text generation AI system G, the text generation AI system G is caused to generate the proposal information D8. The display unit 14 displays the proposal information D8 obtained from the text generation AI system G, the drawing data D2, the image data D3, and the numerical data D4 obtained from the case database 11. Here, this system (the steel frame construction support system 1) is a system added to the BIM system B by an add-on, and the display unit 14 uses the display function installed in the BIM system B.

[0013] <Regarding the case database 11 in component 1> As described above, the specific premise state is the state that is the premise of the case data (especially know-how information), that is, a specific state in which the case data can be adopted. Simply put, the case database 11 accumulates knowledge of "in this case, do this (or, this kind of data applies)". And this "in this case" is the premise of "do this (or, this kind of data applies)", and the specific premise state corresponds to this "in this case". That is, when the current state is the specific premise state (this case), the know-how, etc. (do this, or, this kind of data applies) in the case data can be applied to improve the current situation (when the case data is classified as an appropriate implementation case), or an estimate can be calculated (when the case data is classified as an appropriate estimate case). Whether the current state of the design data is "this case" is inferred by the AI inference unit 12 described later. Also, D2 to D5 indicate the data types of the case data. That is, for a certain specific premise state, at least one of the data types of D2 to D5 can correspond. Also, as the data types in a certain case data, in order to present the necessary information on the display unit 14, the types of data necessary to correspond to the specific premise state need to be included. An example of the specific premise state is shown. The case data in Table 1 are the types of appropriate implementation examples and appropriate estimation examples described later that correspond to each specific premise state.

[0014]

Table 1

[0015] An example when the specific premise state is "the current state has a specific problem" in Table 1 is specifically shown below. In the case of a certain design data (set) a1, assume that a specific premise state b1 of "the state where welding at the joint... becomes difficult" occurs as "the current state has a specific problem". And in the case where know-how information c1 of "welding difficulties can be avoided if..." is accumulated in the case database 11 corresponding to this specific premise state b1. The flow is that when the design data a1 is in the state of b1, do c1. Also, for example, image data before and after improvement may be associated with the know-how information c1. The know-how information c1 in this specific example corresponds to the "example regarding means for improving fit" among the types of appropriate implementation examples of the case data described later, and shows that know-how.

[0016] <Regarding another example of the case database of the present embodiment> An "appropriate implementation example" refers to an example of implementation that is appropriate. Generally, "implementation examples" can include appropriate examples, i.e., examples that solve problems or make improvements, and inappropriate examples, i.e., bad examples, examples that worsen the situation, or examples that should not be done. However, in this collection of implementation examples, we are limiting the accumulation of "appropriate" examples, including examples of appropriate estimates, as the minimum necessary case data. However, as an alternative example to this implementation example, especially regarding implementation examples, it would also be acceptable to accumulate "inappropriate" examples and use them alongside "appropriate" examples to illustrate the "inappropriate" cases and draw attention to them.

[0017] <Regarding the AI ​​inference unit 12 of component 1> As mentioned above, the AI ​​inference unit 12 infers whether the current state of the design data is a specific premise state, that is, "this case". The AI ​​inference unit 12 performs inference using the design data D6, the specific premise state D1 corresponding to the design data D6, and a first learning model which is a model learned from these. Here is an example of building the first learning model. In the example above, if design data a1, a specific presupposition state b1 occurs. Similarly, if design data a2, the same presupposition state b1 occurs. However, if design data a3 and a4 occur, the specific presupposition state b1 does not occur (or a different specific presupposition state occurs). Using multiple training data sets, including those mentioned above, a learning model can be created that infers the specific presupposition state b1 from a given design data set. This is a learning model that focuses solely on the specific presupposition state b1. Furthermore, learning is performed by combining various design data with various specific presuppositions that may or may not occur as a result. This builds a learning model for the AI ​​inference unit 12 that determines whether or not a certain state among the various specific presuppositions occurs (or the likelihood of it occurring) when new design data is input.

[0018] <Regarding the text generation function in component 1> In this embodiment, a RAG (Retrieval Augmented Generation) mechanism is used to generate text by utilizing case data such as special know-how information. RAG searches a proprietary database and outputs text using a generation AI. In this embodiment, to supplement the knowledge that a general-purpose generative AI lacks, such as the unique know-how of specific companies in the steel frame construction industry and specialized terminology of the steel frame industry, external information for the generative AI is searched from a proprietary database (case study database 11). Based on this information, it becomes possible to generate answers (text creation) that include the unique know-how. Since the knowledge to be taught to the general-purpose generative AI is stored in the database, the model itself is not modified, and the model itself is not retrained. Therefore, it is a mechanism that can be implemented inexpensively and quickly. The configuration of RAG in component 1 consists of a case database 11, a generation AI collaboration unit 13, and an external system, the text generation AI system G. For AI text generation systems, AI models such as ChatGPT (Chat Generative Pre-trained Transformer) and GPT are used as text generation AI models. Such text generation models are composed of Large Language Models (LLMs). The generation AI collaboration unit 13 outputs know-how instruction information D7, which consists of know-how information D5 and instruction information, and causes the text generation AI system G to generate proposal information D8. The instruction information is what is commonly called a prompt. By providing the know-how instruction information D7 to the text generation AI system G, proposal information D8, which is text containing special know-how, can be obtained from the text generation AI system G.

[0019] <Another example regarding the generation AI for component 1> As another example of generative AI, it is not limited to text generation AI, but may also utilize other generative AI (for example, image generation AI or image editing AI) to generate new standardized component data (drawing data) as described later, or it may be used to generate drawings (such as implementation drawings) by combining existing drawing data related to drawing components with the standardized component data described later. For image generation models, we use, for example, DALL-E3. Furthermore, the generative AI is an external system, connected to the system of this embodiment, for example, via a network, for use. However, as an alternative, the generative AI may be incorporated into this system as an internal system. Alternatively, a unique generative AI may be constructed and incorporated as an internal system using fine-tuning, that is, a technique of further retraining an already trained model. Fine-tuning has the advantage of potentially yielding high effectiveness if successful, but it also has the disadvantage of incurring computational machine costs as well as the costs of storing and managing the retrained model.

[0020] <Regarding the display unit 14 of component 1> As mentioned above, the display unit 14 utilizes the display functions installed in BIM system B. Therefore, rather than being the part that actually displays the data, it is more of an add-on that has the function of displaying data on BIM system B. In reality, in addition to the design data D6, other necessary data input and output are required between the steel frame construction support system 1 and BIM system B in order to display the data on BIM system B. For simplicity, Figure 1 does not show the input and output of other necessary data.

[0021] [Regarding the effects of component 1] With component 1 in this embodiment, the steel frame construction support system has the function of effectively utilizing accumulated downstream process know-how and other case data for arbitrary design data using the AI ​​inference unit 12. This feature is useful because it reduces the burden of design, allows work to proceed ahead of schedule, and enables increased efficiency and overall optimization. Furthermore, this function is useful because it allows for efficient ordering of fabricators based on streamlined upstream design information, enabling further overall optimization.

[0022] [Component 2 of Steel Frame Construction Support System 1]

[0023] <Configuration of component 2> The examples of proper implementation in this embodiment include, as types of examples, at least one of the following: examples related to value engineering in basic design, detailed design, or production design; examples related to means of improving fit; examples related to implementation; and examples related to safety. Furthermore, examples of appropriate estimating include at least one of the following types of examples: examples related to rough estimates in detailed design, and examples related to procurement estimates in production design.

[0024] <Regarding the types of examples of proper implementation and proper estimation in Component 2> The relationship between these examples of proper implementation and proper estimation, and the specific assumptions, is shown in Table 1 above. Here, the types of appropriate implementation examples and appropriate estimation examples in this invention are merely examples that can be considered in steel frame construction, and are not limited to the above. Any example that can solve the problem of design data, an improvement example, a recommended example, etc., is acceptable. Furthermore, the type may include all or part of another type in component 2, or may consist only of another type.

[0025] <Examples of specific assumptions and appropriate implementation and estimation examples for components 1 and 2> Examples of specific assumptions and corresponding appropriate implementation and estimation examples are shown below. The following examples show everything from design data to specific assumptions, case data, and display. Furthermore, while examples of implementation cases are not provided, they could include, for example, recommended implementations for basic and detailed designs, or examples of areas where there is room for improvement in production design implementation (such as the order of steel frame installation). Figure 2 shows an example of the display screen (BIM screen) of the display unit 14 in this embodiment, and shows the results of the rough estimate (numerical data D4), proposal information D8 regarding value engineering, and image data D3 of examples of changes to the proposal. Figure 3 shows an example of the display screen (BIM screen) of the display unit 14 in this embodiment, and shows proposal information D8 for ordering a quote from a fabricator and image data D3 corresponding to the proposal. [Case Studies on Value Engineering] (In this case, the result of <display> is shown in the section following "AI advice" in Figure 2.) <Design Data> (Detailed Design) Shape: I-shaped beam (or column) Length: α Width:β Thickness: γ Relationship with other steel structures: ... Relationship with other components: ... Location of steel frame: ... Placement: It is adjacent to the exterior wall with an X-shaped clearance. <Specific Assumptions> (The state estimated by the AI ​​inference unit 12 based on the above design data) The current situation is that "the exterior wall and the steel frame are in contact with each other with a certain clearance," and there is room for improvement in that "parts can be standardized by adjusting the clearance between the exterior wall and the steel frame." <Case Data> (Case data corresponding to the above specific premise) This is an example of value engineering, among the types of examples of proper implementation. Know-how information: (The following is proposed) By using design data and empirical data (data stored in the case study database), parts can be shared by defining the clearance as Y. Based on calculations using design data and empirical data (data stored in the case study database), the budget reduction amount is Z. Image data from the case database: Image data of examples of changes to the proposals. <Display> The proposed information will be displayed using numerical data and image data, specifically, The message "If the clearance with the exterior wall is Y'mm, parts can be shared, thus reducing the budget for Z'" is displayed, along with image data of examples of changes to the proposal. [Examples of methods to improve fit and finish] <Design Data> (Detailed Design) Detailed design data for a certain joint (details omitted) <Specific Assumptions> (The state estimated by the AI ​​inference unit 12 based on the above design data) The current situation presents a problem: "Welding is difficult," but there is room for improvement. <Case Data> (Case data corresponding to the above specific premise) Examples of proper implementation, specifically those related to methods for improving the fit and finish. Know-how information: (The following is proposed) Change the ... at the joint to ... Image data from the case database: Image data of examples of changes to the proposals. <Display> The proposed information and image data will be displayed, specifically, The message reads, "Welding is likely to be difficult at the current joint. By changing the part of this joint to , welding will become possible and the fit will be improved," along with image data of an example of the proposed change. [Examples of safety considerations in detailed design] <Design Data> (Detailed Design) Detailed design data for a certain joint (details omitted) <Specific Assumptions> (The state estimated by the AI ​​inference unit 12 based on the above design data) The current situation is problematic because "there are many welded points, which can cause distortion due to welding and potentially compromise safety," but there is room for improvement by "reducing the number of welded points." <Case Data> (Case data corresponding to the above specific premise) Examples of proper implementation, specifically those related to safety. Know-how information: (The following is proposed) By changing the joint from ... to ..., the number of welds is reduced, making distortion less likely (ensuring safety). Image data from the case database: Image data of examples of changes to the proposals. <Display> The proposed information and image data will be displayed, specifically, The text reads, "The current joint has many welds, making it highly susceptible to distortion. By changing the joint to , the number of welds can be reduced, making distortion less likely and increasing safety." This is accompanied by image data of an example of the proposed change.

[0026] [Component 3 of Steel Frame Construction Support System 1]

[0027] <Configuration of component 3> The steel frame construction support system 1 of this embodiment further includes a standardized member database 15 and an AI determination unit 16. The standardized component database 15 stores standardized component data D10, which is created by pre-standardizing sets of component attribute data D11 by grouping them, and assigning a group number D9 to each group. The AI ​​determination unit 16 receives member attribute data D11 as input and outputs group number D9 to the standardized member database 15. Group number D9 is the group number of the standardized member data D10 determined using a second learning model that was trained using member attribute data D11 and the standardized member data D10 corresponding to member attribute data D11. The display unit 14 further displays the standardized component data D10 obtained from the standardized component database 15. Figure 4 is a diagram showing an example of the display screen (BIM screen) of the display unit 14 in this embodiment, and shows the selection of member attribute data D11 and the search results of standardized member data D10 for that selection.

[0028] [Regarding the effects of adding component 3] In this embodiment, component 3 enables the steel frame construction support system 1 to standardize and utilize member data using the accumulated member attribute data D11 and the corresponding member data. This function allows the steel frame construction support system having component 3 to achieve cost reduction, improved design and production efficiency, and is useful in constructing an overall optimized steel frame construction support system. However, component 3 is an additional configuration or function in the steel frame construction support system, and the steel frame construction support system in the present invention is not limited to a system including component 3.

[0029] [Component 4 of Steel Frame Construction Support System 1]

[0030] <Configuration of component 4> Component 4 of this embodiment performs the above grouping in the standardized component database 15 by clustering using machine learning. The group number D9 is the class number resulting from the above clustering. Furthermore, the second learning model in the AI ​​judgment unit 16 uses the learning model learned by the clustering described above.

[0031] <Regarding clustering in component 4> By performing clustering using machine learning, it becomes possible to use the resulting clustered model as a second learning model. The method of clustering using machine learning falls under unsupervised learning within machine learning. In other words, it is a method of learning without being given the answer. Furthermore, clustering is classified into hierarchical clustering and non-hierarchical clustering, but it is generally preferable to use hierarchical clustering. This is because hierarchical clustering has the advantages of not requiring the number of clusters (number of classes) to be given in advance, the classification process can be made clear, and the processing is relatively simple. However, it is not applicable when the number of clusters to be clustered is large, so in that case, it is preferable to use non-hierarchical clustering, which can be used even with a large number of clusters, although it is necessary to give the number of clusters in advance.

[0032] [Regarding the effects of adding component 4] With component 4 in this embodiment, the steel frame construction support system 1 has the function of clustering the grouping in the standardized member database 15 using machine learning, and the function of utilizing the clustering results in the AI ​​judgment unit 16. This function makes the steel frame construction support system having component 4 useful because it can efficiently group using machine learning and efficiently construct the AI ​​judgment unit (second learning model). However, clustering using machine learning is not an essential method for grouping; it is possible to group data without clustering. That is, existing component data can be referenced and classified (grouped). Then, based on that classification, the system can learn to group unknown component data. This is a supervised learning method. Therefore, component 4 is an additional component or function in a steel frame construction support system having component 3, and the steel frame construction support system having component 3 in the present invention is not limited to a system including component 4.

[0033] [Component 5 of Steel Frame Construction Support System 1]

[0034] <Configuration of component 5> The example data in this embodiment uses standardized component data.

[0035] [Regarding the effects of adding component 5] Component 5 in this embodiment is useful because it allows for database compression and further overall optimization by using standardized component data D10 in the case data. However, it is not the case that this steel frame construction support system, which has component 3 or 4, cannot be operated until standardized component data has been applied to all case data; there is no problem in gradually standardizing the entire system. Therefore, component 5 is an additional configuration or function in a steel frame construction support system having component 3 or 4, and the steel frame construction support system having component 3 or 4 in the present invention is not limited to a system including component 5. In other words, using standardized member data in the case data is not an essential configuration for operating a steel frame construction support system having component 3 or 4. [Explanation of Symbols]

[0036] 1. Steel frame construction support system 11 Case Database 12 AI Reasoning Department 13. Generation AI Collaboration Department 14 Display section 15 Standardized Components Database 16 AI judgment section D1 Specific Prerequisites D2 drawing data D3 image data D4 Numerical Data D5 Know-how Information D6 Design Data D7 Know-how Instructions D8 Proposal Information D9 Group Number D10 Standardized component data D11 Component attribute data G Text Generation AI System B BIM System

Claims

1. A steel frame construction support system that assists in the design of steel frame construction and ordering from fabricators, The aforementioned steel frame construction support system is A case database containing case data consisting of drawing data, image data, numerical data, and know-how information, which corresponds to specific preconditions and are classified as appropriate implementation examples or appropriate estimation examples. An AI inference unit inputs design data and outputs the specific premise state, inferred using a first learning model that has been previously learned using the design data and the specific premise state corresponding to the design data, to the case database. A generation AI collaboration unit outputs know-how instruction information to a text generation AI system that generates and outputs text according to input information, the know-how instruction information consisting of know-how information obtained from the case database and instruction information indicating proposed design changes based on the know-how information, and causes the text generation AI system to generate the proposed information. A display unit that displays the proposed information obtained from the text generation AI system, the drawing data obtained from the case database, the image data, and the numerical data. Equipped with, The aforementioned steel frame construction support system is a system added to the BIM system as an add-on, The aforementioned display unit utilizes the display function installed in the BIM system. A steel frame construction support system characterized by the following features.

2. A steel frame construction support system according to claim 1, The aforementioned examples of proper implementation include, as a type of example, at least one of the following: examples related to value engineering in basic design, detailed design, and production design; examples related to means of improving fit; examples related to implementation; and examples related to safety. The aforementioned examples of appropriate estimating include at least one of the following types of examples: examples relating to rough estimates in detailed design, and examples relating to procurement estimates in production design. A steel frame construction support system characterized by the following features.

3. A steel frame construction support system according to either claim 1 or 2, The aforementioned steel frame construction support system further, A standardized member database stores standardized member data, which is created by pre-standardizing sets of member attribute data into groups, and assigning a group number to each of the grouped sets. An AI determination unit inputs the aforementioned member attribute data and outputs the group number of the standardized member data, determined using a second learning model trained with the aforementioned member attribute data and the standardized member data corresponding to the aforementioned member attribute data, to the standardized member database. Equipped with, The display unit further displays the standardized member data obtained from the standardized member database. A steel frame construction support system characterized by the following features.

4. A steel frame construction support system according to claim 3, The grouping in the standardized component database is done by clustering using machine learning, and the group number is the class number resulting from the clustering. The second learning model in the AI ​​judgment unit uses the learning model learned by the clustering. A steel frame construction support system characterized by the following features.

5. A steel frame construction support system according to claim 3, further, The aforementioned example data is based on the standardized component data. A steel frame construction support system characterized by the following features.

6. A steel frame construction support system according to claim 4, further, The aforementioned example data is based on the standardized component data. A steel frame construction support system characterized by the following features.

Images