JP2014241065A - Component selection method, program and system - Google Patents

Abstract

Provided is a component selection method that creates a simple plant model from a prototype, predicts the effect when a component that is a material candidate is mounted, and enables an appropriate T-BOM selection with a small number of component replacements. [MEANS FOR SOLVING PROBLEMS] A variation of parameters is tried for one type of material (402), a candidate for a component having the maximum and minimum properties is selected for each material, a simple plant model is estimated (404), and selected at the time of design. The component combination is tested in the actual test (406), and it is determined whether the result of the actual test and the design target are almost coincident (408). If the difference is within a predetermined difference range, the process is terminated. Select the material that is involved in the direction with a larger absolute value (410), and select the candidate part that has the closest sum of the coefficient norms interpolated at each candidate position to the sum of the coefficient norms of the default model. Then, it is tried in a real test (412). [Selection] Figure 4

Description

  The present invention relates to a part selection technique for optimizing performance in a machine product including a large number of parts such as an automobile.

  A collection of parts for manufacturing a machine product is called a BOM (Bills of materials). The BOM includes a design BOM (E-BOM) determined from design information, a technical BOM (T-BOM) consisting of multiple candidates with variations in part parameters, and manufacturing that is a group of parts that also considers the manufacturing process. BOM (M-BOM) is known.

  In designing automobiles (many other machine products), before making a prototype of a real machine (prototype), a controller (controller) and a plant (control target) are modeled on a computer and (virtual) simulations are performed. Based on the design data, parameters of each part (corresponding to E-BOM) are determined so that a desired design goal can be achieved. For example, the spring constant and oil viscosity for the plant, and the PID coefficients for the controller.

  However, the simulation of the model cannot always reproduce the phenomenon defined by the absolute value of the actual machine. For example, the absolute value of the spring hardness cannot be accurately determined. In other words, it cannot be solved as an optimization problem on the model.

  Therefore, in the actual design and development, the component parameter variations are prepared as T-BOMs (material candidates) within a specified range (multiple), and the combinations can be actually tested to create the expected phenomenon. Will be tried. The problem is that this combination trial is very expensive.

  On the other hand, in the simulation of the model, for example, it is possible to show the tendency of how hard the spring is or how soft it is, so the absolute value does not necessarily match, but the relative value can show the tendency It is desirable to use this method.

通りの組み合わせの中から選択する作業が必要になり、多くの場合、それは組み合わせ的爆発を引き起こすので、実行困難である。 In order to actually manufacture the final product, it is necessary to determine a manufacturing part group (M-BOM) defined by a unique combination from T-BOMs that have multiple part parameter variations for each part. However, if there are N types of materials (1 ≦ k ≦ N) in T-BOM, and each of them is M _k part candidates, to determine M-BOM,

The task of selecting from a combination of streets is required, and in many cases it is difficult to implement because it causes a combinatorial explosion.

In addition, the following is known as a prior art regarding component selection.
First, Japanese Patent Laid-Open No. 2007-140678 is intended to provide a general-purpose component selection support device that makes it easy to select a general-purpose component, shortens the product development period, and reduces the cost of the product. For each bolt that is a general-purpose part, information for identifying the bolt and shape information of the bolt are stored in association with each other, and when a command for displaying a list of bolts is generated, the information and the shape information are stored. It is disclosed that display data to be displayed in a list is created, and a list of bolt information and a list of shape information of each bolt are displayed in one screen.

  Japanese Patent Application Laid-Open No. 2008-2907 discloses a predetermined part information acquisition unit that acquires information about a predetermined part included in a vehicle, an alternative part information generation unit that generates information about an alternative part that can be replaced with the predetermined part, and input information Vehicle performance calculation means for calculating vehicle performance based on the current vehicle performance calculated by the vehicle performance calculation means when the information regarding the predetermined part acquired by the predetermined part information acquisition means is input, When the information on the substitute part created by the substitute part information creating unit is input instead of at least a part of the information on the prescribed part acquired by the predetermined part information obtaining unit, the post-replacement calculated by the vehicle performance calculating unit Disclosing providing information to a user based on a comparison with vehicle performance.

  However, these conventional technologies do not solve the problem of the large number of component combinations when determining the M-BOM from the T-BOM.

JP 2007-140678 A JP 2008-2907 A

  An object of the present invention is to create a simple plant model from a prototype, predict the influence when a component that is a material candidate is mounted, and enable appropriate T-BOM selection with a small number of component replacements.

In order to solve the above problems, the present invention executes the following steps.
(1) In T-BOM, when a simple inspection is performed for the parameter variation width of one type of part (hereinafter also referred to as material) including a plurality of part candidates, the E-BOM design is performed with the other parts remaining as initial parts. A simple simulation is performed based on the model of the time and past design data, and the material part candidates are selected so as to guarantee the performance limit. This is inspected for all materials. However, this inspection is omitted when the range of parameters to be given as T-BOM is clear from past design data.
(2) For each material, select the candidate parts that show the maximum and minimum properties, and for other materials, estimate the simple plant model with the initial parts, and express the sum of the coefficient norms as N Plot on + 1D space. The (N + 1) th dimension is the numerical value of the sum of the coefficient norms. For this reason, 2N simulation results and the results of the simple plant model in the initial setting state are used. Then, the performance prediction hypersurface is constructed by differentiable interpolation.
(3) Test the combination of parts selected at the time of design in the actual test.
(4) For the observation amount selected to show the performance of the entire system, the observation output result of the actual test and the design target are almost the same.
(5) If there is a deviation, the degree of influence on performance at the mounting point, for example, each material axis direction (two directions of Max and Min) and two material axes, preferably crossing 45 degrees (Max. , Min (combined 4 directions)) is calculated on the performance prediction hypersurface, and the material related to the direction with the largest absolute value of influence is selected (however, the direction already selected is excluded).
(6) From the selected materials, select the candidate part whose sum of the coefficient norms interpolated at each candidate position is closest to the sum of the coefficient norms of the initial setting model, and test it with an actual test.
(7) Return to step (4).

  According to the present invention, it is possible to greatly reduce the cost required for increasing the number of conventional combination trials by calculating the influence of parameter change with the gradient of the hypersurface and selecting the material to be changed. It becomes.

It is a schematic diagram which shows the process of the premise of this invention. It is a block diagram of the hardware of the computer for implementing this invention. It is a block diagram for implementing this invention of the functional element for implementing this invention. It is a figure which shows the flowchart of the process of this invention. It is a figure which shows the hyperplane for material selection. It is a figure which shows the direction of the gradient for material selection.

  Embodiments of the present invention will be described below with reference to the drawings. It should be understood that these examples are for the purpose of illustrating preferred embodiments of the invention and are not intended to limit the scope of the invention to what is shown here. Further, throughout the following drawings, the same reference numerals denote the same objects unless otherwise specified.

  First, the premise of the processing of the present invention will be described with reference to the schematic diagram of FIG. In FIG. 1, an initial prototype 102 is manufactured based on past design data, and constitutes a stable system as a whole.

  The simple plant 104 is obtained from a measurement result of the initial prototype 102 and a typical physical law expressing the plant operation. As described above, since the initial prototype 102 constitutes a stable system as a whole, the effects of component adjustment can be predicted to some extent from the behavior of the simple plant 104 obtained therefrom.

  The entire system adjusted based on the information here with respect to the simple plant 104 thus obtained is configured with the plant 106 and the controller 108. The configured entire system is operated, and the output value selected as an observation target obtained as a result is compared with the design value. If there is a deviation, the component in the plant 106 is replaced by the mounting / adjustment step 120 in the entire system. Exchange within the T-BOM range and try to bring the output value from the entire system closer to the design value.

  Next, the configuration of computer hardware for realizing the processing of the present invention will be described with reference to the block diagram of FIG. In FIG. 2, a CPU 204, a main memory (RAM) 206, a hard disk drive (HDD) 208, a keyboard 210, a mouse 212, and a display 214 are connected to the system bus 202. The CPU 204 is preferably based on a 32-bit or 64-bit architecture, such as Intel Core (TM) i3, Core (TM) i5, Core (TM) i7, Xeon (R), AMD Athlon ™, Phenom ™, Sempron ™, etc. can be used. The main memory 206 preferably has a capacity of 8 GB or more, more preferably 16 GB or more.

  The hard disk drive 208 stores an operating system. The operating system may be any suitable for the CPU 204, such as Linux (trademark), Microsoft Windows (trademark) 7, Windows (trademark) 8.

  The keyboard 210 and the mouse 212 are used to operate graphic objects such as icons, task bars, and text boxes displayed on the display 214 in accordance with a graphic user interface provided by the operating system.

  Further, a known interface board 216 such as PCI or USB is connected to the system bus 202, and the whole system 218 is further connected to the interface board 216. The interface board 216 is used to input control data such as a coefficient norm from the entire system 218.

The hard disk drive 208 further includes a main program 302, T-BOM data 304, a nominal model setting routine 306, a material reliability check routine 308, a coefficient norm calculation routine 310, which will be described later with reference to FIG. A material selection routine 312 and a material display routine 314 are stored.
And #d
Main program 302, nominal model setting routine 306, material reliability check routine 308, coefficient norm calculation routine 310, material selection routine 312 and material display routine 314 are C, C ++, C #, Java (R), etc. Can be written in any existing programming language.

  Next, the functional configuration of the present invention will be described with reference to FIG. In FIG. 3, a main program 302 starts or stops processing and performs other controls in response to operator operations using the keyboard 210, mouse 212, and the like.

  The T-BOM 304 stores a plurality of variations for each material that can be used.

  The nominal model setting routine 306 is a routine for setting a nominal model that is a model of the initial prototype 102. Set the observation input and output, and express the initial prototype with a state space model from system identification or past design data.

  The material reliability check routine 308 calculates the maximum (Max) and minimum (Min) values of the candidate part specifications (parameters) of each material in the T-BOM given multiple candidates for each material. Select the result of simple plant model estimation in each case or information based on past data used at the time of E-BOM design, and set the coefficient fluctuation in the state space equation for the nominal model. The performance limit value at a given reliability is obtained from the distribution using a probability algorithm. The probability measure that expresses the shake here may be a uniform distribution centered on the nominal model, with the maximum and minimum parameter variations of the material at both ends, and the amplitude characteristics are known from past experimental data. In such a case, the characteristic distribution may be used. Thus, for example, G.Calafiore, F.Dabbene, R.Tempo, “A survey of randomized algorithms for control synthesis and performance verification,” Journal of Complexity, Vol.23, pp.301-316, 2007. By assigning a probability level and reliability to the evaluation function, and selecting a material variation according to the probability level, a desired reliability can be given. This initial check is performed until the component group falls within the allowable performance range. If the guarantee cannot be made here, a component with too high a component sensitivity is a candidate component, and the selection needs to be reexamined. However, when comparing the difference with the nominal model, if the observed quantities are compared directly, it may be difficult to compare the distribution depending on the rise characteristics due to the component specifications, so mapping as the sum of the coefficient norms of the state space equation We will make a comparison standard.

係数ノルムの和は、

で与えられる。ここで、Aは行列であり、BとCはベクトルである。ベクトルのノルムの定義は自明であるが、行列のノルムには、誘導されたノルム、フロベニウスノルム、トレースノルムなどがあり、どれを用いてもいいが、ここでは、AとAの転置行列を掛けた行列の固有値のうちの最大値の平方根で定義される行列のノルムを用いる。 The coefficient norm calculation routine 310 calculates the sum of coefficient norms as follows. That is, y (t) is the observation output, u (t) is the observation input, x (t) is the internal parameter, and the state space model is expressed by the following equation:

The sum of the coefficient norms is

Given in. Here, A is a matrix, and B and C are vectors. The definition of the vector norm is self-explanatory, but the matrix norm includes the derived norm, the Frobenius norm, and the trace norm, which can be used, but here we multiply the transposed matrix of A and A. The norm of the matrix defined by the square root of the maximum value among the eigenvalues of the matrix is used.

  The material selection routine 312 calculates the nominal value, the point where one of the nominal materials is replaced with the maximum parameter, and the point where the parameter is replaced with the minimum value. A hyperplane is constructed by spline interpolation of the sum of the coefficient norms when expressed as, and the process of selecting the material to be selected is performed using the gradient there. The more detailed processing is shown in FIG. This will be described with reference to a flowchart.

  The material display routine 314 displays the material selected by the material selection routine 312 on the display 214 with reference to the information of the T-BOM 304. The operator looks at the material displayed on the display 214, replaces the material in the entire system 218 with the displayed material, and performs the next test.

  The parameter group 316 requires a material selection routine 312 such as nominal material data, maximum and minimum materials in each material, and the value of the sum of the coefficient norms of the state space model derived from each simple plant model. Data to be included. The data of the parameter group 316 is preferably stored in the hard disk drive 208.

  Next, the operation of the logical configuration shown in the functional block diagram of FIG. 3, mainly the processing of the material selection routine 312, will be described using the flowchart of FIG. 4.

Given the following description, plants Assumed N (N is a predetermined natural number) a Material, Materials of i-th (1 ≦ i ≦ N) is to consist of M _i number of components candidate.

Here, {M _i } = {1,2, ..., M _i } is defined. And the set of parts in the nominal model,
(m ₁ , m ₂ , ..., m _N ), m _i ∈ {M _i } for 1 ≦ i ≦ N. A set of parts in such a nominal model has a reliability set using a probability algorithm by a material reliability check routine 308.

In step 402, the simulation is performed for the M _i part candidates of the i-th material from i = 1 to i = N with the nominal model other than the i-th part remaining, and only the part candidates that guarantee the performance limit are left. , Reorder it, and let {M ' _i } = {1,2, ..., M' _i }. {M ' _i } ⊂ {M _i }. At this time, in {M ′ _i }, the index of the component candidate with the maximum parameter is m ^max _i , and the index of the component candidate with the minimum parameter is m ^min _i . M ′ _i obtained in this way, the index of the component candidate with the largest parameter is m ^max _i , the index of the component candidate with the smallest parameter is m ^min _i , and the value such as the sum of the coefficient norms at that time is The parameters are saved as a parameter group 316 by the main program 302.

As a result, the following 2N N + 1-dimensional points are obtained. In the following, Z ₁ , Z ₂ ,..., Z _{2N + 1} are values of the sum of the corresponding coefficient norms.
(m ^max ₁ , m ₂ , ..., m _N , Z ₁ ), m ^max ₁ ∈ {M ' ₁ }
(m ^min ₁ , m ₂ , ..., m _N , Z ₂ ), m ^min ₁ ∈ {M ' ₁ }
.....
(m ₁ , m ₂ , ..., m ^max _1N , Z _2N-1 ), m ^max _N ∈ {M ' _N }
(m ₁ , m ₂ , ..., m ^min _1N , Z _2N ), m ^min _N ∈ {M ' _N }

In step 404, the material selection routine 312 plots the 2N points in the N + 1 dimension, and for 1 ≦ i ≦ N,
(m ₁ , ..., m _i , ..., m _N , Z ₀ )
(m ₁ , ..., m ^max _i , ..., m _N , Z _2i-1 )
(m ₁ , ..., m ^min _i , ..., m _N , Z _2i )
These three points are interpolated with a cubic function spline. Here, Z ₀ is the nominal height. An example of the hypersurface formed in this way is indicated by reference numeral 502 in FIG. FIG. 5 shows a graph in which the sum of the coefficient norms is made high in coordinates obtained by normalizing the parameters of the material 1 and the material 2 with nominal values.

  In step 406, the nominal part combination selected by the operator at the time of design is tested in an actual test.

  Based on the result, the material selection routine 312 compares the actual test observation output result with the design target for the actual test result, that is, the observation amount selected as indicating the performance of the entire system in Step 408, If it is almost the same, that is, if the difference is within the predetermined difference range, the process ends. Otherwise, the material selection routine 312 calculates in step 410 the gradients in each direction of each material axis and six directions in the direction crossing the two material axes at 45 degrees on the hypersurface, and the gradient absolute Select the material involved in the direction of higher values. FIG. 6 shows six such directions.

  Since there are N types of materials, the above-mentioned gradients are obtained in 2N in the axial direction and N (N-1) / 2 in the cross direction. If the maximum gradient appears in any of these axial directions, the material in that direction is selected. If the maximum gradient appears in the cross direction, the two materials corresponding to that direction are simultaneously selected for adjustment.

  Referring to FIG. 5 as an example, assuming that the mounting point 504 is (0, 0) as a reference, the gradient at the point 506 where the coefficient norm direction of this point and the curved surface intersect is considered. In FIG. 5, if the gradient in the region 508 which is the horizontal axis minus axis is the largest, the material corresponding to the horizontal axis is selected.

  Thus, for example, once the material 1 is selected as a candidate, the material selection routine 312 cuts out the direction of the material 1 in step 412, and selects each of the material 1 component candidates a, b, c, d,. A component candidate whose sum of the coefficient norms interpolated at the candidate position is closest to the nominal is selected and displayed on the display by the material display routine 314. The operator who sees it on the display 214 equips it with the whole system and tries it in the actual test, and returns to Step 408 to compare the result of the actual test with the design target.

  In the above-described embodiment, the gradient in the six directions in the direction crossing the two material axes by 45 degrees is calculated on the hypersurface. However, the crossing angle is not limited to 45 degrees, and there is a degree of freedom, for example, 30 You may make it look in the direction which crosses the direction of 60 degree | times and 60 degree | times. Other angles may be used depending on the application.

  As the interpolation method for forming the hypersurface, any interpolation method other than spline interpolation can be used if a differentiable curve is generated. In spline interpolation, the cubic spline method is generally used. Other B splines, non-uniform rational B splines (NURBS), etc. can also be used.

  Further, the above embodiment relates to the selection of parts for automobiles. However, the scope of application of the present invention is not limited to automobiles, and it has a stable nominal model and can be used for designing arbitrary mechanical products composed of a large number of parts. Applicable.

218 ... Overall system 304 ... T-BOM
306 ... Nominal model setting routine 308 ... Material reliability check routine 310 ... Coefficient norm calculation routine 312 ... Material selection routine

Claims (12)

This is a method that uses a computer to determine the desired combination of component variations to achieve the design target for a product that consists of multiple components, each of which can have multiple variations, including the nominal variations that make up the initial prototype. And
For the plurality of parts, on the basis of a simple model based on input / output measurement of the control target portion in the initial prototype and past design data, selecting a variation that meets a performance limit;
The number of parts including variations is assumed to be N. For each part, the variation showing the maximum property and the variation showing the minimum property are selected, and the simulation using the above simple model is performed for the 2N types, and the above-mentioned nominal variation is used. In conjunction with the case, plotting the coefficient norm of the state space model derived from the simplified model in an N + 1 dimensional space;
Configuring a hypersurface by an interpolation method based on the plotted points so that the derivative is not discontinuous;
Obtaining a test result on an actual device including a controller for a combination corresponding to the initial prototype; and
It is determined whether the difference between the result of the test in the real device and the design target is within a predetermined threshold value, and if not, each of the points corresponding to the parts constituting the real device on the hypersurface For the axis of the part variation, calculate the gradient in the direction affected by the combination of the variation axis direction and the other variation axis, and select the part variation related to the direction with the largest gradient absolute value. Presenting a variation of a part to perform a test in an actual device, and repeating the test until the difference between the test result in the actual device and the design target is within a predetermined threshold,
Method.   The method according to claim 1, wherein the gradient in the direction affected by the combination with the other variation axis is a gradient crossing 45 degrees with respect to the axial direction of the two variations.   The method according to claim 1, wherein the interpolation method in which the differentiation does not become discontinuous is a spline interpolation method.   2. The maximum property and the minimum property are directly the magnitudes of parameters of each component, and indirectly are properties derived when the corresponding parameter is used in the simplified model. the method of. It is a program that uses a computer to determine the desired combination of component variations to achieve the design goal for a product that consists of multiple components, and each component can have multiple variations, including the nominal variations that make up the initial prototype. And
In the computer,
For the plurality of parts, on the basis of a simple model based on input / output measurement of the control target portion in the initial prototype and past design data, selecting a variation that meets a performance limit;
The number of parts including variations is assumed to be N. For each part, the variation showing the maximum property and the variation showing the minimum property are selected, and the simulation using the above simple model is performed for the 2N types, and the above-mentioned nominal variation is used. In conjunction with the case, plotting the coefficient norm of the state space model derived from the simplified model in an N + 1 dimensional space;
Configuring a hypersurface by an interpolation method based on the plotted points so that the derivative is not discontinuous;
Obtaining a test result on an actual device including a controller for a combination corresponding to the initial prototype; and
It is determined whether the difference between the result of the test in the real device and the design target is within a predetermined threshold value, and if not, each of the points corresponding to the parts constituting the real device on the hypersurface For the axis of the part variation, calculate the gradient in the direction affected by the combination of the variation axis direction and the other variation axis, and select the part variation related to the direction with the largest gradient absolute value. Presenting variations of parts to perform a test in an actual device, and executing a step of repeating until a difference between the result of the test in the actual device and the design target is within a predetermined threshold,
program.   The program according to claim 5, wherein the gradient in the direction affected by the combination with the other variation axis is a gradient that crosses 45 degrees with respect to the axial direction of the two variations.   The program according to claim 5, wherein the interpolation method in which the differentiation does not become discontinuous is a spline interpolation method.   6. The maximum property and the minimum property are directly the size of a parameter of each part, and indirectly, a property derived when the corresponding parameter is used in the simplified model. Program. This system consists of multiple parts, and each part uses a computer to determine the desired combination of part variations to achieve the design goals for a product that can have multiple variations, including the nominal variations that make up the initial prototype. And
For the plurality of parts, on the simple model based on the input / output measurement of the part to be controlled in the initial prototype, based on past design data, means for selecting a variation that meets the performance limit,
The number of parts including variations is assumed to be N. For each part, the variation showing the maximum property and the variation showing the minimum property are selected, and the simulation using the above simple model is performed for the 2N types, and the above-mentioned nominal variation is used. In addition to the case, means for plotting the coefficient norm of the state space model derived from the simplified model in an N + 1 dimensional space;
Based on the plotted points, means for constructing a hypersurface by an interpolation method in which differentiation is not discontinuous;
For a combination corresponding to the initial prototype, means for obtaining a test result with an actual device including a controller;
It is determined whether the difference between the result of the test in the real device and the design target is within a predetermined threshold value, and if not, each of the points corresponding to the parts constituting the real device on the hypersurface For the axis of the part variation, calculate the gradient in the direction affected by the combination of the variation axis direction and the other variation axis, and select the part variation related to the direction with the largest gradient absolute value. Presenting variations of parts to perform a test in an actual device, and having means for repeating until the difference between the result of the test in the actual device and the design target is within a predetermined threshold,
system.   The system of claim 9, wherein the gradient in the direction affected by the combination with the other variation axis is a gradient that crosses 45 degrees with respect to the axial direction of the two variations.   The system according to claim 9, wherein the interpolation method in which the differentiation does not become discontinuous is a spline interpolation method.   10. The maximum property and the minimum property are directly the magnitudes of parameters of each component, and indirectly are properties derived when the corresponding parameter is used in the simplified model. System.

Images