JPH1185721A - Optimum design supporting device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To attain the automation of series of work of parameter generation and optimum design and the shortening of time required for the optimum resign by repeating the execution of an application program while generating an argument input value until a satisfactory solution is provided. SOLUTION: An execution condition setting part 1 designates execution conditions such as an application program name for performing the optimum design, the information of an argument and the end conditions of execution and stores them in an execution condition storage part 6. Then, the input values of arguments generated by an input value generating part 2 according to a designated input value generating method are stored in an input value group storage part 5. Next, a parallel execution part 3 performs calculation concerning the optimum design while using the application program stored in the input value group storage part 5 and preserves the found solution in an executed result storage part 8, and an evaluation value showing the quality of the solution is stored in an evaluated result storage part 7 together with the combination of argument input values. When the execution of the application program is finished, based on the information stored in the execution condition storage part 6 and evaluated result storage part 7, an end discriminating part 4 discriminates whether the satisfactory solution can be provided or not, and the optimum design is performed by repeating the generation of argument input values, the execution of the program, the evaluation of the solution and the prescribed procedure of end discrimination until the satisfactory solution can be provided.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an apparatus for supporting, in a distributed parallel environment, an optimum design for finding an optimum combination of a plurality of parameters or a combination of values satisfying a certain evaluation criterion.

[0002]

2. Description of the Related Art FIG. 14 is a diagram showing a process for performing an optimum design for finding an optimum combination of a plurality of parameters or a value satisfying a certain evaluation criterion in the prior art.
In FIG. 14, several combinations of parameters are generated,
The combination of the parameters is given as an input value of an argument to an application program for performing a simulation and executed, and it is determined whether a satisfactory solution is obtained. If satisfied, the process is terminated. If not satisfied, a combination of parameters is generated. , The combination of the parameters is given as an input value of an argument to an application program for performing the simulation, and executed, and it is determined whether a satisfactory solution is obtained again. If satisfied, the process is terminated. If not satisfied, the process is repeated. This was a manual operation. The execution of the application program by the combination of the parameters is independent and has parallelism, so that parallel processing is possible. Therefore, parameter generation / aggregation / evaluation,
Further, automation including generation of the next combination of parameters based on the previous execution result (evaluation of the obtained intermediate result) is expected.

[0003] http: //www.activetoo on the Internet
ActiveTools C shown in ls.com/products.html
luster1.1 is a tool for automating the generation of parameter combinations and the execution of application programs based on them in a distributed parallel environment connected by a network.

[0004] The automatic parameter generation has a function of uniformly generating a value in a range designated for each parameter or randomly generating a value. For each parameter generated, an application in which all combinations are specified. Give to the program,
Execute in parallel.

[0005]

In the conventional Cluster 1.1, a combination of parameters to be given to an application program is automatically generated. However, since there is no index as to which range should be generated and at what width, an appropriate value is initially set. The generation range was generated with an appropriate step size, the application program was executed, the result was judged by the user, and the generation range and the step size had to be changed for each parameter. That is, these changes need to be manually performed by the user, and are not automated for optimal design. Therefore, in order to use a lot of parts that can be automated, the range of parameter generation should be large,
The generation step width is set to a small value. As a result, the number of parameter combinations increases, which causes a problem that the execution time increases.

[0006] Further, regarding parameter generation, since the generation range and the step size of generation are manually specified, optimization, that is, a guideline for searching for an optimal solution is not provided by the system, and the part thereof is at the discretion of the user. Has been entrusted to Therefore,
There is a problem that not all users can efficiently search for an optimal solution.

Further, since the execution of the application program corresponding to each parameter combination is performed independently, even if a satisfactory result is obtained with a certain parameter combination,
Not only can the user not see the result until all the executions have been completed, but the execution by the combination of other parameters is continued.
As a result, there is a problem that a lot of time is spent for optimal design.

The present invention has been made to solve the above problems, and automatically generates parameters based on the execution result of an application program for performing an optimal design.
An object of the present invention is to automate a series of operations for optimal design and reduce the time required for optimal design.

[0009]

According to a first aspect of the present invention, there is provided an optimal design support apparatus, comprising: a name of an application program to be designed based on a given combination of argument input values; information on arguments; Execution condition setting means for setting a plurality of combinations of argument input values to be given to the application program based on the set execution conditions and the evaluation result indicating the quality of the solution obtained by the previous application program. Means,
A parallel execution means for giving a generated combination of a plurality of argument input values to an application program and executing the application program in parallel, and an end determination for judging whether a satisfactory solution has been obtained after execution of all the application programs is completed Means is terminated when a satisfactory solution is obtained, and if no satisfactory solution is obtained, an argument input value is generated again by the argument input value generating means and the execution of the application program is repeated. .

According to a second aspect of the present invention, in the optimal design support apparatus according to the first aspect of the present invention, in the generation of the next argument input value when a satisfactory solution is not obtained, a good evaluation result is obtained in the previous execution. And an argument input value generating means for generating the vicinity of the adopted argument input value in a range finer than the previous time.

According to a third aspect of the present invention, in the optimal design support apparatus according to the second aspect of the present invention, a good evaluation result can be obtained by the previous execution in the generation of the next argument input value when a satisfactory solution is not obtained. In the adoption of argument input values, the number of argument input values set in order of the evaluation result from the argument input value with the best evaluation result is adopted, and the vicinity of those argument input values is narrower than the previous time. And an argument input value generating means for generating the input value.

According to a fourth aspect of the present invention, in the optimal design support apparatus according to the second aspect of the present invention, a good evaluation result can be obtained in the previous execution in the generation of the next argument input value when a satisfactory solution cannot be obtained. Argument input value generation that adopts all the argument input values that have obtained evaluation results that are equal to or higher than the set evaluation criteria when adopting the argument input values that have been set, and generates the vicinity of those argument input values in a smaller range than the previous time Means.

According to a fifth aspect of the present invention, in the optimal design support apparatus according to the second aspect of the present invention, in the next generation of the argument input value when a satisfactory solution is not obtained, the vicinity of the adopted argument input value is set to the previous value. When the generation is performed in a finer range, an argument input value generating means for eliminating duplication of the argument input value of the same value is provided.

According to a sixth aspect, in the optimal design support apparatus according to the fourth aspect, a good evaluation result can be obtained by the previous execution in the generation of the next argument input value when a satisfactory solution cannot be obtained. When adopting all of the argument input values that have obtained an evaluation result equal to or greater than the set evaluation criterion in the adoption of the argument input values, the execution of the application program by one combination of the argument input values is completed, and the evaluation result is obtained. Is greater than or equal to the set evaluation criterion, immediately generate an argument input value in the vicinity of that argument input value in a smaller range than the previous time without waiting for the execution of the application program by combining other argument input values. Then, a combination of the generated argument input values is given to an application program to be executed in parallel.

According to a seventh aspect, in the optimal design support apparatus according to the fourth aspect, a good evaluation result can be obtained by the previous execution in the generation of the next argument input value when a satisfactory solution cannot be obtained. When adopting all the argument input values that have obtained the evaluation results that are equal to or higher than the set evaluation criterion, only one argument input value that satisfies the evaluation result that is equal to or higher than the set evaluation criterion is adopted. If it is set by the user that it does not exist, the execution of the application program by the combination of one argument input value ends, and if the evaluation result is equal to or greater than the set evaluation criterion, the other argument input value Abort all execution of the application program by the combination, immediately generate an argument input value in the vicinity of the argument input value in a finer range than the previous time, and transfer the generated combination of argument input values to the application program. Ete is obtained so as to run in parallel.

According to an eighth aspect of the present invention, in the optimal design support apparatus according to the fourth aspect of the present invention, a good evaluation result can be obtained in the previous execution in the generation of the next argument input value when a satisfactory solution cannot be obtained. In the adoption of the argument input values, when all of the argument input values that obtained the evaluation results equal to or higher than the set evaluation criteria were adopted, the argument input values that satisfied the evaluation results equal to or higher than the set evaluation criteria were set. If the user sets that there is only one in the range, the execution of the application program by the combination of one argument input value ends,
If the evaluation result is equal to or greater than the set evaluation criterion, abort all execution of the application program by the combination of other argument input values within the set range, and immediately check the vicinity of the argument input value more closely than the previous time. In this case, an argument input value is generated within a range, and a combination of the generated argument input values is given to an application program to be executed in parallel.

According to a ninth aspect, in the optimal design support device according to the first aspect, when the generated combination of the argument input values is given to the application program and executed in parallel, all the combinations of the generated argument input values are combined. Is not executed in parallel at once, but is executed in parallel for the set number. If the execution is completed and a satisfactory solution is not obtained, the set number of the remaining combinations of the argument input values is set. Only parallel execution is performed.

According to a tenth aspect, in the optimum design support apparatus according to the ninth aspect, there is provided means for setting an execution priority to the argument input value generated in the generation of the argument input value,
When the generated combinations of the argument input values are given to the application program and executed in parallel, all the combinations of the generated argument input values are set in the set execution priority order instead of being executed at once. When the execution is completed and a satisfactory solution is not obtained, it is necessary to execute in parallel the number of combinations of the remaining argument input values set in the set execution priority order. It is intended to be repeated.

According to an eleventh aspect of the present invention, in the optimal design support apparatus according to the first aspect of the present invention, when a satisfactory solution is not obtained, a next argument input value is generated with respect to a combination of previous argument input values. , A crossover, a mutation, and a selection operation, thereby generating an argument input value.

According to a twelfth aspect, in the optimal design support apparatus according to the second aspect, in the generation of the next argument input value when a satisfactory solution is not obtained, the combination of the argument input values having a good evaluation result is used. Arguments with the same value are fixed to the same value, and the other remaining arguments are provided with an argument input value generation means for generating the vicinity of the adopted argument input value in a smaller range than the last time. It is.

According to a thirteenth aspect, in the optimum design support apparatus according to the first aspect, a function is provided in which a correspondence between the argument input value and the evaluation result of the solution obtained by the corresponding application program is graphically displayed on a terminal. Things.

According to a fourteenth aspect, in the optimal design support apparatus according to the first aspect, a function is provided in which a time transition of the best evaluation result among the evaluation results of the solution obtained by the application program is displayed on a terminal in a graph. It was made.

According to a fifteenth aspect, in the optimum design support apparatus according to the first aspect, a function of dynamically changing and setting a method of generating an argument input value given to an application program even during execution of the application program is provided. Things.

According to a sixteenth aspect, in the optimal design support apparatus according to the first aspect, after the set time has elapsed,
This aborts all running application programs, terminates the solution with the best evaluation result at that time as the final solution, and terminates.

According to a seventeenth aspect, in the optimal design support apparatus according to the first aspect, when a satisfactory solution is not obtained, the best evaluation result in the previous execution is stored, and a satisfactory solution is obtained. In the end determination means for determining whether or not the best evaluation result in the current execution is compared with the stored best evaluation result in the previous execution, the ratio is smaller than the ratio set by the user. Then, it is determined that a satisfactory solution has been obtained, and the process is terminated.

According to an eighteenth aspect, in the optimum design support apparatus according to the first aspect, the end determination means for determining whether a satisfactory solution has been obtained, wherein the best evaluation result in the current execution is set by the user. If the value is equal to or greater than the value set, it is determined that a satisfactory solution has been obtained, and the process is terminated.

According to a nineteenth aspect, in the optimal design support apparatus according to the first aspect, all running application programs are aborted by a user's operation, and a solution having the best evaluation result at that time is finally determined. The solution is terminated.

[0028]

BEST MODE FOR CARRYING OUT THE INVENTION

Embodiment 1 FIG. An embodiment of the present invention will be described with reference to FIGS. FIG. 1 is a diagram showing the overall configuration of the optimal design support device and the flow of information in the present embodiment. In the figure, 1 is an execution condition setting unit for designating an application program, a method of generating an argument input value to the program, an end condition, and the like, and 2 is an execution condition and an evaluation of the previous program execution. An input value generation unit for generating an argument input value for the next program execution based on the result, 3 is a parallel execution unit for executing the application program in parallel, and 4 is for determining whether a satisfactory solution has been obtained. An end determination unit for performing the operation, an input value group storage unit in which the program argument input values are stored, an execution condition storage unit designated by the execution condition setting unit 1, a combination of program argument input values, An evaluation result storage unit that stores a set of evaluation values of the execution results obtained by the corresponding application program, 8 is an execution result storage unit of the application program, and 9 is an execution status display that displays the execution status of the apparatus. , And the solid lines the flow of control, and the dashed lines the flow of information.

Next, the operation of the optimal design support apparatus will be described. First, the execution condition setting unit 1 specifies an application program name for performing an optimal design, a method of generating input values of arguments given to the program, and execution conditions such as an end condition of execution. Can be stored. Then, an input value of an argument to be given to the application program is generated according to the input value generation method specified by the input value generation unit 2 and stored in the input value group storage unit 5. Next, the application program specified by the execution condition setting unit 1 is executed by the parallel execution unit 3.
The processing is executed using the argument input values stored in the input value group storage unit 5. At this time, the application program stores the input value group storage unit 5
Are executed in parallel by combining the values of each argument with respect to a plurality of input values for each argument stored in. And
The application program calculates the optimal design according to the combination of the argument input values, stores the obtained solution in the execution result storage unit 8, and obtains an evaluation value representing the quality of the obtained solution.
The evaluation value is stored in the evaluation result storage unit 7 together with the combination of the argument input values.

When the execution of the application program is completed, the end determination unit 4 determines whether or not a satisfactory solution is obtained based on the information stored in the execution condition storage unit 6 and the evaluation result storage unit 7. If a satisfactory solution is obtained, the process is terminated. If not satisfied, the control is transferred to the input value generation unit 2. Then, the execution condition storage unit 6 and the evaluation result storage unit 7
The input value generation unit 2 again generates an argument input value of the application program based on the information accumulated in the application program, and the parallel execution unit 3 executes the application program in parallel. Here, the generation of the argument input value, the execution of the program, the evaluation of the solution, the evaluation of the termination, and the termination determination are called one generation, and the present apparatus performs the optimal design by repeating this generation until a satisfactory solution is obtained. When the end determination unit 4 finds a mark indicating that the execution condition has been changed in the evaluation result accumulation unit 7, it passes control to the execution condition setting unit 1.

As described above, since the input value of the argument of the next execution is determined based on the input value of the argument of the previous program execution and the evaluation value of the corresponding solution, the search for the optimum solution is efficiently performed. Since the execution of the application program can be performed at high speed and the execution of the application program using each argument input value is performed in parallel, an optimal solution can be obtained at high speed.

Next, a method of specifying an execution condition in the execution condition setting section 1 will be described. The execution condition setting unit 1 designates an application program, an evaluation program, information on arguments given to the application program, a method of proceeding with execution, and an end condition.

First, in specifying an application program,
Specify the name and the number of arguments that the application program receives. Also, the name of an evaluation program for evaluating the quality of the execution result output by the application program is specified. In an environment in which the present apparatus operates, these application programs and evaluation programs have their executable files stored in secondary storage, and can execute each program by designating its name. It must be. The format of the execution result output by the application program handled by this device is arbitrary,
The evaluation program must know the format and be able to evaluate the quality of the execution results.
The evaluation value is a single numerical value, and the larger the value, the better the evaluation.

In the specification of the information on the arguments of the application program, whether each argument is individually specified depends on whether a genetic algorithm (hereinafter referred to as GA) is used to generate the argument input value in the next-generation execution. , All arguments are specified at once. Therefore, in the specification of the information on the arguments of the application program, first, it is specified whether or not the GA is used to generate the argument input value in the next-generation execution.

A case where GA is not used to generate an argument input value in the next-generation execution, that is, a case where information is specified for each argument will be described. In this case, for each argument, the data type, the method of generating the initial input value of the argument, and the method of generating the input value in the next-generation execution are specified.

As the data type, an integer or a floating-point number can be specified. In the specification of the generation method of the initial input value of the argument, the following three types of generation methods can be specified. The first is “step”, which specifies the maximum value, minimum value, and step value of the argument input value. In this case, if the argument input value is from the specified minimum value to the maximum value,
Generated to increase by the specified step value. The second is "sequential", which specifies the maximum value, minimum value and number of argument input values. In this case, the specified number of argument input values are generated at equal intervals from the specified minimum value to the specified maximum value. The third is “random”, which specifies the maximum value, minimum value, and number of argument input values. In this case, a specified number of argument input values are randomly generated from a specified minimum value to a specified maximum value. In the specification of the initial input value of this argument, a plurality of specifications can be made by these three types of specification methods.

FIG. 2 shows an example in which a method of generating an initial input value is specified. 2-1 is an example of steps in specifying an initial input value generation method, 2-2 is a sequential example in specifying an initial input value generation method, and 2-3 is specification of an initial input value generation method. It is an example of random in. Note that the data type in these examples is an integer, and a floating-point number can be similarly specified.

FIG. 3 is an example of an input value group generated by the method for generating the initial input value specified in FIG. 3-1 is an argument input value group generated by step designation in 2-1; 3-2 is an argument input value group generated by sequential designation in 2-2; 3-3 is 2-3; It is an example of an argument input value group generated by random designation.

The input value group of the argument generated in this way is
It is stored in the input value group storage unit 5, and by the parallel execution unit 3,
An application program is executed in parallel using these input values. When the application program takes a plurality of arguments, the application program is executed by combining all the input value groups generated for each argument. For example, if the application program takes three arguments, and in FIG. 2 2-1 designates the first argument, 2-2 designates the second argument, and 2-3 designates the third argument, FIG. As shown in FIG. 7, the first argument has seven input values, the second argument has four input values, and the third argument has four input value groups. Therefore, the number of application programs executed by the parallel execution unit 3 is 7 × 4 × 4 = 112.

Next, the specification of the input value generation method in the next-generation execution will be described. When the execution of one generation is completed, an argument input value that has obtained a good evaluation based on the result stored in the evaluation result storage unit 7 is adopted, and a value near the input value is replaced with the argument input value of the next generation. Generate as The specification of the input value generation method in the next generation execution specifies how to adopt the previous generation argument input value and how to generate a nearby value in the next generation. First,
Regarding the adoption of the input values of the previous generation, it is possible to designate from the input value having the best evaluation value in that generation to some values in descending order of the evaluation value, and to designate all the input values having a certain allowable evaluation value or more. In the former designation, since the input is adopted from those which are relatively better than other inputs, the search range of the solution area is steadily narrowed, and there is a high possibility that the target solution can be searched at high speed. On the other hand, in the latter specification, the decision is made based on the absolute allowable evaluation value rather than relative to other input values, so the search range is not narrowed down steadily, but all solutions of a certain level or more are adopted. It is likely that the number of missed solutions will be reduced. When the adoption is specified by the allowable evaluation value, if it is known that there is only one argument input value exceeding the allowable evaluation value, it is also specified.
This one may be one for the entire input value group, or one in each specified range when a plurality of designations are made in the input value generation method.

Next, a range to be searched in the next generation is designated.
This is equivalent to designating a size near the input value adopted in the previous generation. This is performed by designating what percentage of the search range of the next generation is to be searched for in the next generation. That is, the search range of the previous generation is divided by the designated value, and the region including the input value adopted in the previous generation is set as the search range of the next generation. The search range is a range specified by a method of generating an initial input value, that is, a maximum value and a minimum value in step, sequential, and random specification. Then, the next-generation argument input value group is generated in this range by the method specified by the initial input value generation method. In the case of a step, since it is meaningless to generate an input value group with the same step value as the previous generation in the next generation, the step value in the next generation is calculated by dividing the step value of the previous generation by the search range. Divided by.

When there are a plurality of input values adopted in the previous generation, there is a possibility that neighboring ranges for each input value may overlap. In that case, by combining them as one search range, duplicate argument input value groups are prevented from being generated. This makes it possible to prevent useless execution of the application program in the next generation.

FIG. 4 shows an example of specifying an input value generation method in the next-generation execution. 4-1 specifies that only one having the best evaluation value is adopted, and the search range is divided into two in the next generation. 4-2 specifies that the evaluation value is 80 or more, and specifies that the search range is divided into three in the next generation. No. 4-3 specifies that only one having an evaluation value of 90 or more is adopted, and the search range is divided into four in the next generation.

FIG. 5 is an example showing a change in the argument input value group in each generation, that is, a change in the search range. 5-1
Is the sequence (2) indicated by 2-2 in FIG.
In the previous generation argument input value group generated by (0, 8, 4), four input values “8, 12, 16,
20, "narrow (best (1), 2)" indicated by 4-1 in FIG. 4 is specified by the next generation input value generation method. Then, an input value 1 indicated by 5-1-1
Assume that the evaluation value of 2 is the best and is adopted for the search of the next generation. 5-2 shows the result of dividing the search range of the next generation into two. That is, 8 to 20 are divided by 14, and 8 to 14 on the side including the input value 12 adopted in the previous generation is the next generation search range. 5-3 indicates a next-generation argument input value group, that is, four input value groups “8, 10, 12, 14” ranging from 8 to 14, and in the next generation,
The application program is executed by this argument input value.

As described above, the input value having a good evaluation in the previous generation is adopted and its vicinity is searched in further detail. Therefore, the promising search range is narrowed down little by little for each generation based on the execution result of the previous generation. In addition, a detailed search can be performed, and an efficient optimal solution search, that is, an optimal design can be automatically performed. In addition, since several higher-order argument input values having good evaluation values are adopted in each generation, the number of the input values can be changed to control the narrowing of the search range. Furthermore, when there are a plurality of argument input values adopted in the previous generation and their neighborhoods are duplicated, the duplication is removed and a next-generation argument input value group is generated. Execution of various application programs can be prevented. In addition, in each generation, all argument input values having a certain evaluation value or more can be adopted, so that all argument input values having a solution having a certain quality or more can be adopted instead of the number. It is possible to reduce missing.

Next, a case where information is specified for all arguments at once, that is, a case where GA is used to generate an argument input value in the next-generation execution will be described. In this case, for each argument, a data type and a method of generating an initial input value are specified for each argument. These designation methods are the same as when the argument information is designated for each argument. That is, an integer or a floating-point number can be specified for the data type, and "step", "sequential", or "random" can be specified for the method of generating the initial input value. These designations are used to generate the initial population. Note that the individual in the present device is one combination of the argument input values.

Next, parameters necessary for performing the GA operation are specified. The parameters are the number of individuals, the crossover rate and the mutation rate. The number of individuals is the number of individuals in each generation, that is, the number of combinations of argument input values executed in each generation. The crossover rate is the probability of performing a crossover process when two individuals are selected, and the mutation rate is the probability of performing a mutation.

The initial population is generated by first generating an input value group by the initial input value generation method for each argument.
This is performed by randomly generating combinations for the number of individuals from each input value group. The crossover process is performed between two randomly selected individuals, that is, combinations of argument input values.
This is performed by exchanging input values of corresponding arguments selected at random. It should be noted that in the crossover process, instead of simply exchanging input values to generate two children,
A method is conceivable in which the average value of the input values of the corresponding arguments is used as the input value of the child. In this case, only one child is generated by the crossover process. Mutation processing is
This is performed by replacing the input value of the randomly selected argument of the randomly selected individual with a randomly generated value within the range specified by the initial input value generation method.

FIGS. 6 to 8 show examples of crossover and mutation in an application program with five arguments. In FIG. 6, 6
-1 is one parent to perform crossover, 6-1-1 is crossover target 1,
6-1-2 is the crossover target 2, 6-2 is the other parent that performs crossover, 6-2-1 is the crossover target 1, and 6-2-2 is the crossover target 2. In FIG. 7, 7-1 is one child after crossover,
7-1-1 is the mutation target 1, 7-1-2 is the mutation target 2, and 7-2 is the other child after crossover. In FIG. 8, 8-1 is a child after mutation, 8-1-1 is a mutation target 1, and 8-1-2 is a mutation target 2. The method of generating the initial input value of each argument is as follows.
2, the second argument is 2-2 in FIG. 2, the third argument is 2-3 in FIG. 2, the fourth argument is 2-1 in FIG. 2, and the fifth argument is 2-2 in FIG. It shall be generated by the method specified in 3. In the crossover process, a combination of two argument input values,
That is, “8, 16, 9, 20, 15” and “10,
8, 13, 18, and 19 ”, respectively.
And randomly selected argument "16" shown in 6-2-1
7 and “15” and “19” shown in 6-1-2 and 6-2-2, respectively, the combination of these two argument input values becomes
As shown in 1, 7-2, "8, 8, 9, 20, 19"
And "10, 16, 13, 18, 15". In the mutation, “8, 8, 9, 20, 19” shown in 7-1 of FIG. 7, which is a combination of the input values of the arguments, is selected at random for 7-1-1 and 7-1-1. By replacing the argument values “8” and “9” shown in FIG. 2 with values that are randomly generated within the range specified by the method for generating the initial input value, the combination of the argument input values is changed as shown in FIG. It becomes as shown in.

As described above, by using the GA for generating the argument input value in the next generation, the argument input value is generated in a wide range without narrowing down the generation range, so that the so-called local solution is less likely to occur. In other words, it is possible to efficiently search for an optimal solution in a wide range of application problems having various properties represented by multimodality.

After specifying the information on the arguments to be given to the application program, the next step is to specify how to proceed. In the one-generation execution method in the present apparatus, in addition to the method of giving all of the combinations of the argument input values generated by the input value generation unit 2 to the application program in one generation and executing the same in parallel, A method is provided for executing the programs in parallel. In the latter case, in the next generation, instead of generating a new input value, the input value generation unit 2 performs the next generation by using the input value remaining in the input value group storage unit 5. .

In the specification of the execution procedure, the number of combinations of argument input values to be executed at once in one generation is specified. In this specification, "infinite" can be specified in addition to the numerical value. When “infinity” is designated, the combination of all the argument input values stored in the input value group storage unit 5 is 1
It is executed in parallel each time. Also, if the number of combinations of argument input values is
If the number is less than the specified number, all combinations of argument input values are executed in parallel.

Further, when a plurality of designations are made for one argument in the designation of the method of generating the initial input value of the argument, it is considered that the execution priorities are designated in the designated order, and the execution is executed in the priority order. Is performed. When the application program receives a plurality of arguments, the execution priority is obtained by multiplying the priority for each argument.

FIG. 9 shows two arguments for each of the two arguments.
This is an example in which one input value generation method is specified. FIG.
9-1 is the designation of the input value generation method of the first argument.
Is a designation of the input value generation method of the second argument. 9-
1-1 designates a method of generating a first input value of a first argument, 9-
1-2 designates a method for generating a second input value of the first argument, 9-
2-1 designates the first input value generation method of the second argument, 9-
2-2 is designation of a second input value generation method of the second argument. FIG. 10 shows input value groups generated by the input value generation method designated in FIG. 10, 10-1 is an argument input value group generated by the method specified by 9-1-1, 10-2 is an argument input value group generated by the method specified by 9-1-2, 10-3 is 9-
Argument input value group generated by the method specified in 2-1 is an argument input value group generated by the method specified in 9-2-2. The input value group “8, 10, 12” indicated by 10-1 in FIG.
The input value group “16, 18, 20” indicated by 0-2 corresponds to 9 in FIG.
Input value group "0, 4, 8" indicated by -1-2 to 10-3
Are input value groups “2” indicated by 9-2-1 to 10-4 in FIG.
"0, 24, 28" is generated from 9-2-2 in FIG.
Here, the execution priority of each argument is in the order in which the generation method is specified, so that the input value group “8, 10, 12” indicated by 10-1 is the first priority, The input value group “16, 18, 20” indicated by 2 has the second priority, and the second
For the argument, input value group “0, 4, 8” indicated by 10-3
Is the first priority and the input value group “20, 24, 2” indicated by 10-4
8 "is the second priority. FIG. 11 shows all combinations from the input value group of the two arguments in FIG. In FIG. 11, reference numeral 11-1 denotes a combination of argument input values of the first priority;
-2 is a combination of argument input values of the second priority, and 11-3 is a combination of argument input values of the third priority. Since the actual execution priority is obtained by multiplying the priority of each argument, the priority of the combination of the argument input values is represented by the input value group “80, 84, 88,. .., 12 4,12
8 ”is the first priority, and the input value group“ 82 ”indicated by 11-2
., 20 4, 20 8 ”is the second priority, and the input value group“ 16 20, 16 2 ”indicated by 11-3
,..., 20 24, 20 28 ”is the third priority, and when the number of combinations to be executed in one generation is specified, the number of argument input values specified for each generation according to this priority Is given to the application program and executed in parallel

As described above, by providing a function that does not execute all combinations of the argument input values in one generation and executes only a fixed number at the same time, the used storage capacity and CPU
Control according to the execution environment of the present apparatus, such as resource reduction, can be performed, and an efficient optimal solution search can be performed. Further, by providing a function for designating the execution priority to the argument input value, it becomes possible to give the application program in order from the most probable argument input value and execute it, and it is more likely that a satisfactory solution can be searched at high speed. .

Here, a method of controlling the transition to the next generation in the parallel execution unit 3 will be described. In the parallel execution unit 3,
During execution of each application program, the end of the application program corresponding to each combination of argument input values is detected by periodically monitoring the evaluation results stored in the evaluation result storage unit 7. Then, when all the combinations of the argument input values to be executed for the generation are completed, the control is passed to the end determination unit 4. However, if the adoption method of the next-generation input value is “greater than or equal to the permissible evaluation value”, there is no need to wait for all the executions of that generation. be able to. In this case, the parallel execution unit 3 simultaneously executes a plurality of generations of application programs.

If it is specified that there is only one combination of argument input values equal to or more than the allowable evaluation value, and if a combination of argument input values equal to or more than the allowable evaluation value is found, the other arguments of that generation Abort the execution of the application program corresponding to the combination of input values. A normal operating system in an environment in which the present apparatus operates has a function of aborting a running program by designating a process ID. Therefore, by storing the combination of the combination of each argument input value and the ID of the execution process of the corresponding application program corresponding to the combination, it is easy to abort the execution of the application program other than the adopted combination of argument input values. It is.

Further, when a plurality of input value generation methods are specified for one argument, each specification is treated as one group, and if there is only one combination exceeding the allowable evaluation value in that group. May be specified. In this case, when a combination of argument input values equal to or more than the allowable evaluation value is found, the execution of another application program in the group is aborted, and the group immediately shifts to the next generation execution. Also for this realization method, a unique ID is provided for each group, and a combination of the group ID, the combination of each argument input value, and the ID of the execution process of the application program corresponding to the combination is stored in the group. It is easy to abort the execution of the application program other than the combination of the input argument values.

As described above, when the adoption method of the next-generation input value is “more than the permissible evaluation value”, the process immediately shifts to the next-generation execution for the adopted argument input value, so that the same value is always obtained. Of application programs are executed in parallel, so that the number of application programs is less than the number of CPUs available in the parallel execution environment.
U resources can be utilized, thereby also reducing overall execution time. Also, if it is specified that there is only one combination of argument input values equal to or more than the allowable evaluation value, when a combination of argument input values equal to or more than the allowable evaluation value is found, the combination of other argument input values of that generation By aborting the execution of the application program corresponding to (1), the execution of the application program corresponding to a combination of the argument input values that is consequently wasted can be terminated early, and an optimal design with less waste can be performed. Further, by permitting the specification that there is only one combination of argument input values exceeding the allowable evaluation value in a group, a function of adopting an argument input value combination equal to or more than the allowable evaluation value in a wider range of optimal design problems is provided. This makes it possible to apply the effect of using the function, that is, to apply the optimal design with less waste to a wide range of optimal design problems.

Next, the fixing of the argument input value will be described. In this system, when the quality of the obtained solution converges to some extent, among the arguments received by the application program, the input values of some arguments are fixed without changing for each generation, and multiple input values are set only for the remaining arguments. Generate and continue generation execution. This fixing of the argument input value is performed, for example, as follows before the input value generation unit 2 generates a new argument input value. Sort the contents of the evaluation result storage unit 7, that is, the combination of the argument input value and the corresponding execution evaluation value of the application program in descending order of the evaluation value,
Check if there is an argument whose input value is almost the same in the combination of several upper argument input values, and if such an argument exists, fix the value of that argument from that generation, and The input value is generated as before for only the argument of.

As described above, when the quality of the obtained solution converges to some extent, among the arguments received by the application program, the input values of some arguments are fixed without changing for each generation, and only a plurality of arguments are received for the remaining arguments. Is generated, the number of combinations of argument input values is reduced, so that the number of executions of the corresponding application program is also reduced, and it is possible to perform an optimal design at higher speed.

Next, a method of terminating the present apparatus will be described. Several methods for ending the apparatus are prepared, and the specification is performed by the execution condition setting unit 1. First, the first termination method is to terminate when a specified number of generations have been executed. This can be realized by storing the number of generations in the evaluation result storage unit 7 and checking it by the end determination unit 4.

Second, after the designated time has elapsed, the execution of all application programs is aborted and the process is terminated. This is because when the parallel execution unit 3 periodically monitors the evaluation results stored in the evaluation result storage unit 7 in order to detect the end of all application programs in the generation, it checks the elapsed time from the start. By doing so, it can be realized. According to this termination method, the apparatus is always terminated after a certain period of time. For example, it is possible to start the apparatus in the middle of the night and always terminate the apparatus in the next morning. When the execution environment is used by a plurality of users, a usage method that does not bother other users becomes possible.

The third ending method is to end when the change of the best evaluation value for each generation becomes equal to or less than a specified ratio. This can be realized by storing the two best evaluation values of the previous generation and the current generation in the evaluation result storage unit 7 and comparing the two values in the end determination unit 4. According to this termination method, when the quality of the solution converges, the solution can be automatically terminated, and meaningless execution can be suppressed.

The fourth ending method is to end when the best evaluation value reaches a specified value. this is,
This can be realized by storing the best evaluation value of the current generation in the evaluation result storage unit 7 and comparing the value with a designated value in the end determination unit 4. According to this termination method, when a solution of the quality required by the user is obtained, the termination can be automatically terminated, and the meaningless execution can be suppressed.

The fifth termination method is a manual termination by a user, that is, a termination by an interrupt from a terminal or the like. This is because, when an interrupt is received from the terminal, a mark "whole end" is stored in the evaluation result storage unit 7, and the parallel execution unit 3 periodically detects the end of all the application programs in the generation. Storage unit 7
This can be realized by checking whether or not this mark is present when monitoring the evaluation results stored in.
According to this termination method, the present device can be terminated at any time by the user's will.

Next, the display of the execution status of the apparatus will be described. The execution status display unit 9 displays the execution status on the terminal by periodically referring to the evaluation results stored in the evaluation result storage unit 7 independently of the execution flow of the present apparatus. There are two types of execution status: the argument input value, the corresponding evaluation value, and the best evaluation value for each generation. These two
The type is updated at regular intervals and displayed on the terminal.

FIG. 12 is a display example of the argument input value and the corresponding evaluation value. The horizontal axis represents the argument input value and the vertical axis represents the evaluation value. This display allows the user to see in real time the correlation between the input value and the evaluation value, such as what input value the evaluation value is good, during the execution of the present apparatus.

FIG. 13 is a display example of the best evaluation value for each generation. The horizontal axis indicates the number of generations and the vertical axis indicates the evaluation value. With this display, it is possible to see in real time the convergence state of the solution for each generation, such as how much the execution result is improved for each generation, during the execution of the present apparatus.

Further, in the execution status display section 9, the execution can be interrupted and the execution conditions can be changed by the operation of the user. This is because the execution condition change mark is written in the evaluation result storage unit 7, and the evaluation results stored in the periodic evaluation result storage unit 7 are periodically stored in the parallel execution unit 3 in order to detect the end of all application programs in the generation. When monitoring the execution condition, if a mark indicating that the execution condition has been changed is found, control is passed to the end determination unit 4. If the end determination unit 4 also finds a mark indicating that the execution condition has been changed from the evaluation result storage unit 7, the execution condition setting unit 1 It can be realized by passing control to. When the parallel execution unit 3 finds a mark indicating that the execution condition has been changed, it is also possible to abort the execution of all the application programs.

As described above, by the operation of the user,
Since the setting of the execution condition can be changed, the user can change the optimal design policy such as the method of generating the argument input value while observing the execution status, so that the efficient optimal design can be performed.

[0072]

Since the present invention is configured as described above, it has the following effects. According to the present invention, the name of an application program to be designed based on a given combination of argument input values, information on arguments, how to proceed with execution,
An execution condition setting means for setting an end condition, and an argument input for generating a plurality of combinations of argument input values to be given to the application program based on the set execution conditions and an evaluation result indicating the quality of the solution obtained by the previous application program. Value generation means, parallel execution means for giving a combination of a plurality of generated argument input values to an application program and executing the program in parallel, and whether a satisfactory solution has been obtained after execution of all the application programs has been completed. End determination means for judging the condition. If a satisfactory solution is obtained, the processing is terminated. If a satisfactory solution is not obtained, an argument input value is generated again by the argument input value generation means, and execution of the application program is repeated. Therefore, based on the input value of the argument of the previous program execution and the evaluation value of the corresponding solution,
Argument input values for the next execution can be determined, the optimal solution can be searched efficiently and quickly, and the execution of the application program using each argument input value can be performed in parallel with respect to the execution of the application program. Therefore, an optimal solution can be obtained at high speed.

In the generation of the next argument input value when a satisfactory solution is not obtained, only the argument input values for which a good evaluation result was obtained in the previous execution are adopted, and the neighborhood of the adopted argument input values is determined. Providing an argument input value generation unit that generates a combination of argument input values by dividing it in a smaller range than the previous time, narrows down the promising search range a little at each repetition of each execution based on the previous execution result In addition, a detailed search can be performed, and an efficient optimal solution search, that is, an optimal design can be performed.

In the generation of the next argument input value when a satisfactory solution was not obtained, the best evaluation result was obtained in adopting the argument input value that obtained a good evaluation result in the previous execution. Argument input value that adopts the number of argument input values set in order of good evaluation result from the argument input value and divides the vicinity of those argument input values into a smaller range than the previous time to generate a combination of argument input values By providing the generation unit, the number of the argument input values can be set by the user, so that the search range can be narrowed down.

In the generation of the next argument input value when a satisfactory solution is not obtained, when an argument input value that has obtained a good evaluation result in the previous execution is adopted, an evaluation that is higher than a set evaluation criterion is adopted. Adopted by adopting all the argument input values for which the results were obtained, and by providing an argument input value generation means for generating a combination of the argument input values by dividing the vicinity of those argument input values into a smaller range than the previous time. Instead of the number of argument input values to be executed, all of the argument input values having a certain quality or more can be adopted, and the number of missed solutions can be reduced.

In the next generation of the argument input value when a satisfactory solution is not obtained, when the neighborhood of the adopted argument input value is generated in a smaller range than the previous time, the same argument input value is generated. By providing an argument input value generating means for eliminating duplication of the above, execution of useless application programs can be prevented.

In the generation of the next argument input value when a satisfactory solution is not obtained, when an argument input value that has obtained a good evaluation result in the previous execution is adopted, an evaluation value equal to or higher than the set evaluation criterion is used. When all of the argument input values for which the result has been obtained are adopted and the execution of the application program by one combination of the argument input values is completed and the evaluation result is equal to or greater than the set evaluation criterion, the other argument input values are used. Without immediately waiting for the execution of the application program by the combination of the above, the vicinity of the argument input value is immediately divided into a smaller range than the previous time to generate the argument input value, and the generated combination of the argument input values is given to the application program. By executing the programs in parallel, the same number of application programs are always executed in parallel, so that the number of application programs is less than the number of CPUs available in the parallel execution environment. Ri, it is possible to efficiently utilize the CPU resource, and thereby are shortened overall execution time.

In the generation of the next argument input value when a satisfactory solution is not obtained, when the argument input value for which a good evaluation result was obtained in the previous execution is adopted, an evaluation value that is equal to or higher than the set evaluation criterion is adopted. If it is set by the user that there is only one argument input value that satisfies the evaluation result equal to or higher than the set evaluation criterion by adopting all the argument input values for which the result is obtained, one argument input value When the execution of the application program by the combination is completed and the evaluation result is equal to or greater than the set evaluation criterion, all executions of the application program by the combination of other argument input values are aborted, and immediately near the argument input value. Is divided into smaller ranges than the previous time to generate argument input values, and the generated combination of argument input values is given to an application program and executed in parallel, resulting in wasteful use. Can terminate the execution application program corresponding to the combination of several input values to the early, it is possible to less wasteful optimal design.

In the generation of the next argument input value when a satisfactory solution is not obtained, when the argument input value for which a good evaluation result was obtained in the previous execution is adopted, an evaluation that is higher than the set evaluation criterion is adopted. If the user adopts all of the argument input values for which the results have been obtained and the user has determined that there is only one argument input value within the set range that satisfies the evaluation result that is equal to or greater than the set evaluation criteria, When the execution of the application program by the combination of one argument input value is completed and the evaluation result is equal to or greater than the set evaluation criterion, the execution of the application program by the combination of other argument input values within the set range Are immediately aborted, the neighborhood of the argument input value is immediately divided into smaller ranges than the previous time to generate the argument input value, and the generated combination of the argument input values is given to the application program and executed in parallel. By doing so, it becomes possible to use the function of adopting a combination of argument input values that is equal to or greater than the allowable evaluation value in a wider range of optimal design problems, and the effect of using that function, that is, the optimization with less waste The design can be applied to a wide range of optimal design problems.

When the generated combinations of the argument input values are given to the application program and executed in parallel, not all the generated combinations of the argument input values are executed in parallel at once, but in parallel by the set number. When the execution is completed and a satisfactory solution is not obtained, the parallel execution is repeated by the set number of the remaining combinations of the input values of the arguments, thereby reducing the used storage capacity and the CPU resources. Control according to the execution environment of the present apparatus, such as reduction, can be performed, and efficient optimal solution search can be performed.

A means is provided for setting the execution priority to the argument input value generated in the generation of the argument input value. When the combination of the generated argument input values is given to the application program and executed in parallel, all the generated If the set of argument input values are not executed in parallel at once, but are executed in parallel in the set execution priority order and in the set number of times, and the execution is completed and a satisfactory solution is not obtained, By repeating the parallel execution for the set number of combinations of the remaining argument input values in the order of the set execution priority, it is possible to give the application program and execute it in order from the most probable argument input values. It is more likely that the solution to be searched can be searched at high speed.

In the generation of the next argument input value when a satisfactory solution is not obtained, by performing genetic operations such as crossover, mutation, and selection on the combination of the previous argument input values, By providing the argument input value generation means for generating the input value, the argument input value is generated in a wide range without narrowing the generation range of the argument input value, so that the so-called local solution is reduced, and It is possible to efficiently search for optimal solutions for a wide range of application problems with various properties represented by sex.

In the next generation of the argument input value when a satisfactory solution is not obtained, among the combinations of the argument input values having a good evaluation result, the argument having the same value is fixed to the value.
By providing an argument input value generating means for generating the vicinity of the adopted argument input value in a smaller range than the previous one for other remaining arguments, the number of combinations of the argument input values is reduced, so that The number of executions of the application program is reduced, and the optimum design can be performed at higher speed.

By providing a function for displaying the correspondence between the argument input value and the evaluation result of the solution obtained by the corresponding application program on the terminal in a graph, the user can input what kind of input value evaluation value is good. The correlation between the value and the evaluation value can be seen in real time during the execution of the present apparatus.

By providing a terminal with a function of graphically displaying the time transition of the best evaluation result among the evaluation results of the solution obtained by the application program, it is possible to determine how much the execution result is improved for each generation. The convergence state of the solution for each generation can be viewed in real time during execution of the present apparatus.

By providing a function for dynamically changing and setting a method of generating an argument input value to be given to an application program even during the execution of the application program, the user can view the execution status and optimize the method of generating the argument input value. Since it is possible to change the design policy, efficient optimal design can be performed.

After the set time has elapsed, all running application programs are aborted, the solution with the best evaluation result at that time is determined as the final solution, and the operation is terminated. Since the apparatus is terminated, for example, it is possible to start the apparatus in the middle of the night and always terminate it in the next morning, so that a plurality of users use the parallel execution environment for operating the apparatus. In this case, a usage method that does not bother other users becomes possible.

When a satisfactory solution is not obtained, the best evaluation result in the previous execution is stored, and in the end determining means for determining whether a satisfactory solution has been obtained,
Compare the best evaluation result in the current run with the stored best evaluation result in the previous run, and if the ratio is less than the ratio set by the user,
By judging that a satisfactory solution has been obtained and terminating the operation, if the quality of the solution converges, the solution can be automatically terminated, and meaningless execution can be suppressed.

In the termination judgment means for judging whether or not a satisfactory solution has been obtained, if the best evaluation result in this execution is equal to or more than the value set by the user, it is determined that a satisfactory solution has been obtained. By judging and ending the operation, if a solution of the quality required by the user is obtained, it can be automatically ended, and it is possible to suppress meaningless execution.

By aborting all the application programs being executed by the user's operation, the solution with the best evaluation result at that time is determined as the final solution, and the operation is terminated.
This device can be terminated at any time by the user's will.

[Brief description of the drawings]

FIG. 1 is a block diagram showing the overall configuration and information flow of the present invention.

FIG. 2 is a diagram illustrating an example in which a method of generating an initial input value is specified.

FIG. 3 is a diagram showing an input value group generated by an initial input value generation method specified in FIG. 2;

FIG. 4 is a diagram showing designation of a method of generating an input value in a next-generation execution.

FIG. 5 is a diagram showing a change in an argument input value group, that is, a change in a search range in each generation.

FIG. 6 is a diagram showing before a combination of argument input values is crossed in a five-argument application program.

FIG. 7 is a diagram showing a state after a combination of argument input values is crossed in a five-argument application program.

FIG. 8 is a diagram showing a state after a combination of argument input values is mutated in a five-argument application program.

FIG. 9 is a diagram showing designation of two input value generation methods for two arguments.

FIG. 10 is a diagram illustrating input value groups generated by the input value generation method designated in FIG. 9;

FIG. 11 is a diagram showing all combinations of two arguments from the input value group of FIG. 10;

FIG. 12 is a diagram showing display of an argument input value and an evaluation value corresponding thereto.

FIG. 13 is a diagram showing the display of the best evaluation value for each generation.

FIG. 14 is a diagram showing a procedure of a conventional optimal design work by a human.

[Explanation of symbols]

 1 execution condition setting unit, 2 input value generation unit, 3 parallel execution unit, 4 end judgment unit, 5 input value group storage unit, 6 execution condition storage unit, 7 evaluation result storage unit, 8 execution result storage unit, 9 execution status Display

 ──────────────────────────────────────────────────の Continued on the front page (72) Inventor Masaru Shiraishi 2-3-2 Marunouchi, Chiyoda-ku, Tokyo Mitsubishi Electric Corporation

Claims (19)

[Claims] An apparatus for supporting, in a distributed parallel environment, an optimal design for finding an optimum combination of values of a plurality of parameters or a combination of values satisfying a certain evaluation criterion, based on a combination of given argument input values. An execution condition setting means for setting the name of the application program to be designed, argument information, execution procedure, and termination conditions, and an evaluation result indicating the set execution conditions and the quality of the solution obtained by the previous application program. Input value generating means for generating a plurality of combinations of argument input values to be provided to the application program, parallel execution means for providing the generated combination of the plurality of argument input values to the application program and executing the application programs in parallel, and all application programs After completion of the execution of the above, there is provided end determination means for determining whether or not a satisfactory solution has been obtained. If a satisfactory solution is obtained, the process is terminated, and a satisfactory solution is obtained. Optimal design support apparatus characterized by repeated execution of the generated application program an argument input value by no If again the argument input value generating means. 2. A method for generating a next argument input value when a satisfactory solution is not obtained, adopting only the argument input value which obtained a good evaluation result in the previous execution, and adopting the adopted argument input value. 2. The optimal design support device according to claim 1, further comprising: an argument input value generation unit that divides a neighborhood of the value into a smaller range than the previous time and generates a combination of the argument input values. 3. In the generation of the next argument input value when a satisfactory solution is not obtained, the best evaluation result is obtained by adopting the argument input value that obtained a good evaluation result in the previous execution. Arguments that adopt a set number of argument input values from the argument input values given in order of good evaluation result and divide the neighborhood of those argument input values into a smaller range than the previous time to generate a combination of argument input values 3. The optimal design support device according to claim 2, further comprising an input value generation unit. 4. A method for generating a next argument input value when a satisfactory solution is not obtained, in adopting an argument input value having a good evaluation result obtained in the previous execution, at least a set evaluation criterion. The method further includes an argument input value generating unit that employs all the argument input values obtained as the result of the evaluation, divides the vicinity of the argument input values into a smaller range than the previous time, and generates a combination of the argument input values. 3. The optimal design support device according to claim 2, wherein: 5. In the next generation of argument input values when a satisfactory solution is not obtained, the neighborhood of the adopted argument input values is divided into a smaller range than the previous time to generate a combination of argument input values. 3. The optimal design support device according to claim 2, further comprising: an argument input value generation unit that removes duplication of the argument input value having the same value when performing. 6. In the generation of the next argument input value when a satisfactory solution is not obtained, when adopting the argument input value which obtained a good evaluation result in the previous execution, the set evaluation criterion or more is used. If all of the argument input values for which the evaluation result has been obtained are adopted and the execution of the application program by one combination of the argument input values is completed and the evaluation result is equal to or greater than the set evaluation criterion, the other arguments are used. Without waiting for the end of execution of the application program by the combination of input values, immediately divide the vicinity of the argument input value into a smaller range than the previous time to generate argument input values, and combine the generated combination of argument input values with the application program. 5. The optimum design support apparatus according to claim 4, wherein the optimum design support apparatus is executed in parallel with the program. 7. In the generation of the next argument input value when a satisfactory solution is not obtained, in the adoption of the argument input value for which a good evaluation result was obtained in the previous execution, the set evaluation criterion or more is used. If the user adopts all the argument input values for which the evaluation result has been obtained, and the user sets that there is only one argument input value that satisfies the evaluation result equal to or higher than the set evaluation criterion, one argument input When the execution of the application program by the combination of values is completed and the evaluation result is equal to or greater than the set evaluation criterion, all executions of the application program by the combination of other argument input values are aborted, and the argument input values are immediately 5. The method according to claim 4, wherein an argument input value is generated by dividing the neighborhood of in a smaller range than the previous time, and a combination of the generated argument input values is given to an application program and executed in parallel. Optimal design support equipment. 8. In the generation of the next argument input value when a satisfactory solution is not obtained, when adopting the argument input value that obtained a good evaluation result in the previous execution, the set evaluation criterion or more is used. Is adopted by the user, and the user sets that there is only one argument input value within the set range that satisfies the evaluation result equal to or higher than the set evaluation criterion. In this case, when the execution of the application program by the combination of one argument input value is completed and the evaluation result is equal to or greater than the set evaluation criterion, the application program by the combination of other argument input values within the set range Aborts the execution of all the arguments, immediately divides the neighborhood of the argument input values into smaller ranges than the previous time, generates argument input values, and gives the generated combination of argument input values to the application program to execute in parallel. 5. The optimal design support device according to claim 4, wherein the optimization is performed. 9. When the generated combinations of argument input values are given to an application program and executed in parallel, instead of executing all the generated combinations of argument input values in parallel at once,
A parallel execution is performed for a set number of times, and when those executions are completed and a satisfactory solution is not obtained, the parallel execution is performed repeatedly for the set number of remaining argument input value combinations. The optimal design support device according to claim 1. 10. An apparatus according to claim 1, further comprising means for setting an execution priority to the argument input value generated in the generation of the argument input value. Rather than executing the generated combinations of argument input values in parallel at one time, they were executed in parallel in the set execution priority order by the set number, and the execution was completed and no satisfactory solution was obtained. 10. The optimal design support apparatus according to claim 9, wherein, in the case, the parallel execution is repeated by the number set in the set execution priority order among the remaining combinations of the argument input values. 11. When generating a next argument input value when a satisfactory solution is not obtained, a genetic operation such as crossover, mutation, or selection is performed on a combination of previous argument input values. 2. The optimal design support apparatus according to claim 1, further comprising an argument input value generating means for generating an argument input value. 12. In the next generation of an argument input value when a satisfactory solution is not obtained, among the combinations of argument input values having a good evaluation result, the argument having the same value is fixed to that value. 3. An optimal design according to claim 2, further comprising: an argument input value generating means for generating an argument input value by dividing a neighborhood of the adopted argument input value into a smaller range than the previous one for the remaining arguments. Support equipment. 13. The optimal design support apparatus according to claim 1, further comprising a function of displaying on a terminal a correspondence between the argument input value and the evaluation result of the solution obtained by the application program corresponding to the argument input value. 14. The optimal design support apparatus according to claim 1, further comprising a function of displaying, on a terminal, a temporal transition of the best evaluation result among the evaluation results of the solution obtained by the application program. 15. The optimal design support apparatus according to claim 1, further comprising a function of dynamically changing and setting a method of generating an argument input value given to the application program even during execution of the application program. 16. After a set time has elapsed, all running application programs are aborted, a solution having the best evaluation result at that time is set as a final solution, and the operation is terminated. 1. The optimal design support device according to 1. 17. When the satisfactory solution is not obtained, the best evaluation result in the previous execution is stored, and the end determining means for determining whether a satisfactory solution has been obtained is determined by an end determination unit. The good evaluation result is compared with the stored best evaluation result in the previous execution.If the ratio is less than the ratio set by the user, it is determined that a satisfactory solution has been obtained, and the operation is performed. 2. The optimal design support device according to claim 1, wherein the process is terminated. 18. An end determination means for determining whether a satisfactory solution has been obtained, if the best evaluation result in the present execution is equal to or more than a value set by a user, a satisfactory solution is obtained. 2. The optimal design support apparatus according to claim 1, wherein it is determined that the operation has been performed, and the operation is terminated. 19. The method according to claim 1, wherein all running application programs are aborted by a user operation, a solution having the best evaluation result at that time is determined as a final solution, and the operation is terminated. Optimal design support device as described.