JP6156932B2 - Analysis apparatus, analysis method, and computer program - Google Patents

Description

  The present invention relates to an analysis technique for electromagnetic field analysis using a FDTD (Finite Difference Time Domain method) method.

  In recent years, a method for analyzing an electromagnetic field using the FDTD method which is a finite difference method in a time domain has been proposed. The FDTD method is applied in an analysis space divided into cells. The FDTD method is executed by discretizing the entire analysis space to be calculated into voxels having a size corresponding to the frequency at which the calculation is performed. An analysis space model discretized into voxels is called a voxel model.

  When performing electromagnetic field simulation such as radio wave propagation characteristics analysis using the FDTD method, a voxel model that reproduces the radio wave source (eg, antenna) and surrounding environment in the analysis space with a rectangular parallelepiped micro-block (Yee cell) Need to build. In the FDTD method, the electric field / magnetic field is determined by developing the Maxwell equation into a differential equation in the space / time domain and sequentially calculating it for all voxels constituting the voxel model. Electromagnetic field analysis using the FDTD method is generally computationally intensive and takes time. In order to solve such a problem, a technique for reducing the time required for electromagnetic field analysis using the FDTD method has been proposed (for example, see Patent Document 1).

JP 2008-250956 A

However, in the FDTD method, parameters for each voxel necessary for calculation are held in a memory. Therefore, there is a problem that a large amount of memory access occurs at the time of calculation and the calculation time becomes long.
In view of the above circumstances, an object of the present invention is to provide a technique for reducing the frequency of memory access during calculation and reducing the calculation time in electromagnetic field analysis using the FDTD method.

  One aspect of the present invention includes a reading unit that reads parameters used for calculation related to a plurality of models in a discretized analysis space, a common unit that shares the parameters among the plurality of models, and the plurality of models. A calculation unit that performs a calculation by the FDTD method using the parameters read by the reading unit, and the commoning unit groups parameters common to all models of the plurality of models, and the reading unit The unit is an analysis device that reads parameters necessary for the calculation of the plurality of models collectively in units grouped by the common unit.

  One aspect of the present invention includes a reading step for reading parameters used for calculation related to a plurality of models in a discretized analysis space, a common step for sharing the parameters among the plurality of models, and a discretized analysis. A calculation step of performing calculation by the FDTD method using the parameters read by the reading step for a plurality of models in the space, and the commoning step is common to all the models of the plurality of models In the analysis method, the parameters are grouped and the parameters required for the calculation of the plurality of models are collectively read in the unit grouped in the commonization step in the reading step.

  One embodiment of the present invention is a computer program for causing a computer to function as the analysis apparatus.

  According to the present invention, in the electromagnetic field analysis using the FDTD method, the frequency of memory access at the time of calculation can be reduced and the calculation time can be shortened.

It is a figure showing the specific example of the voxel model of the object analyzed using FDTD method. It is a figure showing the specific example of the voxel model of the object analyzed using FDTD method. It is the schematic diagram which represented the voxel model of the model 1 simply. It is the schematic diagram which represented the voxel model of the model 2 simply. It is a functional block diagram showing the functional structure of the analyzer 100 using the conventional method. It is a figure showing the timing of reading of a medium constant at the time of using the conventional method 1 for the calculation method using the FDTD method. It is a flowchart showing the flow of a process at the time of using the conventional method 1 for the calculation of the electromagnetic field analysis using the FDTD method. It is a figure showing the timing of reading of a medium constant at the time of using the conventional method 2 for the calculation method using the FDTD method. It is a flowchart showing the flow of a process at the time of using the conventional method 2 for the calculation of the electromagnetic field analysis by FDTD method. It is a functional block diagram showing functional composition of analysis device 100a of an embodiment. It is a figure showing reading of a medium constant in analysis device 100a of an embodiment. It is a flowchart showing the flow in which the analyzer 100a of embodiment calculates the calculation of an electromagnetic field analysis. It is a flowchart showing the flow of a reading process of a medium constant.

[Overview]
In the electromagnetic field analysis using the FDTD method, the Maxwell equation is expanded into a difference equation in the space / time domain for all voxels constituting the voxel model. An electric field and a magnetic field are determined by performing sequential calculation based on the developed formula. In the FDTD method, the time to be analyzed is divided into a plurality of unit times (hereinafter referred to as “time steps”). Electric field / magnetic field calculation is performed at every time step for all the voxels constituting the voxel model. When the electric field / magnetic field at a certain time step n (n is an integer of 1 or more) is calculated, the electric field / magnetic field at the next time step (n + 1) is determined based on the electric field / magnetic field determined at the time step n. The Therefore, in the FDTD method, when the processing shifts to the next time step, the medium constants of all the voxels constituting the voxel model to be analyzed are read. The medium constant is a constant representing the ease of propagation of the wave in the substance / object constituting the field where the wave propagates. For example, the medium constant is a dielectric constant or a magnetic permeability. In the FDTD method, a medium constant for each voxel is required. Therefore, an enormous number of memory accesses occur depending on the space to be analyzed.
The analysis apparatus according to the embodiment reduces the calculation time required for the electromagnetic field analysis using the FDTD method by reducing the number of memory accesses that occur when reading the medium constant necessary for the calculation. In order to clarify the effects obtained by the method of the embodiment, first, a method for reading the medium constant in the conventional method will be described.

1 and 2 are diagrams illustrating specific examples of a voxel model to be analyzed using the FDTD method. 1 and 2 are voxel models in which the space inside the vehicle of the same train is the object of analysis. FIG. 1 is a voxel model (hereinafter referred to as “model 1”) in a state where no passengers are present inside the vehicle. FIG. 2 is a voxel model (hereinafter referred to as “model 2”) in a state where passengers are seated in all seats inside the vehicle.
FIG. 3 is a schematic diagram simply showing the voxel model of model 1. FIG. 3 shows that the voxel model of the model 1 is composed only of models (vehicle models) that constitute the vehicle.
FIG. 4 is a schematic diagram simply showing the voxel model of model 2. FIG. 4 shows that the voxel model of model 2 includes a vehicle model and a model (human body model) that constitutes a human body part seated on a seat inside the vehicle.

There are two conventional methods for calculating a plurality of models using the FDTD method. One is a method of performing calculation of each model in series (hereinafter referred to as “conventional method 1”). The other is a method of calculating each model in parallel for each time step (hereinafter referred to as “conventional method 2”).
FIG. 5 is a functional block diagram showing a functional configuration of the analysis apparatus 100 using the conventional method. The analysis apparatus 100 includes a control unit 10, an FDTD calculation unit 11, an analysis space information memory 12, a voxel model storage memory 13, and a medium constant reading unit 14.

The control unit 10 controls the overall operation of the analysis apparatus 100.
The FDTD calculation unit 11 executes various calculation processes for electromagnetic field analysis using the FDTD method. The FDTD calculation unit 11 includes an FDTD calculation unit 11a, an electric field component absorption boundary condition calculation unit 11b, a magnetic field component calculation unit 11c, and a magnetic field component absorption boundary condition calculation unit 11d.

The analysis space information memory 12 is configured using a storage device such as a magnetic hard disk device or a semiconductor storage device. The analysis space information memory 12 stores the configuration of the analysis space together with information on the coordinate system. The analysis space information memory 12 also stores a medium constant value corresponding to each point on the analysis space.
The voxel model storage memory 13 is configured using a storage device such as a magnetic hard disk device or a semiconductor storage device. The voxel model storage memory 13 stores voxel model information.
The medium constant reading unit 14 reads the medium constant corresponding to the calculation target voxel from the analysis spatial information memory 12.

FIG. 6 is a diagram showing the reading timing of the medium constant when the conventional method 1 is used as the calculation method using the FDTD method. In the conventional method 1, the calculation of the model 1 and the model 2 is performed in series. Further, in the conventional method 1, calculation is performed for all time steps for each model.
FIG. 7 is a flowchart showing the flow of processing when the conventional method 1 is used for calculation of electromagnetic field analysis using the FDTD method. First, the control unit 10 reads a target voxel model to be calculated from the voxel model storage memory 13 (step S101).

  The control unit 10 reads information on the absolute value of the coordinates of the three-dimensional space applied to the voxel model to be calculated from the analysis space information memory 12 (step S102). The information on the absolute value of the coordinates in the three-dimensional space is, for example, three-dimensional CAD data (3D-CAD data) that defines the structure of the analysis space. The control unit 10 applies the absolute value of the input coordinates in the three-dimensional space to the voxel model, and generates a three-dimensional array corresponding to the voxel coordinates. When the controller 10 generates an array corresponding to the voxel model, the controller 10 assigns a medium constant corresponding to each coordinate value (step S103).

  The control unit 10 updates the time step at the target time for calculation (step S104). The control unit 10 sets a time obtained by adding a predetermined unit time to the current time step as a new time step. The medium constant reading unit 14 reads the medium constant for the calculation target voxel of the calculation target voxel model (step S105).

  The FDTD calculator 11 reads the electromagnetic field component of the time step immediately before the voxel to be calculated in order to calculate the electromagnetic field component of the current time step (step S106). The FDTD calculation unit 11 calculates the electromagnetic field component at the current time step of the voxel to be calculated based on the electromagnetic field component at the immediately previous time step (step S107).

  When the calculation of the electromagnetic field component in the calculation target voxel is completed, the control unit 10 determines whether the calculation of the electromagnetic field component is completed for all the voxels of the calculation target voxel model in the current time step ( Step S108). If there is a voxel that has not been calculated in the current time step (NO in step S108), the process returns to step S105, and the calculation of the electromagnetic field component is continued for the voxel that has not been calculated.

  On the other hand, when calculation of all voxels is completed in the current time step (step S108-YES), the control unit 10 determines whether or not the calculation of the electromagnetic field component has converged (step S109). If the calculation of the electromagnetic field component has not converged (step S109—NO), the process returns to step S104, and the calculation is continued until the calculation of the electromagnetic field component converges.

  On the other hand, when the calculation of the electromagnetic field component has converged (YES in step S109), the control unit 10 determines whether the calculation has been completed for all the voxel models (step S110). If there is a voxel model that has not been calculated (step S110-NO), the process returns to step S101, and the calculation for the voxel model that has not been calculated is continued. On the other hand, when the calculation is completed for all the voxel models (step S110—YES), the analysis apparatus 100 ends the analysis process and outputs the result.

FIG. 8 is a diagram showing the reading timing of the medium constant when the conventional method 2 is used as the calculation method using the FDTD method. In the conventional method 2, calculation of the model 1 and the model 2 is performed in parallel. In the conventional method 2, calculation is performed for all voxel models at each time step.
FIG. 9 is a flowchart showing the flow of processing when the conventional method 2 is used for calculation of electromagnetic field analysis by the FDTD method. First, the control unit 10 reads all voxel models to be calculated from the voxel model storage memory 13 (step S201).

  The control unit 10 reads information on the absolute value of the coordinates of the three-dimensional space to be applied to each voxel model from the analysis space information memory 12 (step S202). The control unit 10 applies the absolute value of the input coordinate in the three-dimensional space to each voxel model, and generates a three-dimensional array corresponding to the voxel coordinate. When the controller 10 generates an array corresponding to the voxel model, the controller 10 assigns a medium constant corresponding to each coordinate value of each model (step S203).

  The control unit 10 updates the time step at the target time for calculation (step S204). The control unit 10 sets a time obtained by adding a predetermined unit time to the current time step as a new time step. The medium constant reading unit 14 reads the medium constant for the calculation target voxel of each calculation target voxel model (step S205).

  The FDTD calculation unit 11 reads the electromagnetic field component of the time step immediately before the voxel to be calculated in order to calculate the electromagnetic field component of the current time step (step S206). The FDTD calculation unit 11 calculates the electromagnetic field component at the current time step of the voxel to be calculated based on the electromagnetic field component at the immediately preceding time step (step S207).

  When the calculation of the electromagnetic field component in the calculation target voxel is completed, the control unit 10 determines whether or not the calculation of the electromagnetic field component is completed for all the voxels of all the voxel models in the current time step (step). S208). If there is a voxel that has not been calculated in the current time step (NO in step S208), the process returns to step S205, and the calculation of the electromagnetic field component is continued for the voxel that has not been calculated.

  On the other hand, when calculation of all voxels is completed in the current time step (step S208—YES), the control unit 10 determines whether or not the calculation of the electromagnetic field component has converged (step S209). When the calculation of the electromagnetic field component has not converged (step S209—NO), the process returns to step S204, and the calculation of the electromagnetic field component is continued. On the other hand, when the calculation of the electromagnetic field component converges (step S209—YES), the analysis apparatus 100 ends the analysis process and outputs the result.

As described above, in the electromagnetic field analysis using the FDTD method according to the conventional method, the medium constant is read the same number of times in both the conventional method 1 and the conventional method 2. That is, when the time required to calculate one model is T, the number of voxels of the calculation target voxel model is N, and the number of calculation target models is M, the memory by reading the medium constant generated by the calculation of one model The number of accesses is M × T × N.
The analysis apparatus according to the embodiment reduces the calculation time required for the electromagnetic field analysis using the FDTD method by reducing the number of memory accesses that occur when reading the medium constant necessary for the calculation.
Hereinafter, the analysis apparatus of the embodiment will be described.

[Embodiment]
FIG. 10 is a functional block diagram illustrating a functional configuration of the analysis apparatus 100a according to the embodiment. The analysis apparatus 100a is different from the analysis apparatus 100 using the conventional method shown in FIG. 5 in that the medium constant sharing unit 15 is provided.

  The medium constant sharing unit 15 groups voxels having a medium constant common to all voxel models as a common part. Further, the medium constant sharing unit 15 groups voxels other than the common part as non-common parts. For example, in the example of FIGS. 3 and 4, the medium constant sharing unit 15 sets the human body part (human body model) as a non-common part and sets the vehicle part (vehicle model) other than the human body part (human body model) as a common part. Group. The medium constant sharing unit 15 stores the information on the common part and the non-common part grouped in this way in the analysis space information memory 12 as grouping information.

  FIG. 11 is a diagram illustrating reading of a medium constant in the analysis apparatus 100a of the embodiment. In the analysis apparatus 100a of the embodiment, the calculation of the model 1 and the model 2 is performed in parallel as in the conventional method 2, and the calculation is performed for all voxel models at each time step.

  FIG. 12 is a flowchart illustrating a flow in which the analysis apparatus 100a according to the embodiment performs calculation of electromagnetic field analysis. First, the control unit 10 reads all voxel models to be calculated from the voxel model storage memory 13 (step S301).

  The control unit 10 reads information on the absolute value of the coordinates of the three-dimensional space applied to each voxel model from the analysis space information memory 12 (step S302). The control unit 10 applies the absolute value of the input coordinate in the three-dimensional space to each voxel model, and generates a three-dimensional array corresponding to the voxel coordinate. When generating an array corresponding to the voxel model, the control unit 10 assigns a medium constant corresponding to each coordinate value of each model (step S303).

  The medium constant sharing unit 15 shares the medium constant assigned to the voxel in step S303 among the voxel models (step S304). Specifically, the medium constant sharing unit 15 compares the medium constants assigned to voxels located at the same coordinates among the voxel models. As a result of the comparison, the medium constant sharing unit 15 defines voxels having the same medium constant among all voxel models as a common part. Further, the medium constant sharing unit 15 defines voxels that are not defined as common portions as non-common portions for each voxel model. The medium constant sharing unit 15 uses information obtained by combining the definitions of the voxels of the common part and the non-common part and the medium constants corresponding to the common part and the non-common part as the common information of the medium constant as the analysis space information memory 12. Save to.

The control unit 10 updates the time step at the target time for calculation (step S305). The control unit 10 sets a time obtained by adding a predetermined unit time to the current time step as a new time step.
The medium constant reading unit 14 performs a medium constant reading process on the calculation target voxel of each calculation target voxel model (step S306).

FIG. 13 is a flowchart showing the flow of the medium constant reading process.
First, the medium constant reading unit 14 determines whether or not the medium constant corresponding to the calculation target voxel is read for the calculation target voxel model (step S401). If the medium constant has already been read (step S401—YES), the medium constant reading unit 14 ends the process without reading the medium constant. On the other hand, when the medium constant is not read (step S401-NO), the medium constant reading unit 14 refers to the common information of the medium constant stored in the analysis space information memory 12 (step S402).

The medium constant reading unit 14 determines whether the target voxel is a common part from the common information of the medium constant (step S403). When the target voxel is a common part (step S403-YES), the medium constant reading unit 14 reads the medium constant corresponding to the target voxel (step S404). Then, the medium constant reading unit 14 sets the read medium constant as the medium constant of the corresponding voxel of all the common voxel models (step S204).
On the other hand, when the target voxel is not a common part (step S403-NO), the medium constant reading unit 14 reads the medium constant of the non-common part of the voxel model to which the target voxel belongs (step S405).

Returning to the description of FIG. The FDTD calculation unit 11 reads the electromagnetic field component of the time step immediately before the voxel to be calculated in order to calculate the electromagnetic field component of the current time step (step S307).
The FDTD calculation unit 11 calculates the electromagnetic field component at the current time step of the voxel to be calculated based on the electromagnetic field component of the immediately previous time step (step S308).

  When the calculation of the electromagnetic field component in the calculation target voxel is completed, the control unit 10 determines whether or not the calculation of the electromagnetic field component is completed for all the voxels of all the voxel models in the current time step (step). S309). If there is a voxel that has not been calculated at the current time step (step S309—NO), the process returns to step S306, and the calculation of the electromagnetic field component is continued for the voxel that has not been calculated.

On the other hand, when calculation of all voxels is completed in the current time step (step S309—YES), the control unit 10 determines whether or not the calculation of the electromagnetic field component has converged (step S310). If the calculation of the electromagnetic field component has not converged (step S310—NO), the process returns to step S305, and the calculation of the electromagnetic field component is continued.
On the other hand, when the calculation of the electromagnetic field component has converged (step S310—YES), the analysis apparatus 100a ends the analysis process and outputs the result.

  In the analysis apparatus 100a of the embodiment configured as described above, the medium of the voxel model calculated in parallel is obtained by sharing the reading of the medium constant for each voxel necessary for the calculation of the electromagnetic field analysis among the voxel models. Duplicate reading of constants is eliminated. Therefore, the number of memory accesses required for reading the medium constant is reduced. As a result, the calculation speed of electromagnetic field analysis by the FDTD method can be improved.

For example, in the example of FIGS. 3 and 4, the number of voxel models is M, the number of voxels constituting the voxel model is N, the number of voxels in the non-common part among the voxels constituting the voxel model is P, and the FDTD method is used. Consider a case where T is the number of time steps required to complete the calculation.
In the conventional FDTD method, the number of memory accesses generated by updating the medium constant at the time of calculation is M × N × T times, whereas in this embodiment, {(N−P) + P × M} × T times.
M * N * T-{(NP) + P * M} * T = (M-1) * T * P
That is, in the FDTD method of the present embodiment, (M−1) × T × P memory accesses can be reduced compared to the conventional FDTD method, and the higher the voxel models that perform calculations in parallel, the higher the reduction. An effect can be obtained. Further, in the FDTD method of the present embodiment, an effect is not obtained at a limited timing such as only in the initial stage of analysis, but an effect is obtained in the entire processing time.

You may make it implement | achieve the analysis apparatus 100a in embodiment mentioned above with a computer. In that case, a program for realizing this function may be recorded on a computer-readable recording medium, and the program recorded on this recording medium may be read into a computer system and executed. Here, the “computer system” includes an OS and hardware such as peripheral devices. The “computer-readable recording medium” refers to a storage device such as a flexible medium, a magneto-optical disk, a portable medium such as a ROM and a CD-ROM, and a hard disk incorporated in a computer system. Furthermore, the “computer-readable recording medium” dynamically holds a program for a short time like a communication line when transmitting a program via a network such as the Internet or a communication line such as a telephone line. In this case, a volatile memory inside a computer system serving as a server or a client in that case may be included and a program held for a certain period of time. Further, the program may be a program for realizing a part of the above-described functions, and may be a program capable of realizing the functions described above in combination with a program already recorded in a computer system. You may implement | achieve using programmable logic devices, such as FPGA (Field Programmable Gate Array).
The embodiment of the present invention has been described in detail with reference to the drawings. However, the specific configuration is not limited to this embodiment, and includes designs and the like that do not depart from the gist of the present invention.

DESCRIPTION OF SYMBOLS 100, 100a ... Analysis apparatus, 10 ... Control part, 11 ... FDTD calculation part, 11a ... Electric field component calculation part, 11b ... Electric field component absorption boundary condition calculation part, 11c ... Magnetic field component calculation part, 11d ... Magnetic field component absorption boundary condition calculation , 12 ... Analysis space information memory, 13 ... Voxel model storage memory, 14 ... Medium constant reading unit, 15 ... Medium constant sharing unit

Claims (3)

A reading unit for reading in parameters used for calculations on multiple models of the discretized analysis space;
A common unit for sharing the parameters among the plurality of models;
For the plurality of models, a calculation unit that performs calculation by the FDTD method using the parameters read by the reading unit;
With
The common unit groups parameters common to all models of the plurality of models,
The analysis unit, wherein the reading unit collectively reads parameters necessary for the calculation of the plurality of models in units grouped by the commoning unit. An analysis method executed by an analysis device including a reading unit, a common unit, and a calculation unit,
A step of reading the parameters used for calculation related to a plurality of models of the discretized analysis space;
A commonization step in which the commonization unit shares the parameters among the plurality of models;
A calculation step in which the calculation unit calculates a plurality of models of the discretized analysis space by the FDTD method using the parameters read by the reading step;
Have
In the sharing step, the sharing unit groups parameters common to all models of the plurality of models,
In the reading step, the reading unit is a parameter necessary for the calculation of the multiple models, to read collectively in units grouped in said consolidating step, the analysis method.   A computer program for causing a computer to function as the analysis apparatus according to claim 1.

Images