JP2002007999A - Optimizing method with usage of genetic algorithm - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an optimizing method capable of obtaining a solution with sufficient accuracy without falling into a local solution and not requiring time or skill. SOLUTION: In this optimizing method with the usage of genetic algorithm, a quasi-optimizing problem obtained by simplifying a secondary parameter optimizing problem as an object to be optimized according to a prescribed rule is optimized by the genetic algorithm, and the obtained solution is included in an initial group. Then, the secondary parameter optimizing problem is optimized by the genetic algorithm. Thus, optimization in two steps can be performed.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an optimization method using a genetic algorithm used for optimizing a mounting operation of an electronic component.

[0002]

2. Description of the Related Art In a component mounting process, which is one of the manufacturing processes of a module product, a work plan of a mounting machine, such as a component mounting order on a board and an arrangement of a component supply feeder, greatly affects a tact time. For this reason, work plans are created using various automation programs, but the number of conditions is so large that all of them cannot be taken into consideration, and converge to an optimum value is not sufficient. In addition, although the automation program is modified, it is an operation that can be performed only by a skilled person, and takes a lot of time. When the work plan is obtained by brute force, for example, for a product having 621 mounting points and 109 kinds of parts,
Approximately 10 ⁴⁶⁷ years of calculation time required (Pentium II 300MHz)
And this method is not practical.

[0003]

SUMMARY OF THE INVENTION Japanese Patent Application Laid-Open No. Hei 10-2096
No. 81 proposes a method using a genetic algorithm (hereinafter, referred to as GA) as an electronic component mounting optimization method. GA is an optimization algorithm that simulates the evolution of living things, and expresses an optimization target by an individual having a sequence of numbers (genes) (chromosomes). Optimization is realized by repeatedly performing a genetic operation such as selection, crossover, and mutation on a group composed of a plurality of individuals.
GA has an advantage that the effect of optimization is large with respect to the simplicity of the algorithm.

FIG. 8 shows a general flow of GA. Evaluation: Finding the excellence of individuals (closeness to the optimal value) Selection: Remaining many excellent individuals in the next generation and extinguishing individuals far from the optimal value Crossover: Gene replacement operation between two individuals (parent) Mutation: A gene replacement operation performed in an individual.
The search range of the solution is expanded by mutation. When these work synergistically, a wide solution search can be efficiently performed.

However, when a plurality of variables such as “feeder arrangement” and “mounting order on a board” are combined as in a mounting operation, and these variables influence each other, one genetic Even if an attempt is made to solve directly with an algorithm, there is a problem that a solution tends to fall into a local solution, and a solution with sufficient accuracy cannot be obtained. In addition, it is necessary to calculate the GA by increasing the number of generations (the number of calculations), or to determine what processing is effective by trial and error, and there is a problem that time and skill are required.

An object of the present invention is to provide an optimization method using a genetic algorithm that can obtain a solution with sufficient accuracy without falling into a local solution and does not require time and skill.

[0007]

In order to achieve the above object, according to the first aspect of the present invention, an optimization method using a genetic algorithm is an object to be optimized having two variables that influence each other. A step of obtaining a quasi-optimization problem obtained by simplifying the two-variable optimization problem in accordance with a predetermined rule; a step of optimizing the quasi-optimization problem by a genetic algorithm; Optimizing the optimization problem with a genetic algorithm.

[0008] "Arrangement of feeders" and "order of mounting on board"
When optimizing mutually influencing variables as in, optimization of one variable may adversely affect the other.
If such two variables are optimized at once, they often fall into local solutions. Therefore, in the present invention, as a countermeasure against local solutions, a problem in which variables are put together by a certain rule (sub-optimization problem) is solved by GA, and the sub-optimal solution obtained next is included in the initial population and To solve the above problem (two-variable optimization problem) with GA.
Thereby, it is possible to reach the optimum solution in a short time.

If the GA according to the first aspect is applied to the mounting optimization according to the second aspect, the moving distance or the moving time of the supply means and the XY stage can be optimized. That is, the optimization of the second aspect is to determine the “arrangement of the feeder” and the “mounting order on the substrate” so that the moving distance or the moving time of the feeder and the XY stage is minimized (optimization of two variables). Problem). By applying the rule of “mounting electronic components in the order of feeder placement” to this problem, the “order of placement on the board” can be expressed in the “order of feeder placement”. The problem can be simplified to optimization (quasi-optimization problem). In this way, by applying the two-stage GA to the two-variable optimization problem, a solution with higher accuracy, that is, a solution with a shorter moving time or shorter moving distance, than a case where the two-variable optimization problem is directly solved by the GA. Can be obtained.

[0010]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS FIG. 1 is a schematic diagram of a mounting machine targeted for optimization by a GA according to the present invention. FIG.
An example of an operation timing chart of the mounting machine is shown in FIG. The mounting machine has an XY stage 2 on which a substrate 1 is placed and movable in X and Y directions, and a plurality of feeders 4 loaded with components W.
It has a component supply device 3 for moving a desired feeder 4 in the X direction to a component suction position, and a plurality of suction heads 6, and moves from the suction position to the mounting position to transfer the component W to the feeder 4.
And a rotary head 5 mounted on the substrate 1 by suction from the substrate.

[0011] As shown in FIG.
If the supply feeder and the XY stage do not complete the operation during the rotation of the head, a waiting time occurs, which leads to an increase in tact. The moving time of the supply device 3 is a time for moving the feeder 4 loaded with the next component to be supplied to the suction position, and the moving time of the XY stage 2 is a time for moving the next mounting point of the substrate 1 to the mounting position. is there. These moving times are determined by “arrangement of feeders” and “order of mounting components on a board”. The mounting optimization refers to obtaining the “feeder arrangement” and the “component mounting order on the board” that minimize the waiting time. In other words, the “placement of the feeder” and the “order of component mounting on the board” are determined so that the moving distance or the moving time of the supply device 3 and the XY stage 2 is minimized.

[0012] Since the optimization targets "arrangement of feeders" and "mounting order on the board" in the work plan of the mounting machine are mutually influential variables, optimization of one of the variables adversely affects the other. There is. If such two variables are optimized at once, it often falls into a local solution, making it impossible to obtain an optimal solution. Therefore, in the present invention, as a countermeasure against local solutions, a problem in which variables are put together by a certain rule (sub-optimization problem) is solved by GA, and the sub-optimal solution obtained next is included in the initial population and Proposed a two-stage GA that solves the above problem (two-variable optimization problem) with GA.

FIG. 3 shows an example of an optimization method using a two-stage GA according to the present invention. First, the two-variable optimization problem is simplified to a quasi-optimization problem by a rule (step S
1) After an initial group is formed (step S2), evaluation is performed (step S3). Then, it is determined whether or not the end condition is satisfied (step S4). If not, the elite is stored (step S5), and the selection is performed (step S4).
6), crossover (step S7), mutation (step S7)
Steps such as 8) are repeated until the termination condition is satisfied. If the termination condition is satisfied, the obtained solution is included in the initial group (step S9), and the following second-stage GA is performed. That is, evaluation (step S10), determination of termination condition (step S11), elite storage (step S12), selection (step S13), crossover (step S10)
14), steps such as mutation (step S15) are repeated until the termination condition is satisfied. Then, the finally obtained solution becomes the optimal solution (step S16).

Hereinafter, a specific procedure of the two-stage GA will be described. [Semi-optimization problem] Set the following rules. Rule 1: The same type of components are successively mounted. The mounting order within the same type of components is determined in advance. Rule 2: The electronic components are mounted in the order in which the feeders are arranged. In other words, it can be expressed in the “mounting component order”, and the “feeder arrangement” is equal to the “mounting component order” according to Rule 2. In other words, since the "mounting order on the board" can be expressed by the "feeder arrangement" according to the rules 1 and 2, the variables can be combined into one (feeder arrangement). As described above, the optimization target in the quasi-optimization problem is “arrangement order of feeders”, and this is expressed by a gene. A number (part No.) is assigned to each type of component, and these are arranged in the order of arrangement of the feeders and used as a gene expression.

[Two-Variable Optimization Problem] The "feeder arrangement" and the "mounting order on the board" to be optimized are separately expressed by genes, and these are connected to one chromosome.
“Arrangement of feeders” is the same as described above. The “mounting order on the board” is a gene expression by attaching numbers (mounting point Nos.) To the mounting points on the board and arranging them in the mounting order. FIG. 4 shows an example of gene expression.

[Evaluation Function] The quality of the work plan is determined by the size of the mounting time of one board. Therefore, this was used as the evaluation function. The calculation is performed in consideration of the moving time of the rotary head, the XY stage, and the feeder as described above. [Selection] As a selection method, an expected value selection method was used. This selection method is a method of determining the number of selection in proportion to the evaluation value of each individual. [Crossover] In this case, the "placement of feeders" and the "order of mounting on the board" are directly expressed by genes, and duplication of genes within chromosomes is not allowed. For example, duplication of the “mounting order on the substrate” gene corresponds to an unrealizable work plan of mounting twice at the same point. Therefore, a partial matching crossover, which is a crossover method that does not cause overlap, is adopted. FIG. 5 shows an example of partial coincidence crossover. [Mutation] As in the case of crossover, inversion was adopted for mutation because duplication of genes was not allowed. Inversion is an operation that reverses the order of a randomly selected gene sequence. FIG. 6 shows an example of inversion. [Elite preservation] In order to search for a wider solution, the frequency of crossover and mutation is set high this time, so that excellent individuals may be killed only by selection. Therefore,
Elite preservation, an operation that unconditionally leaves the best individuals of the previous generation in the next generation, was added before selection. The saved elite is replaced with an individual with a lower evaluation value in the next generation. Therefore, the number of individuals does not change between generations.

The following parameters (GA parameters) exist for crossover, mutation, and elite preservation and need to be adjusted appropriately. [Crossover rate] Percentage of individuals to be replaced by crossover in all individuals 100%, all individuals are replaced If the rate is 100% or more, the same individual is mutated multiple times. [Elite Conservation Ratio] Percentage of elite-preserving individuals to all individuals

Next, the results of verifying the operation and effect of the present invention under the following conditions will be shown. [Target product] CATV tuner Number of mounting points: 639 points / parent board Component type: 113 types / parent board [GA parameters] Sub-optimization problem Crossover rate: 100% Mutation rate: 200% Elite conservation rate: 20% 2 variables Optimization problem Crossover rate: 100% Mutation rate: 40% (feeder arrangement) 200% (order of mounting) Elite conservation rate: 20% [Other conditions] Number of individuals: 100 Number of generations: 10,000

The evaluation of the above-mentioned target product was as follows. Table 1 shows the results obtained by performing simulations on the time required for mounting using a conventional GA (one-stage) mounting plan and a 2.2-stage GA mounting plan. As a comparison, a case where it is assumed that no waiting time is caused by the movement of the XY stage and the feeder (it cannot be realized) is also shown.

[0020]

[Table 1]

As is apparent from Table 1, when the two-stage GA of the present invention is used, the mounting time can be reduced by 6.7% as compared with the example using the conventional GA. That is, the conventional G
It can be seen that the solution obtained in A is a local solution, while the two-stage GA has obtained the optimum solution. Also, compared to the case where no waiting time occurs, the difference in the mounting time is less than 1%, and the optimization by the two-stage GA of the present invention is considered to be at a level near the limit. The calculation time required for the quasi-optimized GA was about 20 minutes, and the calculation time required for the two-variable optimized GA was about 25 minutes (however, the CPU was Penti).
um® II 300 MHz).

FIG. 7 shows a comparison of the locus of the XY stage. FIG. 7A shows the trajectory of the XY stage in the mounting operation using the conventional GA (one stage), and FIG. 7B shows the movement of the XY stage in the mounting operation using the two-stage GA of the present invention. It shows a moving trajectory. The thin line indicates a locus where no waiting time occurs when the XY stage moves, and the thick line indicates a locus where a waiting time occurs when moving the XY stage. It can be seen that when the optimization according to the GA of the present invention is performed as compared with the conventional GA, the movement of the XY stage with a waiting time is significantly reduced. There was no waiting time for the feeder movement.

In the above embodiment, as shown in FIG. 1, the mounting machine includes an XY stage 2, a component supply device 3 movable in the X direction, and a rotary head 5 having a plurality of suction heads 6. , But is not limited to this. For example, as described in JP-A-10-209681, the present invention is also applied to a mounting machine in which a component supply device and a substrate are fixed, and a suction head moves in the XY directions to transfer components from the feeder to the substrate. It is possible. However, in this case, the rules for obtaining the sub-optimization problem are different from those in the example of FIG. The optimization method of the present invention is as follows. It is not only applicable to mounting electronic components on a board,
It goes without saying that it can be applied to all other tasks.

[0024]

As is apparent from the above description, claim 1
According to the invention described in (1), a quasi-optimization problem obtained by simplifying a two-variable optimization problem according to a predetermined rule is optimized by a genetic algorithm, and the obtained solution is included in an initial population to generate a two-variable optimization problem. Optimization with a genetic algorithm
Since the optimization is performed in stages, compared to the case where the two-variable optimization problem is directly optimized by the genetic algorithm,
An accurate solution can be obtained without falling into a local solution. In addition, since an accurate solution can be obtained without changing the GA parameter by trial and error, time and skill are not required.

[Brief description of the drawings]

FIG. 1 is a schematic diagram of a mounting machine targeted for optimization by a genetic algorithm according to the present invention.

FIG. 2 is an example of an operation timing chart of the mounting machine.

FIG. 3 is a flowchart illustrating an example of a two-stage optimization using a genetic algorithm according to the present invention.

FIG. 4 is an example of a gene expression.

FIG. 5 is an example of a partial match crossover.

FIG. 6 is an example of an inversion.

FIG. 7 is a comparison diagram of the movement locus of the XY stage.

FIG. 8 is a flowchart showing a general flow of a genetic algorithm.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Substrate 2 XY stage 3 Component supply apparatus (supply means) 4 Feeder 5 Rotary head (mounter) 6 Suction head

Claims (2)

[Claims] 1. An optimization method using a genetic algorithm, comprising the steps of obtaining a quasi-optimization problem obtained by simplifying a two-variable optimization problem having two variables interacting with each other according to a predetermined rule; Optimizing an optimization problem with a genetic algorithm, and optimizing a two-variable optimization problem with a genetic algorithm by including the obtained solution in an initial population. . 2. A supply means comprising a plurality of feeders which are movable in one horizontal direction and supply electronic components, a substrate arranged on an XY stage movable in two horizontal directions, and a suction position. A mounting head that reciprocates between a mounting position and a suction head that sucks an electronic component from a feeder of a supply unit at a suction position and mounts the electronic component on a substrate at the mounting position; The variable optimization problem is to find the order of mounting on the board and the order of placement of the feeders. The above rule is to mount electronic components in the order of placement of the feeders. 2. The optimization method according to claim 1, wherein the order is determined.