# ANALYSIS APPARATUS USING LEARNED MODEL AND METHOD THEREFOR

An analysis apparatus is provided. The analysis apparatus includes a model deriver configured to generate an analytic model for predicting a result of a numerical analysis performed multiple times iterations for a component by using a plurality of analytic data used for the numerical analysis for the component, and a model analyzer configured to predict the result of the numerical analysis performed multiple times iterations for a design target component by using the analytic model.

**Description**

**CROSS-REFERENCE TO RELATED APPLICATION**

This application claims priority to Korean Patent Application No. 10-2018-0097539, filed on Aug. 21, 2018, the disclosure of which is incorporated by reference herein in its entirety.

**BACKGROUND**

**Field**

Apparatuses and methods consistent with exemplary embodiments relate to an analysis technology, and more particularly, to an apparatus for optimizing analysis using a learned model and a method therefor.

**Description of the Related Art**

To manufacture high performance/high reliability core components, analysis such as computational fluid analysis/structural analysis/electromagnetic analysis is essential in the design procedure. For example, in the case of a turbine blade, the computational fluid analysis and the structural analysis are required, and in the case of a motor, the electromagnetic analysis is required. However, the conventional analysis methods based on physics are time consuming. Therefore, the analysis conditions are simplified to shorten the analysis time, but in this case, it does not become the sophisticated design. In addition, the analysis is not done only once but should be iterated until proper performance comes out. As a result, it needs a long time to develop the component. Therefore, to shorten the time required for developing the component even while performing the sophisticated design, an analysis method capable of minimizing the analysis time is needed.

**SUMMARY**

Aspects of one or more exemplary embodiments provide an analysis apparatus capable of shortening the analysis time for component design and a method therefor.

Additional aspects will be set forth in part in the description which follows and, in part, will become apparent from the description, or may be learned by practice of the exemplary embodiments.

According to an aspect of an exemplary embodiment, there is provided an analysis apparatus including: a model deriver configured to generate an analytic model for predicting a result of a numerical analysis performed multiple times iterations for a component by using a plurality of analytic data used for the numerical analysis for the component, and a model analyzer configured to predict the result of the numerical analysis performed multiple times iterations for a design target component by using the analytic model.

The model deriver may include an analytic data storage configured to store the analytic data including a plurality of input signals used for the numerical analysis and a plurality of output signals corresponding to each of the plurality of input signals, and an analytic model deriver configured to generate the analytic model for deriving the output signal of the numerical analysis performed multiple times iterations through the analytic data.

The analytic model deriver constitutes a relationship equation of the analytic model where a parameter is not determined, and generates the analytic model by deriving the parameter through learning by using the analytic data.

The model deriver further includes a processor configured to perform preprocessing for correcting or removing the analytic data according to a predetermined condition.

The model deriver further includes a data analyzer configured to derive a relationship between cells and a relationship between data in each cell by analyzing the preprocessed analytic data.

The model analyzer may include a numerical analyzer configured to derive analytic data by performing the numerical analysis for a plurality of cells that divide the space around the design target component, and an analyzer configured to predict an output signal of the numerical analysis performed multiple times iterations by applying the analytic data to the analytic model derived from the analytic model deriver.

The analysis apparatus further includes an optimizer configured to derive an optimized result that optimizes the plurality of output signals derived from the model analyzer.

The optimizer may include a filter configured to remove noise in each of the plurality of output signals, a primary optimizer configured to optimize the output signal from which the noise has been removed primarily, and a secondary optimizer configured to optimize the primarily optimized result secondarily.

The numerical analyzer outputs the analytic data by iterating the numerical analysis based on the optimized result optimized by the optimizer, and the analyzer predicts the output signal of the numerical analysis performed multiple times iterations by applying the analytic data output according to the iterated numerical analysis to the analytic model derived from the analytic model deriver.

The numerical analyzer outputs the analytic data by iterating the numerical analysis based on the optimized result optimized by the optimizer, and the analytic model deriver updates the analytic model for deriving the output signal of the numerical analysis performed multiple times iterations through the analytic data output according to the iterated numerical analysis.

According to an aspect of another exemplary embodiment, there is provided an analysis apparatus including: a model deriver configured to generate an analytic model for simulating a numerical analysis for a component by using analytic data used for the numerical analysis for the component, and a model analyzer configured to perform the numerical analysis for a design target component by using the analytic model.

The analytic model may include at least one of a parametric model including a Transfer Function model and a State Space model and a nonparametric model.

The analytic model can be a model for simulating the numerical analysis for each of a plurality of cells, a model for simulating the numerical analysis for a cell group including a predetermined number of cells adjacent to each other, a model for simulating the numerical analysis for a cell group including cells having the similar characteristics to each other, or a model for simulating the numerical analysis for all of the plurality of cells, when the periphery of the design target component is divided into the plurality of cells.

The analytic model predicts the result of the numerical analysis performed multiple times iterations.

According to an aspect of another exemplary embodiment, there is provided an analysis method including: generating, by a model deriver, an analytic model for predicting a result of a numerical analysis performed multiple times iterations for a component by using a plurality of analytic data used for the numerical analysis for the component, and predicting, by a model analyzer, the result of the numerical analysis performed multiple times iterations for a design target component by using the analytic model.

The generating the analytic model includes storing, by an analytic data storage, the analytic data including a plurality of input signals used for the numerical analysis and a plurality of output signals corresponding to each of the plurality of input signals, and generating, by an analytic model deriver, the analytic model for deriving the output signal of the numerical analysis performed multiple times iterations through the analytic data.

The generating the analytic model includes constituting, by the analytic model deriver, a relationship equation of the analytic model where a parameter is not determined, and generating, by the analytic model deriver, the analytic model by deriving the parameter through learning by using the analytic data.

The analysis method further includes, before the generating the analytic model, performing, by a preprocessor, preprocessing for correcting or removing the analytic data according to a predetermined condition, and deriving, by a data analyzer, the relationship between cells and the relationship between data in each cell by analyzing the learning data.

The predicting the result of the numerical analysis includes deriving, by a numerical analyzer, the analytic data including an input signal and an output signal corresponding to the input signal by performing the numerical analysis, and deriving, by an analyzer, the output signal of the numerical analysis performed multiple times iterations by applying the analytic data to the analytic model derived by the analytic model deriver.

The analysis method further includes, after the deriving the output signal, deriving, by an optimizer, optimization data by optimizing the plurality of output signals derived by the analyzer.

The performing the optimization includes removing, by a filter, noise in each of the plurality of output signals, optimizing, by a primary optimizer, the output signal from which the noise has been removed primarily, and deriving, by a secondary optimizer, optimization data by optimizing the primarily optimized output signal secondarily.

After the deriving the optimization data, the analysis method can iterate the deriving the analytic data, the deriving the output signal, and the deriving the optimization data, if the optimization data does not converge within a predetermined range.

As described above, according to one or more exemplary embodiments, it is possible to shorten the analysis time for component design, thereby shortening the time required for developing the component.

**BRIEF DESCRIPTION OF THE DRAWINGS**

The above and other aspects will become more apparent from the following description of the exemplary embodiments with reference to the accompanying drawings, in which:

**DETAILED DESCRIPTION**

Hereinafter, various modifications and various embodiments will be described in detail with reference to the accompanying drawings so that those skilled in the art can easily carry out the disclosure. It should be understood, however, that the various embodiments are not for limiting the scope of the disclosure to the specific embodiment, but they should be interpreted to include all modifications, equivalents, and alternatives of the embodiments included within the spirit and scope disclosed herein. In order to clearly illustrate the disclosure in the drawings, some of the elements that are not essential to the complete understanding of the disclosure may be omitted, and like reference numerals refer to like elements throughout the specification

The terminology used in the disclosure is for the purpose of describing specific embodiments only and is not intended to limit the scope of the disclosure. The singular expressions “a”, “an”, and “the” are intended to include the plural expressions as well unless the context clearly indicates otherwise. In the disclosure, terms such as “comprises,” “include,” or “have/has” should be construed as designating that there are such features, integers, steps, operations, components, parts, and/or combinations thereof, not to exclude the presence or possibility of adding of one or more of other features, integers, steps, operations, components, parts, and/or combinations thereof.

Further, terms such as “first,” “second,” and so on may be used to describe a variety of elements, but the elements should not be limited by these terms. The terms are used simply to distinguish one element from other elements. The use of such ordinal numbers should not be construed as limiting the meaning of the term. For example, the components associated with such an ordinal number should not be limited in the order of use, placement order, or the like. If necessary, each ordinal number may be used interchangeably.

Expressions such as “at least one of,” when preceding a list of elements, modify the entire list of elements and do not modify the individual elements of the list. For example, the expression, “at least one of a, b, and c,” should be understood as including only a, only b, only c, both a and b, both a and c, both b and c, all of a, b, and c, or any variations of the aforementioned examples.

First, an analytic model according to an exemplary embodiment will be described.

Referring to

Referring to

Dynamics (CFD) can be performed for the analysis. For the numerical analysis according to Computational Fluid Dynamics, the periphery of the component CP is divided into a plurality of cells CE. Then, a nonlinear partial differential equation for the plurality of cells CE is established. Then, an approximate solution to the partial differential equation can be obtained, for example, by a Gaussian elimination method.

Referring to a graph of

Therefore, according to one or more exemplary embodiments, an analytic model is generated to derive an output signal that is the result of the numerical analysis performed multiple times iterations by using the analytic data including a plurality of input signals used for the numerical analysis for component design and a plurality of output signals corresponding to the plurality of input signals. That is, the generated analytical model simulates the result of the numerical analysis performed multiple times iterations. Therefore, it is possible to reduce the time required for obtaining the approximate solution to the partial differential equation, thereby shortening the analysis time. Therefore, it is possible to shorten the time for designing the component.

The analytical model according to an exemplary embodiment may include at least one of a parametric model including a Transfer Function model and a State Space model and a nonparametric model. A Table 1 below illustrates examples of the parametric model and the non-parametric model.

In addition, the analytic model can be derived by using at least one of the optimization algorithms listed in Table 2 below.

In particular, referring back to

Next, an analysis apparatus according to an exemplary embodiment will be described. **10** according to an exemplary embodiment may include a model deriver **100**, a model analyzer **200**, and an optimizer **300**.

The model deriver **100** generates an analytic model for predicting the result of the numerical analysis performed multiple times iterations for the component by using analytic data that include a plurality of input signals used for the numerical analysis for the component and a plurality of output signals corresponding to the plurality of input signals. The analytic model simulates the result of the numerical analysis performed multiple times iterations. As described above, the analytical model can be composed of a plurality of models, and include at least one of a parametric model and a non-parametric model.

The model deriver **100** may include an analytic data storage **110**, a processor **120**, a data analyzer **130**, and an analytic model deriver **140**.

The analytic data storage **110** stores analytic data. The analytic data may be data used for the numerical analysis for a plurality of cells CE that divide the area around the component CP. The analytic data includes a plurality of input signals and a plurality of output signals corresponding to the plurality of input signals. For example, the input signal can be a laminar flow viscosity of the fluid, a turbulent conduction, a time difference between the numerical analysis performed multiple times iterations, etc. in each cell CE. The output signal may be the characteristics of the fluid. For example, the output signal can be a density, a momentum in the x and y directions, an internal energy, etc. in each cell CE.

The processor **120** performs preprocessing for correcting or removing learning data according to a predetermined condition. The preprocessing for the learning data means to remove empty data in the middle, erroneous data, etc. among the learning data, or to convert them into correct numeric values, and to select only the learning data meeting a predetermined requirement. The processor **120** performs preprocessing by correcting or removing the learning data according to a predetermined condition.

The data analyzer **130** derives the relationship between the cells and the relationship between the data in the cell by analyzing the learning data. That is, the data analyzer **130** derives the relationship between the cells CE and the relationship between the data in the cell CE by analyzing the design specification and condition, the relationship between the cells CE, and the data for each cell CE.

The analytic model deriver **140** derives the analytic model for predicting an output signal of the numerical analysis performed multiple times iterations by using the analytic data that include the plurality of input signals used for the numerical analysis and the plurality of output signals corresponding to the plurality of input signals. The analytical model simulates the numerical analysis performed multiple times iterations.

The analytic model deriver **140** constitutes a relationship equation of the analytical model where parameters are not determined, and derives the parameters through an optimization algorithm by putting the analytic data into the relationship equation. Therefore, the analytic model deriver **140** can generate the analytic model by applying the derived parameters to the relationship equation of the analytic model. For example, the analytic model deriver **140** can constitute the relationship equation of the analytical model where the parameters for determining the relationship between the input signal and the output signal of the numerical analysis performed multiple times iterations are unknown, and derive the parameters by learning a plurality of analytic data for the constituted relationship equation. As a result, the analytic model deriver **140** can derive the analytic model.

The model analyzer **200** performs the analysis for the plurality of cells CE that divide the space around the design target component CP by using the analytical model derived from the model deriver **100**. The model analyzer **200** may include a numerical analyzer **210** and an analyzer **220**.

The numerical analyzer **210** performs the numerical analysis for the plurality of cells that divide the space around the design target component. Therefore, the input signal for the numerical analysis and the output signal corresponding to the input signal are derived. Referring to

The analyzer **220** predicts the output signal of the numerical analysis performed multiple times (k+T) iterations by inputting the analytic data derived from the numerical analyzer **210** to the analytic model generated by the analytic model deriver **140**. Referring to

However, according to an exemplary embodiment, because the output signal Ŷ(k+T), which is the result of the numerical analysis performed the analytic model multiple times (k+T) iterations, can be obtained from the k^{th }numerical analysis of the numerical analyzer **210**, it is not necessary to perform the numerical analysis multiple times T iterations, such that it is possible to shorten the time required for the analysis by the time performed the numerical analysis multiple times T iterations. Therefore, it is possible to shorten the time required for developing the component.

An optimizer **300** is for optimizing the analysis result derived from the model analyzer **200**. The analysis result converges to a specific value as the iteration of the numerical analysis is performed. Therefore, it is possible to optimize the result (i.e., the plurality of output signals) predicted by the model analyzer **200** through the optimizer **300**. Referring to **300** may include a filter **310**, a primary optimizer **320**, and a secondary optimizer **330**.

The filter **310** is for removing noise of the output signal derived from the model analyzer **200**. The filter **310** can use a filter technology to remove noise. The filter can be, for example, at least one of an averaging filter, a moving average filter, a low-pass filter such as an exponentially weighted moving average filter, a high-pass filter, a band-pass filter, and a Kalman filter.

The primary optimizer **320** is for optimizing the output signal that is the result of the analysis of the model analyzer **200** primarily. The primary optimizer **320** outputs a primary optimization value through a primary optimization operation for the plurality of output signals that are outputs of the model analyzer **200**. For example, the primary optimizer **320** outputs primary optimization data by calculating the average of the predetermined number of output signals among the plurality of output signals.

The secondary optimizer **330** is for optimizing the result primarily optimized by the primary optimizer **320** secondarily. The secondary optimizer **330** outputs a secondary optimization value through a secondary optimization operation for the plurality of primary optimization data that are outputs of the primary optimizer **320**. For example, the secondary optimizer **330** outputs optimum data by calculating the average of the predetermined number of the primary optimization data among the plurality of primary optimization data.

Meanwhile, the optimum data is fed back to the numerical analyzer **210** again, the numerical analyzer **210** again performs the numerical analysis based on the optimum data, and the analyzer **220** can predict an output signal of the numerical analysis performed multiple times iterations according to the analytic model. This procedure is iterated until the output signal predicted by the analyzer **220** converges within a predetermined range.

Meanwhile, when the numerical analyzer **210** outputs the analytic data by iterating the numerical analysis based on the optimum data, the analytic model deriver **140** can update the analytic model for deriving the output signal of the numerical analysis performed multiple times iterations based on the analytic data output by iterating the numerical analysis based on the optimum data. Then, the updated analytic model can be again provided to the analyzer **220**.

Next, an analysis method of the analysis apparatus **10** according to an exemplary embodiment will be described.

Referring to **100** generates the analytic model for performing the numerical analysis for the plurality of cells CE that divide the space around the target component CP by using the analytic data (operation S**110**). Herein, the analytic data includes a plurality of input signals used for the numerical analysis performed multiple times iterations and a plurality of output signals corresponding to the plurality of input signals. That is, the analytical model simulates the result of the numerical analysis performed multiple times iterations computationally.

The model analyzer **200** performs the numerical analysis multiple times iterations for the plurality of cells CE in the space around the target component CP through the analytic model derived from the model deriver **100** (operation S**120**).

The above-described operations S**110** and S**120** will be described in more detail.

**110**) according to an exemplary embodiment.

Referring to **110** stores the analytic data including the plurality of input signals used for the numerical analysis and the plurality of output signals corresponding to each of the plurality of input signals, and outputs the analytic data (operation S**210**).

The processor **120** preprocesses the analytic data (operation S**220**). The processor **120** removes empty data in the middle, erroneous data, etc. among the learning data, or converts them into correct numeric values, and selects and outputs only the analytic data meeting a predetermined requirement. The data analyzer **130** can derive the relationship between the cells CE and the relationship between the data in each cell CE by analyzing the analytic data (operation S**230**). That is, the data analyzer **130** derives the relationship between the cells CE and the relationship between the data in each cell CE by analyzing the design specification and condition, the relationship between the cells CE, and the data for each cell CE. The above-described operations S**220** and S**230** can be selectively omitted.

The analytic model deriver **140** constitutes the relationship equation of the analytic model where parameters for determining the relationship between the input signal and the output signal are not determined (operation S**240**). That is, the analytic model deriver **140** constitutes the relationship equation where the parameters for determining the relationship between the input signal and the output signal of the numerical analysis are unknown. The analytic model deriver **140** derives the parameters through the optimization algorithm by putting the analytic data into the relationship equation (operation S**250**). That is, the analytic model deriver **140** performs learning for the analytic data through the optimization algorithm. This learning can be, for example, map learning, non-map learning, etc. The analytic model deriver **140** derives the analytic model by applying the derived parameters to the relationship equation (operation S**260**). This analytical model predicts the output signal of the numerical analysis performed multiple times iterations.

Next, a method for performing the analysis by using the above-described analytical model will be described.

Referring to **210** of the model analyzer **200** derives the analytic data including the input signal and the output signal by performing the numerical analysis (operation S**310**). For example, as shown in **210** can derive the first and second analytic data ({circle around (**1**)}, {circle around (**2**)}).

The analyzer **220** of the model analyzer **200** predicts the output signal (i.e., the predicted data) of the numerical analysis performed multiple times (k+T) iterations by reflecting the analytic data of the numerical analyzer **210** to the analytic model (operation S**320**). For example, the analyzer **220** can predict the output signals (i.e., the predicted data) ({circle around (a)}, {circle around (b)}) of the numerical analysis performed multiple times (k+T) iterations from the first and second analytic data ({circle around (**1**)}, {circle around (**2**)}).

The optimizer **300** performs optimization for the plurality of output signals that are the predicted result (operation S**330**). For example, the average value of the output signals (i.e., the predicted data) ({circle around (a)}, {circle around (b)}) of the numerical analysis can be derived as optimization data.

The optimizer **300** determines whether the optimization data has converged to a predetermined range as a result of the optimization (operation S**340**). If it does not converge within the predetermined range, the above-described operations S310 to S**340** are iterated. If it converges to the predetermined range, the flow advances to operation S**350** to terminate the analysis. For example, by repeating the operations S**310** to S**340**, the numerical analyzer **210** has derived the third and fourth analytic data ({circle around (**3**)}, {circle around (**4**)}) based on the optimum data of optimizing the predicted data ({circle around (a)}, {circle around (b)}), the analyzer **220** has predicted the output signals (i.e., the predicted data) ({circle around (c)}, {circle around (d)}) of the numerical analysis performed multiple times (k+T) iterations from the third and fourth analytic data ({circle around (**3**)}, {circle around (**4**)}), and the optimizer **300** has calculated the average of the predicted data ({circle around (c)}, {circle around (d)}). At this time, if the value of the optimization data calculated by the optimizer **300** is within the predetermined range, the analysis can be terminated.

**130**) according to an exemplary embodiment.

Referring to **310** removes noise in each of the plurality of output signals derived from the model analyzer **200** (operation S**410**). Herein, the filter can be, for example, at least one of an averaging filter, a moving average filter, a low-pass filter such as an exponentially weighted moving average filter, a high-pass filter, a band-pass filter, and a Kalman filter.

The primary optimizer **320** outputs the primary optimization data by optimizing the plurality of output signals from which the noise has been removed primarily according to a primary optimization operation (operation S**420**). For example, the primary optimizer **320** can output the average value of the predetermined number of output signals among the plurality of output signals as the primary optimization data through the primary optimization operation.

The secondary optimizer **330** receives a plurality of primary optimization data from the primary optimizer **320**, and outputs secondary optimization data by optimizing the plurality of input primary optimization data secondarily (operation S**430**). For example, the secondary optimizer **330** can output the average value of the predetermined number of the primary optimization data among the plurality of primary optimization data through the secondary optimization operation as the secondary optimization data.

**100** can be the apparatus described in the present specification (e.g., the analysis apparatus, etc.).

Referring to **100** can include at least one processor TN**110**, a transceiver TN**120**, and a memory TN**130**. In addition, the computing apparatus TN**100** can further include a storage device TN**140**, an input interface TN**150**, and an output interface TN**160**. The components included in the computing apparatus TN**100** can be connected by a bus TN**170** and communicate with each other.

The processor TN**110** can execute a program command stored in at least one of the memory TN**130** and the storage device TN**140**. The processor TN**110** can include a central processing unit (CPU), a graphics processing unit (GPU), or a dedicated processor on which the methods according to an exemplary embodiment are performed. The processor TN**110** can be configured to implement the procedures, functions, methods, etc. described in connection with an exemplary embodiment. The processor TN**110** can control each component of the computing apparatus TN**100**.

Each of the memory TN**130** and the storage device TN**140** can store various information related to an operation of the processor TN**110**. Each of the memory TN**130** and the storage device TN**140** can be composed of at least one of a volatile storage medium and a nonvolatile storage medium. For example, the memory TN**130** can be composed of at least one of a read only memory (ROM) and a random access memory (RAM).

The transceiver TN**120** can transmit and/or receive a wired signal or a wireless signal. The transceiver TN**120** can be connected to a network to perform communication.

Meanwhile, various methods according to an exemplary embodiment described above can be implemented in the form of a readable program through various computer means and recorded in a computer-readable recording medium. Herein, the recording medium can include program commands, data files, data structures, etc. alone or in combination thereof. The program commands to be recorded on the recording medium can be those specially designed and constructed for the present disclosure or can also be those known and available to those skilled in the art of computer software. For example, the recording medium can be 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 the program commands such as ROMs, RAMs, and flash memory. Examples of the program commands can include not only machine language wires such as those produced by a compiler but also high-level language wires that can be executed by a computer by using an interpreter, etc. This hardware device can be configured to operate as one or more software modules in order to perform the operation of the present disclosure, and vice versa.

While one or more exemplary embodiments have been described with reference to the accompanying drawings, it is to be understood by those skilled in the art that various modifications and changes in form and details can be made therein without departing from the spirit and scope as defined by the appended claims. Therefore, the description of the exemplary embodiments should be construed in a descriptive sense only and not to limit the scope of the claims, and many alternatives, modifications, and variations will be apparent to those skilled in the art.

## Claims

1. An analysis apparatus, comprising:

- a model deriver configured to generate an analytic model for predicting a result of a numerical analysis performed multiple times iterations for a component by using a plurality of analytic data used for the numerical analysis for the component; and

- a model analyzer configured to predict the result of the numerical analysis performed multiple times iterations for a design target component by using the analytic model.

2. The analysis apparatus of claim 1,

- wherein the model deriver comprises:

- an analytic data storage configured to store the analytic data comprising a plurality of input signals used for the numerical analysis and a plurality of output signals corresponding to each of the plurality of input signals; and

- an analytic model deriver configured to generate the analytic model for deriving the output signal of the numerical analysis performed multiple times iterations through the analytic data.

3. The analysis apparatus of claim 2,

- wherein the analytic model deriver constitutes a relationship equation of the analytic model where a parameter is not determined, and

- generates the analytic model by deriving the parameter through learning by using the analytic data.

4. The analysis apparatus of claim 2,

- wherein the model deriver further comprises a processor configured to perform preprocessing for correcting or removing the analytic data according to a predetermined condition.

5. The analysis apparatus of claim 4,

- wherein the model deriver further comprises a data analyzer configured to derive a relationship between cells and a relationship between data in each cell by analyzing the preprocessed analytic data.

6. The analysis apparatus of claim 1,

- wherein the model analyzer comprises:

- a numerical analyzer configured to derive analytic data by performing the numerical analysis for a plurality of cells that divide the space around the design target component; and

- an analyzer configured to predict an output signal of the numerical analysis performed multiple times iterations by applying the analytic data to the analytic model derived from the analytic model deriver.

7. The analysis apparatus of claim 6, further comprising:

- an optimizer configured to derive an optimized result that optimizes the plurality of output signals derived from the model analyzer.

8. The analysis apparatus of claim 7,

- wherein the optimizer comprises:

- a filter configured to remove noise in each of the plurality of output signals;

- a primary optimizer configured to optimize the output signal from which the noise has been removed primarily; and

- a secondary optimizer configured to optimize the primarily optimized result secondarily.

9. The analysis apparatus of claim 7,

- wherein the numerical analyzer outputs the analytic data by iterating the numerical analysis based on the optimized result optimized by the optimizer, and

- wherein the analyzer predicts the output signal of the numerical analysis performed multiple times iterations by applying the analytic data output according to the iterated numerical analysis to the analytic model derived from the analytic model deriver.

10. The analysis apparatus of claim 7,

- wherein the numerical analyzer outputs the analytic data by iterating the numerical analysis based on the optimized result optimized by the optimizer, and

- wherein the analytic model deriver updates the analytic model for deriving the output signal of the numerical analysis performed multiple times iterations through the analytic data output according to the iterated numerical analysis.

11. An analysis apparatus, comprising:

- a model deriver configured to generate an analytic model for simulating a numerical analysis for a component by using analytic data used for the numerical analysis for the component; and

- a model analyzer configured to perform the numerical analysis for a design target component by using the analytic model.

12. The analysis apparatus of claim 11,

- wherein the analytic model comprises at least one of a parametric model comprising a Transfer Function model and a State Space model and a nonparametric model.

13. The analysis apparatus of claim 11,

- wherein the analytic model is a model for simulating the numerical analysis for each of a plurality of cells, a model for simulating the numerical analysis for a cell group comprising a predetermined number of cells adjacent to each other, a model for simulating the numerical analysis for a cell group comprising cells having the similar characteristics to each other, or a model for simulating the numerical analysis for all of the plurality of cells, when the periphery of the design target component is divided into the plurality of cells.

14. The analysis apparatus of claim 11,

- wherein the analytic model predicts a result of the numerical analysis performed multiple times iterations.

15. An analysis method, comprising:

- generating, by a model deriver, an analytic model for predicting a result of a numerical analysis performed multiple times iterations for a component by using a plurality of analytic data used for the numerical analysis for the component; and

- predicting, by a model analyzer, the result of the numerical analysis performed multiple times iterations for a design target component by using the analytic model.

16. The analysis method of claim 15,

- wherein the generating the analytic model comprises:

- storing, by an analytic data storage, the analytic data comprising a plurality of input signals used for the numerical analysis and a plurality of output signals corresponding to each of the plurality of input signals; and

- generating, by an analytic model deriver, the analytic model for deriving the output signal of the numerical analysis performed multiple times iterations through the analytic data.

17. The analysis method of claim 16,

- wherein the generating the analytic model comprises constituting, by an analytic model deriver, a relationship equation of the analytic model where a parameter is not determined, and generating, by an analytic model deriver, the analytic model by deriving the parameter through learning by using the analytic data.

18. The analysis method of claim 16, further comprising:

- before the generating the analytic model,

- performing, by a preprocessor, preprocessing for correcting or removing the analytic data according to a predetermined condition; and

- deriving, by a data analyzer, the relationship between cells and the relationship between data in each cell by analyzing the learning data.

19. The analysis method of claim 16,

- wherein the predicting the result of the numerical analysis comprises:

- deriving, by a numerical analyzer, the analytic data comprising an input signal and an output signal corresponding to the input signal by performing the numerical analysis; and

- deriving, by an analyzer, the output signal of the numerical analysis performed multiple times iterations by applying the analytic data to the analytic model derived by the analytic model deriver.

20. The analysis method of claim 17, further comprising

- after the deriving the output signal,

- deriving, by an optimizer, optimization data by optimizing the plurality of output signals derived by the analyzer.

**Patent History**

**Publication number**: 20200065441

**Type:**Application

**Filed**: May 23, 2019

**Publication Date**: Feb 27, 2020

**Inventors**: Jaehyeon PARK (Hwaseong-si), Sangjin LEE (Yongin-si), Jeehun PARK (Gwangmyeong-si), Hyunsik KIM (Gimpo-si)

**Application Number**: 16/420,167

**Classifications**

**International Classification**: G06F 17/50 (20060101); G06N 3/08 (20060101); G06N 7/00 (20060101); G06F 17/13 (20060101);