KR20210122533A - Deep learning based Computer-Aided Engineering analysis result generation method, and computer program therefor - Google Patents

Abstract

The present invention relates to a deep learning based computer-aided engineering (CAE) analysis result generation method and a computer program therefor. According to an embodiment of the present invention, disclosed are the deep learning based CAE analysis result generation method and the computer program therefor configured to learn visualized information of a CAE analysis result generated based on shapes and analysis conditions of a plurality of products through a deep learning model and input a shape and/or an analysis condition of a product which needs CAE analysis to the learned deep learning model to generate an analysis result.

Description

Deep learning based Computer-Aided Engineering analysis result generation method, and computer program therefor}

The present invention relates to a method for calculating a CAE analysis result based on deep learning and a computer program therefor, wherein the visualization information of the CAE analysis result calculated based on the shape and analysis conditions of multiple products is learned through a deep learning model, and CAE analysis is performed. It relates to a method for calculating CAE analysis results based on deep learning configured to calculate analysis results by inputting the shape and/or analysis conditions of this required product into a learned deep learning model, and to a computer program therefor.

CAE (Computer-Aided Engineering) is to perform simulation, verification, and optimization from the viewpoint of product function and performance based on a three-dimensional model of a product created on a computer through CAD (Computer-Aided Design).

CAE is applied in fields such as machinery, aviation, electricity, and automobiles, and is generally based on finite element analysis (FEA).

In general, CAE analysis proceeds in three steps: pre-processing, analysis solving, and post-processing.

The pre-processing is a step in which CAD data is converted to finite element modeling, and boundary conditions (constraint conditions) and load conditions are applied. In finite element modeling, a finite element constituting a mesh is the minimum unit constituting the shape of an object, and a point forming the vertex of each element is defined as a node.

Analysis is a step in which the nodes and elements constructed through preprocessing are calculated using numerical equations.

Post-processing is a step of calculating visualization information (shape, color distribution, and legend information) of the CAE analysis result by visualizing the result value of the analysis step.

As an example of the prior art related to CAE analysis, Korean Patent Registration No. 10-0863075 (registered on October 06, 2008) relates to a web-based automatic CEA analysis method and system, and three-dimensional three-dimensional shape data created by a CAD program. When importing from CAE program, by automatically mapping the grid size value and boundary condition value corresponding to the property values of each grid, CAE analysis is automatically possible without the manipulation of complicated menu functions provided in the CAE program. The configuration provided was proposed.

However, when CAE analysis based on the conventional numerical analysis is performed on products with similar shape and material conditions and similar problem occurrence factors, similar characteristics such as shape, boundary condition (constraint condition) and load condition of the product are obtained. In spite of this, a complex and time-consuming numerical analysis has to be performed one time and repeatedly for each product, which increases the product development period and has limitations in that it is difficult to quickly diagnose abnormalities.

Republic of Korea Patent Registration 10-0863075 (Registered on October 06, 2008)

The present invention was devised in consideration of the above problems, and it learns the visualization information of the CAE analysis result calculated based on the shape and analysis conditions of multiple products through a deep learning model, and learns the shape and / Alternatively, an object of the present invention is to provide a method for calculating CAE analysis results based on deep learning configured to calculate analysis results by inputting analysis conditions to a learned deep learning model, and a computer program for this purpose.

According to one aspect of the present invention in consideration of the above object, as a CAE (Computer-Aided Engineering) analysis result calculation method executed by a deep learning-based computing means, 1) deep learning the CAE analysis results of a plurality of products as learning data generating a CAE analysis model; 2) inputting basic analysis information of a product requiring CAE analysis to the CAE analysis model; and 3) calculating a CAE analysis result for the analysis basic information by the CAE analysis model, wherein the learning data, the basic analysis information, and the CAE analysis result include outer edge line information indicating the shape of the product, the Deep, characterized in that it comprises at least one of color distribution map information that classifies and displays the analysis result for shape by color, and legend information that describes the maximum and minimum numerical values of colors included in the color distribution information. A method for calculating a CAE analysis result based on learning is disclosed.

Preferably, the deep learning of step 1) is supervised learning, the input value of the learning data is information about the outer edge line indicating the shape of the product, and the output value of the learning data is the color classification of the analysis result for the shape. color distribution information and legend information describing the maximum and minimum numerical values of colors included in the color distribution information.

Preferably, the deep learning in step 1) is unsupervised learning, and the input value of the learning data is outer edge line information indicating the shape of the product, color distribution information for dividing and displaying the analysis result for the shape by color, and the Legend information that describes the maximum and minimum numerical values of colors included in the color distribution information.

Preferably, in step 1), the outer edge line information and color distribution information included in the learning data are pre-processed to be input as a condition for looking at the product from a preset direction, and in step 2), the analysis basic information is The included outer edge line information and color distribution information are pre-processed to be input as a condition for looking at the product from a preset direction.

Preferably, in the color distribution information, an analysis result for any one of a stress distribution, a pressure distribution, and a fluid pressure distribution is displayed by color.

According to one aspect of the present invention, computer-readable to execute a deep learning-based CAE analysis result calculation method in combination with hardware including a memory for storing one or more instructions and a processor for executing the one or more instructions stored in the memory As a computer program stored in a medium, the deep learning-based CAE analysis result calculation method includes: 1) generating a CAE analysis model by deep learning the CAE analysis results of a plurality of products as learning data; 2) inputting basic analysis information of a product requiring CAE analysis to the CAE analysis model; and 3) calculating a CAE analysis result for the analysis basic information by the CAE analysis model, wherein the learning data, the basic analysis information, and the CAE analysis result include outer edge line information indicating the shape of the product, the A computer characterized in that it comprises at least one of color distribution map information that classifies and displays the analysis result for shape by color and legend information that describes the maximum and minimum numerical values of colors included in the color distribution information. A computer program stored in a readable medium is disclosed.

As such, the present invention is configured to calculate the analysis result by inputting the shape and/or analysis conditions of the product requiring CAE analysis to the learned deep learning model, so that similar problem occurrence factors are inherent while having similar shape and material conditions. CAE analysis is possible in a short time without complex and time-consuming numerical analysis of products, and it has the advantage of shortening the product development period and enabling rapid abnormal diagnosis.

1 is a configuration diagram of a deep learning-based CAE analysis result calculation system according to an embodiment of the present invention;
2 is a schematic diagram from a hardware perspective of a deep learning-based CAE analysis result calculation system according to an embodiment of the present invention;
3 is a flowchart of a method for calculating a CAE analysis result based on deep learning according to an embodiment of the present invention;
4 is a flowchart of a method for calculating CAE analysis results based on deep learning of a supervised learning method according to an embodiment of the present invention;
5 is a flowchart of a CAE analysis result calculation method based on deep learning of an unsupervised learning method according to an embodiment of the present invention;
6 is a schematic diagram for explaining load conditions and restraint conditions in CAE analysis;
7 is a schematic diagram for explaining learning data according to an embodiment of the present invention;
8 is a schematic diagram for explaining information about an outer edge line and color distribution map according to an embodiment of the present invention;
9 is a schematic diagram for explaining legend information according to an embodiment of the present invention;
10 is a schematic diagram for explaining a CAE analysis model of a supervised learning method according to an embodiment of the present invention;
11 is a schematic diagram for explaining a CAE analysis model of an unsupervised learning method according to an embodiment of the present invention;
12 is another schematic diagram for explaining a CAE analysis model of an unsupervised learning method according to an embodiment of the present invention.

The present invention may be embodied in various other forms without departing from its technical spirit or main characteristics. Accordingly, the embodiments of the present invention are merely illustrative in all respects and should not be construed as limiting.

The terms 1st, 2nd, etc. are used only for the purpose of distinguishing one component from another component. For example, without departing from the scope of the present invention, 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.

When a component is referred to as being “connected” or “connected” to another component, it may be directly connected or connected to the other component, but another component may exist in between.

As used herein, the singular expression includes the plural expression unless the context clearly dictates otherwise. In the present application, terms such as "comprise" or "comprising", "have" and the like are intended to represent the presence of elements or combinations thereof described in the specification, and the possibility that other elements or features may be present or added. It is not precluded.

Hereinafter, preferred embodiments according to the present invention will be described in detail with reference to the accompanying drawings.

1 is a configuration diagram of a deep learning-based CAE analysis result calculation system according to an embodiment of the present invention, and FIG. 2 is a schematic diagram from a hardware perspective of a deep learning-based CAE analysis result calculation system according to an embodiment of the present invention. .

The CAE analysis result calculation system 100 of this embodiment is a computing means in which a computer program for executing a deep learning-based CAE analysis result calculation method is driven.

The CAE analysis result calculation system 100 of this embodiment, from a functional point of view, inputs and pre-processes the CAE analysis results of a plurality of products as learning data, the learning data input unit 102, and input analysis basic information of products requiring CAE analysis and The analysis basic information input unit 104 to pre-process, the CAE analysis model 106 that performs deep learning using the input learning data and calculates the CAE analysis result for the input analysis basic information, the calculated CAE analysis result is visually or It includes a CAE analysis result output unit 108 for outputting numerical information, and a storage unit (not shown) for storing various information calculated as intermediate or final results in the execution process of each functional element.

Referring to FIG. 2 , from a hardware point of view, the CAE analysis result calculation system 100 of this embodiment includes a memory 2 that stores one or more instructions and a processor 4 that executes the one or more instructions stored in the memory 2 . ), and is a computing device in which a computer program stored in a medium is executed to execute a method for calculating a CAE analysis result based on deep learning. The CAE analysis result calculation system 100 of the present embodiment may include a data input/output interface 6 , a communication interface 8 , a data display unit 3 , and a data storage unit 5 .

3 is a flowchart of a method for calculating a CAE analysis result based on deep learning according to an embodiment of the present invention, FIG. 6 is a schematic diagram for explaining a load condition and a constraint condition in CAE analysis, and FIG. 7 is an embodiment of the present invention Fig. 8 is a schematic diagram for explaining the learning data according to the present invention, Fig. 8 is a schematic diagram for explaining the outer edge line information and color distribution information according to an embodiment of the present invention, Fig. 9 is a schematic diagram for explaining the legend information according to an embodiment of the present invention It is a schematic diagram for

In step 1), the CAE analysis result calculation system 100 generates a CAE analysis model by deep learning the CAE analysis results of a plurality of products as learning data.

The product of this embodiment is a product having a three-dimensional shape that requires analysis of stress distribution, pressure distribution or fluid pressure distribution, for example, a fan or propeller that requires noise analysis, parts for home appliances that require stress analysis, mechanical parts, etc. this can be

6 is a schematic diagram for explaining a process of calculating a stress distribution for a kimchi refrigerator pedestal (bracket) for commercial use, and illustrates a state in which a load condition and a boundary condition (constraint condition) are set. The constraint condition is entered into the wheel axle joint as shown in FIG. 6 (assuming it is fixed and does not move), and the load condition is 1/2 of the refrigerator load on the top surface directly receiving the weight of the refrigerator (a condition in which two pedestals are installed) is assigned a value corresponding to

Taking the stress analysis of a structure as an example, the deep learning of this embodiment, when a plurality of products have the following analysis conditions, input the corresponding products as learning data and/or analysis basic information for one CAE analysis model, resulting in analysis results can get

First, products to be analyzed have similar common shape conditions in terms of basic shape. For example, in the case of a kimchi refrigerator pedestal (bracket) for commercial use, the left and right sides are bent to support the load on both sides of the elongated body surface and to accommodate the wheels. shape condition.

Next, the products to be analyzed have similar load conditions. For example, in the case of a kimchi refrigerator pedestal (bracket) for commercial use, it has a load condition in which a load due to the weight of the kimchi refrigerator is applied from the upper side over the entire upper surface of the elongated body surface.

Next, the products to be analyzed have similar constraint conditions. For example, in the case of a kimchi refrigerator pedestal (bracket) for commercial use, there is a constraint condition in which the protrusion for coupling the wheels at two points is constrained.

In addition, boundary conditions (eg, temperature, etc.) have similar conditions.

Taking the flow analysis of a blower fan as another example, the deep learning of this embodiment inputs the products as training data and/or analysis basic information for one CAE analysis model when a plurality of products have the following conditions Thus, the analysis result can be obtained.

First, in terms of the shape condition of the product, there is a motor connection part in the center of the fan, there is a shroud on the edge, and a plurality of blades are connected and installed in the middle between the motor connection part and the shroud.

Next, in terms of load conditions, the rotational driving force of the motor shaft is applied to the central motor connection part.

In addition, boundary conditions (eg, temperature, etc.) have similar conditions.

In the case of flow analysis, there are no constraints.

The CAE analysis model of this embodiment has common analysis conditions for products with specific functions as described above, but some changes in shape (e.g., changes in the size of the body and the size of the bent part) and load conditions for each design condition of each product On the premise that changes in constraint conditions (e.g., spacing between wheels) occur (e.g., kimchi refrigerator capacity change), etc., it is particularly useful to obtain analysis results when some design conditions for the same functional product are changed. have.

If the purpose or function of the product is similar, it can be analyzed with a single CAE analysis model even in cases where the shape conditions are partially different (eg, a product with two wheels, a product with four wheels).

The CAE analysis model of this embodiment may be used for accurate analysis results, and may also be used for trend analysis of products having similar shapes.

As an example, the training data of this embodiment may be CAE analysis result data using a general finite element method. CAE analysis using the finite element method converts 3D CAD data (3D modeling data) to perform finite element modeling, a pre-processing step of applying boundary conditions (e.g., flow velocity, pressure, constraint conditions) and load conditions, the pre-processing steps A CAE analysis result is generated through an analysis step of calculating the nodes and elements constituted through numerical equations, and a post-processing step of visualizing the result value of the analysis step to calculate visualization information (color distribution diagram) of the CAE analysis result. Typically, CAE analysis results using the finite element method are calculated in the order of generation of outer edge line information -> generation of color distribution information -> generation of legend information.

When the CAE analysis result as described above is used, the learning data is the outer edge line information indicating the shape of the product, the color distribution map information for classifying and displaying the analysis result for the shape by color, and the maximum number of colors included in the color distribution map information It is configured to include at least one of legend information describing a value and a minimum numerical value. There may be some differences in the configuration of the learning data depending on whether it is a supervised learning method or an unsupervised learning method, which will be described later.

The color distribution map information is to display an analysis result for any one of a stress distribution, a pressure distribution, and a fluid pressure distribution by color.

The outer edge line information, color distribution information, and legend information are input to the CAE analysis model as learning data and analysis basic information to be described later in the form of visual image information calculated as a CAE analysis result.

For example, in the deep learning process of step 1), legend information (color-numerical value matching table, maximum/minimum numerical values) of training data is recognized -> product shape recognition through outer edge line information -> color distribution information recognition Learning can be performed in sequence.

Preferably, in step 1), the outer edge line information and color distribution information included in the learning data are pre-processed to be input as a condition for looking at the product from a preset direction.

The pre-processing is performed based on a plurality of different directions preset for one product, and the outer edge line information and color distribution information are input for each direction.

Preferably, a plurality of different directions may constitute a perspective view, a six-view (front view, a rear view, a top view, a bottom view, a left view, a right view) direction as a set, but is not limited thereto. The perspective direction may be, for example, an angle rotated by 45 degrees along the x, y, and z axes, preferably in the same direction for each product. As the input data for one product, if the dataset can be distinguished, the size of each image constituting the input dataset does not necessarily have to be the same.

In step 2), the CAE analysis result calculation system 100 inputs basic analysis information of a product requiring CAE analysis to the CAE analysis model.

The analysis basic information is the outer edge line information indicating the shape of the product, the color distribution map information for classifying and displaying the analysis results for the shape by color, and the maximum numerical value and minimum numerical value of the color included in the color distribution information. It is configured to include at least one of the legend information. Depending on whether the method is a supervised learning method or an unsupervised learning method, there may be differences in the composition of the basic interpretation information, which will be described later.

In step 2), the outer edge line information and the color distribution information included in the basic analysis information are pre-processed to be input as a condition for looking at the product from a preset direction.

The pre-processing is performed based on a plurality of different directions preset for one product, and the outer edge line information and color distribution information are input for each direction.

In step 3), the CAE analysis result calculation system 100 calculates a CAE analysis result for the analysis basic information by the CAE analysis model.

The CAE analysis result is the outer edge line information indicating the shape of the product, the color distribution map information that classifies the analysis results for the shape by color, and the maximum and minimum numerical values of colors included in the color distribution information. It is configured to include at least one of the legend information. The composition of CAE analysis results may differ depending on whether it is a supervised learning method or an unsupervised learning method, which will be described later.

For example, the outer edge line information may be configured in such a way that the outer shape of the product is indicated by a black line.

The color distribution map information may be composed of a rainbow-colored contour that displays different numerical values of an analysis result (eg, a stress distribution, a pressure distribution, or a fluid pressure distribution) by dividing them into ranges.

The legend information may be configured in the form of a color-numeric value matching table indicating which numerical value range corresponds to each color of the rainbow-colored contour included in the color distribution information, and also included in the color distribution information It also includes numerical value information including the maximum numerical value and the minimum numerical value of the color.

Referring to the legend information of FIG. 9 , a numerical value corresponding to red is a maximum numerical value, and a numerical value corresponding to blue at the bottom is a minimum numerical value. The maximum numerical value may have a different value for each product. In consideration of this point, in deep learning, numerical values in the form of text data next to the legend information are also included as input data.

4 is a flowchart of a method for calculating a CAE analysis result based on deep learning of a supervised learning method according to an embodiment of the present invention, and FIG. 10 is a schematic diagram for explaining a CAE analysis model of a supervised learning method according to an embodiment of the present invention am.

Preferably, a deep learning model based on CNN (Convolution Neural Networks) is used for the CAE analysis model of this embodiment, and the deep learning of step 1) consists of supervised learning.

Convolutional neural networks (CNNs) are classified as deep neural networks in deep learning, and are most commonly used for visual image analysis. The CAE analysis model of this embodiment learns and interprets based on visual image information such as legend information (color-numeric value matching table), outer edge line information, and color distribution information, so deep learning based on CNN (Convolutional Neural Network) model is preferred.

The CAE analysis model of this embodiment is a CAE analysis model learned based on the analysis results according to various shape changes of the product. is a model 10 illustrates the concept of input and output of the CAE analysis model of the present embodiment.

The input value of the learning data of the present embodiment is outer edge line information indicating the shape of the product.

The output value of the learning data of this embodiment is color distribution information for classifying and displaying the analysis result for the shape by color, and legend information for explaining the maximum and minimum numerical values of colors included in the color distribution information.

The learning data may be those calculated as a result of numerical analysis or those calculated as a result of deep learning of this embodiment, and preferably, those calculated as a result of numerical analysis may be used.

For example, in the deep learning process of step 1) of this embodiment, learning is performed in the order of recognizing the legend information (maximum/minimum numerical values) of the learning data -> product shape recognition through outer edge line information -> color distribution information recognition can be executed

In this embodiment, the analysis basic information of step 2) includes information about the outer edge line indicating the shape of the product.

In addition, in this embodiment, the CAE analysis result of step 3) is the outer edge line information indicating the shape of the product, the color distribution map information that classifies the analysis results for the shape by color, and the maximum number of colors included in the color distribution information Contains legend information describing values and minimum numerical values.

By the CAE analysis model of this embodiment, it is possible to analyze how color distribution such as stress will occur when a new product shape (CAD 3D data) is input.

5 is a flowchart of a CAE analysis result calculation method based on deep learning of an unsupervised learning method according to an embodiment of the present invention, and FIG. 11 is an unsupervised learning method CAE analysis model according to an embodiment of the present invention. It is a schematic diagram for

Preferably, a deep learning model based on a generative adversarial network (GAN) is used for the CAE analysis model of this embodiment, and the deep learning of step 1) consists of unsupervised learning (autonomous learning).

A generative adversarial neural network (GAN) is an artificial intelligence algorithm used in unsupervised learning. It consists of a generator that generates a virtual data sample and a discriminator that determines whether the input data sample is real data. It is a deep learning model built through adversarial training between a generator and a discriminator.

The purpose of the generator is to extract and generate new data from a GAN-based deep learning model after learning the probability distribution of real data. The GAN-based deep learning model learns various shapes and their analysis results with the discriminator first, and then repeats the process of learning the generator while giving and receiving. By inputting the fake data created by the generator into the discriminator, and training the generator to generate data similar to real data enough to classify fake data as real, new optimized shapes that were not in existing data can be created infinitely. The discriminator discriminates using data accumulated through a shape obtained from a parametric design or a large-scale analysis result.

For the above learning, the generator receives random noise and generates a virtual data sample. When a real data sample or a virtual data sample is input, the discriminator determines whether the input data sample is real (or virtual) data. In more detail, the discriminator outputs a confidence score corresponding to the input data sample corresponding to the actual data, and determines whether the input data sample is real data based on the confidence score. To this end, the discriminator can be trained on real data samples or virtual data samples. The generator is trained by back-propagating the error of the discriminator determined through the error judgment process.

12 is an example of flow analysis of a fan for a blower, and illustrates examples of optimization shapes that can be generated by a generator of a CAE analysis model based on GAN and input data of a fan for a blower.

The CAE analysis model of this embodiment proposes a shape that reduces this maximum value for the part of the product where the maximum value (eg, red) occurs in the input CAE analysis result when the CAE analysis result is input based on the numerical analysis. is a model Taking the commercial kimchi refrigerator pedestal (bracket) of FIG. 7 as an example, the CAE analysis model of this embodiment proposes an optimized shape in which stress concentration is reduced in the part of the product with the highest stress in red. 11 illustrates the concept of input and output of the CAE analysis model of the present embodiment.

The input value of the learning data of this embodiment is the outer edge line information indicating the shape of the product, the color distribution map information for classifying and displaying the analysis result for the shape by color, and the maximum and minimum values of colors included in the color distribution information Legend information that describes the value.

All of the learning data calculated as a result of numerical analysis or calculated as a result of deep learning of the present embodiment may be used.

When these training data are input into the GAN-based deep learning model and trained to generate a shape (optimized shape) that reduces this maximum value for the product part where the maximum value (eg, red) occurs in the CAE analysis result, the CAE of this embodiment An analytical model is obtained.

In this embodiment, the analysis basic information of step 2) includes outer edge line information indicating the shape of the product, color distribution information indicating the analysis result for the shape by color, and the maximum numerical value of a color included in the color distribution information and legend information describing the minimum numerical value.

In this embodiment, the CAE analysis result of step 3) is the outer edge line information indicating the shape of the product, the color distribution map information for classifying and displaying the analysis results for the shape by color, and the maximum numerical value of the color included in the color distribution information and legend information describing the minimum numerical value, in particular, the CAE analysis result is information for generating a shape more optimized than the analysis basic information.

The more optimized shape is a shape modified so that the maximum numerical value of the color included in the color distribution information is further reduced based on the color distribution information included in the basic analysis information of step 2).

With the CAE analysis model of this embodiment, when a CAE analysis result is input based on a numerical analysis, an optimization shape that reduces this maximum value for the part of the product where a maximum value (eg, stress) occurs in the input CAE analysis result is selected. can be created and proposed.

The CAE analysis model of this embodiment may be additionally applied to the CAE analysis results calculated as the results of FIGS. 4 and 10 described above, and may also be applied to the CAE analysis results calculated as the general numerical analysis results.

Embodiments of the present invention include a program for performing various computer-implemented operations and a computer-readable recording medium recording the same. The computer-readable recording medium may include program instructions, data files, data structures, etc. alone or in combination. The media may be specially designed and configured for the present invention, or may be known and available to those skilled in the art of computer software. Examples of computer-readable recording media include hard disks, magnetic media such as floppy disks and magnetic tapes, optical recording media such as CD-ROMs, DVDs, and USB drives, magneto-optical media such as floppy disks, and ROM, RAM, Included are hardware devices specially configured to store and execute program instructions, such as flash memory and the like. Examples of program instructions include not only machine language codes such as those generated by a compiler, but also high-level language codes that can be executed by a computer using an interpreter or the like.

100: CAE analysis result calculation system
102: learning data input unit
104: analysis basic information input unit
106: CAE analysis model
108: CAE analysis result output unit

Claims (11)

A method for calculating CAE (Computer-Aided Engineering) analysis results executed by deep learning-based computing means, the method comprising:
1) deep learning the CAE analysis results of a plurality of products as learning data to generate a CAE analysis model;
2) inputting basic analysis information of a product requiring CAE analysis to the CAE analysis model; and
3) calculating a CAE analysis result for the analysis basic information by the CAE analysis model;
The learning data, analysis basic information, and CAE analysis result are the outer edge line information indicating the shape of the product, the color distribution map information that classifies the analysis results for the shape by color, and the maximum numerical value of the color included in the color distribution map information And Deep learning-based CAE analysis result calculation method, characterized in that each comprises at least any one of legend information describing the minimum numerical value.
According to claim 1,
The deep learning in step 1) is supervised learning,
The input value of the learning data is outer edge line information indicating the shape of the product,
The output value of the learning data is color distribution map information that classifies and displays the analysis result for the shape by color, and legend information that describes the maximum and minimum numerical values of colors included in the color distribution information. Based CAE analysis result calculation method.
3. The method of claim 2,
The analysis basic information of step 2) includes information about the outer edge line indicating the shape of the product,
The CAE analysis result of step 3) is the outer edge line information indicating the shape of the product, the color distribution map information for classifying and displaying the analysis results for the shape by color, and the maximum and minimum values of colors included in the color distribution information A method for calculating CAE analysis results based on deep learning, characterized in that it includes legend information explaining the value.
3. The method of claim 2,
The CAE analysis model is a deep learning-based CAE analysis result calculation method, characterized in that it is a deep learning model based on CNN (Convolution Neural Networks).
According to claim 1,
The deep learning in step 1) is unsupervised learning,
The input value of the learning data is the outer edge line information indicating the shape of the product, the color distribution map information for classifying and displaying the analysis results for the shape by color, and the maximum and minimum numerical values of colors included in the color distribution information. Deep learning-based CAE analysis result calculation method, characterized in that it is legend information to explain.
6. The method of claim 5,
The analysis basic information of step 2) includes outer edge line information indicating the shape of the product, color distribution map information for classifying and displaying the analysis results for the shape by color, and the maximum and minimum values of colors included in the color distribution information Contains legend information that describes the value,
The CAE analysis result of step 3) is the outer edge line information indicating the shape of the product, the color distribution map information for classifying and displaying the analysis results for the shape by color, and the maximum and minimum values of colors included in the color distribution information Including legend information describing the value, wherein the CAE analysis result is information for generating a shape more optimized than the analysis basic information - The more optimized shape is included in the analysis basic information of step 2) A method for calculating CAE analysis results based on deep learning, characterized in that, based on the color distribution information, the shape is deformed so that the maximum numerical value of the color included in the color distribution information is further reduced.
6. The method of claim 5,
The CAE analysis model is a deep learning-based CAE analysis result calculation method, characterized in that it is a deep learning model based on a generative adversarial network (GAN).
According to claim 1,
In step 1) above,
The outer edge line information and color distribution information included in the learning data are pre-processed to be input as a condition for looking at the product from a preset direction,
In step 2) above,
The method for calculating CAE analysis results based on deep learning, characterized in that the outer edge line information and color distribution map information included in the analysis basic information are pre-processed to be input as a condition for looking at the product from a preset direction.
9. The method of claim 8,
The pretreatment is
A method for calculating CAE analysis results based on deep learning, characterized in that it is executed based on a plurality of different directions set in advance for one product, and the outer edge line information and color distribution information are input for each direction.
According to claim 1,
The color distribution map information is a deep learning-based CAE analysis result calculation method, characterized in that the analysis results for any one of a stress distribution, a pressure distribution, and a fluid pressure distribution are displayed by color.
A computer program stored in a computer-readable medium to execute a deep learning-based CAE analysis result calculation method in combination with hardware comprising a memory for storing one or more instructions and a processor for executing the one or more instructions stored in the memory,
The deep learning-based CAE analysis result calculation method is,
1) deep learning the CAE analysis results of a plurality of products as learning data to generate a CAE analysis model;
2) inputting basic analysis information of a product requiring CAE analysis to the CAE analysis model; and
3) calculating a CAE analysis result for the analysis basic information by the CAE analysis model;
The learning data, analysis basic information, and CAE analysis result are the outer edge line information indicating the shape of the product, the color distribution map information that classifies the analysis results for the shape by color, and the maximum numerical value of the color included in the color distribution map information and at least any one of legend information describing the minimum numerical value.

Images