JP2009258879A - Design support device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To satisfactorily reduce a retrieval period by eliminating unnecessary calculation without deteriorating the accuracy of retrieving the optimal shape in fluid analysis concerning a design support device. <P>SOLUTION: This design support device for calculating the optimal shape of an intake port satisfying a target evaluation index (tambour ratio or the like) by numerical fluid calculation is provided with a similarity decision part 30 for deciding whether the shapes are similar between a new analytic shape and a past analytic shape based on the deformation quantity of design variables selected by an optimization retrieval part 26. Furthermore, this design support device is provided with a design variable sensitivity decision part 32 for, when it is decided that the new analytic shape is similar to the past analytic shape by the similarity decision part 30, deciding whether the new analytic shape has low sensitivity to the analytic result of the evaluation index. When it is decided that the new analytic shape has low sensitivity, the new analytic shape is excluded from the object of analytic calculation. <P>COPYRIGHT: (C)2010,JPO&INPIT

Description

  The present invention relates to a design support apparatus for supporting a design of an intake port and the like by a computer.

  Conventionally, for example, Patent Document 1 discloses a design support method and a design support system for obtaining an optimal combination of design variables for the structure optimization of a design object. More specifically, in this conventional method (and system), design variables are arbitrarily set for a design object, analysis calculation is performed using an analysis mesh model based on the set design variables, and the analysis calculation is performed. The optimization calculation for determining the analysis result is repeated.

  Further, in the above-described conventional technology, a set of executed analysis results and parameters (design variables and analysis conditions) is stored in a database. Then, when a new analysis calculation is performed, based on the parameter combination in the new analysis calculation, a search is made as to whether there is a pair of parameter pairs that can be applied to the past analysis result in the database. When a pair of parameters is extracted from the database that is the same as or more similar to the set of parameters in the new analysis calculation, and the pair of parameters is extracted, the extracted past analysis result is used for the new calculation result. This is applied as a result of analysis. As a result, the calculation efficiency in the optimization calculation is improved (calculation time is shortened).

JP 2007-164393 A JP 2007-122205 A JP 2006-344096 A

  However, for example, in a fluid analysis such as a flow analysis in an intake port of an internal combustion engine, even if analysis parameters such as design variables are similar, the analysis results are not necessarily close. This is because the difference in shape (difference in analysis parameters such as design variables) greatly affects the analysis results when calculating vortex information such as tumble ratio due to the influence of fluid turbulence in the flow field. is there. Therefore, in fluid analysis, when a new analysis calculation is performed, if the analysis parameter such as a design variable is similar to the past analysis parameter, the new analysis parameter is simply excluded from the calculation target. There is a concern that the search accuracy of the optimum shape will deteriorate.

  The present invention has been made to solve the above-described problems. In the fluid analysis, useless calculation can be omitted without deteriorating the search accuracy of the optimum shape, thereby favorably reducing the search period. An object of the present invention is to provide a design support apparatus that can be used.

A first invention is a design support apparatus,
A design support device for obtaining an optimum shape of a structure that satisfies a target evaluation index by numerical fluid calculation,
Cell information acquisition means for acquiring a calculation result according to a predetermined analysis parameter and a predetermined rule for each of a plurality of cells constituting the structure model to be analyzed;
Analysis execution means for analyzing the evaluation index based on the calculation results for the plurality of cells;
Analysis history storage means for storing the analysis result of the evaluation index in association with the corresponding analysis parameter;
Similarity determination means for determining whether a new analysis parameter to be newly analyzed and a past analysis parameter that has already been analyzed are similar,
Whether the new analysis parameter is an analysis parameter having low sensitivity to the analysis result of the evaluation index when the similarity determination unit determines that the new analysis parameter is similar to the past analysis parameter Parameter sensitivity determination means for determining whether or not
Calculation exclusion determination means for excluding the new analysis parameter from the target of analysis calculation when the parameter sensitivity determination means determines that the new analysis parameter is an analysis parameter having low sensitivity to the analysis result of the evaluation index; ,
It is characterized by providing.

The second invention is the first invention, wherein
Cell deformation amount setting means for setting the deformation amounts of a plurality of cells based on the analysis parameters;
Morphing execution means for performing node movement of the plurality of cells based on the deformation amount set by the cell deformation amount setting means,
The similarity determination means includes node movement information in the node movement target area by the morphing execution means, and node movement information in an interpolation area between the node movement target area and the node movement non-target area. Based on this, it is determined whether the new analysis parameter is similar to the past analysis parameter.

  According to the first aspect, when searching for the optimum shape, the new analysis parameter is analyzed by performing the parameter sensitivity determination for determining the sensitivity of the new analysis parameter together with the similarity determination for comparing the new analysis parameter with the past analysis parameter. It is determined whether or not to exclude from the calculation target. Only when it is determined that the sensitivity of the new analysis parameter to the analysis result of the evaluation index is low and it is safe to exclude the new analysis parameter from the analysis calculation target, the new analysis parameter is excluded from the analysis calculation target. I try to exclude it. For this reason, according to the present invention, in the fluid analysis, useless calculation can be omitted without causing deterioration of the search accuracy of the optimum shape, and thus the search period can be favorably shortened.

  According to the second invention, the similarity between the new analysis parameter and the past analysis parameter is evaluated based on not only the node movement information in the target area of the node movement by the morphing execution means but also the node movement information in the interpolation area. I am doing so. For this reason, according to the present invention, since the interpolation region is also a sample node extraction target, high-precision similarity determination can be realized. Further, since a wide range search can be performed at the initial stage of the optimized search, the search for the same or similar analysis parameter is reduced at the initial stage of the search, and it is possible to suppress falling into a local optimal solution at the initial stage of the search.

Embodiment 1 FIG.
[Description of system configuration]
FIG. 1 is a diagram for explaining a hardware configuration of a design support system according to Embodiment 1 of the present invention. The design support system of this embodiment is a system for supporting the design of a structure (intake port here) that is a design target (analysis target), and a general-purpose computer 10 as shown in FIG. 1 is used as hardware. It is feasible.

  A computer 10 shown in FIG. 1 includes an input device 12 that receives input of predetermined design conditions (design variables and the like), a CPU 14 that executes a predetermined program based on the input design conditions, and an output device that outputs a calculation result by the CPU 14. 16 and the basic components such as the storage device 18 in which various programs executed by the CPU 14 and various data necessary for arithmetic processing are stored.

  FIG. 2 is a diagram for explaining a software configuration of the design support system according to the first embodiment of the present invention. As shown in FIG. 2, the system of this embodiment includes a base mesh model creation unit 20, a condition setting unit 22, a calculation solver 24, an optimization search unit 26, a morphing processing unit 28, a similarity determination unit 30, and a design variable. A sensitivity determination unit 32 is provided.

  Here, an example in which the present design support system is applied will be described as a system for performing fluid analysis for optimizing the shape of the intake port of the internal combustion engine. Specifically, as an evaluation index for optimizing the intake port, the in-cylinder tumble (vertical vortex) ratio TR and the intake port flow coefficient CF are used. The aim of this system is to search for and obtain an optimal port shape that satisfies both the tumble ratio TR and the flow coefficient CF which are in a trade-off relationship.

  The storage device 18 stores three-dimensional shape data (CAD (Computer Aided Design) data) of the intake port created in advance by the user of the present design support system. Further, the storage device 18 stores information on each cell of the mesh model and various other work data. The base mesh model creation unit 20 has a function of reading the CAD data from the storage device 18 and then dividing the CAD data into a plurality of cells (finite elements) to create a base mesh model. . The base mesh model is basically automatically created by the base mesh model creating unit 20, but may be created by the user's hand.

  In the condition setting unit 22, boundary conditions and calculation conditions input from the user by the input device 12 are set. The boundary conditions here are the boundary conditions of the mesh model (intake port wall surface position and temperature information, etc.), and the calculation conditions are the various calculation conditions (intake port) required for the mesh model flow calculation. Design variables such as the diameter of each part and port height, and the flow velocity at the mesh model inlet).

  In the calculation solver 24, the flow calculation is executed on the mesh model under the environmental condition and the calculation condition obtained from the condition setting unit 22. FIG. 3 is a simplified view of the mesh model 40 of the intake port. As shown in FIG. 3, information on the flow velocity at each position is given to the volume center position of each cell 42 of the mesh model 40 in association with each cell 42. The calculation solver 24 calculates the flow velocity for each cell 42 of the mesh model 40.

  The calculation solver 24 calculates the tumble ratio TR in the cylinder of the internal combustion engine and the flow coefficient CF of the intake port according to a predetermined relational expression based on the calculation result of the flow velocity for each cell 42. Here, the information on the flow velocity is given to the center position of each cell 42. However, the present invention is not limited to this, and the information on the flow velocity may be given to each node of each cell 42. Further, the information given to each cell 42 may not be the gas flow velocity in the intake port but the pressure in the intake port or the kinetic energy of the gas flowing in the intake port.

  In the system of the present embodiment, the above calculation is repeated while changing the shape of the mesh by the morphing processing unit 28 until both the tumble ratio TR calculated by the calculation solver 24 and the flow coefficient CF satisfy the target value. I am doing so. For such a purpose, the calculation result by the calculation solver 24 is sent to the optimization search unit 26.

  Further, the optimization search unit 26 uses the calculation result of the calculation solver 24 and the estimated value (described later) by interpolation of each evaluation index (such as a tumble ratio) by the design variable sensitivity determination unit 32 of the base mesh model 40. The amount of deformation of the design variables (the diameter of each part of the intake port, the port height, etc.) (that is, the amount of deformation of the mesh model 40 by the morphing processing unit 28) is selected to be the optimal solution.

  In the morphing processing unit 28, processing for changing the shape of the mesh model 40, that is, morphing is performed in accordance with the deformation amount of the design variable selected by the optimization searching unit 26. More specifically, the shape of the mesh model 40 is changed so that the diameter and port height of each part of the intake port become values selected by the optimization search unit 26.

  The calculation solver 24 morphs the appropriate design variables selected by the optimization search unit 26 until a calculation result is obtained that can determine that an optimal intake port shape that satisfies both the target tumble ratio TRm and the target flow coefficient CFm is obtained. While giving to the processing unit 28, the calculation is repeated again on the mesh model 40 whose shape has been changed by the morphing processing unit 28.

FIG. 4 is a diagram illustrating an example of a shape change of the mesh model 40 accompanying execution of morphing.
More specifically, FIG. 4A shows a part of the mesh model 40 before execution of morphing, and FIG. 4B shows the mesh model 40 in the direction of the arrow in FIG. A part of the mesh model 40 after the morphing is executed so that the cell group 42 to be morphed is enlarged.

  Morphing is a process of directly changing the node of each cell 42 of the mesh model 40, whereby the shape of each cell 42 is changed. When performing morphing, as shown in FIG. 4 (B), the mesh model 40 has a main nodal movement target region in which the nodal movement amount based on the design variable is defined and controlled, and the nodal movement non-target. A region is distinguished from an interpolation region that is smoothly deformed so as to interpolate between them, and the movement of the node of each cell 42 in morphing is performed with respect to the target region of the node movement and the interpolation region. become. Note that the non-target region of the node movement mentioned here corresponds to a portion (port throat portion) on the outlet side (combustion chamber side) of the intake port.

  According to the above method of changing the shape of the mesh model 40 by performing morphing, the mesh model 40 can be acquired in a short time without having to recreate the mesh model 40 on the CAD one by one. Further, according to morphing, it is possible to change the shape of the mesh model 40 while maintaining the coarse and dense distribution of the entire mesh model 40 without largely changing from the base mesh model 40 before the shape change.

  As a result, the deviation amount of the model calculation result with respect to the actual machine evaluation does not change greatly every time the shape is changed, so that the model calculation result can be relatively evaluated. Furthermore, according to the above method, in addition to the automatic correction of the mesh model 40 by morphing and evaluation of the result of the model calculation, the optimum shape according to the purpose can be automatically calculated by combining the optimization search described above. It becomes possible.

  In the similarity determination unit 30 shown in FIG. 2, the shape of the mesh model 40 after the shape change based on the deformation amount of the design variable selected by the optimization search unit 26 (hereinafter, referred to as “new analysis shape”). ) And the shape of the mesh model 40 before the shape change (hereinafter, sometimes referred to as “past analysis shape”), it is determined whether or not the shapes are similar.

  Further, in the design variable sensitivity determination unit 32, when the similarity determination unit 30 determines that the new analysis shape and the past analysis shape are similar, the new analysis shape becomes the analysis result of the evaluation index such as the tumble ratio. It is determined whether the shape has a low sensitivity.

  In this embodiment, when the design variable sensitivity determination unit 32 determines that the new analysis shape is a shape with low sensitivity, the new analysis shape is excluded from the target of the analysis calculation. . The system according to the present embodiment is characterized in that the similarity determination unit 30 and the design variable sensitivity determination unit 32 are provided, and the configuration of each of the determination units 30 and 32. The characteristic configuration will be described in detail below with reference to FIGS.

FIG. 5 is a conceptual diagram showing a state in which the search by the optimization search unit 26 is performed. Note that the evaluation indexes 1 and 2 in FIG. 4 correspond to the flow coefficient CF and the tumble ratio TR in the case of the flow analysis of the intake port of the present embodiment.
As indicated by the black circles in FIG. 4, an optimization search is performed to obtain a solution (analysis result). In the present embodiment, analysis results (evaluation results) obtained in the search process, design variable values, and analysis conditions are stored in a database as needed in association with each other.

  In this embodiment, useless calculation can be omitted without degrading the search accuracy of the optimum shape, and thus the similarity determination unit 30 and the design variable described above in the search process can be shortened in order to favorably shorten the search period. The sensitivity determination unit 32 is used. First, the similarity determination unit 30 more specifically evaluates the similarity between the new analysis shape and the past analysis shape according to the method shown in FIG. 6 below.

FIG. 6 is a diagram for explaining a similarity determination method performed by the similarity determination unit 30 illustrated in FIG. 2.
FIG. 6A shows an example of shape change by morphing (change of port diameter). 6B and 6C show a cross section of the intake port in a state where the mesh is not shown. FIG. 6B shows the shape before the shape change by morphing, and FIG. 6C shows the shape after the shape change by morphing.

In the similarity determination unit 30, each time a new analysis shape is created by an optimization search, the reference coordinates based on the deformation direction of the port diameter (design variable) as shown in FIG. A system is being constructed. Then, as shown in FIG. 6B, arbitrary n sample nodes are provided from the nodes of the cell 42 on the shape surface of the intake port, and the vector value u from the reference coordinates is set for each sample node. _i is stored in the database.

Further, as shown in FIG. 6C, the similarity determination unit 30 also accumulates in the database vector values v _i for each of the sample nodes whose coordinate values have changed due to shape change by morphing. Further, in the present embodiment, as shown in FIGS. 6B and 6C, sample nodes are extracted not only from the target area of the node movement by morphing but also from the interpolation area.

The similarity determination unit 30 performs similarity determination between the new analysis shape and the past analysis shape using the similarity S obtained by substituting the vector values u _i and v _i into the following equation (1). ing.

According to the above equation (1), the similarity S approaches 1 as the two vector values u _i and v _i approach. Therefore, in the present embodiment, the similarity S is calculated between the vector value u _i of the past analysis shape referenced from the database and the vector value v _i of the new analysis shape.

  If the predetermined similarity characteristic value | S-1 | is smaller than the predetermined threshold value γ, it is determined that both shapes are similar, and the process proceeds to the process of the design variable sensitivity determination unit 32. . On the other hand, when the similar characteristic value | S-1 | is equal to or greater than the threshold value γ, it is determined that both shapes are different shapes, and a predetermined analysis calculation is performed on the new analysis shape this time. Yes.

FIG. 7 is a diagram for explaining a method for calculating an estimated value of each evaluation index for a new analysis shape in the design variable sensitivity determination unit 32 illustrated in FIG. 2, and illustrates three evaluation indices 1 to 3. Yes.
In the design variable sensitivity determination unit 32, as shown in FIG. 7, for each evaluation index (tumble ratio TR, flow coefficient CF, etc.), design variables (shapes a to d) analyzed in the past and each evaluation index are calculated. In the meantime, an nth-order evaluation approximate expression is obtained by a least square method or the like. Then, using the evaluation approximation formula, as shown in FIG. 7, the estimated value of each evaluation index corresponding to the design variable value of the new analysis shape is calculated.

  If the estimated value of each evaluation index calculated as described above is a value with a small error from the actual value, using this estimated value and omitting the execution of a new analysis calculation will optimize the search. The search time in can be shortened satisfactorily.

  FIG. 8 is a diagram illustrating an example in which an error exists between the approximate curve and the true evaluation index curve. However, for example, as shown in FIG. 8, in the case where the slope of the true evaluation index curve (solid line) changes abruptly, the estimated value of each evaluation index is calculated from the approximate curve (broken line) based on the above evaluation approximation formula. If calculated, the true value of the evaluation index and the estimated value may deviate greatly. Even in such a case, if the new analysis shape is similar to the past analysis shape and the analysis calculation for the new analysis shape is excluded, the search accuracy of the optimization search deteriorates.

  Therefore, the design variable sensitivity determination unit 32 determines whether or not analysis calculation should be performed on a new analysis shape similar to the past analysis shape by performing the following processing. .

FIG. 9 is a diagram for explaining a sensitivity determination method for a new analysis shape by the design variable sensitivity determination unit 32 shown in FIG. That is, in the sensitivity determination, first, the solutions of the shapes b and c near the estimated value (diamond mark) of the new analysis shape are extracted as sample points (triangle mark and square mark). Then, tangent slopes k _b and k _c (sensitivity) of the approximate curve (broken line) by the evaluation approximation formula at each sample point are calculated.

Here, sensitivity determination is performed using the rate of change of the slopes k _b and k _c between these sample points. Specifically, when the shape b in the vicinity of the estimated value of the new analysis shapes, the discrete width of the design variables c and [Delta] x _i, the rate of change of slope _{k b, k c, | k} b -k c | / it can be calculated as Δx _i.

The design variable sensitivity determination unit 32 determines whether or not the determination expression | k _b −k _c | / Δx _i ≦ ζ is satisfied with ζ as a threshold value. When this determination formula | k _b −k _c | / Δx _i ≦ ζ holds, the change from the slope k _b to the slope k _c is gentle. In this case, the design variable sensitivity determination unit 32 has a sufficient error from the true value of the evaluation index even if the estimated value calculated using the evaluation approximate expression as described above is used as the solution of the new analysis shape. The new analysis shape is excluded from the target of the analysis calculation by using the estimated value.

On the other hand, when the determination formula | k _b −k _c | / Δx _i > ζ is satisfied, the change from the inclination k _b to the inclination k _c is abrupt, that is, the approximate curve has an uneven shape. I can judge. In this case, the design variable sensitivity determination unit 32 determines that the approximation error is likely to be large unless analysis calculation is performed on the new analysis shape, stops using the estimated value, and creates a new analysis shape. On the other hand, a predetermined analysis calculation is executed.

FIG. 10 is a flowchart showing a design support program (routine) representing a series of processes of fluid analysis calculation with optimization search in the system of the present embodiment including the similarity determination unit 30 and the design variable sensitivity determination unit 32 described above. It is.
In the design support program shown in FIG. 10, first, the target tumble ratio TRm and the target flow coefficient CFm are selected as target values (a target performance value desired by the user is received by the input device 12), and the base mesh model creation unit 20 As a result, a target base mesh model 40 is created (step 101).

  Next, after predetermined boundary conditions and calculation conditions are set by the condition setting unit 22, the flow velocity of each cell 42 in the base mesh model 40 is calculated by the calculation solver 24, and the tumble is calculated based on the calculation result. The ratio TR and the flow coefficient CF are calculated (step 102). More specifically, the flow velocity, the tumble ratio TR, and the flow coefficient CF are repeatedly calculated every predetermined calculation step (predetermined time).

Next, the optimization search unit 26 uses a design of experiment (DOE) or optimization algorithm to change the shape deformation amount due to morphing (design variables such as the intake port diameter and port height in the base mesh model 40). A deformation amount X _i is calculated (step 103).

Next, based on the shape deformation amount X _i calculated as described above, the morphing processing unit 28 changes the shape of the mesh model 40 by morphing (step 104).

Next, as described above with reference to FIG. 6, the similarity determination unit 30 sets a reference coordinate system for similarity determination and any n sample nodes (step 105). Next, the vector value u _i of the past analysis shape and the vector value v _i of the new analysis shape referenced from the database are calculated (step 106).

At this time, as described above, the sample values are extracted not only from the target area of the node movement by morphing but also from the interpolation area, and the vector values u _i and v _i are calculated. Next, using these vector values u _i and v _i , the similarity S is calculated according to the above equation (1) (step 107).

  Next, it is determined whether or not the similarity characteristic value | S-1 | based on the similarity S is smaller than the threshold value γ (step 108). As a result, if it is determined that the similar characteristic value | S-1 | is equal to or greater than the threshold value γ, it can be determined that the new analysis shape this time is not similar to the past analysis shape stored in the database. In this case, the process proceeds to step 116, and analysis calculation is executed for the new analysis shape.

  On the other hand, when it is determined in step 108 that the similar characteristic value | S-1 | is smaller than the threshold value γ, the new analysis shape this time is a shape similar to the past analysis shape stored in the database. I can judge. In this case, the design variable sensitivity determination unit 32 first refers to the database of the calculation result of the past analysis shape, and compares the design variables (shapes a to d) analyzed in the past and each evaluation index. In the meantime, an nth-order evaluation approximate expression is obtained by the least square method or the like (step 109).

  Next, the estimated values TRf and CFf of each evaluation index (tumble ratio and flow coefficient) are calculated for the new analysis shape by the design variable sensitivity determination unit 32 based on the evaluation approximate expression (step 110). . Next, solutions of two past analysis shapes in the vicinity of the estimated values TRf and CFf of the new analysis shape are set as sample points (step 111).

Next, the design variable sensitivity determination unit 32 calculates the change rate (tangential slope) k _i−1 , k _{i + 1} of the approximate curve according to the evaluation approximation formula at the two sample points (step 112). Next, the rate of change | k _b −k _c | / Δx _{i of the} two slopes k _i−1 and k _{i + 1} calculated by the design variable sensitivity determination unit 32 with the discrete width as Δx _i is equal to or less than the threshold ζ _i. Is determined (step 113).

As a result, when it is determined that the inclination change rate | k _b −k _c | / Δx _i ≦ ζ _i is satisfied, the current new analysis shape is similar to the past analysis shape, and It can be determined that the sensitivity of the design variable is low and the error from the true evaluation index value is small even if the estimated value is used. For this reason, in this case, it is determined that the analysis calculation for the new analysis shape should be excluded (step 114). Then, the estimated values TRf and CFf of each evaluation index calculated in the above step 110 are stored in the database in a state associated with the current design variable in the form of being substituted as the analysis result regarding the current new analysis shape ( Step 115). And the process after the said step 103 is repeatedly performed.

On the other hand, if it is determined in step 113 that the slope change rate | k _b −k _c | / Δx _i ≦ ζ _i is not established, the new analysis shape of this time is similar to the past analysis shape. However, since the sensitivity of the current design variable is high, it can be determined that the use of the estimated value is likely to increase the approximation error with the true evaluation index value. Therefore, in this case, the analysis calculation for the new analysis shape is executed as usual (step 116). In this step 116, the boundary condition and the calculation condition are set, and the flow velocity for each cell 42 is calculated for the new analysis shape after the morphing is executed.

  Next, the tumble ratio TRs and the flow coefficient CFs in a state where the tumble ratio TR is stable in the new analysis shape after execution of the morphing are acquired (step 117).

  Next, it is determined whether or not the tumble ratio TRs and the flow coefficient CFs acquired in step 117 are equal to or greater than the target tumble ratio TRm and the target flow coefficient CFm in step 101 (step 115).

  As a result, when both the tumble ratio TRs and the flow coefficient CFs have not reached the target values, the analysis result of the new analysis shape this time is stored in the database in a state associated with the current design variable (step 119), In addition, the processing after step 103 is repeatedly executed. On the other hand, when the tumble ratio TRs and the flow coefficient CFs both reach the target values, the processing of the design support program shown in FIG. 10 is ended.

  In the fluid analysis such as the flow analysis in the intake port of the present embodiment, even if the analysis parameters such as the design variables are similar, the analysis results are not necessarily close. This is because, when calculating vortex information such as the tumble ratio TR, the difference in shape (difference in design variables) greatly affects the analysis result due to the influence of fluid disturbance in the flow field. Therefore, in fluid analysis, when a new analysis calculation is performed, if the analysis parameters such as design variables are similar to the past analysis shape, the new analysis shape is simply excluded from the calculation target. There is a concern that the search accuracy of the optimum shape will deteriorate.

  On the other hand, according to the design support program shown in FIG. 10 described above, when searching for the optimum shape, the sensitivity of the design variable of the new analysis shape is determined together with the similarity determination comparing the new analysis shape with the past analysis shape. The design variable sensitivity is determined to determine whether or not to exclude the new analysis shape from the target of the analysis calculation. Only when the sensitivity of the design variable to the analysis result of each evaluation index is low and it is determined that it is safe to exclude the new analysis shape from the analysis calculation target, the new analysis shape is excluded from the analysis calculation target. In addition, the estimated value of each evaluation index simply obtained from the evaluation approximate expression is used as a substitute for the analysis result, so that the analysis calculation regarding a similar shape and a shape having almost the same result is not performed. For this reason, in the fluid analysis, useless calculation can be omitted without causing deterioration of the search accuracy of the optimum shape, and thus the search period can be favorably shortened.

  In addition, if many similar shapes with similar solutions are searched in the early stage of the optimization search, a situation in which the search concentrates on a local solution in the vicinity thereof, and as a result, a wide range of optimization searches cannot be performed. There is a problem. Further, if similarity determination is performed only with a shape change portion (the above-mentioned node movement target region) based on design variables (analysis parameters), similarity determination accuracy is achieved in several k dimensions of design variables, and sufficient similar shape comparison is performed. I can't do it.

On the other hand, according to the design support program described above, sample values are extracted not only from the target area of node movement by morphing but also from the interpolation area, and the vector values u _i and v _i are calculated, and the new analysis shape is obtained. And the past analysis shape are evaluated for similarity. As described above, in this method, since the interpolation region is also a sample node extraction target, the number of design variables is k.times.the number of sample nodes is n, and high-precision similarity determination can be realized. . In addition, since a wide-range search can be performed at the initial stage of the optimized search, searches for the same or similar shape can be reduced at the initial stage of the search, and it is possible to suppress falling into a local optimal solution at the initial stage of the search. In addition, as the optimization progresses, the probability of being excluded from calculation will increase, so that the final optimal solution can be obtained with only the solution that is truly desired, and the search performance can be improved. it can.

  By the way, in the above-described first embodiment, when determining the estimated value of each evaluation index of the new analysis shape in the sensitivity determination of the design variable, the nth-order evaluation approximate expression is obtained using the least square method. ing. However, such an evaluation approximate expression is not limited to the one obtained using the least square method, and may be obtained using another general approximation method such as use of Taylor expansion, for example. . In addition, when an optimization search is performed, a response surface is created and used for the search, and the estimated value may be obtained using the response surface.

In the first embodiment described above, the sensitivity of the design variable is determined using the change rates of the slopes k _b and k _c between the sample points near the estimated value of the new analysis shape. However, the sensitivity determination method in the present invention is not limited to the above method, and for example, using the error value itself (difference value) between the estimated value and the result of the neighborhood solution, the error value is a predetermined threshold value. The sensitivity of the design variable may be determined by comparison.

In the first embodiment described above, the mesh model 40 corresponds to the “structure model” in the first invention. Further, when the CPU 14 executes the process of the above step 102 or 116, the “cell information acquisition means” in the first invention executes the processes of the above steps 117 and 118. The "analysis execution means" executes the process of step 119, so that the "analysis history storage means" in the first invention executes the processes of steps 105 to 108, so that the "similarity" The “determination means” executes the processing of the above-described steps 109 to 113 so that the “parameter sensitivity determination means” in the first invention executes the processing of the above-described step 114 when the determination of the above-mentioned step 113 is established. Thus, the “calculation exclusion judging means” in the first invention is realized.
Further, the “cell deformation amount setting means” in the second invention is executed by the CPU 14 executing the process in step 103, and the “morphing executing means in the second invention is executed by executing the process in step 104. "Is realized. The vector values u _i and v _i correspond to the “node movement information in the node movement target area” and the “node movement information in the interpolation area” in the second invention, respectively.

It is a figure for demonstrating the hardware constitutions of the design support system in Embodiment 1 of this invention. It is a figure for demonstrating the software configuration of the design support system in Embodiment 1 of this invention. It is the figure which expressed the mesh model of the intake port simplified. It is a figure which shows an example of the shape change of the mesh model accompanying execution of morphing. It is the conceptual diagram showing a mode that the search by the optimization search part was performed. It is a figure for demonstrating the similarity determination method by the similarity determination part shown in FIG. FIG. 3 is a diagram for explaining a method for calculating an estimated value of each evaluation index for a new analysis shape in the design variable sensitivity determination unit shown in FIG. 2. It is the figure which showed the example in which an error exists between an approximated curve and a true evaluation index curve. It is a figure for demonstrating the sensitivity determination method of the new analysis shape by the design variable sensitivity determination part shown in FIG. It is a flowchart of the design support program performed in Embodiment 1 of this invention.

Explanation of symbols

10 Computer 12 Input Device 14 CPU
16 output device 18 storage device 20 base mesh model creation unit 22 condition setting unit 24 calculation solver 26 optimization search unit 28 morphing processing unit 30 similarity determination unit 32 design variable sensitivity determination unit 40 mesh model 42 cell CF, CFm, CFs flow coefficient CFf Estimated value of flow coefficient S Similarity TR, TRm, TRs Tumble ratio TRf Estimated value of tumble ratio u _i , v _i vector value X _i Shape deformation amount

Claims (2)

A design support device for obtaining an optimum shape of a structure that satisfies a target evaluation index by numerical fluid calculation,
Cell information acquisition means for acquiring a calculation result according to a predetermined analysis parameter and a predetermined rule for each of a plurality of cells constituting the structure model to be analyzed;
Analysis execution means for analyzing the evaluation index based on the calculation results for the plurality of cells;
Analysis history storage means for storing the analysis result of the evaluation index in association with the corresponding analysis parameter;
Similarity determination means for determining whether a new analysis parameter to be newly analyzed and a past analysis parameter that has already been analyzed are similar,
Whether the new analysis parameter is an analysis parameter having low sensitivity to the analysis result of the evaluation index when the similarity determination unit determines that the new analysis parameter is similar to the past analysis parameter Parameter sensitivity determination means for determining whether or not
Calculation exclusion determination means for excluding the new analysis parameter from the target of analysis calculation when the parameter sensitivity determination means determines that the new analysis parameter is an analysis parameter having low sensitivity to the analysis result of the evaluation index; ,
A design support apparatus comprising: Cell deformation amount setting means for setting the deformation amounts of a plurality of cells based on the analysis parameters;
Morphing execution means for performing node movement of the plurality of cells based on the deformation amount set by the cell deformation amount setting means,
The similarity determination means includes node movement information in the node movement target area by the morphing execution means, and node movement information in an interpolation area between the node movement target area and the node movement non-target area. The design test apparatus according to claim 1, wherein it is determined whether or not the new analysis parameter and the past analysis parameter are similar to each other.

Images