Method, device, and system for providing CFD-based prediction solution by using artificial intelligence model

Abstract

According to one embodiment, a method by which a device provides a CFD-based prediction solution by using an artificial intelligence model is provided, the method comprising the steps of: acquiring a 3D image of a first region; extracting, as a first image, a cross-sectional view of a first height from the 3D image of the first region; generating, on the basis of the first image, first shape information, which is building shape information at the first height; when a wind speed is set to a first speed and a wind direction is set to a first direction, generating a first input variable including the first shape information, the first speed and the first direction; inputting the first input variable into a first artificial intelligence model, which is a U-Net model trained to predict wind speed and wind pressure on the basis of CFD by using building shape information, wind speed, and wind direction as input variables; and, when the wind power and the wind pressure for the first height of the first region are predicted through the first input variable, acquiring, from the first artificial intelligence model, a first wind speed prediction result, which is the result of predicting the wind speed for the first height of the first region, and a first wind pressure prediction result, which is the result of predicting the wind pressure for the first height of the first region.

Description

Method, device, and system for providing CFD-based prediction solutions using artificial intelligence models

[0001] The examples below relate to a technology that provides predictive solutions based on CFD by utilizing an artificial intelligence model.

[0002] CFD (Computational Fluid Dynamics) is a computer simulation technique used to numerically solve fluid dynamics problems. It is a technique that mathematically models fluid flow, heat transfer, chemical reactions, etc. and numerically analyzes them using a computer. It can predict the distribution of fluid velocity, pressure, temperature, etc. by solving the basic equations of fluid dynamics, including the Navier-Stokes equations.

[0003] These CFDs are used in various fields such as aerospace, automotive, architecture, energy, and environmental engineering, and can be used, for example, to analyze aerodynamic performance during aircraft design, optimize the thermal management system of automobile engines, or improve airflow inside buildings.

[0004] In particular, CFD has established itself as a powerful tool for solving complex fluid problems, and its scope of application is expected to expand further with continued technological advancements.

[0005] Accordingly, interest in CFD-based prediction solutions is increasing, and research and development of related technologies is required.

[0006] [Prior Art Literature]

[0007] [Patent Document]

[0008] (Patent Document 1) Korean Patent Publication No. 10-2023-0094686

[0009] (Patent Document 2) Korean Patent No. 10-1927629

[0010] (Patent Document 3) Korean Patent No. 10-1409359

[0011] (Patent Document 4) Korean Patent No. 10-1142832

[0012] According to one embodiment, the purpose is to provide a prediction solution based on CFD by utilizing an artificial intelligence model.

[0013] The purpose of the present invention is not limited to the purposes mentioned above, and other purposes not mentioned can be clearly understood from the description below.

[0014] According to one embodiment, a method for providing a CFD-based prediction solution utilizing an artificial intelligence model, which is performed by a device, comprises the steps of: acquiring a 3D image of a first region; extracting a cross-section of a first height from the 3D image of the first region as a first image; generating first shape information, which is building shape information of the first height, based on the first image; generating a first input variable including the first shape information, the first speed, and the first direction when the wind speed is set to a first speed and the wind direction is set to a first direction; inputting the first input variable into a first artificial intelligence model, which is a U-Net model trained to predict wind speed and wind pressure based on CFD using the building shape information, wind speed, and wind direction as input variables; And when wind speed and wind pressure for the first height in the first region are predicted through the first input variable, a method for providing a CFD-based prediction solution utilizing an artificial intelligence model is provided, including a step of obtaining a first wind speed prediction result, which is a result of predicting wind speed for the first height in the first region, and a first wind pressure prediction result, which is a result of predicting wind pressure for the first height in the first region, from the first artificial intelligence model.

[0015] The method for providing a CFD-based prediction solution utilizing the artificial intelligence model may further include a step of, when a wind speed prediction result, which is a result of predicting wind speed for each height of the first region, is obtained for each height, collecting the wind speed prediction results obtained for each height to generate a wind speed prediction image that predicts the wind speed of the first region when the wind speed is the first speed and the wind direction is the first direction; and a step of, when a wind pressure prediction result, which is a result of predicting wind pressure for each height of the first region, is obtained for each height, collecting the wind pressure prediction results obtained for each height to generate a wind pressure prediction image that predicts the wind pressure of the first region when the wind speed is the first speed and the wind direction is the first direction.

[0016] The method for providing a CFD-based prediction solution utilizing the artificial intelligence model may further include, after the step of generating the wind speed prediction image, a step of classifying an area in which the wind speed predicted on the wind speed prediction image is faster than a first preset reference speed as a wind speed abnormality area; a step of classifying an area with a larger area than a first preset reference area among the areas classified as the wind speed abnormality area as a wind speed danger area; and a step of controlling the wind speed danger area to be highlighted and displayed when the wind speed prediction image is displayed, and may further include, after the step of generating the wind pressure prediction image, a step of classifying an area in which the wind pressure predicted on the wind pressure prediction image is higher than a first preset reference pressure as a wind pressure abnormality area; a step of classifying an area with a larger area than the first preset reference area among the areas classified as the wind pressure abnormality area as a wind pressure danger area; and a step of controlling the wind pressure danger area to be highlighted and displayed when the wind pressure prediction image is displayed.

[0017] According to one embodiment, by providing a prediction solution based on CFD using an artificial intelligence model, the accuracy and user convenience of the prediction solution can be increased.

[0018] Meanwhile, the effects according to the embodiments are not limited to those mentioned above, and other effects not mentioned can be clearly understood by a person having ordinary skill in the art from the description below.

[0019] Figure 1 is a schematic diagram showing the configuration of a system according to one embodiment.

[0020] Figure 2 is a flowchart illustrating a process for providing a CFD-based prediction solution using an artificial intelligence model according to an embodiment.

[0021] Figure 3 is a diagram showing a 3D image and CFD data according to an embodiment.

[0022] Figure 4 is a drawing showing a method of extracting images for each height according to one embodiment.

[0023] Figure 5 is a diagram showing the wind speed prediction results obtained for each height and the wind pressure prediction results obtained for each height according to an embodiment.

[0024] Figure 6 is a drawing showing a 3D image generated by compiling prediction results according to an embodiment.

[0025] Figure 7 is a flowchart for explaining a process for highlighting a wind speed hazard zone according to an embodiment.

[0026] Figure 8 is a flowchart for explaining a process for highlighting a wind pressure risk area according to an embodiment.

[0027] Figure 9 is a flowchart illustrating an artificial intelligence model learning process for CFD-based prediction using multi-modality according to an embodiment.

[0028] Figure 10 is a diagram showing the operation process of the first artificial intelligence model according to an embodiment.

[0029] Figure 11 is a diagram showing the operation process of a second artificial intelligence model according to an embodiment.

[0030] Figure 12 is a flowchart for explaining the process of updating wind speed prediction results and wind pressure prediction results according to an embodiment.

[0031] Figure 13 is a diagram for explaining a process of predicting wind speed and wind pressure using a first artificial intelligence model and a second artificial intelligence model according to an embodiment.

[0032] Figure 14 is a flowchart for explaining a process for setting a reference number according to an embodiment.

[0033] Figure 15 is a diagram illustrating an artificial intelligence model network structure for CFD-based prediction using multi-modality according to an embodiment.

[0034] Figure 16 is a flowchart illustrating a process for providing a CFD-based prediction solution utilizing an artificial intelligence model for smart air conditioning according to an embodiment.

[0035] Figure 17 is a diagram showing temperature prediction results obtained for each height, flow rate prediction results obtained for each height, and pressure prediction results obtained for each height according to an embodiment.

[0036] Figure 18 is a flowchart illustrating a process for obtaining temperature prediction results, flow rate prediction results, and pressure prediction results by utilizing multi-modality according to an embodiment.

[0037] Figure 19 is a diagram showing the operation process of the 1-1 artificial intelligence model according to an embodiment.

[0038] Figure 20 is a diagram showing the operation process of the 2-1 artificial intelligence model according to an embodiment.

[0039] Figures 21 to 24 are flowcharts for explaining a process for additionally obtaining prediction results for an intermediate height according to an embodiment.

[0040] Figures 25 and 26 are flowcharts for explaining a process for determining whether there is a problem with the shape of a building according to one embodiment.

[0041] Figure 27 is an example diagram of the configuration of a device according to one embodiment.

[0042] Hereinafter, embodiments are described in detail with reference to the attached drawings. However, the embodiments may be modified in various ways, and the scope of the patent application is not limited or restricted by these embodiments. It should be understood that all modifications, equivalents, or alternatives to the embodiments are included within the scope of the patent application.

[0043] Specific structural or functional descriptions of the embodiments are disclosed for illustrative purposes only and may be modified and implemented in various forms. Accordingly, the embodiments are not limited to the specific disclosed form, and the scope of this specification includes modifications, equivalents, or alternatives that fall within the technical concept.

[0044] Although terms such as "first" or "second" may be used to describe various components, these terms should be interpreted solely to distinguish one component from another. For example, a first component may be referred to as a second component, and similarly, a second component may also be referred to as a first component.

[0045] When it is said that a component is "connected" to another component, it should be understood that it may be directly connected or connected to that other component, but there may also be other components in between.

[0046] The terms used in the examples are for illustrative purposes only and should not be construed as limiting. Singular expressions include plural expressions unless the context clearly dictates otherwise. In this specification, terms such as "comprise" or "have" are intended to indicate the presence of a feature, number, step, operation, component, part, or combination thereof described in the specification, but should be understood to not preclude the presence or addition of one or more other features, numbers, steps, operations, components, parts, or combinations thereof.

[0047] Unless otherwise defined, all terms used herein, including technical or scientific terms, have the same meaning as commonly understood by a person of ordinary skill in the art to which the embodiments pertain. Terms defined in commonly used dictionaries should be interpreted as having a meaning consistent with their meaning in the context of the relevant technology, and shall not be interpreted in an idealized or overly formal sense unless explicitly defined herein.

[0048] In addition, when describing with reference to the attached drawings, identical components will be assigned the same reference numerals regardless of the drawing numbers, and redundant descriptions thereof will be omitted. When describing embodiments, if a detailed description of a related known technology is judged to unnecessarily obscure the gist of the embodiment, the detailed description will be omitted.

[0049] The embodiments may be implemented in various forms of products, such as personal computers, laptop computers, tablet computers, smart phones, televisions, smart home appliances, intelligent vehicles, kiosks, and wearable devices.

[0050] In practice, an artificial intelligence (AI) system is a computer system that demonstrates human-level intelligence. Unlike existing rule-based smart systems, it is a machine-based system that learns and makes decisions on its own. As AI systems become more widely used, their recognition rates improve and their ability to more accurately understand seller preferences increases. As a result, existing rule-based smart systems are gradually being replaced by deep learning-based AI systems.

[0051] AI technology consists of machine learning and its underlying technologies. Machine learning is an algorithmic technology that autonomously classifies and learns the characteristics of input data. Elementary technologies utilize machine learning algorithms, such as deep learning, to mimic the cognitive and judgmental functions of the human brain. These technologies encompass areas such as linguistic understanding, visual understanding, inference / prediction, knowledge representation, and motion control.

[0052] The various fields in which AI technology is applied are as follows. Linguistic understanding refers to the technology that recognizes, applies, and processes human language / text, including natural language processing, machine translation, dialogue systems, question-answering, and speech recognition / synthesis. Visual understanding refers to the technology that recognizes and processes objects similar to human vision, including object recognition, object tracking, image search, person recognition, scene understanding, spatial understanding, and image enhancement. Inference prediction refers to the technology that logically infers and predicts information by judging it, including knowledge / probability-based inference, optimization prediction, preference-based planning, and recommendations. Knowledge representation refers to the technology that automatically processes human experience information into knowledge data, including knowledge construction (data creation / classification) and knowledge management (data utilization). Motion control refers to the technology that controls the movement of autonomous vehicles and robots, including movement control (navigation, collision, driving), and manipulation control (behavior control).

[0053] Typically, applying machine learning algorithms to real-world applications requires a trial-and-error approach, due to the fundamental nature of machine learning methodology. Deep learning, in particular, requires hundreds of thousands of iterations. Because this is impossible to implement in an actual physical environment, learning is performed through simulations, where the actual physical environment is virtually implemented on a computer.

[0054] Figure 1 is a schematic diagram showing the configuration of a system according to one embodiment.

[0055] Referring to FIG. 1, a system according to one embodiment may include a user terminal (100) and a device (200) capable of communicating with each other through a communication network.

[0056] First, the communication network can be configured regardless of the communication mode, such as wired or wireless, and can be implemented in various forms to perform communication between servers and between servers and terminals.

[0057] The user terminal (100) may be implemented as a computing device having a communication function, and may be implemented as, for example, a mobile phone, a desktop PC, a laptop PC, a tablet PC, a smart phone, etc., but is not limited thereto, and may be implemented as various types of communication devices that can be connected to an external server.

[0058] The user terminal (100) may be configured to perform all or part of the computational functions, storage / reference functions, input / output functions, and control functions of a typical computer, and may be configured to communicate with the device (200) via wired or wireless means.

[0059] The user terminal (100) may access a web page operated by a person or organization providing a service using the device (200), or may install an application developed and distributed by a person or organization providing a service using the device (200). The user terminal (100) may be linked to the device (200) through a web page or application.

[0060] That is, the user terminal (100) can access the device (200) through a web page, application, etc. provided by the device (200).

[0061] The device (200) may be a private server owned by an entity or organization providing services using the device (200), a cloud server, or a peer-to-peer (p2p) collection of distributed nodes. The device (200) may be configured to perform all or part of the computational functions, storage / reference functions, input / output functions, and control functions of a typical computer. The device (200) may be equipped with at least one artificial intelligence model that performs an inference function.

[0062] The device (200) can be configured to communicate with the user terminal (100) via wired or wireless communication, control the operation of the user terminal (100), and control which information to display on the screen of the user terminal (100).

[0063] The device (200) is implemented as a server that provides a prediction solution based on CFD, and thus can provide a platform for providing a CFD-based prediction service. Here, CFD stands for Computational Fluid Dynamics, which can refer to computational fluid dynamics that calculates the dynamic movement of a fluid using a computer through a numerical analysis method.

[0064] According to one embodiment, the device (200) can predict wind speed, wind pressure, etc. by utilizing an artificial intelligence model, and a detailed description related thereto will be described later with reference to FIG. 2.

[0065] In the present invention, artificial intelligence (AI) refers to a technology that mimics human learning, reasoning, and perception abilities and implements them on a computer. It may include concepts such as machine learning and symbolic logic. Machine learning (ML) is an algorithm technology that classifies or learns the characteristics of input data on its own. AI technology analyzes input data using machine learning algorithms, learns the results of that analysis, and makes judgments or predictions based on the results of that learning. Furthermore, technologies that utilize machine learning algorithms to mimic human brain functions such as cognition and judgment can also be understood as falling under the category of AI. For example, this may include technical fields such as linguistic understanding, visual understanding, inference / prediction, knowledge representation, and motion control.

[0066] Machine learning can refer to the process of training a neural network model using data processing experience. Through machine learning, computer software can improve its data processing capabilities. Neural network models are built by modeling correlations between data, and these correlations can be expressed by multiple parameters. Neural network models extract and analyze features from given data to derive correlations between data. This process of iteratively optimizing the parameters of a neural network model can be defined as machine learning. For example, a neural network model can learn the mapping (correlation) between inputs and outputs for data presented as input-output pairs. Alternatively, a neural network model can learn the relationships between inputs and outputs by deriving regularities between the given data, even when presented with only input data.

[0067] An artificial intelligence learning model or neural network model can be designed to implement the structure of the human brain on a computer, and can include multiple network nodes that simulate the neurons of a human neural network and have weights. The multiple network nodes can have connections with each other by simulating the synaptic activity of neurons that exchange signals through synapses. In the artificial intelligence learning model, the multiple network nodes can be located at layers of different depths and exchange data according to convolutional connections. The artificial intelligence learning model can be, for example, an artificial neural network model, a convolutional neural network (CNN), etc. In one embodiment, the artificial intelligence learning model can be machine-learned according to a method such as supervised learning, unsupervised learning, or reinforcement learning. Machine learning algorithms that can be used to perform machine learning include decision trees, Bayesian networks, support vector machines, artificial neural networks, Ada-boost, perceptrons, genetic programming, and clustering.

[0068] CNNs are a type of multilayer perceptron designed to utilize minimal preprocessing. They consist of one or more convolutional layers stacked on top of regular artificial neural network layers, with additional weight and pooling layers. This structure allows CNNs to fully utilize two-dimensional input data. Compared to other deep learning architectures, CNNs demonstrate excellent performance in both image and audio domains. CNNs can also be trained using standard backpropagation. Compared to other feedforward artificial neural network techniques, CNNs are easier to train and have fewer parameters.

[0069] Convolutional networks are neural networks that contain sets of nodes with bounded parameters. The increasing availability of training data and computational power, combined with advances in algorithms such as piecewise linear units and dropout training, have led to significant improvements in many computer vision tasks. With the massive datasets available for many tasks today, overfitting is less of a concern, and increasing network size improves test accuracy. Optimal utilization of computing resources becomes a limiting factor. To address this, distributed, scalable implementations of deep neural networks can be utilized.

[0070] FIG. 2 is a flowchart for explaining a process for providing a CFD-based prediction solution using an artificial intelligence model according to an embodiment, FIG. 3 is a diagram showing a 3D image and CFD data according to an embodiment, and FIG. 4 is a diagram showing a method for extracting images for each height according to an embodiment.

[0071] Referring to FIG. 2, first, in step S201, the device (200) can acquire a 3D image of a first region. Here, the first region may refer to a specific region where buildings are arranged, and the 3D image of the first region may refer to a 3D image of buildings arranged within the first region, and may be acquired from an external server or a database of the device (200).

[0072] For example, referring to FIG. 3, the image shown on the left is a 3D image showing buildings arranged within a specific area, and the 3D image of the first area may be any one of the 3D images shown on the left of FIG. 3.

[0073] Referring to Fig. 3, the image shown on the right shows CFD data that was analyzed by considering buildings located within a specific area. The CFD data shown in Fig. 3 can be generated based on the wind environment assessment guidelines using wind tunnel experiments and computational analysis techniques (CFD). For example, if the data generation conditions are set to 10 types of buildings, 3 types of wind speeds, and 8 types of wind directions, a total of 240 CFD data can be generated through (10 X 3 X 8).

[0074] CFD data is 3D data and contains a lot of numerical data within the 3D fluid domain, making it difficult to learn. Therefore, the 3D image can be cut at a certain height to extract a 2D image, and the extracted 2D image can be used to proceed with the prediction process.

[0075] In step S202, the device (200) can extract a cross-sectional view of a first height from a 3D image of a first region as a first image. Here, the first height can be set differently depending on the embodiment.

[0076] For example, referring to FIG. 4, the device (200) can extract images at 3m intervals from a 3D image of a first region, extract a cross-section of 0m from the 3D image of the first region as an n-th image, extract a cross-section of 3m from the 3D image of the first region as an n+1-th image, and extract a cross-section of 6m from the 3D image of the first region as an n+2-th image.

[0077] In step S203, the device (200) may generate first shape information, which is building shape information of a first height, based on the first image. At this time, the first shape information may include information indicating which area is inside the building, which area is outside the building, and the boundary between the inside and outside of the building when the first height is present. For example, in the first shape information, an area inside the building may be indicated as 0, and an area outside the building may be indicated as 1.

[0078] In step S204, the device (200) can generate a first input variable including first shape information, a first speed, and a first direction when the wind speed is set to a first speed and the wind direction is set to a first direction. At this time, the first speed and the first direction can be set through the user terminal (100).

[0079] In step S205, the device (200) may input a first input variable to the first artificial intelligence model. Here, the first artificial intelligence model may be a U-Net model trained to predict wind velocity and wind pressure based on CFD using building shape information, wind speed, and wind direction as input variables.

[0080] That is, the first artificial intelligence model may be an algorithm that, when input variables such as building shape information, wind speed, and wind direction are input, predicts wind speed within a specific area based on CFD, considers the building shape information, wind speed, and wind direction, and then predicts wind pressure within the specific area, and then outputs the wind speed prediction results and wind pressure prediction results. A detailed description of the first artificial intelligence model will be provided below with reference to Fig. 10.

[0081] In step S206, when the device (200) predicts the wind speed and wind pressure for the first height in the first region through the first input variable, the device can obtain the first wind speed prediction result and the first wind pressure prediction result from the first artificial intelligence model. Here, the first wind speed prediction result is a result of predicting the wind speed for the first height in the first region, and can be generated as an image in which each coordinate is displayed in a different color according to the speed value of the wind speed predicted for each coordinate, and the first wind pressure prediction result is a result of predicting the wind pressure for the first height in the first region, and can be generated as an image in which each coordinate is displayed in a different color according to the pressure value of the wind pressure predicted for each coordinate.

[0082] FIG. 5 is a diagram showing wind speed prediction results obtained for each height and wind pressure prediction results obtained for each height according to an embodiment, and FIG. 6 is a diagram showing a 3D image generated by compiling prediction results according to an embodiment.

[0083] According to one embodiment, the device (200) can obtain wind speed prediction results, which are results of predicting wind speed for each height in the first region, for each height in the same manner as the method of obtaining the first wind speed prediction result, which is a result of predicting wind speed for the first height in the first region.

[0084] That is, as shown in Fig. 5, wind speed prediction results can be obtained for each height.

[0085] For example, the device (200) can obtain a result of predicting the wind speed for 0 m in the first region, a result of predicting the wind speed for 3 m in the first region, a result of predicting the wind speed for 6 m in the first region, etc.

[0086] According to one embodiment, the device (200) can obtain a wind pressure prediction result, which is a result of predicting wind pressure for each height in the first region, for each height in the same manner as the method of obtaining the first wind pressure prediction result, which is a result of predicting wind pressure for the first height in the first region.

[0087] That is, as shown in Fig. 5, wind pressure prediction results can be obtained for each height.

[0088] For example, the device (200) can obtain a result of predicting wind pressure at 0 m in the first region, a result of predicting wind pressure at 3 m in the first region, a result of predicting wind pressure at 6 m in the first region, etc.

[0089] When the device (200) obtains wind speed prediction results for each height in the first region, the device can compile the wind speed prediction results obtained for each height to generate a wind speed prediction image that predicts the wind speed in the first region when the wind speed is the first speed and the wind direction is the first direction. At this time, the wind speed prediction image can be configured as a 3D image displayed in different colors according to the wind speed value, as illustrated in FIG. 6.

[0090] When a wind speed prediction image is generated, the device (200) can control the wind speed prediction image to be transmitted to the user terminal (100) so that the wind speed prediction image is displayed on the screen of the user terminal (100).

[0091] When the device (200) obtains wind pressure prediction results for each height in the first region, the device (200) can compile the wind pressure prediction results obtained for each height to generate a wind pressure prediction image that predicts the wind pressure in the first region when the wind speed is the first speed and the wind direction is the first direction. At this time, the wind pressure prediction image can be configured as a 3D image displayed in different colors according to the wind speed value, as illustrated in FIG. 6.

[0092] When a wind pressure prediction image is generated, the device (200) can control the wind pressure prediction image to be transmitted to the user terminal (100) so that the wind pressure prediction image is displayed on the screen of the user terminal (100).

[0093] Figure 7 is a flowchart for explaining a process for highlighting a wind speed hazard zone according to an embodiment.

[0094] Referring to FIG. 7, first, in step S701, the device (200) generates a wind speed prediction image, and then can classify an area in which the wind speed predicted on the wind speed prediction image is faster than a first reference speed as an abnormal wind speed area. Here, the first reference speed may be set differently depending on the embodiment.

[0095] That is, the device (200) can check the predicted wind speed value by coordinates on the wind speed prediction image, and connect points where the speed value is faster than the first reference speed to classify them as an abnormal wind speed zone.

[0096] In step S702, the device (200) can classify an area, among areas classified as wind speed abnormality areas, as a wind speed risk area, having an area larger than a first reference area. Here, the first reference area may be set differently depending on the embodiment.

[0097] In step S703, the device (200) can control the wind speed hazard zone to be highlighted and displayed when the wind speed prediction image is displayed.

[0098] That is, the device (200) can control the wind speed hazard zone to be highlighted and displayed when a wind speed prediction image is displayed on the screen of the user terminal (100), and at this time, the wind speed hazard zone can be highlighted and displayed using a predetermined special effect.

[0099] Figure 8 is a flowchart for explaining a process for highlighting a wind pressure risk area according to an embodiment.

[0100] Referring to FIG. 8, first, in step S801, the device (200) generates a wind pressure prediction image, and then can classify an area in which the wind pressure predicted on the wind pressure prediction image is higher than a first reference pressure as an abnormal wind pressure area. Here, the first reference pressure may be set differently depending on the embodiment.

[0101] That is, the device (200) can check the predicted wind pressure value on the wind pressure prediction image by coordinates, and connect points where the pressure value is higher than the first reference pressure to distinguish them as an abnormal wind pressure zone.

[0102] In step S802, the device (200) can classify an area among the areas classified as wind pressure abnormality areas, whose area is larger than the first reference area, as a wind pressure risk area.

[0103] In step S803, the device (200) can control the wind pressure risk area to be highlighted and displayed when the wind pressure prediction image is displayed.

[0104] That is, the device (200) can control the wind pressure risk area to be highlighted and displayed when a wind pressure prediction image is displayed on the screen of the user terminal (100), and at this time, the wind pressure risk area can be highlighted and displayed using a predetermined special effect.

[0105] Figure 9 is a flowchart illustrating an artificial intelligence model learning process for CFD-based prediction using multi-modality according to an embodiment.

[0106] According to one embodiment, each step illustrated in FIG. 9 may be performed after each step illustrated in FIG. 2 is performed.

[0107] Referring to FIG. 9, first, in step S901, the device (200) can match the first wind speed prediction result and the first wind pressure prediction result to generate the nth matching result.

[0108] In step S902, the device (200) can input the nth matching result into the second artificial intelligence model. Here, the second artificial intelligence model may be an MMTM (Multimodal Transfer Module for CNN Fusion) model trained to predict wind speed and wind pressure by considering the relationship between the wind speed prediction result and the wind pressure prediction result.

[0109] That is, the second artificial intelligence model may be an algorithm that, when the matching result of the wind speed prediction result and the wind pressure prediction result are input as input variables, predicts the wind speed in a specific area by considering the relationship between the wind speed prediction result and the wind pressure prediction result, and then predicts the wind pressure in a specific area, and then outputs the updated wind speed prediction result and the updated wind pressure prediction result. A detailed description of the second artificial intelligence model will be described later with reference to Fig. 11.

[0110] In step S903, if the device (200) predicts the wind speed and wind pressure for the first height in the first region through the n-th matching result, it can obtain the n+1-th wind speed prediction result and the n+1-th wind pressure prediction result from the second artificial intelligence model. Here, the n+1-th wind speed prediction result is a result of predicting the wind speed for the first height in the first region, and can be generated as an updated result from the first wind speed prediction result input as an input variable, and the n+1-th wind pressure prediction result is a result of predicting the wind pressure for the first height in the first region, and can be generated as an updated result from the first wind pressure prediction result input as an input variable.

[0111] Figure 10 is a diagram showing the operation process of the first artificial intelligence model according to an embodiment.

[0112] Referring to Fig. 10, the first artificial intelligence model can be configured with the existing structure of U-Net, and can learn the entire image data to separately predict wind pressure and wind velocity.

[0113] As an example, U-Net is a structure proposed in the paper “U-Net: Convolutional Networks for Biomedical Image Segmentation, MICCAI 2015” and can be used to perform accurate image segmentation with a very small amount of training data.

[0114] The learning device where the first artificial intelligence model is trained may be the same device as the device (200) or may be a separate device. The process by which the first artificial intelligence model is trained is described below.

[0115] The first AI model includes an input layer that receives training samples and an output layer that outputs training outputs, and can be learned based on the differences between the training outputs and first labels. Here, the first labels can be defined based on cross-sectional images of each region at different heights. The first AI model is connected into groups of multiple nodes and is defined by weights between the connected nodes and an activation function that activates the nodes.

[0116] The learning device can train the first artificial intelligence model using the Gradient Descent (GD) technique or the Stochastic Gradient Descent (SGD) technique. The learning device can use a loss function designed based on the outputs and labels of the first artificial intelligence model.

[0117] The learning device can calculate the training error using a predefined loss function. The loss function can be predefined as input variables, including labels, outputs, and parameters, where the parameters can be set by weights within the first AI model. For example, the loss function can be designed in the form of a Mean Square Error (MSE) or entropy. Various techniques or methods can be employed in the design of the loss function.

[0118] The learning device can use the backpropagation technique to identify weights that influence training errors. Here, the weights represent relationships between nodes within the first AI model. The learning device can utilize the SGD technique, which utilizes labels and outputs, to optimize the weights identified through the backpropagation technique. For example, the learning device can update the weights of a loss function defined based on the labels, outputs, and weights using the SGD technique.

[0119] According to one embodiment, the learning device can acquire labeled training cross-sectional images from a database. The learning device can acquire pre-labeled information for each cross-sectional image set for each height in each region, and the cross-sectional images can be labeled according to the pre-classified cross-sectional images for each height in each region.

[0120] According to one embodiment, the learning device can acquire a plurality of labeled training cross-sectional images and generate first training cross-sectional vectors based on the labeled training cross-sectional images. Various methods can be employed to extract the first training cross-sectional vectors.

[0121] According to one embodiment, a learning device can obtain first training outputs by applying first training cross-sectional vectors to a first artificial intelligence model. The learning device can train the first artificial intelligence model based on the first training outputs and the first labels. The learning device can calculate training errors corresponding to the first training outputs and optimize the connection relationship between nodes in the first artificial intelligence model to minimize the training errors, thereby training the first artificial intelligence model. Through this, the first artificial intelligence model can separately distinguish and predict wind speed and wind pressure on a cross-sectional image.

[0122] Figure 11 is a diagram showing the operation process of a second artificial intelligence model according to an embodiment.

[0123] Referring to Figure 11, the second AI model is an MMTM model that improves performance by fusing multimodal data. It exchanges and integrates information between different modalities to improve the learning process and effectively combines the features of different modalities to predict better results. Here, the different modalities may refer to wind speed prediction results and wind pressure prediction results.

[0124] In one embodiment, MMTM, a structure proposed in the paper “MMTM: Multimodal Transfer Module for CNN Fusion, CVPR 2020,” can be used to recalibrate channel-specific features of each CNN stream by leveraging knowledge of multiple modalities using squeeze and excitation operations.

[0125] Specifically, the second artificial intelligence model exchanges and integrates information between the wind speed prediction result and the wind pressure prediction result as illustrated in (a) of FIG. 11, then separately obtains the update result for the wind speed prediction result and the wind pressure prediction result as illustrated in (b) of FIG. 11, and then repeatedly obtains each update result as illustrated in (c) of FIG. 11.

[0126] A detailed description of the structure of the second artificial intelligence model will be provided later with reference to Fig. 15.

[0127] Figure 12 is a flowchart for explaining the process of updating wind speed prediction results and wind pressure prediction results according to an embodiment.

[0128] According to one embodiment, each step illustrated in FIG. 12 may be performed after each step illustrated in FIG. 9 is performed.

[0129] Referring to FIG. 12, first, in step S1201, the device (200) can match the n+1-th wind speed prediction result and the n+1-th wind pressure prediction result to generate the n+1-th matching result.

[0130] At step S1202, the device (200) can input the n+1-th matching result into the second artificial intelligence model.

[0131] At step S1203, when the wind speed and wind pressure for the first height in the first region are predicted through the n+1 matching result, the device (200) can obtain an updated n+1 wind speed prediction result and an updated n+1 wind pressure prediction result from the second artificial intelligence model.

[0132] At step S1204, the device (200) can check whether steps S1201 to S1203 have been repeated a reference number of times. Here, the reference number of times may be set differently depending on the embodiment, and a detailed description of the process of setting the reference number of times will be described later with reference to FIG. 14.

[0133] If it is confirmed in step S1204 that steps S1201 to S1203 have not been repeated a standard number of times, the process may be repeated again from step S1201, where the updated n+1-th wind speed prediction result and the updated n+1-th wind pressure prediction result are matched to generate the n+1-th matching result.

[0134] That is, the device (200) can repeat steps S1201 to S1203 a standard number of times.

[0135] Figure 13 is a diagram for explaining a process of predicting wind speed and wind pressure using a first artificial intelligence model and a second artificial intelligence model according to an embodiment.

[0136] Referring to Fig. 13, by using the first artificial intelligence model and the second artificial intelligence model, a composite model of U-net and MMTM can be provided, and by recognizing each wind speed and wind pressure as multimodal in the existing simple separate prediction, performance can be improved through data fusion, information exchange by modality, etc.

[0137] That is, the device (200) obtains the first wind speed prediction result and the first wind pressure prediction result from the first artificial intelligence model, and then inputs the first wind speed prediction result and the first wind pressure prediction result into the second artificial intelligence model, thereby repeatedly obtaining the updated n+1-th wind speed prediction result and the updated n+1-th wind pressure prediction result, thereby helping to improve the accuracy of the wind speed prediction result and the wind pressure prediction result.

[0138] Figure 14 is a flowchart for explaining a process for setting a reference number according to an embodiment.

[0139] Referring to FIG. 14, first, in step S1401, the device (200) can calculate an average value of wind speed for a first height in a first region as a first average speed based on the first wind speed prediction result.

[0140] That is, the device (200) can check the wind speed value for the first height in the first region for each coordinate based on the first wind speed prediction result, and calculate the average value of the speed values ​​checked for each coordinate as the first average speed. At this time, since the speed value inside the building is predicted to be 0, the device (200) can calculate the average value of the speed values ​​checked in the remaining coordinates excluding the inside of the building as the first average speed.

[0141] At step S1402, the device (200) can calculate the difference between the first speed and the first average speed as the second speed.

[0142] That is, when the device (200) compares the first speed and the first average speed and determines that the first speed is faster than the first average speed, the device calculates the second speed by subtracting the first average speed from the first speed, and when the device (200) determines that the first speed is slower than the first average speed, the device calculates the second speed by subtracting the first speed from the first average speed, and when the device (200) determines that the first speed and the first average speed are the same, the device can calculate the second speed as 0.

[0143] At step S1403, the device (200) can set the first value to a higher value within the reference range as the second speed increases. Here, the reference range can be set differently depending on the embodiment, and can be set within the range of 0 to 10, for example.

[0144] For example, if the second speed is calculated as 5 m / s, the device (200) can set the first value to 1, and if the second speed is calculated as 10 m / s, the device (200) can set the first value to 2.

[0145] At step S1404, the device (200) can set the first number as the reference number.

[0146] Figure 15 is a diagram illustrating an artificial intelligence model network structure for CFD-based prediction using multi-modality according to an embodiment.

[0147] Referring to Figure 15, first, the second artificial intelligence model can distinguish between the wind pressure feature map A and the wind speed feature map B in the n-th matching result.

[0148] Specifically, for input feature collection, the second artificial intelligence model is divided into two branches from the features extracted through the encoder of the U-Net structure, so that each wind pressure feature map (Pressure, A) and wind speed feature map (Velocity, B) can be processed in each branch.

[0149] Next, the second AI model can compress A to generate S_A and compress B to generate S_B.

[0150] Specifically, for compression (Squeeze), the second AI model can compress the spatial information of each branch into a smaller form (S_A, S_B) by performing a global average pooling operation on the feature map of each branch. At this time, the features of the two branches that have undergone the compression step can be created as a single feature map through vector concatenation.

[0151] Next, the second AI model can generate Z by combining S_A and S_B, generate E_A, an activation signal of a wind pressure branch, by passing Z through one connection layer, and generate E_B, an activation signal of a wind speed branch, by passing Z through another connection layer, and generate a rescaled A by multiplying A and E_A, and generate a rescaled B by multiplying B and E_B.

[0152] Specifically, to generate a multimodal excitation representation, the second AI model can combine the compressed feature maps into a single comprehensive representation, and calculate the importance of each channel of each branch to readjust the existing feature map. At this time, the second AI model can generate a new comprehensive representation Z that includes information of both branches by passing the combined vector through a fully connected layer, and can generate E_A, the activation signal of the wind pressure branch, by passing Z through two independent fully connected layers, and generate E_B, the activation signal of the wind speed branch, by passing it through another fully connected layer. Here, the output of the fully connected layer is converted to a value between 0 and 1 using a sigmoid activation function to calculate the importance of each channel, and readjustment can be performed by multiplying the calculated importance of the channel of each branch by the existing wind pressure feature map A and the wind speed feature map B.

[0153] This allows for more accurate wind tunnel environment predictions by comprehensively utilizing important information from both streams (wind speed and wind pressure), and enables better prediction performance than processing data from each stream individually using MMTM.

[0154] The second AI model can calculate the information entropy of each of A and B, and based on the calculated information entropy, generate mutual information, which is a value that measures the dependency between two variables.

[0155] According to one embodiment, mutual information is a value that measures the dependency between two variables, and may mean a value indicating how much information about one variable influences another variable. To calculate mutual information, feature maps of two branches can be used, information entropy can be calculated from the feature maps of each branch, and mutual information can be calculated based on this.

[0156] The second artificial intelligence model may include a mutual information maximization module, and the mutual information maximization module may input the value of mutual information into the loss function to maximize mutual information, thereby adjusting the mutual information to be maximized.

[0157] According to one embodiment, the Mutual Information Maximization Module can adjust the model during the learning process to maximize the mutual information of each branch by adding a mutual information value to the loss function to maximize the mutual information, and to integrate the Mutual Information Maximization Module into MMTM, the feature map of each branch can be provided as input to the Mutual Information Module, and the mutual information calculated by this module can be processed to be reflected in the loss function.

[0158] This can facilitate the sharing of useful information between the two branches by maximizing mutual information, and improve the model's predictive performance by effectively integrating the complementary information of the two branches.

[0159] The loss function can be formulated as follows:

[0160]

[0161] In terms of the loss function, feature maps are extracted from each of the wind pressure brunch (A) and the wind speed brunch (B), a neural network T_θ is trained that receives A and B as inputs and estimates mutual information, and the total loss function is calculated by adding the mutual information loss (L_{MI}) to the existing prediction loss (L_{MSE}), and the model can be trained through this.

[0162] Figure 16 is a flowchart illustrating a process for providing a CFD-based prediction solution utilizing an artificial intelligence model for smart air conditioning according to an embodiment.

[0163] Referring to FIG. 16, first, in step S1601, the device (200) can acquire a 3D image of a first space. Here, the first space may refer to a specific space in which structures are arranged, and the 3D image of the first space may refer to a 3D image in which structures are arranged within the first space, and may be acquired from an external server or a database of the device (200).

[0164] At step S1602, the device (200) can extract a cross-sectional view of the first height from the 3D image of the first space as the first-1 image.

[0165] At step S1603, the device (200) may generate 1-1 shape information, which is structure shape information of the first height, based on the 1-1 image. At this time, the 1-1 shape information may include information indicating which area is where the structure is located and which area is not where the structure is located when the height is the first height, and information indicating the boundary line of the structure.

[0166] At step S1604, the device (200) can generate a first input variable including first shape information, first speed, and first direction when the wind speed is set to the first speed and the wind direction is set to the first direction.

[0167] At step S1605, the device (200) can input the first input variable into the first artificial intelligence model. Here, the first artificial intelligence model may be a U-Net model trained to predict temperature, flow rate, and pressure based on CFD using structure shape information, wind speed, and wind direction as input variables.

[0168] That is, the 1-1 artificial intelligence model may be an algorithm that, when structure shape information, wind speed, and wind direction are input as input variables, predicts the temperature within a specific space, predicts the flow velocity within a specific space, and predicts the pressure within a specific space based on CFD, and then outputs the temperature prediction results, flow velocity prediction results, and pressure prediction results. The 1-1 artificial intelligence model may be designed and operated in the same manner as the first artificial intelligence model.

[0169] In step S1606, when the temperature, flow rate, and pressure for the first height of the first space are predicted through the first-1 input variable, the device (200) can obtain the first temperature prediction result, the first flow rate prediction result, and the first pressure prediction result from the first-1 artificial intelligence model. Here, the first temperature prediction result is a result of predicting the temperature for the first height of the first space, and can be generated as an image in which each coordinate is displayed in a different color according to the temperature value of the predicted temperature for each coordinate, the first flow rate prediction result is a result of predicting the flow rate for the first height of the first space, and can be generated as an image in which each coordinate is displayed in a different color according to the speed value of the predicted flow rate for each coordinate, and the first pressure prediction result is a result of predicting the pressure for the first height of the first space, and can be generated as an image in which each coordinate is displayed in a different color according to the pressure value of the predicted pressure for each coordinate.

[0170] Figure 17 is a diagram showing temperature prediction results obtained for each height, flow rate prediction results obtained for each height, and pressure prediction results obtained for each height according to an embodiment.

[0171] According to one embodiment, the device (200) can obtain a temperature prediction result, which is a result of predicting the temperature for each height of the first space, for each height in the same manner as the method of obtaining the first temperature prediction result, which is a result of predicting the temperature for the first height of the first space.

[0172] That is, as shown in Fig. 17, temperature prediction results can be obtained for each height.

[0173] When the device (200) obtains the temperature prediction results for each height of the first space for each height, it can compile the temperature prediction results obtained for each height and generate a temperature prediction image that predicts the temperature of the first space when the wind speed is a first speed and the wind direction is a first direction. At this time, the temperature prediction image can be composed of a 3D image displayed in different colors according to the temperature value.

[0174] When a temperature prediction image is generated, the device (200) can control the temperature prediction image to be displayed on the screen of the user terminal (100) by transmitting the temperature prediction image to the user terminal (100).

[0175] According to one embodiment, the device (200) can obtain a velocity prediction result, which is a result of predicting a velocity for each height of the first space, for each height in the same manner as the method of obtaining the first velocity prediction result, which is a result of predicting a velocity for the first height of the first space.

[0176] That is, as shown in Fig. 17, the flow rate prediction result can be obtained for each height.

[0177] When the device (200) obtains the results of predicting the flow velocity for each height of the first space for each height, it can compile the results of predicting the flow velocity obtained for each height and generate a velocity prediction image predicting the flow velocity of the first space when the wind velocity is the first velocity and the wind direction is the first direction. At this time, the velocity prediction image can be composed of a 3D image displayed in different colors according to the velocity value of the flow velocity.

[0178] When a velocity prediction image is generated, the device (200) can control the velocity prediction image to be transmitted to the user terminal (100) so that the velocity prediction image is displayed on the screen of the user terminal (100).

[0179] According to one embodiment, the device (200) can obtain pressure prediction results, which are results of predicting pressure for each height of the first space, for each height in the same manner as the method of obtaining the first pressure prediction result, which is a result of predicting pressure for the first height of the first space.

[0180] That is, as shown in Fig. 17, pressure prediction results can be obtained for each height.

[0181] When the device (200) obtains the pressure prediction results for each height of the first space, the device can compile the pressure prediction results obtained for each height and generate a pressure prediction image that predicts the pressure of the first space when the wind speed is a first speed and the wind direction is a first direction. At this time, the pressure prediction image can be composed of a 3D image displayed in different colors according to the pressure value.

[0182] When a pressure prediction image is generated, the device (200) can control the pressure prediction image to be displayed on the screen of the user terminal (100) by transmitting the pressure prediction image to the user terminal (100).

[0183] The device (200) can operate smart air conditioning by utilizing a temperature prediction image, a flow rate prediction image, and a pressure prediction image.

[0184] Figure 18 is a flowchart illustrating a process for obtaining temperature prediction results, flow rate prediction results, and pressure prediction results by utilizing multi-modality according to an embodiment.

[0185] According to one embodiment, each step illustrated in FIG. 18 may be performed after each step illustrated in FIG. 16 is performed.

[0186] Referring to FIG. 18, first, in step S1801, the device (200) can match the first temperature prediction result, the first flow rate prediction result, and the first pressure prediction result to generate the m matching result.

[0187] In step S1802, the device (200) can input the m matching result into the 2-1 artificial intelligence model. Here, the 2-1 artificial intelligence model may be an MMTM (Multimodal Transfer Module for CNN Fusion) model trained to predict temperature, flow rate, and pressure by considering the relationship between the temperature prediction result, the flow rate prediction result, and the pressure prediction result.

[0188] That is, the 2-1 artificial intelligence model may be an algorithm that, when the matching results of the temperature prediction results, the flow rate prediction results, and the pressure prediction results are input as input variables, considers the relationship between the temperature prediction results, the flow rate prediction results, and the pressure prediction results, predicts the temperature within a specific space, predicts the flow rate within a specific space, and predicts the pressure within a specific space, and then outputs the updated temperature prediction results, the updated flow rate prediction results, and the updated pressure prediction results. The 2-1 artificial intelligence model may be designed and operated in the same manner as the 2nd artificial intelligence model.

[0189] In step S1803, if the temperature, velocity, and pressure for the first height of the first space are predicted through the m-th matching result, the device (200) can obtain the m+1-th temperature prediction result, the m+1-th velocity prediction result, and the m+1-th pressure prediction result from the 2-1 artificial intelligence model. Here, the m+1-th temperature prediction result is a result of predicting the temperature for the first height of the first space, and can be generated as an updated result from the first temperature prediction result input as an input variable, the m+1-th velocity prediction result is a result of predicting the velocity for the first height of the first space, and can be generated as an updated result from the first velocity prediction result input as an input variable, and the m+1-th pressure prediction result is a result of predicting the pressure for the first height of the first space, and can be generated as an updated result from the first pressure prediction result input as an input variable.

[0190] Figure 19 is a diagram showing the operation process of the 1-1 artificial intelligence model according to an embodiment.

[0191] Referring to Figure 19, the 1-1 artificial intelligence model can be configured with the existing structure of U-Net, and can learn the entire image data to separately predict pressure, velocity, and temperature.

[0192] Figure 20 is a diagram showing the operation process of the 2-1 artificial intelligence model according to an embodiment.

[0193] Referring to Figure 20, the MMTM model, a second-generation artificial intelligence model, is designed to improve performance by fusing multimodal data. It exchanges and integrates information between different modalities to improve the learning process and effectively combines the features of different modalities to predict better results. Here, the different modalities may refer to temperature prediction results, flow rate prediction results, and pressure prediction results.

[0194] Specifically, the 2-1 artificial intelligence model exchanges and integrates information between the temperature prediction result, the flow rate prediction result, and the pressure prediction result, and then separately obtains the update result for the temperature prediction result, the update result for the flow rate prediction result, and the update result for the pressure prediction result, and then repeatedly obtains each update result.

[0195] Figures 21 to 24 are flowcharts for explaining a process for additionally obtaining prediction results for an intermediate height according to an embodiment.

[0196] According to one embodiment, each step illustrated in FIGS. 21 to 24 may be performed after each step illustrated in FIG. 2 is performed.

[0197] Referring to FIGS. 21 to 24, first, in step S2101, the device (200) can extract a cross-section of a second height from a 3D image of a first region as a second image.

[0198] For example, the device (200) can extract a cross-section of 0 m from a 3D image of a first region as a first image, and can extract a cross-section of 3 m from a 3D image of the first region as a second image.

[0199] At step S2102, the device (200) can generate second shape information, which is building shape information of a second height, based on the second image.

[0200] At step S2103, the device (200) can generate a second input variable including second shape information, a first speed, and a first direction.

[0201] At step S2104, the device (200) can input a second input variable into the first artificial intelligence model.

[0202] In step S2105, when the device (200) predicts the wind speed and wind pressure for the second height in the first region through the second input variable, the device can obtain the second wind speed prediction result and the second wind pressure prediction result from the first artificial intelligence model. Here, the second wind speed prediction result is a result of predicting the wind speed for the second height in the first region, and can be generated as an image in which each coordinate is displayed in a different color according to the speed value of the wind speed predicted for each coordinate, and the second wind pressure prediction result is a result of predicting the wind pressure for the second height in the first region, and can be generated as an image in which each coordinate is displayed in a different color according to the pressure value of the wind pressure predicted for each coordinate.

[0203] After step S2105, in step S2201, the device (200) can calculate an average value of wind speed for a first height in a first region as a first average speed based on the first wind speed prediction result, and can calculate an average value of wind speed for a second height in the first region as a second average speed based on the second wind speed prediction result.

[0204] That is, the device (200) can check the wind speed value for the first height in the first region for each coordinate based on the first wind speed prediction result, and calculate the average value of the speed values ​​checked for each coordinate as the first average speed.

[0205] In addition, the device (200) can check the wind speed value for the second height in the first region for each coordinate based on the second wind speed prediction result, and calculate the average value of the speed values ​​checked for each coordinate as the second average speed.

[0206] At step S2202, the device (200) can calculate the difference between the first average speed and the second average speed as the first difference speed.

[0207] That is, when the device (200) compares the first average speed and the second average speed and determines that the first average speed is faster than the second average speed, the device calculates the first difference speed by subtracting the second average speed from the first average speed, and when the device (200) determines that the first average speed is slower than the second average speed, the device calculates the first difference speed by subtracting the first average speed from the second average speed, and when the device (200) determines that the first average speed and the second average speed are the same, the device can calculate the first difference speed as 0.

[0208] At step S2203, the device (200) can check whether the first difference speed is faster than the second reference speed. Here, the second reference speed may be set differently depending on the embodiment.

[0209] If it is determined in step S2203 that the first difference speed is faster than the second reference speed, then in step S2204, the device (200) may determine that an additional prediction result for an intermediate height between the first height and the second height is required.

[0210] After step S2105, in step S2301, the device (200) can calculate an average value of wind pressure for a first height in a first region as a first average pressure based on the first wind pressure prediction result, and can calculate an average value of wind pressure for a second height in the first region as a second average pressure based on the second wind pressure prediction result.

[0211] That is, the device (200) can check the pressure value of the wind pressure for the first height in the first region for each coordinate based on the first wind pressure prediction result, and calculate the average value of the pressure values ​​checked for each coordinate as the first average pressure.

[0212] In addition, the device (200) can check the pressure value of the wind pressure for the second height in the first region for each coordinate based on the second wind pressure prediction result, and calculate the average value of the pressure values ​​checked for each coordinate as the second average pressure.

[0213] At step S2302, the device (200) can calculate the difference between the first average pressure and the second average pressure as the first difference pressure.

[0214] That is, when the device (200) compares the first average pressure and the second average pressure and determines that the first average pressure is higher than the second average pressure, the device can calculate the first difference pressure by subtracting the second average pressure from the first average pressure, and when the device (200) determines that the first average pressure is lower than the second average pressure, the device can calculate the first difference pressure by subtracting the first average pressure from the second average pressure, and when the device (200) determines that the first average pressure and the second average pressure are the same, the device can calculate the first difference pressure as 0.

[0215] At step S2303, the device (200) can check whether the first differential pressure is higher than the second reference pressure. Here, the second reference pressure may be set differently depending on the embodiment.

[0216] If it is determined in step S2303 that the first differential pressure is higher than the second reference pressure, then in step S2204, the device (200) may determine that a prediction result for an intermediate height between the first height and the second height is additionally required.

[0217] After step S2204 or step S2304, in step S2401, the device (200) can calculate the median value of the first height and the second height as the third height.

[0218] At step S2402, the device (200) can extract a cross-section of a third height from a 3D image of the first region as a third image.

[0219] At step S2403, the device (200) can generate third shape information, which is building shape information of a third height, based on the third image.

[0220] At step S2404, the device (200) can generate a third input variable including third shape information, a first speed, and a first direction.

[0221] At step S2405, the device (200) can input a third input variable into the first artificial intelligence model.

[0222] In step S2406, when the wind speed and wind pressure for the third height in the first region are predicted through the third input variable, the device (200) can obtain the third wind speed prediction result and the third wind pressure prediction result from the first artificial intelligence model. Here, the third wind speed prediction result is a result of predicting the wind speed for the third height in the first region, and can be generated as an image in which each coordinate is displayed in a different color according to the speed value of the wind speed predicted for each coordinate, and the third wind pressure prediction result is a result of predicting the wind pressure for the third height in the first region, and can be generated as an image in which each coordinate is displayed in a different color according to the pressure value of the wind pressure predicted for each coordinate.

[0223] As described above, when the device (200) obtains a result of predicting wind speed for each height in the first region, if an additional prediction result for an intermediate height is required, it can additionally obtain a third wind speed prediction result, and when obtaining a result of predicting wind pressure for each height in the first region, if an additional prediction result for an intermediate height is required, it can additionally obtain a third wind pressure prediction result.

[0224] According to one embodiment, when the device (200) obtains the results of predicting the flow rate for each height of the first space, if an additional prediction result for an intermediate height is required, it can additionally obtain a third flow rate prediction result, and when obtaining the results of predicting the pressure for each height of the first space, if an additional prediction result for an intermediate height is required, it can additionally obtain a third pressure prediction result. At this time, since the process of obtaining the third flow rate prediction result and the third pressure prediction result is performed in the same manner as the process of obtaining the third wind speed prediction result and the third wind pressure prediction result, a detailed description thereof will be omitted.

[0225] Figures 25 and 26 are flowcharts for explaining a process for determining whether there is a problem with the shape of a building according to one embodiment.

[0226] According to one embodiment, each step illustrated in FIGS. 25 and 26 may be performed after each step illustrated in FIG. 24 is performed.

[0227] Referring to FIG. 25, first, in step S2501, the device (200) can calculate the average value of the wind speed for the third height in the first region as the third average speed based on the third wind speed prediction result.

[0228] At step S2502, the device (200) can calculate the difference between the first average speed and the third average speed as a second difference speed, and calculate the difference between the second average speed and the third average speed as a third difference speed.

[0229] At step S2503, the device (200) can check whether the second difference speed is faster than the third reference speed. Here, the third reference speed may be set differently depending on the embodiment.

[0230] If it is confirmed in step S2503 that the second difference speed is faster than the third reference speed, then in step S2505, the device (200) may determine that there is a problem with the shape of the building located within the first area.

[0231] At step S2504, the device (200) can check whether the third difference speed is faster than the third reference speed.

[0232] If it is confirmed in step S2504 that the third difference speed is faster than the third reference speed, then in step S2505, the device (200) may determine that there is a problem with the shape of the building located within the first area.

[0233] That is, if the device (200) determines that at least one of the second difference speed and the third difference speed is faster than the third reference speed, it can determine that there is a problem with the shape of a building located within the first area.

[0234] Referring to FIG. 26, first, in step S2601, the device (200) can calculate the average value of the wind pressure for the third height in the first region as the third average pressure based on the third wind pressure prediction result.

[0235] In step S2602, the device (200) can calculate the difference between the first average pressure and the third average pressure as the second differential pressure, and can calculate the difference between the second average pressure and the third average pressure as the third differential pressure.

[0236] At step S2603, the device (200) can check whether the second differential pressure is higher than the third reference pressure. Here, the third reference pressure can be set differently depending on the embodiment.

[0237] If it is confirmed in step S2603 that the second differential pressure is higher than the third reference pressure, then in step S2605, the device (200) can determine that there is a problem with the shape of the building located within the first region.

[0238] At step S2604, the device (200) can check whether the third differential pressure is higher than the third reference pressure.

[0239] If it is confirmed in step S2604 that the third differential pressure is higher than the third reference pressure, then in step S2605, the device (200) may determine that there is a problem with the shape of the building located within the first region.

[0240] That is, if the device (200) determines that at least one of the second differential pressure and the third differential pressure is higher than the third reference pressure, it can determine that there is a problem with the shape of a building located within the first region.

[0241] If the device (200) determines that there is a problem with the shape of a building located within the first region, when a wind speed prediction image is displayed on the screen of the user terminal (100), a notification message can be controlled to be displayed as a pop-up to inform that there is a problem with the shape of the building located within the first region.

[0242] According to one embodiment, the device (200) can determine whether there is a problem with the shape of a structure located within a first space in the same manner as determining whether there is a problem with the shape of a building located within a first area.

[0243] Figure 27 is an example diagram of the configuration of a device according to one embodiment.

[0244] A device (200) according to an embodiment includes a processor (210) and a memory (220). The processor (210) may include at least one of the devices described above with reference to FIGS. 1 to 26, or may perform at least one method described above with reference to FIGS. 1 to 26. A person or organization using the device (200) may provide services related to some or all of the methods described above with reference to FIGS. 1 to 26.

[0245] The memory (220) can store information related to the methods described above or store a program in which the methods described below are implemented. The memory (220) can be a volatile memory or a non-volatile memory.

[0246] The processor (210) can execute a program and control the device (200). The code of the program executed by the processor (210) can be stored in the memory (220). The device (200) can be connected to an external device (e.g., a personal computer or a network) through an input / output device (not shown in the drawing) and exchange data through wired or wireless communication.

[0247] The device (200) can be used to train an artificial intelligence model or utilize a trained artificial intelligence model. The memory (220) can include an artificial intelligence model that is being trained or has been trained. The processor (210) can train or execute an artificial intelligence model algorithm stored in the memory (220). The training device that trains an artificial intelligence model and the device (200) that utilizes the trained artificial intelligence model may be the same or separate.

[0248] The embodiments described above may be implemented using hardware components, software components, and / or a combination of hardware components and software components. For example, the devices, methods, and components described in the embodiments may be implemented using one or more general-purpose computers or special-purpose computers, such as, for example, a processor, a controller, an arithmetic logic unit (ALU), a digital signal processor, a microcomputer, a field programmable gate array (FPGA), a programmable logic unit (PLU), a microprocessor, or any other device capable of executing instructions and responding to them. The processing device may execute an operating system (OS) and one or more software applications running on the operating system. The processing device may also access, store, manipulate, process, and generate data in response to the execution of the software. For ease of understanding, the processing device is sometimes described as being used alone; however, one of ordinary skill in the art will recognize that the processing device may include multiple processing elements and / or multiple types of processing elements. For example, a processing unit may include multiple processors, or a processor and a controller. Other processing configurations, such as parallel processors, are also possible.

[0249] The method according to the embodiment may be implemented in the form of program commands that can be executed through various computer means and recorded on a computer-readable medium. The computer-readable medium may include program commands, data files, data structures, etc., alone or in combination. The program commands recorded on the medium may be those specially designed and configured for the embodiment or may be those known and available to those skilled in the art of computer software. Examples of the computer-readable recording medium include magnetic media such as hard disks, floppy disks, and magnetic tapes, optical media such as CD-ROMs and DVDs, magneto-optical media such as floptical disks, and hardware devices specially configured to store and execute program commands, such as ROMs, RAMs, and flash memories. Examples of the program commands include not only machine language codes generated by a compiler, but also high-level language codes that can be executed by a computer using an interpreter, etc. The hardware devices described above may be configured to operate as one or more software modules to perform the operations of the embodiment, and vice versa.

[0250] Software may include a computer program, code, instructions, or a combination of one or more of these, which may configure a processing device to perform a desired operation or may, independently or collectively, command the processing device. The software and / or data may be permanently or temporarily embodied in any type of machine, component, physical device, virtual equipment, computer storage medium or device, or transmitted signal wave, for interpretation by the processing device or for providing instructions or data to the processing device. The software may also be distributed over networked computer systems and stored or executed in a distributed manner. The software and data may be stored on one or more computer-readable recording media.

[0251] Although the embodiments described above have been described with limited drawings, those skilled in the art will appreciate that various technical modifications and variations can be applied based on the above. For example, appropriate results can still be achieved even if the described techniques are performed in a different order than described, and / or components of the described systems, structures, devices, circuits, etc. are combined or combined in a different manner than described, or are replaced or substituted with other components or equivalents.

[0252] Therefore, other implementations, other embodiments, and equivalents to the claims also fall within the scope of the claims described below.

Claims

1. A method for providing a CFD-based prediction solution utilizing an artificial intelligence model performed by a device, Step of acquiring a 3D image of the first region; A step of extracting a cross-section of a first height from a 3D image of the first region as a first image; A step of generating first shape information, which is building shape information of the first height, based on the first image; A step of generating a first input variable including the first shape information, the first speed, and the first direction when the wind speed is set to a first speed and the wind direction is set to a first direction; A step of inputting the first input variable into a first artificial intelligence model, which is a U-Net model trained to predict wind speed and wind pressure based on CFD using building shape information, wind speed, and wind direction as input variables; and When the wind power and wind pressure for the first height in the first region are predicted through the first input variable, a step of obtaining a first wind speed prediction result, which is a result of predicting the wind speed for the first height in the first region, and a first wind pressure prediction result, which is a result of predicting the wind pressure for the first height in the first region, is included. A step of extracting a cross-section of a second height from a 3D image of the first region as a second image; A step of generating second shape information, which is shape information of a building of the second height, based on the second image; A step of generating a second input variable including the second shape information, the first speed, and the first direction; A step of inputting the second input variable into the first artificial intelligence model; A step of obtaining a second wind speed prediction result, which is a result of predicting wind speed for the second height in the first region, and a second wind pressure prediction result, which is a result of predicting wind pressure for the second height in the first region, from the first artificial intelligence model, when wind speed and wind pressure for the second height in the first region are predicted through the second input variable; A step of calculating an average value of wind speed for the first height in the first region as a first average speed based on the first wind speed prediction result, and calculating an average value of wind speed for the second height in the first region as a second average speed based on the second wind speed prediction result; A step of calculating the difference between the first average speed and the second average speed as a first difference speed; If it is confirmed that the first difference speed is faster than the preset second reference speed, a step of determining that a prediction result for an intermediate height between the first height and the second height is additionally required; A step of calculating an average value of wind pressure for the first height in the first region as a first average pressure based on the first wind pressure prediction result, and calculating an average value of wind pressure for the second height in the first region as a second average pressure based on the second wind pressure prediction result; A step of calculating the difference between the first average pressure and the second average pressure as the first difference pressure; If the first differential pressure is confirmed to be higher than the preset second reference pressure, a step of determining that a prediction result for an intermediate height between the first height and the second height is additionally required; If it is determined that an additional prediction result for an intermediate height between the first height and the second height is required, a step of calculating an intermediate value between the first height and the second height as a third height; A step of extracting a cross-section of the third height from the 3D image of the first region as a third image; A step of generating third shape information, which is information on the shape of a building of the third height, based on the third image; A step of generating a third input variable including the third shape information, the first speed, and the first direction; a step of inputting the third input variable into the first artificial intelligence model; and If the wind speed and wind pressure for the third height in the first region are predicted through the third input variable, the method further includes a step of obtaining a third wind speed prediction result, which is a result of predicting the wind speed for the third height in the first region, and a third wind pressure prediction result, which is a result of predicting the wind pressure for the third height in the first region, from the first artificial intelligence model. A method for providing a CFD-based prediction solution using an artificial intelligence model.

2. In paragraph 1, When a wind speed prediction result, which is a result of predicting the wind speed for each height of the first region, is obtained for each height, a step of collecting the wind speed prediction results obtained for each height and generating a wind speed prediction image that predicts the wind speed of the first region when the wind speed is the first speed and the wind direction is the first direction; and When the wind pressure prediction result, which is the result of predicting the wind pressure for each height of the first region, is obtained for each height, the method further includes a step of generating a wind pressure prediction image that predicts the wind pressure of the first region when the wind speed is the first speed and the wind direction is the first direction by collecting the wind pressure prediction results obtained for each height. A method for providing a CFD-based prediction solution using an artificial intelligence model.

3. In paragraph 2, After the step of generating the above wind speed prediction image, A step of dividing an area in which the wind speed predicted on the above wind speed prediction image is faster than a preset first reference speed into an abnormal wind speed area; A step of classifying an area among the areas classified as the above wind speed abnormality area into a wind speed risk area, the area having an area larger than a preset first reference area; and When the above wind speed prediction image is displayed, the step of controlling the wind speed risk zone to be highlighted and displayed is further included. After the step of generating the above wind pressure prediction image, A step of classifying an area in which the result of predicting wind pressure on the above wind pressure prediction image is higher than a preset first reference pressure as an abnormal wind pressure area; A step of classifying an area among the areas classified as the above wind pressure abnormality area into an area with an area larger than the first reference area as a wind pressure risk area; and When the above wind pressure prediction image is displayed, the step of controlling the wind pressure risk area to be highlighted and displayed is further included. A method for providing a CFD-based prediction solution using an artificial intelligence model.

Images