Optimal condition output device, optimal condition output method, and program

Abstract

To obtain an optimal condition of many input factors with respect to a plurality of output characteristics.SOLUTION: An optimal condition output device according to one embodiment includes an acquisition unit that acquires sample data consisting of samples including the values of a plurality of input factors and the values of a plurality of output characteristics, an integration unit that integrates, for each of the samples, a correlation among the values of the input factors included in the sample, and creates the values of a plurality of integrated input factors having a correlation with the values of the output characteristics included in the sample, an optimizing unit that calculates an optimal condition for taking the value of an optimal output characteristic by using the relationship between the values of the integrated input factors and the values of the output characteristics, and an output unit that outputs, as an optimal input factor condition under which the value of the optimal output characteristic is taken, the values of the input factors corresponding to the optimal condition.SELECTED DRAWING: Figure 3

Description

[Technical field]

[0001] The present invention relates to an optimum condition output device, an optimum condition output method, and a program. [Background technology]

[0002] In recent years, there has been an increasing need for multi-objective optimization, such as process optimization and design optimization. Multi-objective optimization involves simultaneously optimizing multiple quality items (also called "output characteristics"). In multi-objective optimization, it is generally required to simultaneously optimize multiple output characteristics by adjusting a large number of input factors that affect the output characteristics.

[0003] For this reason, it has been conventional to perform experiments under conditions of several input factors and measure the output characteristics at that time to determine the conditions of the input factors under which the optimal output characteristics can be obtained (hereinafter also referred to as "optimal conditions"). At this time, particularly when there are many input factors, in order to facilitate evaluation of the experimental results, the input factors are narrowed down based on knowledge such as intuition and experience before the experiment is performed. The optimal output characteristics are output characteristics desired by the user (for example, output characteristics that achieve a maximum value, output characteristics that achieve a minimum value, output characteristics that achieve a certain specific value, etc.). [Prior art documents] [Patent documents]

[0004] [Patent Document 1] JP 2019-86817 A [Patent Document 2] JP 2006-285381 A Summary of the Invention [Problem to be solved by the invention]

[0005] However, in conventional experiments, it is necessary to narrow down the input factors, especially when there are many input factors, so it is difficult to comprehensively evaluate all the conditions of the input factors or to simultaneously optimize all the output characteristics. In addition, since there are multiple output characteristics for many input factors, it is difficult to grasp how the multiple output characteristics change simultaneously in response to changes in the input factors, and it is difficult to simultaneously optimize all the output characteristics.

[0006] The present invention has been made in view of the above-mentioned points, and has an object to obtain optimal conditions of a large number of input factors for a plurality of output characteristics. [Means for solving the problem]

[0007] In order to achieve the above-mentioned object, an optimal condition output device of one embodiment has an acquisition unit that acquires sample data consisting of samples including values ​​of multiple input factors and values ​​of multiple output characteristics; an aggregation unit that aggregates the correlation of the values ​​of the multiple input factors included in each sample and creates multiple aggregated input factor values ​​that are correlated with the values ​​of the multiple output characteristics included in the sample; an optimization unit that calculates optimal conditions for obtaining optimal output characteristic values ​​using the relationship between the values ​​of the multiple aggregated input factors and the values ​​of the multiple output characteristics; and an output unit that outputs the values ​​of the multiple input factors corresponding to the optimal conditions as optimal conditions for the input factors that obtain the optimal output characteristic value. Effect of the Invention

[0008] The optimum of multiple input factors for multiple output characteristics can be obtained. [Brief description of the drawings]

[0009] [Figure 1] FIG. 2 is a diagram illustrating an example of a hardware configuration of the optimum condition output device according to the first embodiment. [Diagram 2] FIG. 11 is a diagram illustrating an example of sample data. [Diagram 3]FIG. 2 is a diagram illustrating an example of a functional configuration of an optimum condition output device according to the first embodiment. [Figure 4] 10 is a flowchart showing an example of the flow of an optimal condition output process according to the first embodiment. [Diagram 5] FIG. 4 is a diagram illustrating an example of a visualization result according to the first embodiment. [Figure 6] FIG. 11 is a diagram illustrating an example of a functional configuration of an optimum condition output device according to a second embodiment. [Figure 7] 13 is a flowchart (part 1) illustrating an example of the flow of an optimum condition output process according to the second embodiment. [Figure 8] FIG. 11 is a diagram (part 1) showing an example of a visualization result according to the second embodiment; [Figure 9] 13 is a flowchart (part 2) illustrating an example of the flow of the optimum condition output process according to the second embodiment. [Figure 10] FIG. 11 is a diagram (part 2) showing an example of a visualization result according to the second embodiment; DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0010] An embodiment of the present invention will be described below. In this embodiment, an optimum condition output device 10 is described that can obtain optimum conditions of a large number of input factors for a large number of output characteristics when sample data consisting of a large number of input factors (particularly a large number of input factors) and a large number of output characteristics corresponding to the input factors are given. The sample data is obtained by performing an experiment under various input factor conditions and measuring the output characteristics at that time.

[0011] [First embodiment] The first embodiment will be described below.

[0012] <Hardware configuration of the optimum condition output device 10> First, the hardware configuration of an optimum condition output device 10 according to the present embodiment will be described with reference to Fig. 1. Fig. 1 is a diagram showing an example of the hardware configuration of the optimum condition output device 10 according to the first embodiment.

[0013] 1, the optimum condition output device 10 according to this embodiment is realized by the hardware configuration of a general computer or computer system, and has an input device 101, a display device 102, an external I / F 103, a communication I / F 104, a processor 105, and a memory device 106. Each of these hardware components is connected to each other via a bus 107 so as to be able to communicate with each other.

[0014] The input device 101 is, for example, a keyboard, a mouse, a touch panel, etc. The display device 102 is, for example, a display, etc. Note that the optimum condition output device 10 does not necessarily have to have at least one of the input device 101 and the display device 102, for example.

[0015] The external I / F 103 is an interface with an external device such as a recording medium 103a. The optimum condition output device 10 can read and write data from and to the recording medium 103a via the external I / F 103. The recording medium 103a may be, for example, a compact disc (CD), a digital versatile disc (DVD), a secure digital memory card (SD memory card), or a universal serial bus (USB) memory card.

[0016] The communication I / F 104 is an interface for connecting the optimum condition output device 10 to a communication network. The processor 105 is, for example, various arithmetic devices such as a CPU (Central Processing Unit) and a GPU (Graphics Processing Unit). The memory device 106 is, for example, various storage devices such as a HDD (Hard Disk Drive), an SSD (Solid State Drive), a RAM (Random Access Memory), a ROM (Read Only Memory), and a flash memory.

[0017] The optimum condition output device 10 according to this embodiment has the hardware configuration shown in Fig. 1, and can realize the optimum condition output process described below. Note that the hardware configuration shown in Fig. 1 is an example, and the optimum condition output device 10 may have other hardware configurations. For example, the optimum condition output device 10 may have multiple processors 105, or multiple memory devices 106.

[0018] <About the sample data> Sample data is data obtained by conducting experiments under various input factor conditions and measuring the output characteristics at those times. The set of input factor conditions and output characteristics at those times in each experiment is called a "sample," and each of the multiple input factors is also called an "input variable," and each of the multiple output characteristics is also called an "output variable."

[0019] Let the number of samples be K and the input factors be x1, , x n , the output characteristics are y1, ,y p As the input factor x at sample number k (1≦k≦K), i (1≦i≦n) value x i (k) , output characteristic y j (1≦j≦p) is the value of y j (k) In addition, the sample number k (1≦k≦K) is the row number, and the input factors x1, ,x n The index i of x is the column number, and the (k,i) element is x i (k) Similarly, let X be the matrix where k (1≦k≦K) is the row number, and y1, ,y p The index j of y is the column number, and the (k,j) element is y j (k) Let Y be a matrix such that

[0020] The sample data is shown in Figure 2. Hereinafter, this sample data is also referred to as (X, Y), where X is also referred to as input factor data and Y is also referred to as output characteristic data. Note that (x1 (k) ,x n (k) ,y1(k) ,···,y p (k) ) is the kth sample, and represents, for example, the input factor conditions in the kth experiment and the output characteristics at that time.

[0021] Here, specific examples of input factors and output characteristics in manufacturing processes, substance and material design, and structural design of structures and machines are shown below.

[0022] Manufacturing process Input factors: Amount of raw materials input, distribution ratio, input timing, etc. Output characteristics: recording density and recording / erasing performance, yield, density, grain size, cost, etc. ·Substance / material design Input factors: molecular structure, blending ratio, catalyst amount, timing of input, etc. Output characteristics: physical properties, viscosity, density, taste, smell, touch, friction, adhesion, boiling point, melting point, particle shape, etc. ·Structural design Input factors: Dimensions, composition and distribution of materials and ingredients, values ​​related to mechanical design, etc. Output characteristics: flow rate, flow rate, temperature, pressure, electromagnetic field, strength (tensile, bending, rigidity, flexibility, etc.), material usage, cost, etc. <Functional configuration of the optimum condition output device 10> Next, the functional configuration of the optimum condition output device 10 according to the present embodiment will be described with reference to Fig. 3. Fig. 3 is a diagram showing an example of the functional configuration of the optimum condition output device 10 according to the first embodiment.

[0023] 3, the optimum condition output device 10 according to this embodiment includes an acquisition unit 201, a compression unit 202, a visualization unit 203, an optimization unit 204, and an output unit 205. Each of these units is realized by, for example, processing executed by the processor 105 of one or more programs installed in the optimum condition output device 10.

[0024] Moreover, the optimum condition output device 10 according to the present embodiment has a storage unit 301. The storage unit 301 is realized, for example, by the memory device 106. However, the storage unit 301 may also be realized, for example, by a storage device (such as a database server) connected to the optimum condition output device 10 via a communication network.

[0025] The storage unit 301 stores the sample data (X, Y) given to the optimum condition output device 10 .

[0026] The acquisition unit 201 acquires sample data (X, Y) stored in the storage unit 301 .

[0027] The compression unit 202 compresses input factors x1, . . . , x of the input factor data X included in the sample data (X, Y) acquired by the acquisition unit 201. n z1,...,z m (where n>>m) and compress the dimensions to create compressed data Z. Hereinafter, z1,...,z m is also called the "aggregated input factor" or "aggregated input variable". In addition, the aggregated input factor z i (1≦i≦m) i (k) In addition, the compressed data Z has the (k,i) element z i (k) The compression unit 202 may be called, for example, an aggregation unit or the like.

[0028] More specifically, the compression unit 202 uses, for example, PLS2 (Partial Least Squares 2) or CCA (Canonical Correlation Analysis) to calculate the input factors x1, . . . , x n The correlation of each z i (1≦i≦m) and each z i (1≦i≦m) is the output characteristic y1, ,y p Compressed data Z is created so as to have a correlation with

[0029] The visualization unit 203 visualizes the relationship between the compressed data Z and the output characteristic data Y as a graph or the like. This enables the user to confirm the relationship between the aggregated input factors obtained by aggregating the input factors and the output characteristics, and for example, enables the user to confirm how a plurality of output characteristics change simultaneously in response to a change in the conditions of the aggregated input factors.

[0030] The optimization unit 204 calculates the optimal condition Z of the aggregated input factors for the output characteristic. opt In addition, the input factor z i The optimal value of (1≦i≦m) is z i (opt) If we write it as Z opt =(z1 (opt) ,...,z m (opt) )

[0031] The output unit 205 outputs the optimal condition Z opt The compression unit 202 converts the input factor X opt is output as the optimal condition for the input factor. Note that the input factor x i The optimal value of (1≦i≦n) is x i (opt) If we write it as X opt =(x1 (opt) ,x n (opt) )

[0032] <Flow of optimal condition output process> Next, the flow of the optimum condition output process according to this embodiment will be described with reference to Fig. 4. Fig. 4 is a flowchart showing an example of the flow of the optimum condition output process according to the first embodiment. It is assumed that the storage unit 301 stores sample data (X, Y) given to the optimum condition output device 10.

[0033] First, the acquiring unit 201 acquires the sample data (X, Y) stored in the storage unit 301 (step S101).

[0034] Next, the compression unit 202 creates compressed data Z from the input factor data X included in the sample data (X, Y) acquired in the above step S101 (step S102). Here, this compressed data Z can be created by, for example, PLS2 or CCA. Specifically, the compression unit 202 aggregates the input factor data X by PLS2 (or CCA) to create input factors z1,...,z m (where n>>m), and also obtains a coefficient matrix R with n rows and m columns that estimates output characteristic data Y from the compressed data Z. That is, the compression unit 202 obtains a coefficient matrix R with n rows and m columns and a coefficient matrix B with m rows and p columns such that Z=XR and ^Y=ZB using PLS2 (or CCA). Here, "^Y" represents an estimated value of output characteristic data Y. In this way, compressed data Z is obtained in which the input factors are aggregated and each row is a set (condition) of values ​​of the aggregated input factors that have a correlation with the output characteristic.

[0035] The k-th row of the compressed data Z is the set of aggregated input factor values ​​for the k-th sample (z1 (k) ,...,z m (k) ) As can be seen from the relationship Z=XR, the input factor z i (k) (1≦i≦m) is the input factor x1, ,x n It is expressed as a linear combination of the above.

[0036] Next, the visualization unit 203 visualizes the relationship between the compressed data Z created in the above step S102 and the output characteristic data Y included in the sample data (X, Y) acquired in the above step S101 as a graph or the like (step S103). That is, the visualization unit 203 visualizes each (z1 (k) ,...,z m (k) ,y1 (k) ,···,y p (k) ) is the point indicated by (z1, ,z m ,y1,...,y pThe visualization unit 203 plots each (z1 (k) ,...,z m (k) ,y1 (k) ,···,y p (k) ) is also approximated by a plane (generally a hyperplane) called (z1, ,z m ,y1,...,y p ) space. This plane (hereinafter also referred to as the "approximate plane") can be created and drawn from the compressed data Z and the coefficient matrix B using the relationship ^Y=ZB. Note that, since a space of four or more dimensions is generally not visible, the scatter plot and its approximate plane are projected into a space of three or less dimensions when presented to the user.

[0037] As an example, a graph of the visualization result when m=2, p=3 is shown in Fig. 5. The left diagram in Fig. 5 is a scatter plot in the (z1,z2,y1) space and its approximated plane, the middle diagram is a scatter plot in the (z1,z2,y2) space and its approximated plane, and the right diagram is a scatter plot in the (z1,z2,y3) space and its approximated plane. Note that the circles in Fig. 5 represent the scatter plots, and the mesh-like plane represents the approximated plane.

[0038] By checking the graph visualized on the display device 102 by the visualization unit 203, the user can know how to change the post-aggregation input factors to obtain the desired output characteristic (e.g., an output characteristic that takes a desired value, maximum value, or minimum value).

[0039] Next, the optimization unit 204 calculates the optimal condition Z opt (Step S104). Here, as described above, each (z1 (k) ,...,z m (k) ,y1 (k) ,···,y p (k) ) is approximated by a plane, so the function y j =h j (z1, ,z m )=b 1jz1+···+b mj z m (where 1≦j≦p) is the objective function, and the optimal condition Z opt In this case, since linear programming generally involves constraints (for example, upper and lower limits determined by the physical constraints of the target, equality constraints, etc.), the optimal condition Z is calculated while taking these constraints into consideration. opt Note that the original input factors x1, ,x n The constraints on the aggregated input factors z1, , z m is converted into a constraint.

[0040] Next, the compression unit 202 performs the compression on the optimum condition Z opt The optimal condition for the original input factors is X opt (Step S105). This is done by using the relationship Z=XR. opt R + =X opt It is possible to convert it by + is the pseudoinverse of R (i.e., a matrix that approximates the inverse of R). This optimal condition X opt are the conditions (set of input factor values) for obtaining the optimal output characteristics.

[0041] Then, the output unit 205 outputs the optimal condition X obtained in step S105. opt (Step S106). The output unit 205 outputs the optimum condition X opt Such an output destination may be, for example, a display device 102 such as a display, a memory device 106, or other devices or apparatuses connected via a communication network.

[0042] For example, steps S101 to S103 may be executed without executing steps S104 to S106, or steps S101 to S102 and steps S104 to S106 may be executed without executing step S103.

[0043] [Second embodiment] The second embodiment will be described below. Note that in the second embodiment, differences from the first embodiment will be mainly described, and descriptions of components similar to those in the first embodiment will be omitted.

[0044] In this embodiment, a graph or the like is presented to the user, which not only shows how the aggregated input factors should be changed to increase or decrease the output characteristics, but also shows whether the direction of change is close to the optimum condition. For this purpose, in this embodiment, the relationship between the aggregated input factors and the output characteristics, or the relationship between the input factors and the output characteristics, is approximated by a nonlinear model and presented to the user.

[0045] <Functional configuration of the optimum condition output device 10> First, the functional configuration of the optimum condition output device 10 according to the present embodiment will be described with reference to Fig. 6. Fig. 6 is a diagram showing an example of the functional configuration of the optimum condition output device 10 according to the second embodiment.

[0046] 6, the optimum condition output device 10 according to this embodiment has a nonlinear model creation unit 206 in addition to the units described in the first embodiment. The nonlinear model creation unit 206 is realized, for example, by a process executed by the processor 105 of one or more programs installed in the optimum condition output device 10.

[0047] The nonlinear model creation unit 206 approximates the relationship between the compressed data Z and the output characteristic data Y, or the relationship between the input factor data X and the output characteristic data Y, using a nonlinear model.

[0048] <Flow of optimal condition output process (part 1)> Next, the flow of the optimum condition output process in the case where the relationship between the aggregated input factors and the output characteristics is approximated by a nonlinear model will be described with reference to Fig. 7. Fig. 7 is a flowchart (part 1) showing an example of the flow of the optimum condition output process according to the second embodiment. It is assumed that the storage unit 301 stores sample data (X, Y) given to the optimum condition output device 10.

[0049] First, the acquiring unit 201 acquires the sample data (X, Y) stored in the storage unit 301 (step S201), similarly to step S101 in FIG.

[0050] Next, the compression unit 202 creates compressed data Z from the input factor data X included in the sample data (X, Y) acquired in the above step S201, similarly to step S102 in FIG. 4 (step S202).

[0051] Next, the nonlinear model creation unit 206 calculates y j A nonlinear model g that approximates j (z1, ,z m ) (step S203). That is, the nonlinear model creation unit 206 creates ^y j (k) =g j (z1 (k) ,...,z m (k) ) a nonlinear model g j (z1, ,z m ), where "^y j (k) " is y j (k) This represents an approximation of the nonlinear model. Such a nonlinear model can be created by a known method. Examples of nonlinear models include SVR (Support Vector Regression), random forest, XGBoost, neural network, polynomial model, etc.

[0052] Next, the visualization unit 203 visualizes the relationship between the compressed data Z created in the above step S202 and the output characteristic data Y included in the sample data (X, Y) acquired in the above step S201 as a graph or the like (step S204). That is, the visualization unit 203 visualizes the relationship between the compressed data Z created in the above step S202 and the output characteristic data Y included in the sample data (X, Y) acquired in the above step S201 as a graph or the like (step S204). (k) ,...,z m (k) ,y1 (k) ,···,y p(k) ) is the point indicated by (z1, ,z m ,y1,...,y p ) space. The visualization unit 203 also plots the nonlinear model y j =g j (z1, ,z m ) also represents a surface (generally a hypersurface) represented by (z1, ,z m ,y1,...,y p ) space. This surface is also called an approximated surface, where j = 1, , p. Generally, a space with four or more dimensions cannot be visually recognized, so when presenting the scatter plot and its approximated surface to the user, they are projected into a space with three or less dimensions.

[0053] As an example, a graph of the visualization result when m=2 and p=3 is shown in Fig. 8. The left diagram in Fig. 8 is a scatter plot in the (z1,z2,y1) space and its approximated surface, the middle diagram is a scatter plot in the (z1,z2,y2) space and its approximated surface, and the right diagram is a scatter plot in the (z1,z2,y3) space and its approximated surface. Note that the circles in Fig. 8 represent the scatter plots, and the mesh-like surface represents the approximated surface.

[0054] By checking the graph visualized on the display device 102 by the visualization unit 203, the user can know how to change the post-aggregation input factors to obtain a desired output characteristic (e.g., an output characteristic that takes a desired value, a maximum value, or a minimum value). Furthermore, by checking the change in the inclination of the approximation surface (i.e., the gradient), it is possible to know whether the direction of change in the post-aggregation input factors indicates that the optimal condition (optimum point) is close. This is because, generally, the gradient is often 0 at the point on the approximation surface where the optimal output characteristic is obtained. In the example shown in FIG. 8, it can be known that there is a high possibility that the optimal point is located around the large circle.

[0055] Next, the optimization unit 204 calculates the optimal condition Z opt (Step S205). Here, as described above, each (z1 (k) ,...,zm (k) ,y1 (k) ,···,y p (k) ) is the nonlinear model y j =g j (z1, ,z m ) is approximated by y j =g j (z1, ,z m ) is used as the objective function, and the optimal condition Z is obtained by using a nonlinear programming method such as the Newton method. opt If there are constraints (for example, upper and lower limits or equality constraints determined by the physical constraints of the target), the optimum condition Z opt Note that the original input factors x1, ,x n The constraints on the aggregated input factors z1, , z m is converted into a constraint.

[0056] Next, the compression unit 202 performs the compression process using the optimal condition Z calculated in the above step S205 in the same manner as in step S105 of FIG. opt The optimal condition for the original input factors is X opt (step S206).

[0057] 4, the output unit 205 outputs the optimal condition X obtained in the above step S206. opt (Step S207).

[0058] For example, steps S201 to S204 may be executed without executing steps S205 to S207, or steps S201 to S203 and steps S205 to S207 may be executed without executing step S204.

[0059] <Flow of optimal condition output process (part 2)> Next, the flow of the optimum condition output process in the case where the relationship between the input factors and the output characteristics is approximated by a nonlinear model will be described with reference to Fig. 9. Fig. 9 is a flowchart (part 2) showing an example of the flow of the optimum condition output process according to the second embodiment. It is assumed that the storage unit 301 stores sample data (X, Y) given to the optimum condition output device 10.

[0060] First, the acquiring unit 201 acquires the sample data (X, Y) stored in the storage unit 301 (step S301), similarly to step S101 in FIG.

[0061] Next, the compression unit 202 creates compressed data Z from the input factor data X included in the sample data (X, Y) acquired in the above step S301, similarly to step S102 in FIG. 4 (step S302).

[0062] Next, the nonlinear model creation unit 206 calculates y j A nonlinear model f that approximates j (x1,...,x n ) (step S303). That is, the nonlinear model creation unit 206 creates ^y j (k) =f j (x1 (k) ,x n (k) ) a nonlinear model f j (x1,...,x n Such a nonlinear model can be created by known methods.

[0063] Next, the visualization unit 203 visualizes the relationship between the compressed data Z created in the above step S302 and the output characteristic data Y included in the sample data (X, Y) acquired in the above step S301 as a graph or the like (step S304). That is, the visualization unit 203 visualizes each (z1 (k) ,...,z m (k) ,y1 (k) ,···,y p(k) ) is the point indicated by (z1, ,z m ,y1,...,y p ) space. The visualization unit 203 also plots the nonlinear model y j =f j (x1,...,x n ) and (z1, ,z m )=(x1, ,x n )R, the surface (approximated surface) represented by (z1, ,z m ,y1,...,y p ) space. Since spaces with four or more dimensions are generally not visible, scatter plots and their approximated surfaces are projected onto a space with three or fewer dimensions when presented to the user.

[0064] The i-th column vector of the coefficient matrix R is i If we write it as: the approximate surface above is {(z1, ,z m ,y1,...,y p )|z i =(x1, ,x n )R i ,y j =f j (x1,...,x n ), 1≦i≦m, 1≦j≦p}.

[0065] As an example, a graph of the visualization result when m=2, p=3 is shown in Fig. 10. The left figure in Fig. 10 is a scatter plot in the (z1,z2,y1) space and its approximated surface, the middle figure is a scatter plot in the (z1,z2,y2) space and its approximated surface, and the right figure is a scatter plot in the (z1,z2,y3) space and its approximated surface. Note that the circles in Fig. 10 represent the scatter plots, and the mesh-like surface represents the approximated surface.

[0066] As described above, by checking the graph visualized by the visualization unit 203, the user can know how to change the aggregated input factors to obtain the desired output characteristics, and can also know whether the direction of change in the aggregated input factors is close to the optimal condition. In the example shown in Fig. 10, it can be seen that there is a high possibility that the optimal point is around the large circle.

[0067] Next, the optimization unit 204 calculates the optimal condition X opt (Step S305). Here, as described above, each (x1 (k) ,x n (k) ,y1 (k) ,···,y p (k) ) is the nonlinear model y j =f j (x1,...,x n ) is approximated by y j =f j (x1,...,x n ) is used as the objective function, and the optimal condition X is obtained by using a nonlinear programming method such as the Newton method. opt If there are constraints, the optimum condition X should be calculated while taking those constraints into account. opt Calculate.

[0068] 4, the output unit 205 outputs the optimal condition X obtained in the above step S305. opt (Step S306). At this time, the output unit 205 outputs the optimal condition X opt After aggregation, Z is converted into an input factor. opt You can also output the optimum condition X by using the relationship Z=XR. opt After aggregating, the optimal condition for the input factor Z opt It is possible to convert it into

[0069] For example, steps S301 to S304 alone may be executed without executing steps S305 to S306, or steps S301 to S303 and steps S305 to S306 may be executed without executing step S304.

[0070] <Summary> As described above, the optimum condition output device 10 according to the first embodiment aggregates the correlations of many input factors so that they are correlated with the output characteristics by using techniques such as PLS2 and CCA, and visualizes the relationship between the aggregated variables (aggregated input factors) and the output characteristics. This makes it possible to simultaneously capture the changes in the multiple output characteristics using a smaller number of aggregated input factors, even in cases where multiple output characteristics change simultaneously in response to changes in many input factors. This makes it possible to know in which direction the aggregated input factors should be changed in order to obtain the desired output characteristics.

[0071] In addition, the optimum condition output device 10 according to the first embodiment calculates and outputs the optimum input factor conditions using the relationship between the aggregated input factors and the output characteristics. This allows the user to know the optimum input factor conditions for obtaining the desired output characteristics.

[0072] Furthermore, the optimum condition output device 10 according to the second embodiment approximates the aggregated input factors or the relationship between the input factors and the output characteristics with a nonlinear model, visualizes this nonlinear model, and calculates and outputs the optimum input factor conditions using this nonlinear model. This makes it possible to know how much the output characteristics change with respect to changes in the aggregated input factors, and therefore makes it possible to know not only the direction in which the aggregated input factors are changed, but also where the optimum point is located in that direction.

[0073] The present invention is not limited to the above-described embodiments specifically disclosed, and various modifications, changes, combinations with known technologies, etc. are possible without departing from the scope of the claims. [Explanation of symbols]

[0074] 10 Optimal Condition Output Device 101 Input Device 102 Display device 103 External I / F 103a Recording media 104 Communication I / F 105 processors 106 Memory Device 107 Bus 201 Acquisition Department 202 Compression section 203 Visualization section 204 Optimization Department 205 Output section 206 Nonlinear Model Creation Department 301 Storage section

Claims

1. an acquisition unit that acquires sample data consisting of samples including values ​​of a plurality of input factors and values ​​of a plurality of output characteristics; an aggregation unit that aggregates, for each sample, correlations between values ​​of a plurality of input factors included in the sample and creates a plurality of aggregated input factor values ​​that are correlated with values ​​of a plurality of output characteristics included in the sample; an optimization unit that calculates an optimal condition for obtaining an optimal output characteristic value by using a relationship between the values ​​of the plurality of aggregated input factors and the values ​​of the plurality of output characteristics; a first visualization unit that visualizes points represented by pairs of the values ​​of the output characteristics and the values ​​of the aggregated input factors, and a plane that approximates the points for each sample; an output unit that outputs values ​​of the plurality of input factors corresponding to the optimum condition as an optimum condition of the input factors that takes the value of the optimum output characteristic; An optimum condition output device having the above-mentioned.

2. The optimization unit is 2. The optimum condition output device according to claim 1, wherein the optimum conditions are calculated by a linear programming method using linearity between the values ​​of the plurality of output characteristics and the values ​​of the plurality of aggregated input factors.

3. An acquisition unit that acquires sample data consisting of samples including values ​​of a plurality of input factors and values ​​of a plurality of output characteristics; an aggregation unit that aggregates, for each sample, correlations between values ​​of a plurality of input factors included in the sample and creates a plurality of aggregated input factor values ​​that are correlated with values ​​of a plurality of output characteristics included in the sample; an optimization unit that calculates an optimal condition for obtaining an optimal output characteristic value by using a relationship between the values ​​of the plurality of aggregated input factors and the values ​​of the plurality of output characteristics; a first nonlinear model creation unit that creates, for each of the output characteristics, a nonlinear model that approximates a value of the output characteristic with respect to values ​​of the plurality of aggregated input factors; a second visualization unit that visualizes points represented by pairs of the values ​​of the output characteristics and the values ​​of the aggregated input factors, and a curved surface represented by the nonlinear model; an output unit that outputs values ​​of the plurality of input factors corresponding to the optimum condition as an optimum condition of the input factors that takes the value of the optimum output characteristic; An optimum condition output device having the above-mentioned.

4. The optimization unit is 4. The optimum condition output device according to claim 3, wherein the optimum conditions are calculated by a nonlinear programming method using the nonlinear model.

5. An acquisition unit that acquires sample data consisting of samples including values ​​of a plurality of input factors and values ​​of a plurality of output characteristics; an aggregation unit that aggregates, for each sample, correlations between values ​​of a plurality of input factors included in the sample and creates a plurality of aggregated input factor values ​​that are correlated with values ​​of a plurality of output characteristics included in the sample; an optimization unit that calculates an optimal condition for obtaining an optimal output characteristic value by using a relationship between the values ​​of the plurality of aggregated input factors and the values ​​of the plurality of output characteristics; a second nonlinear model creation unit that creates, for each of the output characteristics, a nonlinear model that approximates a value of the output characteristic with respect to values ​​of the plurality of input factors; a third visualization unit that visualizes points represented by pairs of values ​​of the output characteristics and values ​​of the aggregated input factors, and a surface represented by the aggregated input factors corresponding to the input factors and the nonlinear model; an output unit that outputs values ​​of the plurality of input factors corresponding to the optimum condition as an optimum condition of the input factors that takes the value of the optimum output characteristic; An optimum condition output device having the above-mentioned.

6. The optimization unit is Using the nonlinear model, an optimal condition of the plurality of input factors that achieves an optimal output characteristic value is calculated by a nonlinear programming method; The output unit is 6. An optimum condition output device according to claim 5, further comprising: an output device for outputting the optimum condition.

7. The aggregation section is 7. The optimum condition output device according to claim 1, wherein the values ​​of the plurality of aggregated input factors are generated by PLS2 or CCA.

8. an acquisition step for acquiring sample data comprising samples including values ​​of a plurality of input factors and values ​​of a plurality of output characteristics; an aggregation step for each sample, aggregating correlations of values ​​of a plurality of input factors included in the sample and generating a plurality of aggregated input factor values ​​that are correlated with values ​​of a plurality of output characteristics included in the sample; an optimization procedure for calculating an optimal condition for obtaining an optimal output characteristic value by using a relationship between the values ​​of the plurality of aggregated input factors and the values ​​of the plurality of output characteristics; a first visualization step of visualizing points represented by pairs of values ​​of the plurality of output characteristics and values ​​of the plurality of aggregated input factors, and a plane approximating the points for each sample; an output step of outputting values ​​of the plurality of input factors corresponding to the optimum condition as an optimum condition of the input factors that takes the value of the optimum output characteristic; The computer executes the optimum condition output method.

9. An acquisition step of acquiring sample data consisting of samples including values ​​of a plurality of input factors and values ​​of a plurality of output characteristics; an aggregation step for each sample, aggregating correlations of values ​​of a plurality of input factors included in the sample and generating a plurality of aggregated input factor values ​​that are correlated with values ​​of a plurality of output characteristics included in the sample; an optimization procedure for calculating an optimal condition for obtaining an optimal output characteristic value by using a relationship between the values ​​of the plurality of aggregated input factors and the values ​​of the plurality of output characteristics; a first nonlinear model creation step of creating, for each of the output characteristics, a nonlinear model that approximates a value of the output characteristic with respect to values ​​of the plurality of aggregated input factors; a second visualization step of visualizing points represented by pairs of the values ​​of the output characteristics and the values ​​of the aggregated input factors, and a curved surface represented by the nonlinear model; an output step of outputting values ​​of the plurality of input factors corresponding to the optimum condition as an optimum condition of the input factors that takes the value of the optimum output characteristic; The computer executes the optimum condition output method.

10. An acquisition step of acquiring sample data comprising samples including values ​​of a plurality of input factors and values ​​of a plurality of output characteristics; an aggregation step for each sample, aggregating correlations of values ​​of a plurality of input factors included in the sample and generating a plurality of aggregated input factor values ​​that are correlated with values ​​of a plurality of output characteristics included in the sample; an optimization procedure for calculating an optimal condition for obtaining an optimal output characteristic value by using a relationship between the values ​​of the plurality of aggregated input factors and the values ​​of the plurality of output characteristics; a second nonlinear model creation step of creating, for each of the output characteristics, a nonlinear model that approximates a value of the output characteristic with respect to values ​​of the plurality of input factors; a third visualization step of visualizing points represented by pairs of values ​​of the output characteristics and values ​​of the aggregated input factors, and a surface represented by the aggregated input factors corresponding to the input factors and the nonlinear model; an output step of outputting values ​​of the plurality of input factors corresponding to the optimum condition as an optimum condition of the input factors that takes the value of the optimum output characteristic; The computer executes the optimum condition output method.

11. an acquisition step for acquiring sample data comprising samples including values ​​of a plurality of input factors and values ​​of a plurality of output characteristics; an aggregation step for each sample, aggregating correlations of values ​​of a plurality of input factors included in the sample and generating a plurality of aggregated input factor values ​​that are correlated with values ​​of a plurality of output characteristics included in the sample; an optimization procedure for calculating an optimal condition for obtaining an optimal output characteristic value by using a relationship between the values ​​of the plurality of aggregated input factors and the values ​​of the plurality of output characteristics; a first visualization step of visualizing points represented by pairs of values ​​of the plurality of output characteristics and values ​​of the plurality of aggregated input factors, and a plane approximating the points for each sample; an output step of outputting values ​​of the plurality of input factors corresponding to the optimum condition as an optimum condition of the input factors that takes the value of the optimum output characteristic; A program that causes a computer to execute the following.

12. An acquisition step of acquiring sample data comprising samples including values ​​of a plurality of input factors and values ​​of a plurality of output characteristics; an aggregation step for each sample, aggregating correlations of values ​​of a plurality of input factors included in the sample and generating a plurality of aggregated input factor values ​​that are correlated with values ​​of a plurality of output characteristics included in the sample; an optimization procedure for calculating an optimal condition for obtaining an optimal output characteristic value by using a relationship between the values ​​of the plurality of aggregated input factors and the values ​​of the plurality of output characteristics; a first nonlinear model creation step of creating, for each of the output characteristics, a nonlinear model that approximates a value of the output characteristic with respect to values ​​of the plurality of aggregated input factors; a second visualization step of visualizing points represented by pairs of the values ​​of the output characteristics and the values ​​of the aggregated input factors, and a curved surface represented by the nonlinear model; an output step of outputting values ​​of the plurality of input factors corresponding to the optimum condition as an optimum condition of the input factors that takes the value of the optimum output characteristic; A program that causes a computer to execute the following.

13. An acquisition step of acquiring sample data comprising samples including values ​​of a plurality of input factors and values ​​of a plurality of output characteristics; an aggregation step for each sample, aggregating correlations of values ​​of a plurality of input factors included in the sample and generating a plurality of aggregated input factor values ​​that are correlated with values ​​of a plurality of output characteristics included in the sample; an optimization procedure for calculating an optimal condition for obtaining an optimal output characteristic value by using a relationship between the values ​​of the plurality of aggregated input factors and the values ​​of the plurality of output characteristics; a second nonlinear model creation step of creating, for each of the output characteristics, a nonlinear model that approximates a value of the output characteristic with respect to values ​​of the plurality of input factors; a third visualization step of visualizing points represented by pairs of values ​​of the output characteristics and values ​​of the aggregated input factors, and a surface represented by the aggregated input factors corresponding to the input factors and the nonlinear model; an output step of outputting values ​​of the plurality of input factors corresponding to the optimum condition as an optimum condition of the input factors that takes the value of the optimum output characteristic; A program that causes a computer to execute the following.

Images