JP2602906B2 - Analysis model automatic generation method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Industrial Application Field] The present invention divides an object to be analyzed into shapes according to shape characteristics, and analyzes the object by combining a plurality of analysis programs corresponding to each of the divided partial shapes. The present invention relates to an analysis model automatic generation method in which an analysis model of an object is automatically generated with high generation efficiency.

[Prior Art] In numerical analysis for product design, a mathematical model of the shape of a design target is generally used. By the way, according to the latest CAE system, for example, the first edition of “Mold Technology-Plastic Injection Mold Design Data Book-” (Vol. 2, No. 11, October 20, 1987, published by Nikkan Kogyo Shimbun) As shown in Chapter 2 (Runners and Gates), pages 16-19,
Modeling of a product whose shape is defined (mesh automatic division when using the finite element method) can be automatically performed.

[Problems to be Solved by the Invention] However, since the conventional model generation method is unique to an analysis program using an analysis method such as a finite element method or a difference method, analysis is performed by combining a plurality of analysis programs. No consideration has been given to generating a model. One part of the analysis object is solved by the finite element method and the other part is solved by the finite element method, or the other part is analyzed by the analysis program (1) by the difference method, and the other part is analyzed by the difference program (2). ), The analysis model is not generated by combining a plurality of analysis programs, and the model is created exclusively by the analyst. For this reason, there are individual differences in the created analysis models depending on differences in model creation experience and knowledge of analysis programs, and the creation of such models generally requires a lot of time.

An object of the present invention is to provide an analysis model automatic generation method capable of automatically and quickly generating an analysis model of an entire analysis target as a combination of different shape feature correspondence analysis programs.

[Means for Solving the Problems] The object of the present invention is to provide an analysis program corresponding to a predetermined shape feature, and output data of one analysis program between the analysis programs in correspondence with the analysis programs. And an input / output data conversion program for converting the data into an object, and when the shape of the object to be analyzed is input, the object to be analyzed is divided into shapes by the predetermined shape feature, and each of the divided An analysis program corresponding to the partial shape is extracted and each of the partial shapes is modeled, while an input / output data conversion program corresponding to the analysis program having an adjacent relationship on the partial shape is extracted and the extracted analysis program is extracted. This is achieved by matching input / output data between analysis programs that are adjacent to each other.

[Operation] The point is that the correspondence of the analysis program to each of the shape features is obtained in advance based on the empirical knowledge of the analysis engineer, and when the shape of the object to be analyzed is input, the relationship is calculated based on the calculated shape features. After dividing the object to be analyzed,
An analysis program is made to correspond to each partial shape so as to partially model. If the input / output data is consistent between these partial models, the modeling as a whole is finished, but the positional relationship of the divided parts is separately determined, and hence the input / output data between the partial models is Depending on the matching performed by the corresponding input / output data conversion program, an analysis model as a whole for the analysis object can be automatically generated.

[Example] Hereinafter, the present invention will be described with reference to FIG. 1 to FIG.

First, the numerical analysis system according to the present invention will be described. FIG. 2 illustrates a configuration on hardware, and FIG. 3 illustrates a configuration on software. As shown in the figure, the hardware configuration is such that a central processing unit 13 and a disk device 18 are connected to a multi-bus 11 under the control of a bus control device 12 via a disk control device 14. The processing unit 13 also includes a main storage device 15, a display device 16 and a keyboard 17
Is accommodated. As a result, the data from the keyboard 17 is transmitted to the main memory 15 via the central processing unit 13.
At the same time, the data is displayed on the display device 16, and the data on the main storage device 15 is stored in the central processing unit 13,
Disk device via multi-bus 11 and disk controller 14
Data is arbitrarily transferred, for example, stored in 18.

Further, from the software configuration, the control unit 20, the command interpretation unit 21, the input unit 22, the output unit 23, the analysis execution unit 24, the analysis program registration unit 25, the shape registration unit 26, the shape feature registration unit 27,
Shape division rule registration unit 28, division partial positional relationship registration unit 29,
Partial shape registration unit 30, shape feature calculation unit 31, analysis program allocation unit 32, model creation unit 33, input / output data matching unit 34,
The configuration is made up of a shape division unit 35, a partial shape feature determination unit 36, an analysis model integration unit 39, a shape feature determination rule registration unit 37, a model registration unit 38, and the like.

FIGS. 4 (a) to 4 (f) show the contents of the main part registration sections shown in FIG. 3, and the outline of the present invention will be described below.

That is, the analysis program registration unit 25 (see FIG. 4 (f))
Include analysis programs P ₁ , P ₂ , P ₃ and modeling programs M ₁ , M
₂ and the data conversion programs H ₁ and H ₂ , and the shape feature registration unit 27 (see FIG. 4A) stores the shape features T _{1 to} T _n and their calculation formulas (dimensions of the analysis object (shape An operation formula) using the original parameter) as input data, a division rule for the shape feature is further stored in the shape division rule registration unit 28 (see FIG. 4 (b)), and a shape feature determination rule registration unit 37 (the fourth In FIG. 7 (c), a shape feature estimation rule as a determinant of an analysis program is registered in advance by a skilled model analyst.

Now, assuming that the shape of the analysis target A and the analysis item α are designated by the operator's command input from the keyboard 17, the shape of the analysis target A is stored in the shape registration unit 26 as a shape dimension value. First, it is registered. After that, the shape feature registration unit 31
Each feature value is calculated according to a calculation formula corresponding to the shape features T _{1 to} T _n indicated by 27, and the feature values are registered in the shape feature registration unit 27. Next, the shape a 'is stored in the partial shape registration unit 30 in accordance with the division rule registered in the shape division rule registration unit 28 in the form of IF to THEN and indicating how to divide the object shape into units that can be analyzed. , b ', c' and portions a _1, defined from, a ", b", moiety as defined from c "
Two-part name of A ₂ is adapted to be registered. At that time, it is determined whether there is a surface on which the portion A _1, A ₂ each contact is, the positional relationship between the portion A _1, A ₂ each when there is contact with the surface, i.e., in contact The registered surface information is registered in the divisional partial positional relationship registration unit 29. Thereafter, based on the rules registered in the shape feature determination rule registration unit 37 for each of the portions A ₁ and A ₂ , the shape feature serving as a reference for the analysis program is calculated by the partial shape feature determination unit 36, and the partial shape the shape feature field in the registration unit 30, for example, part A, wherein S ₁ is the _1, S _3, for the portion A ₂ are registered in and so on, wherein S _2,.

After that, the analysis item α input from the keyboard 17 and
As part A _1, A _2, wherein S ₁ for each, S _3, S ₂ and the search of the reference value, so that the matching analysis program name in the portion A _1, A ₂ from the analysis program registration unit 25 is searched It has become. As a result, the program P ₁ is stored in the part A ₁
A program P ₂ is assigned to A _2, and is registered in the matching program column in the partial shape registration unit 30. At that time, the analysis programs P ₁ and P ₂ have their own modeling methods M ₁ and M ₂ , respectively, which are also registered in the analysis program registration unit 25. It is registered together with the partial shape registration unit 30. In response, the model creation unit 33
In running the program M ₁ for the part A ₁ part model MO
D ₁ is partial model MOD ₂ also portions A ₂ and further similarly
Is created. (here,
These partial models MOD ₁ and MOD ₂ are groups of numerical values that are a part of the input data of the analysis programs P ₁ and P ₂ , respectively).

As described above, since the analysis program and the model for each part are determined, if the input / output data between the models is matched, the entire modeling is completed. In order to match input / output data, first, it is determined whether or not the portions A ₁ and A ₂ are in an adjacent state by referring to the divided portion positional relationship registration section 29. If part A ₁ , A ₂
Of when adjacent is confirmed, the program P ₁ by output data matching unit 34, P ₂ of input, the input data IN program P ₂ output data OUT1 program P ₁ from the output data
Although the program to be converted into 2 to be searched, since conversion program H ₁ is found as a result, during the analysis for the portion A _1, A _2, respectively, in which steps to perform the program H ₁ is determined.

According to the above procedure, an analysis model based on a combination of different analysis programs can be automatically generated.

The outline of the present invention has been described above, and FIG. 1 shows an outline flow as an example of the processing procedure. The case where this is applied to the design of a mold for a semiconductor plastic package will be described below as an example.

That is, first, the background will be described. A semiconductor plastic package is placed in a mold cavity.
Place the semiconductor chip with the lead frame attached,
It is made by injecting and curing a resin material as a plastic sealing material from a gate (inflow port) of the cavity. At this time, by performing a flow analysis of how the viscous resin material flows in the mold cavity, it is possible to evaluate the mold specifications to be designed. However, this resin material has a thermosetting property in which heat is absorbed from the mold wall surface during the flow and its viscosity is greatly changed, and its characteristics cannot be explained by a clear theoretical formula. Experimentally, a flow analysis model has been developed for a simple mold flow path shape that can measure viscosity, but the actual mold cavity structure is complicated, and its complexity is low. The flow analysis model considered cannot be created by only one kind of calculation method. Therefore, by combining a plurality of analysis programs (models), it is necessary to create an analysis model that takes into account the complexity of the flow path shape.

By the way, when the analyst generates a model to perform the flow analysis in the mold cavity, prior to the analysis, the analysis program registration unit 25 stored in the disk device 18 includes:
Resin flow analysis programs P ₁ , P ₂ , P ₃ , these resin flow analysis programs P ₁ , P ₂ , P ₃ unique model creation method programs M ₁ , M ₂ , M ₃ and input / output data conversion program H ₍₁₂₎ , H
_{(23), H (31)} , H (21), H (32), H (13) (H (ij): Resin flow analysis Puraguramu P _i resin flow analysis program output data
A program for converting the input data into _Pj input data) is registered in advance. Here, it will be described referring to FIG. 5 the contents of the resin flow analysis program P ₁ to P ₃ and modeling techniques program M ₁ ~M _3, resin flow analysis program
P ₁ is referred to as circular pipe flow is for analyzing a flow streamline parallel receiving evenly heat from 360 °, and can apply to the flow portion having a heat source in the circumferential direction . The program P ₁ is because of the use of the finite difference method, the model converts the shape (shape of the portion where the resin flows) of interest in a circular tube is equal cross sectional flow area that is created from Modeling program M ₁ Made Resin flow analysis program
P ₂ is for analyzing the flow under conditions called flat flow, which receive heat only in the vertical direction and negligible heat transfer from the horizontal direction (flat flow with a small height). , because it is created using a finite element method, there is no need to specify particular the direction of flow line (direction of flow) is automatically analyzed Unlike difference method, the modeling method program M ₂ for this The model of the flowing part shape is created by dividing it into triangular meshes. The resin flow analysis program P ₃ has a what is flowing from the outlet of one point may be used under conditions which proceeds on the radiation, its Modeling program M ₃ are adapted by difference method.

Also, a shape feature registration unit 27 and a shape division rule registration unit
In the item 28, items for calculating the characteristics of the shape of the mold cavity 41 shown in FIG. 6 (and the calculation formula) and a method of dividing the cavity shape corresponding to the characteristics are registered. Here, the items for calculating the cavity shape characteristics and the content of the shape division method will be described. The main factors that govern the resin curing reaction are heat transfer from the mold wall surface and the internal structure (chip). Action. Therefore, it is necessary to divide the shape at the point where the heat transfer state inside the mold cavity 41 is different, that is, at the point where the resin flow is separated and flowed by the internal structure (chip 42) to obtain a portion where the heat transfer state is constant. is there. The basis of the shape division is the shape and position of the chip 42 as an internal structure. Where cavity 41
Considering the ratio of the width g of the chip 42 to the width h of the chip 42, when the width h of the chip 41 is substantially the same as the width g of the cavity 41, it can be considered that the flow is divided into the upper and lower portions of the chip 41, and conversely. Chips
When the width h of the tip 41 is smaller than the width g of the cavity 41 by a certain ratio α or more, it is considered that the flow is divided into four parts, that is, the top, bottom, left and right. Therefore, the tip 42 is stored in the shape feature registration unit 27.
The ratio α of the width h of the cavity 41 to the width g of the cavity 41 is α = (= h / g), and the shape division rule registration unit 28 determines whether the division is divided into the upper and lower parts of the chip 42 depending on the value of the ratio α. 42 If the description of whether to divide into four, upper, lower, left and right, is registered in the IF-THEN- format, the cavity shape can be divided according to the shape registration. In addition, as a shape feature serving as a reference for division, a ratio of the length b of the chip 42 to the length 1 of the cavity 41 is given.

Other characteristic item as a reference of the division of the cavity shaped in the shape feature registration unit 27, has been registered together also feature plane serving as a reference to assign the analysis program in divided portions, analysis program P _1, Items that determine P ₂ and P ₃ include streamline direction (determining factor of program P ₁ and P ₃ ) and shape (fluid height that determines program P ₂ , heat source determining program P ₁ ) Uniformity, that is, the aspect ratio of the shape). Among these, the features that serve as criteria for determining the direction of the streamline include the width g of the cavity 41 and the width (diameter) i of the gate 43.
Ratio is considered. When the ratio g / i of the width g of the cavity 41 to the width i of the gate 43 is large, the flow can be regarded as a transition from a narrow portion to a wide portion, and the streamline can be determined to be radial. is there. Conversely, when g / i is small, it can be determined that the streamlines are parallel. As described above, the feature items and their formulas serving as criteria for determining the analysis program are registered in the shape feature registration unit 27, and the feature serving as the reference for determining the analysis program is stored in the shape feature determination rule registration unit 37. The feature criterion (α, β, etc.) for the calculation result is IF
~ THEN ~ format.

Now, the processing procedure shown in FIG. 1 will be described. If the analyst inputs “flow analysis” as an analysis item from the keyboard 17 together with the mold cavity name X to be analyzed, the input analysis request is input to the input unit. The data is taken into the system by 22 and passed to the command interpreter 21 (process 100). The command interpreter 21 analyzes the analysis request, and the mold cavity name X as an analysis result is transmitted to the controller 20 so that the controller 20 can control the “X” in the disk device 18 via the shape registration unit 26. Of the mold cavity shape are searched. In this example, since the mold cavity shape data for “X” is not registered, the input of the shape data is requested via the output unit 23, and based on this, the analyst inputs the mold data for “X”. If you enter the cavity shape data, the shape data will be
Starting from 20, the information is registered in the disk device 18 via the shape registration unit 26 (process 110).

Thereafter, the control unit 20 instructs the shape feature calculation unit 31 to calculate the feature of the input shape. Based on this instruction, the shape feature calculation unit 31 obtains the shape feature item and its calculation formula from the disk device 18 via the shape feature registration unit 27, and further obtains the shape specifications for “X” from the disk device 18 via the shape registration unit 26. After the reading, the feature value corresponding to each feature item is calculated, and the calculation result is registered in the shape feature registration unit 27 (process 120). After the features are calculated in this manner, the shape dividing unit 35, based on the mold cavity shape dividing instruction from the control unit 20, performs the process of registering each feature value registered in the shape feature registering unit 27 with the shape dividing rule registering unit 28. Is read, and how to divide the shape of the mold cavity is determined. As a result, for example, as shown in FIG. 7, the mold cavity is divided into parts A to D (step 130). Further, based on this division result, the shape division unit 3
In 5, the shape data of each of the divided parts A to D is calculated from the mold cavity shape data registered in the shape registration unit 26, and the part name and its shape data are sent to the partial shape registration unit 30. (Process 140). In response to this, the partial shape registration unit 30 registers the partial shape specifications with the name of the part in the disk device 18. The shape division unit 35 further calculates the positional relationship between the divided parts A to D and registers the positional relationship in the divided part positional relationship registration unit 29 (process 150). As shown in FIG. 7 as an example,
The positional relationship between the divided parts A to D is registered assuming that the divided parts B and C are located next to the divided part A in parallel, and the divided part D is located next to the divided parts B and C. Things.

Subsequent to the shape division, the control unit 20 instructs the shape feature calculation unit 31 to calculate a feature serving as a reference for analysis program allocation for each of the divided parts A to D. Based on this, the shape feature calculation unit 31 The partial shapes of the divided parts A to D registered in the registration unit 30 and the shape feature registration unit 27
After reading the feature items and their calculation formulas registered in, calculating the feature values according to the calculation formulas, the shape feature registration unit
27 is registered (process 160). Subsequently, the control unit 20 sends the divided parts A to D to the partial shape feature determination unit 36.
In this case, an instruction is given to determine a characteristic in the analysis program allocation. The partial shape feature determination unit 36 uses the shape feature determination rule registration unit 37 to estimate a factor for determining the analysis program from the feature value, and to define a rule in the IF-THEN-form. ＤD are read, and based on these, the factors serving as references in the analysis program allocation of the divided parts A〜D are estimated, and the estimated result is registered in the partial shape registration unit 30. (Process 170). For example, for the divided portion A, the flow line is “radial” and the flow height is “high” based on the ratio of the width of the entrance gate to the flow width. For the divided portions B and C, the flow height is “low”. Further, the divided portion D is determined as a streamline “parallel” and a flow height “high”.

When the feature determination for all the divided parts A to D is completed as described above (step 180), the control unit 20 allocates the analysis program to the analysis program allocating unit 32 for each of the divided parts A to D. Is to be instructed. In the analysis program allocating unit 32, the partial shape registration unit
The features of the divided parts A to D registered in 30 are read,
Among the flow analysis programs registered in the analysis program registration unit 25, those whose characteristics of each of the divided parts A to D match the program use conditions, and a model creation program name unique to the flow analysis program are selected. (Step 190). As a result, the analysis program P ₃ and the model creation program M ₃
Divided portion B, and also analysis program P ₂ and model creation program M ₂ is and C, even more divided portion D in which the analysis program P ₁ and the model creation program M ₁ are allocated, respectively. Subsequently, the control unit 20 instructs the model creation unit 33 to model each of the divided parts A to D.
At 33, the divided parts A to
The model creation program of D and its shape data are read, and a model is created for each of the divided parts A to D (process 200). The created model is stored in the model registration section 3.
8, the part name is registered in the disk device 18 with the part name as a heading.

When the model creation for all the divided parts A to D is completed (processing 210), the control unit 20 sends the input / output data matching unit 34 the data required for continuously analyzing the divided parts A to D. Output data matching is instructed. The input / output data matching unit 34 reads the positional relationship of the divided parts A to D from the divided part positional relationship registration unit 29, and further reads the analysis program name corresponding to the divided parts A to D from the partial shape registration unit 30, and inputs the analysis data. A continuous section of the analysis program that requires output data matching, that is, A → B, A
A data conversion program for matching input / output data of → C, B → D and C → D is searched by the analysis program registration unit 25. As a result, the data conversion program H ₃₂ to A → B is likewise data conversion program H ₃₂ The A → C is B → D, since the data conversion program H ₂₁ corresponds respectively to the C → D, these Each conversion program name is registered in the divided part positional relationship registration unit 29 (process 220). After that, the control unit 20 integrates the analysis program and the partial model corresponding to each of the divided parts A to D, the input / output data conversion program between the analysis programs, and the one model as a whole. Instructions to be created are provided. In the analysis model integration unit 39, the analysis program names corresponding to the divided parts A to D are transmitted from the partial shape registration unit 29 to the model registration unit.
Modeling results corresponding to the divided parts A to D are read from 38, and data conversion program names corresponding to adjacent parts are read from the divided part positional relationship registration unit 29 and edited as an analysis execution procedure. I have. Based on this analysis execution procedure, the analysis execution unit 24 automatically generates a flow analysis model as a whole in the mold cavity in a form in which the analysis programs P ₁ , P ₂ , and P ₃ are mixed. Things.

As described above, according to the present embodiment, a plurality of existing analysis programs are combined when performing an analysis for which an analysis method has not been established, such as a flow analysis in a mold cavity associated with semiconductor plastic package design. As a result, an analysis model enabling the flow analysis as a whole is automatically generated.

[Effects of the Invention] As described above, according to the present invention, an analysis model as a whole is automatically generated by combining a plurality of analysis programs according to the shape of an analysis target. Therefore, the analysis system is not limited to a specific analysis method.If an analysis program that can analyze a specific portion of the analysis target with higher accuracy or an entirely new analysis method is developed, the analysis system will be significantly reduced. These programs and analysis methods can be reflected in the analysis without making any changes.

In addition, by combining multiple analysis programs without requiring specialized knowledge of existing analysis programs, there is a variety of analysis objects with complicated shapes, objects with undefined analysis methods, and shapes. The analysis can be performed while flexibly dealing with the analysis target.

In addition to this, not only is there no individual difference in the analysis result, but also the effect of greatly reducing the time required for creating the analysis model is obtained.

[Brief description of the drawings]

FIG. 1 is a flowchart showing an outline of a processing procedure in an example of a numerical analysis system according to the present invention, and FIGS. 2 and 3 show a hardware configuration and a software configuration in an example of the numerical analysis system, respectively. FIG. 4 (a)-(f)
Is a diagram showing the contents of each table of the main part registration unit shown in FIG. 3, and FIGS. 5, 6, and 7 illustrate an analysis model generation process in the case of taking a semiconductor plastic package as an example. FIG. 20: Control unit, 21: Command interpreting unit, 22: Input unit,
23 output section, 24 analysis execution section, 25 analysis program registration section, 26 shape registration section, 27 shape feature registration section,
28 ... shape division rule registration unit, 29 ... division partial positional relationship registration unit, 30 ... partial shape registration unit, 31 ... shape feature calculation unit, 32 ... analysis program allocation unit, 33 ... model creation unit, 34: Input / output data matching unit, 35: Shape division unit, 36
... A partial shape feature determination unit, 37... A shape feature determination rule registration unit, 38.

Claims (1)

(57) [Claims] 1. A method for automatically generating an analysis model by numerical analysis, comprising: an analysis program corresponding to a predetermined shape feature;
In correspondence with the analysis programs, an input / output data conversion program for converting output data of one analysis program into input data of the other analysis program between the analysis programs is stored in advance, and the shape of the object to be analyzed is stored. When input, the object to be analyzed is divided into shapes by the predetermined shape feature, and an analysis program corresponding to each of the divided partial shapes is extracted to model each of the partial shapes. An analysis characterized by extracting an input / output data conversion program corresponding to an analysis program having an adjacent relationship and matching input / output data between analysis programs having an adjacent relationship among the extracted analysis programs. Automatic model generation method.