KR101775824B1 - Metric Optimization Method and Apparatus in Inverse Simulation by Genetic Algorithm - Google Patents

Abstract

The present invention relates to a technique related to computational fluid dynamics (CFD), geostatistics, and inverse simulation. More particularly, the present invention relates to a method and an apparatus for forming an optimum measurer for inverse simulation in which topographical characteristics are reflected in a short period of time through dispersing agent inverse simulation using a genetic algorithm. The method comprises the steps of: generating virtual data; extracting virtual detection region information; extracting virtual prediction region information; and evaluating suitability.

Description

TECHNICAL FIELD [0001] The present invention relates to a metric optimization method,

The present invention relates to techniques related to computational fluid dynamics (CFD), geostatistics, and inverse simulation, and more particularly, The present invention relates to a method and an apparatus for forming an optimum measurer for backtrace simulation in which topographical characteristics are reflected in a short time.

When biochemical agents are spread or environmental pollutants are spread by terrorist attacks and chemical attacks, knowing the distribution of these dispersants plays an important role in minimizing human injury or minimizing environmental damage. However, the number of regions where we can know the concentration of the diffusing agent at each time point is extremely limited compared to the rapidly diffusing diffusing agent.

Based on this limited information, geostatistical methods have been developed to predict local concentrations that have not been measured. This process can be considered as a kind of backtracking process because it can predict the point where the diffusing agent is generated. However, these geostatistical methods have disadvantages in that they do not fully reflect the physical, chemical, or geographical characteristics of chemicals.

In addition, in order to reflect various characteristics of chemical agents, there are backtracking methods using CFD (Computational Fluid Dynamics), but it takes a long time to execute detailed CFD. CFD is a discipline that implements the equations of fluid mechanics through computer simulation in certain boundary conditions and situations.

Therefore, it is necessary to reverse the distribution of the rapidly changing diffuser within a short time reflecting the chemical and topographic characteristics. In order to solve these problems, research on backtracking technology is needed.

1. Korean Patent Publication No. 10-2010-0062327 2. Korean Patent Publication No. 10-2010-0105117

The present invention has been proposed in order to solve the problems described in the above background art, and it is an object of the present invention to provide a method and apparatus for optimizing a measurer in a spreading factor tracking using a genetic algorithm capable of backtracking simulation, The purpose is to provide a method.

The present invention addresses the above problems and provides a method for forming an optimum measurer in a spreading factor tracking using a genetic algorithm.

The optimum measurer forming method includes:

(a) generating virtual data for reflecting the physicochemical properties of the contaminant diffusing agent and the spatio-temporal characteristics of the contaminant diffusion agent through diffusion simulation of the contaminant diffusing agent in the diffusion simulation section;

(b) setting a virtual detection area using the virtual data and extracting virtual detection area information for the virtual detection area;

(c) extracting virtual predicted area information for a virtual predicted area excluding the virtual detected area information; And

(d) evaluating the fitness by tracing back the virtual predicted area information using the generated genetic algorithm additionally generated genetic algorithm.

The step (d) includes the steps of: (d-1) generating a first plurality of gratings; (d-2) performing kriging based on each of the first plurality of grids and the virtual detection area information; (d-3) calculating a kriging result obtained by kriging and an average error value of the virtual predicted area information as an evaluation index; And (d-4) generating another second plurality of gratings using the evaluation index and repeating steps (d-1) to (d-3) until the average error value reaches a predetermined value The method comprising the steps of:

The step (d-1) may include setting a transition method required for the generated genetic algorithm and a storage method for storing generated data.

In addition, the first plurality of gratings or the second plurality of gratings may be characterized by an equidistant regular grid used for kriging or a warped grid used in the generation genetic algorithm.

Also, the virtual detection area information may be a contamination diffusion agent concentration in the virtual detection area, and the virtual prediction area information may be a contamination diffusion agent concentration in the virtual prediction area.

The first plurality of gratings and the second plurality of gratings may each be characterized by having generated genetic information of the generated genetic algorithm.

The generation of the second plurality of grids may be performed by selecting two grids at random among the grids of the upper 50%, and changing the probability of the two to 25% And repeating the number of gratings by the number of the gratings of the second plurality in a manner of crossing two gratings.

In addition, the virtual data may be calculated through a computational fluid dynamics program.

According to another embodiment of the present invention, there is provided a diffusion simulator for generating virtual data for reflecting physicochemical characteristics of the contaminant-diffusing agent and time-space characteristics of the contaminant-diffusing agent through diffusion simulation of the contaminant-diffusing agent; A virtual detection area is set using the virtual data, virtual detection area information on the virtual detection area is extracted, and the information extraction part extracts virtual prediction area information on a virtual prediction area excluding the virtual detection area information An information extracting unit; And a generation genetic algorithm unit for evaluating the fitness by backtracking the virtual predicted region information using the generated genetic algorithm. Can be provided.

According to the present invention, the concentration of the diffusing agent can be efficiently predicted in an area that is not actually measured by reflecting the physicochemical characteristics of the diffusing agent through the pre-simulation.

Another advantage of the present invention is that it is possible to predict the concentration of a region not actually measured in a short time.

FIG. 1 is a block diagram of an apparatus for forming an optimum measurer in a spreading factor tracking using a genetic algorithm according to an embodiment of the present invention.
FIG. 2 is a flowchart illustrating a process of forming an optimal measurer in a spreading factor tracking using a genetic algorithm according to an embodiment of the present invention.
3 is a detailed flowchart of the fitness evaluation execution step (S250) shown in FIG.
Figure 4 is an illustration of a twisted grid generation in accordance with one embodiment of the present invention.
5 is a conceptual diagram showing kriging through a warped grid according to an embodiment of the present invention.
FIG. 6 is a graph of experimental results showing that the error is reduced as the generation progresses in the generation genetic algorithm according to an embodiment of the present invention.

Hereinafter, specific embodiments of the present invention will be described with reference to the drawings. However, this is merely an example and the present invention is not limited thereto.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS In the following description of the present invention, a detailed description of known functions and configurations incorporated herein will be omitted when it may make the subject matter of the present invention rather unclear. The following terms are defined in consideration of the functions of the present invention, and may be changed according to the intention or custom of the user, the operator, and the like. Therefore, the definition should be based on the contents throughout this specification.

The technical idea of the present invention is determined by the claims, and the following embodiments are merely a means for effectively explaining the technical idea of the present invention to a person having ordinary skill in the art to which the present invention belongs.

FIG. 1 is a block diagram of an apparatus 100 for optimizing a measurer in a spreading factor tracking using a genetic algorithm according to an embodiment of the present invention. Referring to FIG. Referring to FIG. 1, an apparatus 100 for forming an optimum measurer includes a diffusion simulation unit 10, an information extraction unit 110, a generation genetic algorithm unit 120, and the like.

The diffusion simulation section 10 generates virtual data for reflecting the physicochemical characteristics of the contaminant diffusion agent and / or the spatial characteristics of the space in which the contaminant diffusion agent is spread.

The information extraction unit 110 includes a virtual detection area information extraction unit 111 for extracting a diffusion agent concentration in a virtual detection area by setting a virtual detection area, And a predicted area information extracting unit 113 for extracting the concentration of the diffusing agent in the predicted area.

Genetic Algorithm Unit 210 includes a warped grid generator 121 for generating arbitrary distorted grids and a predicted region kriging unit 123 for performing fitness evaluation through predicted region kriging using a kriging technique ). The Kriging technique is a geostatistical technique that estimates information from all regions with information from some regions.

In addition, the distorted grid generation unit 121 is set with a suitable mutation method and a storage method of generated data necessary for a generative genetic algorithm.

The term "part" or the like in Fig. 1 means a unit for processing at least one function or operation, which can be implemented by hardware, software, or a combination of hardware and software.

FIG. 2 is a flowchart illustrating a process of forming an optimal measurer in a spreading factor tracking using a genetic algorithm according to an embodiment of the present invention. Referring to FIG. 2, the diffusion simulation unit (10 in FIG. 1) generates virtual data for reflecting the physicochemical characteristics of the contaminant diffusion agent and / or the time-space characteristics of the contaminant diffusion agent through diffusion simulation of the contaminant / (Step S210).

Then, the information extracting unit 110 of FIG. 1 sets the virtual detection area using the virtual data, and extracts the contamination diffusion agent concentration (i.e., the virtual detection area information) in the virtual detection area (steps S220 and S230) .

Then, the information extracting unit 110 extracts the concentration of contaminant diffusing agent (i.e., virtual predicted area information) in the virtual predicted area (step S240).

Thereafter, the generation genetic algorithm unit 120 evaluates the fitness by backtracking the concentration of the contaminant diffusing agent in the virtual detection area using the generated genetic algorithm (step S250).

3 is a detailed flowchart of the fitness evaluation execution step (S250) shown in FIG. Referring to FIG. 3, in progressing the generated genetic algorithm, a plurality of distorted gratings are generated (step S310).

Thereafter, kriging is performed based on each of the plurality of warped grids and the virtual detection area information (step S320)

Thereafter, kriging results obtained by performing kriging and an average error value of the virtual predicted area information are calculated and used as evaluation indices, and the fitness evaluation is performed using the evaluation indices (step S330).

Thereafter, a plurality of other warped lattices are generated using the evaluation index, and steps S310 to S330 are repeated until the average error value reaches a predetermined value. A detailed description thereof will be described later.

Figure 4 is an illustration of a twisted grid generation in accordance with one embodiment of the present invention. Referring to FIG. 4, a generic lattice, which is generally used in kriging, shows an example of a warped lattice 420 for use in an isochronous general lattice 410 and a generated genetic algorithm.

5 is a conceptual diagram showing kriging through a warped grid according to an embodiment of the present invention. Referring to FIG. 5, (a) of FIG. 5 shows the result of kriging 510-2 through the virtual detection information 510-1 in the general lattice. 5 (b) shows the result of kriging 520-2 kriging in the warped grid after converting the same information to the virtual detection information 520-1 in the warped grid. Therefore, as shown in Fig. 5, even the same information shows a different result.

<Examples>

The result of the pre-simulation by the pre-simulation unit 10 is effectively stored through the warped grid (420 in FIG. 4), and kriging (520-2 in FIG. 5) is performed in which the result of the pre-simulation is reflected in a short time in an emergency To be shown as an embodiment.

First, as an embodiment, a commercially available computational fluid dynamics program is executed. The result of executing the diffusion simulation through the computational fluid dynamics program in the diffusion simulation section (10 in FIG. 1) is stored in various ways. In one embodiment of the present invention, only a small amount of data was extracted to demonstrate the efficiency of the genetic algorithm. That is, the virtual data which is the result of diffusion simulation at 12 × 64 points is extracted.

It can be inferred from geostatistics using a method called kriging technique that other information can be guessed using geographical information while knowing only some of the information related to the location.

Some of the results obtained by the diffusion simulation of the spread simulator (10 in FIG. 1) are assumed to be the virtual detection area information extracted from the virtual detection area, and information excluding the detected virtual detection area information is stored in the virtual prediction area Information. The value used as the virtual detection area information was only 3 × 11 pieces of information of 12 × 64 pieces.

Since the kriging technique is a technique for estimating the value of a virtual prediction region, which is a non-measuring region, in relation to the distance between two regions, the distance between the virtual detection region information and the virtual prediction region plays an important role. Therefore, if a twisted grid 420 is used in place of the regular equidistant regular grid 410 of FIG. 4, it can be expected that another predicted value will be obtained.

Figure 5 compares these results. Referring to FIG. 5, the result of kriging 520-1 through the virtual detection area information at equal intervals and the predicted value of kriging result 510-2 through the virtual detection information in the twisted grid, It can be confirmed that this edition is different.

Generally, kriging through randomly generated warped lattices (410 in FIG. 4) does not give a more accurate result than using an equidistant normal grating (420 in FIG. 4). But it is clear that there will be the most efficient twisted grid. In order to find such a distorted lattice, a generation genetic algorithm was used in the present invention.

The generated genetic algorithm used in one embodiment of the present invention proceeded as follows. To generate fifty random twisted grids. The information of each twisted grid can be regarded as an individual who has genetic information generated in the genetic algorithm. Kriging is performed on the basis of the virtual detection area information (111 in FIG. 1) extracted from the spreading simulation of the twisted grid and the diffusion simulation part (10 in FIG. 1), which is genetic information of each individual.

The information of the predicted area information extracting unit (113 in FIG. 1) can be considered as a true value that is not used when kriging. As a criterion for judging whether or not the result obtained by performing kriging is effective, an average error value between the result of the kriging performed with kriging and the virtual predicted region information extracted by the predicted local information extracting unit (113 in FIG. 1) I will. Generation of the next generation of individuals (about 50 individuals) utilizing the genetic information of the top 50% individuals with the best metrics (ie, the smallest average error value).

Here's how to create a new generation of individuals: First, select any two of the top 50% of individuals. Selected genetic information causes a variation of 25%. Soon, a part of the twisted grid is transformed. Genes are generated by combining some of the genetic information of some of these genes with some of the other genes. This process is called breeding in the genetic algorithm.

This crossing is carried out 50 times to get 50 individuals of a new generation. This new individual repeats the process of creating a new generation through evaluation and mating. This process allows us to approach the most efficient twisted grid.

FIG. 6 is a graph of experimental results showing that the error is reduced as the generation progresses in the generation genetic algorithm according to an embodiment of the present invention. Referring to FIG. 6, the value of the horizontal axis is the number of generations to be increased as the generation genetic algorithm is implemented, and the value of the vertical axis is the average error rate. As the number of households increases, the error rate decreases sharply.

Further, the abscissa represents each generation, and the value of the ordinate represents the error value of the individual having the best evaluation index in each generation. It can be seen that the error rate decreases as the number of households increases, and soon it can be seen to approach the twisted grid.

While the present invention has been particularly shown and described with reference to exemplary embodiments thereof, it is clearly understood that the same is by way of illustration and example only and is not to be taken by way of limitation, I will understand.

Therefore, the scope of the present invention should not be limited to the above-described embodiments, but should be determined by equivalents to the appended claims, as well as the appended claims.

10: diffusion simulation section
100: Optimum measuring device
110:
111: The virtual detection area information extracting unit
113: Predictive area information extracting unit
120: Genetic Genetic Algorithm
121: Twisted grid generator
123: Prediction Area Kriging
410: Isometric regular grid
420: Twisted grid

Claims (9)

(a) generating virtual data for reflecting the physicochemical properties of the contaminant diffusing agent and the spatio-temporal characteristics of the contaminant diffusion agent through diffusion simulation of the contaminant diffusing agent in the diffusion simulation section;
(b) setting a virtual detection area using the virtual data and extracting virtual detection area information for the virtual detection area;
(c) extracting virtual predicted area information for a virtual predicted area excluding the virtual detected area information; And
(d) evaluating the fitness by tracing back the virtual predicted area information using the generated genetic algorithm additionally generated genetic algorithm,
The step (d)
(d-1) generating a first plurality of gratings;
(d-2) performing kriging based on each of the first plurality of grids and the virtual detection area information;
(d-3) calculating a kriging result obtained by kriging and an average error value of the virtual predicted area information as an evaluation index; And
(d-4) repeating the steps (d-1) to (d-3) until the average error value reaches a predetermined value by generating another second plurality of grids using the evaluation index The method comprising the steps of: (a) selecting a probe from a plurality of probes;
delete The method according to claim 1,
Wherein the step (d-1) comprises the steps of: setting a transition method necessary for the generated genetic algorithm and a storage method for storing generated data; / RTI &gt;
The method according to claim 1,
Characterized in that said first plurality of gratings or said second plurality of gratings are a regular spacing grid used for kriging or a warped grid for use in said generation genetic algorithm. Way.
The method according to claim 1,
Wherein the virtual detection area information is a contamination diffusion agent concentration in the virtual detection area and the virtual prediction area information is a contamination diffusion agent concentration in the virtual prediction area. Method of forming a measurer.
The method according to claim 1,
Wherein the first plurality of lattices and the second plurality of lattices each have occurrence genetic information of the generation genetic algorithm.
The method according to claim 6,
Wherein generation of the second plurality of grids is performed by selecting at least two grids of the top 50% of the grids, selecting the two of the two grids so that the two are mutated at a probability of 25% Wherein the number of gratings of the second plurality of gratings is determined by repeating the number of gratings of the second plurality of gratings. The method according to claim 1,
Wherein said virtual data is computed through a Computational Fluid Dynamics program. &Lt; RTI ID = 0.0 &gt; 8. &lt; / RTI &gt;
A diffusion simulation unit for generating virtual data for reflecting physicochemical characteristics of the contaminant diffusion agent and time-space characteristics of the contaminant diffusion agent through diffusion simulation of the contaminant-diffusion agent;
An information extraction unit for setting a virtual detection area using the virtual data, extracting virtual detection area information for the virtual detection area, and extracting virtual prediction area information for a virtual prediction area excluding the virtual detection area information; And
And generating generic algorithms for evaluating fitness through backtracking of the virtual predicted region information using an occurrence genetic algorithm,
Wherein the generation genetic algorithm unit comprises:
Generating a first plurality of grids, performing kriging based on each of the first plurality of grids and the virtual detection area information, calculating an average error value of the kriging result and the virtual predicted area information And generating a second plurality of gratings by using the evaluation index and repeating the process until the average error value reaches a predetermined value. Measuring device.

Images