WO2006048919A1 - 最適設計管理装置、最適設計計算システム、最適設計管理方法、最適設計管理プログラム - Google Patents
最適設計管理装置、最適設計計算システム、最適設計管理方法、最適設計管理プログラム Download PDFInfo
- Publication number
- WO2006048919A1 WO2006048919A1 PCT/JP2004/016227 JP2004016227W WO2006048919A1 WO 2006048919 A1 WO2006048919 A1 WO 2006048919A1 JP 2004016227 W JP2004016227 W JP 2004016227W WO 2006048919 A1 WO2006048919 A1 WO 2006048919A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- explanatory variable
- variable values
- optimization
- design management
- explanatory
- Prior art date
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
Definitions
- Optimal design management device optimal design calculation system, optimal design management method, optimal design management program
- the present invention relates to an optimum design management device, an optimum design calculation system, an optimum design management method, and an optimum design management program for performing optimum design for explanatory variables and objective variables.
- explanatory variables are variables to be designed such as vertical, horizontal, and height dimensions
- objective variables are variables to be optimized such as stress.
- the relationship between explanatory variables and objective variables can be investigated by examining objective variable values for combinations of explanatory variable values assuming a large number of cases, and explanatory variables that optimize objective variable values. Optimal design calculation systems that investigate combinations of values have been realized.
- Patent Document 1 As a related art related to the present invention, for example, Patent Document 1 shown below is known.
- This equipment reliability design support device assigns design variables of equipment and parts to an orthogonal table based on Taguchi method, and analyzes the design analysis model or inverse problem analysis model based on this orthogonal table. Based on this analysis result, a response surface is obtained, and optimization design is performed using this response surface.
- Patent Document 1 JP 2001-125933 A (Page 3-10, Fig. 1)
- response surfaces representing the relationship between explanatory variables and objective variables
- one or more response surfaces are usually created, and the one that best represents the relationship between explanatory variables and objective variables is selected. And used for optimization.
- selection of the response surface requires specialized knowledge on statistical analysis, which limits the number of users.
- the response surface is usually selected so that the relationship between the explanatory variable and the objective variable is appropriately expressed over the entire range of explanatory variable values. It is not always appropriate in the area, and there may be a problem with the accuracy of the response surface. Furthermore, by selecting one response surface and not selecting force, the search range for optimization is narrowed, and optimization accuracy may be insufficient.
- the present invention has been made to solve the above-described problems, and an optimum design management device, an optimum design calculation system, and an optimum design that improve the accuracy of the optimum design without increasing the degree of difficulty for the user.
- the purpose is to provide a management method and an optimal design management program.
- the present invention analyzes an objective variable value from a set of explanatory variable values using at least one external computation node, and performs optimization using the analysis result.
- a management device that transmits a set of explanatory variable values to be used for analysis to a calculation node and sends an analysis instruction; an objective variable reception unit that receives an objective variable value of an analysis result from the calculation node;
- a response surface calculation unit that calculates at least one response surface from a set of explanatory variable values and objective variable values, and a set of explanatory variable values that approach a predetermined optimization condition on the response surface are set as new explanatory variable values.
- An optimization calculation unit that is updated as a set, and a control unit that repeatedly executes the processing that continues from the explanatory variable transmission unit until a predetermined end condition is satisfied.
- a clustering calculation unit that further combines a set of explanatory variable values calculated by the optimization calculation unit into a predetermined number of representative variable value sets. It is characterized by providing.
- the response surface calculation unit calculates a plurality of response surfaces using different algorithms.
- the response surface is an error-free response surface that models the relationship between the explanatory variable and the objective variable so that there is no error. is there.
- the response surface calculation unit calculates a response surface using a plurality of combinations of explanatory variable values and objective variable values that are allowed to overlap.
- a plurality of response surfaces are calculated by performing a plurality of processes while changing the way of overlapping, and a response surface represented by a representative value of the plurality of response surfaces is output.
- the response surface calculation unit calculates a plurality of response surfaces using different algorithms, and is represented by a representative value of the plurality of response surfaces. A response surface is output.
- the optimization condition is a set of explanatory variable values that optimizes the objective variable value.
- the optimization condition is a set of explanatory variable values such that the distribution of the objective variable values is uniform.
- the optimization condition is a set of explanatory variable values such that a deviation between the explanatory variable value and the objective variable value is reduced.
- the first condition is that the objective variable value is a set of optimal variable values
- the distribution of the objective variable value and the explanatory variable value is
- the second condition is a set of explanatory variable values that are uniform, and the optimization condition satisfies the first condition and the second condition at a predetermined ratio.
- the response surface calculation unit calculates a response surface by omitting a set of non-optimal objective variable values and explanatory variable values corresponding to the objective variable values. It is characterized by this.
- the optimization calculation unit has a predetermined set of explanatory variables used by the explanatory variable transmission unit and a set of explanatory variable values approaching the predetermined optimization condition. If the degree of similarity is satisfied, the set of explanatory variable values approaching the predetermined optimization condition is set as the set of explanatory variable values used by the explanatory variable transmitting unit, and the objective variable value of the corresponding analysis result is set to the value of the next analysis result. Substituting, and for the substituted objective variable value, the next explanation variable transmitting unit and the objective variable receiving unit are not executed.
- the present invention is an optimal design management apparatus that analyzes an objective variable value from a set of explanatory variable values using at least one external calculation node and performs optimization using the analysis result.
- An explanatory variable transmission unit that transmits a set of explanatory variable values used for analysis to the calculation node and instructs the analysis, and an objective variable reception unit that receives the target variable value of the analysis result from the calculation node;
- a statistical information calculation unit that calculates regression tree data from a set of explanatory variable values and objective variable values, and a display screen generation unit that generates a display screen based on the regression tree data.
- an explanatory variable value such that the objective variable value satisfies a predetermined objective variable range using the regression tree data is changed to a new explanatory variable value. It is characterized by comprising a centralized calculation unit that is updated as a set, and a control unit that repeatedly executes the processing that continues from the explanation variable transmission unit until a predetermined end condition is satisfied.
- the present invention is an optimal design management device that analyzes an objective variable value from a set of explanatory variable values using at least one external calculation node, and performs optimization using the analysis result.
- An explanatory variable transmitting unit that transmits a pair of explanatory variable values to the calculation node and instructing analysis
- an objective variable receiving unit that receives the objective variable value from the calculation node
- a principal component from the combination of the explanatory variable value and the objective variable value
- a statistical information calculation unit that calculates analysis data and a display screen generation unit that generates a display screen based on the principal component analysis data are provided.
- an explanatory variable value such that the objective variable value satisfies a predetermined objective variable range using the principal component analysis data is changed to a new explanatory variable value. It is characterized by comprising a centralized calculation unit that is updated as a set, and a control unit that repeatedly executes the processing from the explanatory variable transmission unit until a predetermined end condition is satisfied.
- the present invention is an optimal design management apparatus that analyzes an objective variable value from a set of explanatory variable values using at least one external calculation node and performs optimization using the analysis result.
- the objective variable receiving unit that receives the objective variable value from the calculation node, and the combination of the explanatory variable value and the objective variable value
- a statistical information calculation unit for calculating the local network data
- a display screen generation unit for generating a display screen based on the -eural network data.
- the present invention is an optimal design calculation system that performs analysis of a target variable value from a set of explanatory variable values using at least one calculation node and performs optimization using the analysis result.
- Send the explanatory variable value set used in the analysis to the calculation node, and send an analysis variable based on the explanatory variable value of the explanatory variable transmitting unit and the explanatory variable transmitting unit for instructing the analysis, and the objective variable of the analysis result A calculation node for calculating a value, an objective variable receiving unit for receiving the objective variable value of the analysis result from the calculation node, and a response surface calculating unit for calculating at least one response surface from a set of explanatory variable values and objective variable values;
- a control unit that repeatedly executes the process.
- the present invention is an optimal design management method in which an objective variable value is analyzed from a set of explanatory variable values using at least one external calculation node, and optimization is performed using the analysis result.
- the calculation step and the control step of repeatedly executing the process of the explanatory variable transmission step force until the predetermined end condition is satisfied are executed.
- the present invention is an optimal design management method for analyzing an objective variable value from a set of explanatory variable values using at least one external calculation node and performing optimization using the analysis result.
- a statistical information calculation step for calculating regression tree data from a set of variable values and a display screen generation step for generating a display screen based on the regression tree data are executed.
- the present invention is an optimal design management method for analyzing an objective variable value from a set of explanatory variable values using at least one external calculation node and performing optimization using the analysis result.
- An explanatory variable transmission step for sending a set of explanatory variable values to the calculation node and instructing analysis
- an objective variable receiving step for receiving the objective variable value from the calculation node
- an explanatory variable value A statistical information calculation step for calculating principal component analysis data from a set of objective variable values and a display screen generation step for generating a display screen based on the main component analysis data are executed.
- the present invention is an optimal design management method for analyzing an objective variable value from a set of explanatory variable values using at least one external calculation node and performing optimization using the analysis result, From the explanatory variable sending step to send the explanatory variable value pair to the calculation node and instruct analysis, the objective variable reception step to receive the objective variable value from the calculation node, and the explanatory variable value and objective variable value pair- A statistical information calculation step for calculating the neural network data and a display screen generation step for generating a display screen based on the neural network data are executed.
- the present invention provides a computer with an optimal design management method for analyzing a target variable value from a set of explanatory variable values using at least one external calculation node and performing optimization using the analysis result.
- An optimal design management program to be executed which transmits a set of explanatory variable values to be used for analysis to a calculation node and sends an explanatory variable instruction for instructing the analysis, and a purpose for receiving the target variable value of the analysis result from the calculation node
- the variable reception step, the response surface calculation step for calculating at least one response surface from the combination of the explanatory variable value and the objective variable value, and the explanation variable value set approaching a predetermined optimization condition on the response surface are newly explained.
- An optimization calculation step for updating as a set of variable values and a control step for repeatedly executing the explanatory variable transmission step force until the predetermined end condition is satisfied. The is intended to be executed by the Konbiyu data.
- the present invention provides a computer with an optimal design management method for analyzing a target variable value from a set of explanatory variable values using at least one external calculation node and performing optimization using the analysis result.
- An optimal design management program to be executed which transmits a set of explanatory variable values to be used for analysis to a calculation node and sends an explanatory variable instruction for instructing the analysis, and a purpose for receiving the target variable value of the analysis result from the calculation node
- a computer executes a variable reception step, a statistical information calculation step for calculating regression tree data from a set of explanatory variable values and objective variable values, and a display screen generation step for generating a display screen based on the regression tree data.
- ⁇ 3 ⁇ 4 Things.
- the present invention provides a computer with an optimal design management method for analyzing a target variable value from a set of explanatory variable values using at least one external calculation node and performing optimization using the analysis result.
- An optimal design management program to be executed an explanatory variable sending step for sending a set of explanatory variable values to a calculation node and instructing analysis, an objective variable receiving step for receiving an objective variable value from the calculation node, and an explanatory variable
- a computer executes a statistical information calculation step for calculating principal component analysis data from a set of values and objective variable values, and a display screen generation step for generating a display screen based on the principal component analysis data.
- the present invention provides a computer with an optimal design management method for analyzing an objective variable value from a set of explanatory variable values using at least one external calculation node and performing optimization using the analysis result.
- An optimal design management program to be executed an explanatory variable sending step for sending a set of explanatory variable values to a calculation node and instructing analysis, an objective variable receiving step for receiving an objective variable value from the calculation node, and an explanatory variable
- the computer executes a statistical information calculation step for calculating the universal network data from the set of the value and the objective variable value, and a display screen generation step for generating a display screen based on the neural network data.
- FIG. 1 is a block diagram showing an example of the configuration of an optimal design calculation system according to the present invention.
- FIG. 2 is a flowchart showing an example of the operation of the optimum design management apparatus according to the present invention.
- FIG. 3 is a table showing an example of a set of explanatory variable values.
- FIG. 4 is a table showing an example of objective variable values.
- FIG. 5 is a table showing an example of a set of explanatory variable values of optimization results.
- FIG. 6 is a diagram showing an example of a display screen based on regression tree data according to the present invention.
- FIG. 7 is a diagram showing an example of a display screen based on principal component analysis data according to the present invention.
- FIG. 8 is a diagram showing an example of a display screen based on neural network data according to the present invention.
- FIG. 1 shows the optimum according to the present invention. It is a block diagram which shows an example of a structure of a design calculation system.
- This optimal design calculation system includes an optimal design management device 1, a plurality of calculation nodes 2, and an operation terminal 3.
- the optimal design calculation system is realized by, for example, a grid computing system or a cluster computer system.
- the optimal design management device 1 is a management server
- the operation terminal 3 is a user terminal
- the calculation node 2 is a PC (Personal Computer). )
- workstations which are connected via a network.
- the optimal design management device 1 includes a control unit 11, a storage unit 12, a response surface calculation unit 21, an optimization calculation unit 22, a statistical information calculation unit 31, a centralized calculation unit 32, a clustering calculation unit 41, an analysis result omitting unit 42, An analysis result substitution unit 43 is provided.
- the operation terminal 3 includes an input unit 51 and a display unit 52.
- FIG. 2 is a flowchart showing an example of the operation of the optimum design management apparatus according to the present invention.
- the control unit 11 stores the set value of the optimum design calculation in the storage unit 12 according to the input from the user in the input unit 51 (Sl l).
- the set values are, for example, the explanatory variable range that represents the range of the explanatory variable value, the objective variable range that represents the range of the objective variable value, the explanatory variable value, the objective variable value, the termination condition, the number of omissions, the number of outputs, the optimization and model The ratio etc. Individual setting values will be described later.
- the control unit 11 creates a predetermined number of sets of explanatory variable values by sampling initial explanatory variable values in the explanatory variable range, and stores them in the storage unit 12 (S 12 ).
- the control unit 11 uses the set of explanatory variable values of the initial value stored in the storage unit 12 or the set of explanatory variable values of the optimization result as a set of explanatory variable values used for the analysis. Allotted to each computation node 2 together with an analysis instruction (S13).
- FIG. 3 is a table showing an example of a set of explanatory variable values.
- the explanatory variables are vertical, horizontal, and height dimensions, and 8 sets are sampled.
- each calculation node 2 Upon receiving the analysis instruction and explanatory variable value pairs, each calculation node 2 calculates an objective variable value using an analysis program distributed in advance, and controls the control unit 11 as an objective variable value of the analysis result.
- FIG. 4 is a table showing an example of the objective variable value.
- the objective variable is stress
- the control unit 11 receives the objective variable value of the analysis result from each calculation node 2, associates it with the explanatory variable value used for the analysis, and sets it as a set of the explanatory variable value of the analytical result and the objective variable value.
- the controller 11 determines whether or not to perform response surface calculation (S15).
- the control unit 11 passes a set of explanatory variable values and objective variable values of the analysis result and the number of omissions that are set values to the analysis result omitting unit 42, and the analysis result An omission instruction is given (S21).
- the analysis result omitting unit 42 omits pairs of explanatory variable values and objective variable values corresponding to the number of omissions from the case where the objective variable value of the analysis result is not optimal, and passes it to the control unit 11.
- the control unit 11 updates the combination of the explanatory variable value and the objective variable value of the analysis result in the storage unit 12 to the combination of the explanatory variable value and the objective variable value from the analysis result omitting unit 42.
- control unit 11 passes the data such as a set of explanatory variable values of the analysis result and the objective variable stored in the storage unit 12 to the response surface calculation unit 21, and instructs response surface calculation (S22). ).
- the response surface calculation unit 21 calculates a predetermined number of response surfaces from the set of explanatory variable values and objective variable values of the analysis result, and passes the obtained response surfaces to the control unit 11.
- the control unit 11 stores the obtained response curved surface in the storage unit 12.
- FIG. 5 is a table showing an example of a set of explanatory variable values of the optimization result.
- the explanatory variables are vertical, horizontal, and height dimensions, and four sets are sampled.
- control unit 11 uses the statistical information calculation unit to store data such as a set of explanatory variable values and objective variables stored in the storage unit 12. 31 and give instructions for statistical information calculation (S31).
- the statistical information calculation unit 31 generates statistical information from the combination of the explanatory variable value of the analysis result and the objective variable value, and passes the obtained statistical information to the control unit 11.
- the control unit 11 stores the obtained statistical information in the storage unit 12.
- the control unit 11 sets the explanatory variable value and the objective variable value of the analysis result stored in the storage unit 12, statistical information, the explanatory variable range that is the set value, the objective variable range, the similarity threshold.
- An intermediate calculation unit 32 is instructed to perform optimization calculation (S32).
- the centralized calculation unit 32 performs an optimization calculation that samples a set of explanatory variable values that satisfy a predetermined condition within the range of explanatory variables, and passes the result to the control unit 11 as a set of explanatory variable values of the optimization result.
- the control unit 11 stores the set of explanatory variable values of the optimization result in the storage unit 12.
- the control unit 11 passes the set of explanatory variable values of the optimization result and the number of outputs as the set value to the clustering calculation unit 41, and instructs clustering calculation (S41).
- the clustering calculation unit 41 performs clustering calculation to combine a large number of sets of explanatory variable values of the optimization results into a set of representative explanatory variable values for the number of outputs, and the combined set of explanatory variable values to the control unit 11 To pass.
- the calculation method of clustering for example, the farthest neighbor method or the Ward method is used.
- the control unit 11 updates the set of explanatory variable values of the optimization result in the storage unit 12 to the set of explanatory variable values from the clustering calculation unit 41. By using this clustering calculation, it is possible to reduce the resources for the next analysis and improve the accuracy of optimization when performing a wide area search.
- the control unit 11 stores, from the storage unit 12, a set of explanatory variable values and objective variable values as analysis results, a set of explanatory variable values and objective variable values as optimization results, a response surface, and a set value.
- the explanatory variable range and the similarity threshold are passed to the analysis result substitute unit 43, and an instruction for substitution of the analysis result is given (S42).
- the analysis result substitute unit 43 calculates, for each combination, a pair of explanatory variable values of the analysis result, a pair of explanatory variable values of the optimization result, and the similarity.
- the similarity between two explanatory variable values is calculated, for example, as the reciprocal of the sum of squares of the difference between the two explanatory variable values.
- the optimization calculation unit 22 sets the combination of the explanatory variable value and the objective variable value of the optimization result corresponding to the maximum similarity value, that is, the next analysis.
- the pair of explanatory variable values to be used and the objective variable value that is the result of the analysis are replaced with the pair of explanatory variable values and objective variable values that have already been used for the analysis. Do not perform. By substituting this analysis result, it is possible to create a response surface that emphasizes the explanatory variable value area of the optimization result, and to reduce the resources for analysis.
- the control unit 11 determines whether or not the end condition is satisfied (S51).
- the end condition is, for example, the upper limit of the number of iterations or the similarity between the explanatory variable value pair of the analysis result and the explanatory variable value pair of the optimization result! Set the saddle value.
- the termination condition is met ( S51, Y)
- this flow ends.
- the termination condition is not satisfied (S51, ⁇ )
- the process returns to S13.
- the response surface calculation unit 21 uses any one of the following four response surface calculation methods.
- the plurality of response surface calculation units 21 may be provided to calculate in parallel.
- the control unit 11 inputs a set of explanatory variable values and objective variable values of analysis results of ⁇ set ( ⁇ is an integer) to the response surface calculation unit 21.
- the response surface calculation unit 21 performs response surface calculation using algorithms with different S (L is an integer), and calculates L response surfaces. For example, multiple regression analysis or -Eural network is used to calculate the response surface.
- the response surface calculation unit 21 outputs L response surfaces to the control unit 11.
- This response surface calculation method calculates an error-free response surface that models the relationship between the explanatory variable and the objective variable so that there is no error.
- A1 Y1 / (F2 (XI) — ⁇ 2) / (F3 (XI) — ⁇ 3)
- ⁇ 2 ⁇ 2 / (F3 ( ⁇ 2) ⁇ 3) / (Fl ( ⁇ 2) Y1)
- A3 ⁇ 3 ⁇ (Fl ( ⁇ 3) Yl) / (F2 ( ⁇ 3)- ⁇ 2)
- F (X) Al * (F2 (X) —Y2) * (F3 (X) —Y3)
- the response surface calculation unit 21 outputs the response surface without error to the control unit 11 as a response surface.
- M is an integer
- the response surface calculation unit 21 extracts M sets of explanatory variable values and objective variable values while changing the overlapping method, and repeats the calculation of the response surface M times to calculate M response surfaces. . Finally, the response surface calculation unit 21 calculates a response surface represented by the representative values of the M response surfaces, and outputs this to the control unit 11. According to this response surface calculation method, it is possible to reduce the influence of unreliable pairs of explanatory variable values and objective variable values, and to improve optimization accuracy when there are a small number of analysis results. That's right.
- the control unit 11 inputs a set of K sets of explanation variable values and objective variable values, the number of algorithms N (N is an integer), and the type of N algorithms to the response surface calculation unit 21.
- the response surface calculation unit 21 calculates N response surfaces using N different algorithms.
- the response surface calculation unit 21 calculates a response surface represented by a representative value of N response surfaces, and outputs this to the control unit 11. According to this response surface calculation method, the response surface and the accuracy of optimization using the response surface can be improved.
- the optimization calculation unit 22 uses any of the following four optimization calculation methods.
- the second optimization calculation method or the third optimization calculation method is selected at the initial stage of the optimal design, and the first optimization calculation method is selected at the final stage of the optimal design.
- the fourth optimization calculation method is selected when optimal design is performed with a balance between the two.
- the first optimization calculation method is used when the user narrows down the set of explanatory variable values.
- the control unit 11 inputs the explanatory variable range and the response surface to the optimization calculation unit 22.
- the optimization calculation unit 22 calculates a set of explanatory variable values for which the objective variable value is optimal on the response surface by a steepest descent method or a genetic algorithm, and the control unit 11 sets the set of explanatory variable values for the optimization result.
- the second optimization calculation method will be described.
- This optimization calculation method is a model of objective variables, and aims to sample a set of explanatory variable values so that the distribution of objective variable values is uniform, and the user grasps the state of the entire objective variable range.
- the control unit 11 inputs the explanatory variable range, the response surface, and the objective variable value of the analysis result to the optimization calculation unit 22.
- the optimization calculation unit 22 sets the intermediate value of the interval with the largest interval between the objective variable values as the optimization target value.
- the optimization calculation unit 22 performs the optimization calculation with the objective variable value calculated using the response surface being close to the optimization target value, so that the target variable value becomes close to the optimization target value.
- Such a set of explanatory variable values is passed to the control unit 11 as a set of explanatory variable values for the optimization result.
- This optimization calculation method is a model of explanatory variables, and aims to sample a pair of explanatory variable values so that the deviation between the explanatory variable value and the objective variable value is reduced. It is used when grasping.
- the control unit 11 inputs a set of the explanatory variable range, the response surface, and the explanatory variable value to the optimization calculation unit 22.
- the optimization calculator 22 samples the explanatory variable value pairs in the explanatory variable range, calculates the objective variable value using the response surface, and calculates the similarity between the explanatory variable value pair and the objective variable value. To do.
- the similarity between the pair of explanatory variable values and the objective variable value is calculated as, for example, the reciprocal of the sum of squares of the difference between each explanatory variable value and the objective variable value.
- the optimization calculation unit 22 passes a set of explanatory variable values that minimizes the maximum value of the similarity to the control unit 11 as a set of explanatory variable values of the optimization result.
- the control unit 11 inputs an explanatory variable range, a response surface, a set of explanatory variable values, and a ratio of optimization and modeling to the optimization calculation unit 22.
- optimization is the first optimization calculation method, which is to find a set of explanatory variable values that optimizes the objective variable value.
- modeling is the second optimization calculation method or the third optimization calculation method, and it seeks a set of explanatory variable values that makes the distribution of objective variable values and explanatory variable values uniform.
- the optimization calculation unit 22 executes both optimization and model so as to satisfy the specified optimization and model ratio, and sets the obtained explanatory variable values as the optimization result.
- the statistical information calculation unit 31 uses any of the following three statistical information calculation methods.
- the first statistical information calculation method the statistical information calculation unit 31 generates regression tree data from a set of explanatory variable values and objective variable values, and outputs them to the control unit 11 as statistical information.
- the statistical information calculation unit 31 generates principal component analysis data from a set of explanatory variable values and objective variable values, and outputs the data as statistical information to the control unit 11.
- the statistical information calculation unit 31 generates neural network data from a set of explanatory variable values and objective variable values, and outputs the neural network data to the control unit 11 as statistical information.
- the central calculation unit 32 calculates a set of explanatory variable values of the optimization result using the statistical information of regression tree data, principal component analysis data!
- the control unit 11 inputs the explanatory variable range, the objective variable range, the regression tree data, and the similarity threshold value to the concentration calculation unit 32.
- the centralized calculation unit 32 is a set of explanatory variable values that satisfies the objective variable range, and the input explanatory variable value that is input as the set of explanatory variable values to be calculated.
- An explanatory variable value whose similarity with the set of values is equal to or less than the similarity threshold is sampled, and passed to the control unit 11 as a set of explanatory variable values of the optimization result.
- the control unit 11 inputs the explanatory variable range, each principal component value range, the principal component analysis data, and the similarity threshold value to the concentration calculation unit 32.
- the centralized calculation unit 32 satisfies the respective principal component value ranges, and the explanatory variable values such that the similarity between the calculated explanatory variable value set and the input explanatory variable value set is equal to or lower than the similarity threshold value.
- the control unit 11 accepts display instructions from the input unit 51 at any time. When there is a display instruction, the control unit 11 generates a display screen based on the statistical information and outputs it to the display unit 52.
- the statistical information includes regression tree data, principal component analysis data, and neural network data. Specific examples of display screens based on the respective statistical information are shown.
- FIG. 6 is a diagram showing an example of a display screen based on regression tree data according to the present invention.
- vertical, horizontal, and height dimensions are represented as explanatory variables, and stress is represented as an objective variable.
- stress is represented as an objective variable.
- FIG. 7 is a diagram showing an example of a display screen based on principal component analysis data according to the present invention. This figure shows the relationship between the first principal component and the objective variable, and the relationship between the second principal component and the objective variable. According to this display screen, when the number of explanatory variables is large, it can be represented by a small number of explanatory variables. Further, FIG.
- FIG. 8 is a diagram showing an example of a display screen based on neural network data according to the present invention.
- vertical, horizontal, and height dimensions which are explanatory variables, are represented as input nodes, two intermediate nodes are represented, and stress, an objective variable, is represented as an output node.
- stress an objective variable
- the influence of each node can be expressed by expressing the strength of the relationship by the thickness of the arrows connecting the nodes.
- the response surface calculation unit 21, the optimization calculation unit 22, the statistical information calculation unit 31, the centralized calculation unit 32, the clustering calculation unit 41, the analysis result omitting unit 42, and the analysis result substitute unit 43 are not necessarily all It is not necessary, and individual functions may be omitted depending on the purpose of calculation.
- the optimal design calculation system is realized by a grid computing system or a cluster type computer system, and each of the calculation nodes 2 is realized by a computer.
- Type computer computation node 2 may be a CPU
- the optimal design computation system may be one computer.
- the objective variable value is calculated from the explanatory variable value set by executing the analysis program in the calculation node 2.
- the objective variable value is calculated from the explanatory variable value set by the experimental apparatus. Also good.
- the present invention can be applied to optimum design in the fields of structural analysis, fluid analysis, collision analysis, electromagnetic field analysis, acoustic analysis, etc. by using an analysis program according to the purpose.
- a program that causes a computer that constitutes the optimum design management apparatus to execute the above steps can be provided as an optimum design management program.
- the computer constituting the optimum design management apparatus can be executed.
- Recording media that can be read by a computer include portable storage media such as CD-ROMs, flexible disks, DVD disks, magneto-optical disks, and IC cards, databases that hold computer programs, or other computers and their databases.
- a transmission medium on a line is also included.
- explanatory variable transmission unit corresponds to the control unit in the embodiment.
- the response surface calculation unit corresponds to the response surface calculation unit and the analysis result omission unit in the embodiment.
- the optimization calculation unit corresponds to the optimization calculation unit and the analysis result substitution unit in the embodiment.
- the explanatory variable set or the model of the objective variable can be appropriately performed.
- the accuracy of optimization can be improved when the analysis results are small.
- the accuracy of optimization can be improved by limiting the explanatory variable range to a region in the vicinity where the objective function value is optimal.
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Human Resources & Organizations (AREA)
- Economics (AREA)
- Strategic Management (AREA)
- Marketing (AREA)
- Game Theory and Decision Science (AREA)
- Entrepreneurship & Innovation (AREA)
- Development Economics (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Tourism & Hospitality (AREA)
- Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Complex Calculations (AREA)
Abstract
Description
Claims
Priority Applications (3)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
JP2006542189A JP4643586B2 (ja) | 2004-11-01 | 2004-11-01 | 最適設計管理装置、最適設計計算システム、最適設計管理方法、最適設計管理プログラム |
PCT/JP2004/016227 WO2006048919A1 (ja) | 2004-11-01 | 2004-11-01 | 最適設計管理装置、最適設計計算システム、最適設計管理方法、最適設計管理プログラム |
US11/785,902 US7991617B2 (en) | 2004-11-01 | 2007-04-20 | Optimum design management apparatus from response surface calculation and method thereof |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
PCT/JP2004/016227 WO2006048919A1 (ja) | 2004-11-01 | 2004-11-01 | 最適設計管理装置、最適設計計算システム、最適設計管理方法、最適設計管理プログラム |
Related Child Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
US11/785,902 Continuation US7991617B2 (en) | 2004-11-01 | 2007-04-20 | Optimum design management apparatus from response surface calculation and method thereof |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2006048919A1 true WO2006048919A1 (ja) | 2006-05-11 |
Family
ID=36318939
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/JP2004/016227 WO2006048919A1 (ja) | 2004-11-01 | 2004-11-01 | 最適設計管理装置、最適設計計算システム、最適設計管理方法、最適設計管理プログラム |
Country Status (3)
Country | Link |
---|---|
US (1) | US7991617B2 (ja) |
JP (1) | JP4643586B2 (ja) |
WO (1) | WO2006048919A1 (ja) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2015037133A1 (ja) * | 2013-09-13 | 2015-03-19 | 株式会社日立製作所 | 対応情報生成システム及び対応情報生成方法 |
JP2020144690A (ja) * | 2019-03-07 | 2020-09-10 | 株式会社ジェイテクト | 機械学習装置及び機械学習方法 |
JP7452443B2 (ja) | 2021-01-07 | 2024-03-19 | 東芝三菱電機産業システム株式会社 | 圧延モデルの学習方法 |
Families Citing this family (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US8805012B2 (en) * | 2009-11-27 | 2014-08-12 | Nec Corporation | Monitoring state display apparatus, monitoring state display method, and monitoring state display program |
CN103617305A (zh) * | 2013-10-22 | 2014-03-05 | 芜湖大学科技园发展有限公司 | 一种自适应的电力仿真云计算平台作业调度算法 |
WO2016160539A1 (en) | 2015-03-27 | 2016-10-06 | Equifax, Inc. | Optimizing neural networks for risk assessment |
CA3039182C (en) * | 2016-11-07 | 2021-05-18 | Equifax Inc. | Optimizing automated modeling algorithms for risk assessment and generation of explanatory data |
US11468315B2 (en) | 2018-10-24 | 2022-10-11 | Equifax Inc. | Machine-learning techniques for monotonic neural networks |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JPH11259433A (ja) * | 1998-03-09 | 1999-09-24 | Fujitsu Ltd | 並列実行システム |
JP2001306999A (ja) * | 2000-02-18 | 2001-11-02 | Fujitsu Ltd | データ解析装置およびデータ解析方法 |
JP2003114713A (ja) * | 2001-10-09 | 2003-04-18 | Nippon Steel Corp | 品質不良の原因解析方法 |
JP2004110470A (ja) * | 2002-09-19 | 2004-04-08 | Fujitsu Ltd | 最適設計計算装置及びそのプログラム記憶媒体 |
Family Cites Families (20)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US5155677A (en) * | 1989-11-21 | 1992-10-13 | International Business Machines Corporation | Manufacturing process optimizations |
US5781430A (en) * | 1996-06-27 | 1998-07-14 | International Business Machines Corporation | Optimization method and system having multiple inputs and multiple output-responses |
EP0865890A4 (en) * | 1996-08-08 | 1999-05-26 | Bridgestone Corp | MULTI-COMPONENT MATERIAL DESIGN METHOD, OPTIMIZATION ANALYZER, AND INFORMATION MEDIUM ON WHICH THE MULTI-COMPONENT MATERIAL OPTIMIZATION PROGRAM IS RECORDED |
US5966527A (en) * | 1996-10-28 | 1999-10-12 | Advanced Micro Devices, Inc. | Apparatus, article of manufacture, method and system for simulating a mass-produced semiconductor device behavior |
US6606612B1 (en) * | 1998-08-13 | 2003-08-12 | The United States Of America As Represented By The Administrator Of The National Aeronautics And Space Administration | Method for constructing composite response surfaces by combining neural networks with other interpolation or estimation techniques |
US6259389B1 (en) * | 1998-11-20 | 2001-07-10 | General Electric Company | System for designing and testing a sigma-delta modulator using response surface techniques |
US6219649B1 (en) * | 1999-01-21 | 2001-04-17 | Joel Jameson | Methods and apparatus for allocating resources in the presence of uncertainty |
US6636862B2 (en) * | 2000-07-05 | 2003-10-21 | Camo, Inc. | Method and system for the dynamic analysis of data |
US7047505B2 (en) * | 2000-10-17 | 2006-05-16 | Pdf Solutions, Inc. | Method for optimizing the characteristics of integrated circuits components from circuit specifications |
US6980939B2 (en) * | 2001-06-18 | 2005-12-27 | Ford Motor Company | Method and system for optimizing the design of a mechanical system |
US6934666B2 (en) * | 2001-10-15 | 2005-08-23 | General Electric Company | Method for optimizing strategy for electric machines |
US20030086520A1 (en) * | 2001-11-07 | 2003-05-08 | Russell William Earl | System and method for continuous optimization of control-variables during operation of a nuclear reactor |
DE60233312D1 (de) * | 2002-06-21 | 2009-09-24 | Honda Res Inst Europe Gmbh | Autonome und experimentelle Entwurfsoptimierung |
US7469200B2 (en) * | 2003-05-14 | 2008-12-23 | Ford Global Technologies, Llc | Method and apparatus for predicting belt separation failure in aging tires by computer simulation |
US20050004833A1 (en) * | 2003-07-03 | 2005-01-06 | Reaction Design, Llc | Method and system for integrated uncertainty analysis |
US7904281B2 (en) * | 2003-08-07 | 2011-03-08 | Panasonic Corporation | Mounting process simulation system and method thereof |
US7676390B2 (en) * | 2003-09-04 | 2010-03-09 | General Electric Company | Techniques for performing business analysis based on incomplete and/or stage-based data |
US7269517B2 (en) * | 2003-09-05 | 2007-09-11 | Rosetta Inpharmatics Llc | Computer systems and methods for analyzing experiment design |
US7516423B2 (en) * | 2004-07-13 | 2009-04-07 | Kimotion Technologies | Method and apparatus for designing electronic circuits using optimization |
US7389212B2 (en) * | 2004-09-22 | 2008-06-17 | Ford Motor Company | System and method of interactive design of a product |
-
2004
- 2004-11-01 WO PCT/JP2004/016227 patent/WO2006048919A1/ja active Application Filing
- 2004-11-01 JP JP2006542189A patent/JP4643586B2/ja not_active Expired - Fee Related
-
2007
- 2007-04-20 US US11/785,902 patent/US7991617B2/en not_active Expired - Fee Related
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JPH11259433A (ja) * | 1998-03-09 | 1999-09-24 | Fujitsu Ltd | 並列実行システム |
JP2001306999A (ja) * | 2000-02-18 | 2001-11-02 | Fujitsu Ltd | データ解析装置およびデータ解析方法 |
JP2003114713A (ja) * | 2001-10-09 | 2003-04-18 | Nippon Steel Corp | 品質不良の原因解析方法 |
JP2004110470A (ja) * | 2002-09-19 | 2004-04-08 | Fujitsu Ltd | 最適設計計算装置及びそのプログラム記憶媒体 |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2015037133A1 (ja) * | 2013-09-13 | 2015-03-19 | 株式会社日立製作所 | 対応情報生成システム及び対応情報生成方法 |
JP2020144690A (ja) * | 2019-03-07 | 2020-09-10 | 株式会社ジェイテクト | 機械学習装置及び機械学習方法 |
JP7196696B2 (ja) | 2019-03-07 | 2022-12-27 | 株式会社ジェイテクト | 機械学習装置及び機械学習方法 |
JP7452443B2 (ja) | 2021-01-07 | 2024-03-19 | 東芝三菱電機産業システム株式会社 | 圧延モデルの学習方法 |
Also Published As
Publication number | Publication date |
---|---|
JP4643586B2 (ja) | 2011-03-02 |
US20070198252A1 (en) | 2007-08-23 |
US7991617B2 (en) | 2011-08-02 |
JPWO2006048919A1 (ja) | 2008-05-22 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Frutos et al. | A memetic algorithm based on a NSGAII scheme for the flexible job-shop scheduling problem | |
CN115456160A (zh) | 一种数据处理方法和数据处理设备 | |
CN109993299A (zh) | 数据训练方法及装置、存储介质、电子装置 | |
US20220036231A1 (en) | Method and device for processing quantum data | |
WO2018025706A1 (ja) | テーブル意味推定システム、方法およびプログラム | |
CN104794527A (zh) | 基于卷积神经网络的分类模型构建方法和设备 | |
US7991617B2 (en) | Optimum design management apparatus from response surface calculation and method thereof | |
CN113254716B (zh) | 视频片段检索方法、装置、电子设备和可读存储介质 | |
WO2010048758A1 (en) | Classification of a document according to a weighted search tree created by genetic algorithms | |
JPWO2021033338A5 (ja) | ||
CN104679754B (zh) | 用于数据预测的模型选择设备和方法 | |
Xu et al. | Automatic perturbation analysis on general computational graphs | |
CN116976461A (zh) | 联邦学习方法、装置、设备及介质 | |
CN117273167A (zh) | 医疗模型的训练方法、医疗方法、装置、设备及存储介质 | |
Wang et al. | Multiobjective optimization algorithm with objective-wise learning for continuous multiobjective problems | |
KR102574434B1 (ko) | 사용자가 요청하는 전문화된 경량 신경망 모델을 실시간으로 생성하는 방법 및 장치 | |
CN116598022A (zh) | 智能辅助线上问诊方法和系统、计算机设备及存储介质 | |
van den Boom et al. | A modeling framework for model predictive scheduling using switching max-plus linear models | |
Hong et al. | Fast Discrete-Event Simulation of Markovian Queueing Networks through Euler Approximation | |
Xue et al. | AOME: Autonomous optimal mapping exploration using reinforcement learning for NoC-based accelerators running neural networks | |
Fister et al. | Continuous optimizers for automatic design and evaluation of classification pipelines | |
CN114510351B (zh) | 超大规模分布式机器学习装置 | |
JP5942998B2 (ja) | 線形制約条件生成装置及び方法、半正定値最適化問題求解装置、計量学習装置、並びにコンピュータ・プログラム | |
Phan et al. | TF-MOPNAS: training-free multi-objective pruning-based neural architecture search | |
JP2020030702A (ja) | 学習装置、学習方法及び学習プログラム |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
AK | Designated states |
Kind code of ref document: A1 Designated state(s): AE AG AL AM AT AU AZ BA BB BG BR BW BY BZ CA CH CN CO CR CU CZ DE DK DM DZ EC EE EG ES FI GB GD GE GH GM HR HU ID IL IN IS JP KE KG KP KR KZ LC LK LR LS LT LU LV MA MD MG MK MN MW MX MZ NA NI NO NZ OM PG PH PL PT RO RU SC SD SE SG SK SL SY TJ TM TN TR TT TZ UA UG US UZ VC VN YU ZA ZM ZW |
|
AL | Designated countries for regional patents |
Kind code of ref document: A1 Designated state(s): BW GH GM KE LS MW MZ NA SD SL SZ TZ UG ZM ZW AM AZ BY KG KZ MD RU TJ TM AT BE BG CH CY CZ DE DK EE ES FI FR GB GR HU IE IS IT LU MC NL PL PT RO SE SI SK TR BF BJ CF CG CI CM GA GN GQ GW ML MR NE SN TD TG |
|
121 | Ep: the epo has been informed by wipo that ep was designated in this application | ||
WWE | Wipo information: entry into national phase |
Ref document number: 2006542189 Country of ref document: JP |
|
WWE | Wipo information: entry into national phase |
Ref document number: 11785902 Country of ref document: US |
|
NENP | Non-entry into the national phase |
Ref country code: DE |
|
WWP | Wipo information: published in national office |
Ref document number: 11785902 Country of ref document: US |
|
122 | Ep: pct application non-entry in european phase |
Ref document number: 04799433 Country of ref document: EP Kind code of ref document: A1 |