JP4494909B2 - Feeder layout optimization method for component mounters - Google Patents

Description

  The present invention relates to a feeder placement optimization method for a component mounting machine, and more specifically, includes a suction head equipped with a plurality of suction nozzles, and sucks the components supplied from the feeder with the suction nozzles and mounts them on a circuit board. The present invention relates to a feeder arrangement optimization method for a component mounting machine.

  In an electronic component mounter that mounts electronic components (hereinafter simply referred to as components) on a circuit board, board production (component mounting) is performed by creating a production program for producing the board for each board type. Each production program includes various data for producing a board on a mounting machine, for example, data about a board, data about a mounting position, data about a part (for example, height and width dimensions), data about a suction position, image recognition, etc. Information, data on application of adhesive, and the like.

  Conventionally, production programs have been optimized in order to improve substrate production efficiency. For example, the order of component adsorption and mounting has been optimized to shorten production tact. In this production program optimization process, as shown in FIG. 13, after the production program is read (step S1) and the feeder arrangement is determined (step S2), the suction loading order is generated for the arrangement (step S3). ), And the result is an execution path of outputting the result as a cycle table to a file (step S4).

  In addition, the production tact time is also shortened by linearizing the path through which the head unit that picks up the component moves from the component supply unit to a predetermined position on the circuit board (Patent Document 1).

Also, multiple random arrays of feeders that supply parts are generated, and each of these arrays is used as an individual.Each individual is evaluated using a genetic algorithm, and each individual is evaluated. As an optimal feeder arrangement, component mounting is also optimized (Patent Documents 2 and 3).
JP 2001-94295 A JP 10-209681 A JP 2000-261190 A

  In the case of the conventional processing, the feeder arrangement processing and the generation of the suction mounting order are performed independently. For this reason, there is a case where the suction placement order as intended by the feeder placement is not generated, or the feeder placement is not optimal for the suction placement order. Further, each of the feeder arrangement and suction loading order processing is executed only once, and the feeder arrangement is not reviewed again after the generation of the suction loading order is completed.

  Therefore, there is room to arrange feeders to increase the number of simultaneous adsorptions and reduce the total number of adsorptions when the human thinks about replacing feeders from the output result of the cycle table after optimization is completed. Was.

  In addition, when all permutations are searched in the brute force order, the best arrangement for the suction loading order can be obtained, but such an algorithm causes a combination explosion. For example, if there are 40 8mm tape feeders, the number of permutations is 40! Exists on a 100 MIPS computer! The calculation time is as shown in the table of FIG. 14 (1 air age = 15 billion years).

  The enormous number of combinations as described above, and the method of searching all combinations and searching for the best solution has a problem that the processing does not end within an appropriate execution time.

  Accordingly, the present invention has been made to solve such problems, and it is an object of the present invention to provide a feeder placement optimization method for a component mounter capable of improving the production cycle of a board.

The present invention
A feeder placement optimization method for a component mounting machine that includes a suction head equipped with a plurality of suction nozzles, and sucks the components supplied from the feeder with the suction nozzles and mounts them on a circuit board.
After determining the feeder arrangement according to a predetermined algorithm, to determine the adsorption mounting order of components with respect to their arrangement, for the feeder arrangement adsorption mounting order is determined in part, each suction nozzle of the suction head during component pickup Find the suction pattern that shows the position of the feeder that supplies the parts to be picked up.
A matrix column is formed according to the type of the suction pattern, the feeder position is set as a matrix row, and the arrangement of feeder positions is expanded into a matrix according to the type of the suction pattern.
When replacing the matrix rows and calculating the number of suctions by each suction pattern for each replacement, and finding the optimal feeder placement based on the results ,
The number of times of adsorption by each adsorption pattern is obtained from the matrix, the matrix rows are replaced so that the sum is minimized, and the feeder arrangement corresponding to the feeder position of the matrix row is set as the optimum feeder arrangement. .

In this case, the exchange of matrix rows, the genetic algorithm is used.

  In the present invention, a matrix column is formed according to the type of the suction pattern, the feeder position is set as a matrix row, and the arrangement of the feeder positions is expanded into a matrix according to the type of the suction pattern. Since the number of suctions is calculated for each change and the optimum feeder placement is determined based on the result, the optimum feeder placement can be obtained with a simple configuration, and the total number of suctions by the suction head can be reduced to reduce the substrate. The production tact can be improved.

  The present invention optimizes the arrangement of feeders for supplying components in a component mounting machine and improves the production tact of a substrate. The present invention will be described in detail below based on the embodiments shown in the drawings.

  FIG. 1 illustrates a configuration in which an electronic component mounting apparatus (component mounting machine) 30 is controlled by a host computer (control apparatus) 26 to produce a board. A plurality of feeders 23 a are attached to the feeder bank 23 of the component mounting apparatus 30. These feeders have (40), (42),. . . . . (60) is attached, and in the case of this embodiment, 11 feeders 23a are arranged. Each feeder stores various electronic components to be mounted on the boards 20 and 25 that are transported along the transport path 24 in the direction of the arrow, and the suction head of the component mounting machine 30 sucks the components from these feeders. It is mounted at predetermined locations on the substrates 20 and 25 to produce a substrate. Usually, since the production of the board is divided and produced by a plurality of mounting apparatuses, another electronic component mounting apparatus 30 ′ is shown in FIG. 1 by a virtual line.

  FIG. 2 shows a detailed configuration of the component mounting apparatus 30, and other component mounting apparatuses installed in the line have the same configuration. The component mounting apparatus 30 includes a CPU 31a that controls the entire component mounting, a ROM 31c that stores various control programs and data, and a control unit (control means) 31 that includes a RAM 31b that stores control data and processing data and provides a work area. have. Further, the component mounting apparatus 30 is provided with a data transmission / reception unit 36 capable of transmitting / receiving data to / from the host computer 26, and a production program transmitted from the host computer 26 is transmitted via the data transmission / reception unit 36. Are received and stored in the data storage unit 35. The control unit 31 includes an X / Y drive unit 32 and a suction nozzle that move the suction head in the X and Y directions according to the production program data transmitted from the host computer 26 and the data input via the data input unit 37. The other drive unit 33 that rotates around the Z-axis direction (height direction) and the suction axis (θ) is controlled.

  As shown in FIG. 3, the suction head 70 includes a plurality of suction nozzles 70a to 70d, is driven by the X / Y drive unit 32, moves to the feeder 23a, and a component 71a supplied from the feeder there. To 71d are sucked simultaneously or sequentially by the suction nozzles 70a to 70d. The sucked posture of each sucked component is recognized by the image recognition unit 34 provided with the camera 34a, and the positional deviation between the component center position and the sucked position and the picked angle deviation are corrected. It is mounted at a predetermined location on the substrates 20 and 25 to be conveyed, and the substrate is produced.

  This board production is executed by repeating a cycle in which the operation from the component suction to the next component suction is one mounting cycle according to the production program. The component data recorded in the data storage unit 35 includes data relating to the substrate, data relating to the mounting position (mounting point), data relating to the component (for example, vertical and horizontal height dimensions), data relating to the suction position, and component mounting accuracy. Data, information for image recognition, and other data necessary for component mounting are recorded for each component.

  FIG. 4 shows the overall flow of the production program optimization execution process. Steps S11 to S13 are the same as steps S1 to S3 in FIG. 13, and the production program is read (step S11). After determining the feeder arrangement according to the algorithm (step S12), the suction mounting order is generated for the arrangement (step S13).

  Next, in step S14, feeder arrangement rearrangement (replacement of feeder positions) using a genetic algorithm (hereinafter referred to as GA) is performed as a method of completing the processing within the execution time and obtaining an approximate solution from the viewpoint of overall optimization. I do. This process is performed after determining the order of picking and mounting parts in step S13, thereby improving the initial solution.

  The flow of feeder rearrangement by GA is shown on the right side of FIG. 4. First, in step S21, a cycle table that is the result of the current empirical optimization method is analyzed, and what kind of adsorption is performed. Extract whether a pattern exists. Here, the cycle table is a table that lists all the mounting cycles until all the planned components are mounted on the board, assuming that the operation from component suction to the next component suction is one mounting cycle. By analyzing the above, the type of the adsorption pattern can be obtained. The suction pattern is a pattern indicating the arrangement (arrangement) of the positions of the feeders that supply the parts to be picked up by the suction nozzles of the suction head at the time of picking up the parts in each mounting cycle.

  FIG. 5 shows an example of the suction pattern extracted from the cycle table. The suction pattern (1) is a pattern when the suction nozzles 70a to 70d of the suction head 70 simultaneously suck the parts at the feeder positions (40), (42), (44), and (46). In the substrate production related to No. 1, this adsorption pattern (1) is shown to occur nine times. In the next suction pattern (2), the suction nozzles 70a to 70c simultaneously suck the parts at the feeder positions (42), (44), and (46), and then the remaining suction nozzle 70d is fed to the feeder position (52). This is a pattern for picking up the parts, and means that this suction pattern (2) occurs once. Hereinafter, the same reading is performed up to the adsorption pattern (7).

  Subsequently, as a result of analyzing the cycle table, if an adsorption pattern as shown in FIG. 5 exists, an adsorption pattern matrix as shown in FIG. 6 is constructed in step S22. In this matrix, matrix columns are formed according to the types of suction patterns (patterns (1) to (7)), the feeder positions are matrix rows, and the feeder positions are arranged in the matrix according to the types of suction patterns. Yes. For example, in the suction pattern (1), since the parts at the feeder positions (40), (42), (44), and (46) are sucked, the feeder position is “1”, and other feeders The position is “0”. In the suction pattern (2), the parts at the feeder positions (42), (44), (46), (52) are sucked, so that the feeder position is “1” and the other positions are “0”. ing. Read the same way below.

  Further, in this matrix, the number of components (the number of mounting points) supplied from the feeder at each position and mounted on the board is described, and nine parts are supplied from the feeder position 40 and mounted on the board, Ten parts are supplied from the feeder position 42 and mounted on the substrate. . . . . It is shown. At this time, the number of times each suction pattern is repeated and suction is performed before all parts are mounted on the board (by completion of board production) is parenthesized as the number of repetitions after each suction pattern number. ing. For example, it is shown that the suction pattern (1) is repeated nine times until the substrate production is completed.

  From this suction pattern matrix, the number of times of suction necessary for producing the substrate, that is, necessary for mounting all components on the substrate is calculated. As a method of calculating the number of times of adsorption, the number of times of adsorption is calculated for each column of the matrix and multiplied by the number of repetitions. In that case, the number of times of suction for each column is calculated, but since suction from adjacent feeders can be performed simultaneously, the numerical value of the continuous part for each part where the numerical value is continuously present in the column The maximum value of is added. In the adsorption pattern (1) of the above example, there is only one continuous portion of a numerical value other than 0, and the maximum value is 1. Therefore, the total number of adsorption times = MAX (1, 1, 1, 1) × 9 = 9 times It becomes. In the adsorption pattern (2), {MAX (1,1,1) + MAX (1)} × 1 = 2 times, and in the adsorption pattern (3), {MAX (1,1) + MAX (1,1)} × 3 = 6. The total number of times of adsorption is calculated in the same manner for other adsorption patterns.

  The total number of adsorptions of the matrix in the above example is 9 + 2 + 6 + 6 + 2 + 6 + 1 = 32 times because it is the total number of adsorptions of each adsorption pattern.

  When rearranging the feeder arrangement, replace the matrix rows, calculate the number of adsorptions of the matrix after replacement by the above method, and select a matrix with a small total number of adsorptions. Make improvements. For example, as shown in FIG. 7, when the matrix rows are exchanged, the adsorption of only the simultaneous adsorption is performed completely in each adsorption pattern, and the total number of adsorption described above is 23 times, and improvement is seen.

  Various arrangements (replacement of matrix rows) of the feeder arrangement as described above are evaluated by the GA, and an optimum solution search is executed. This GA solution improvement (step S24) is performed for each feeder bank (23 in FIG. 1), and this is shown by the loop processing between steps S23 and S25 in FIG.

  FIG. 8 and FIG. 9 show the general flow and processing of the GA processing performed in step S24. In the GA process, first, an initial population is generated (step S31) and generated in a gene pool 80 of sufficient individuals. The initial value of the chromosome is randomly determined according to the problem specific coding. Subsequently, two individuals to be left in the next generation are determined from the gene pool 80 according to a certain selection criterion (step S32). For this selection criterion, for example, roulette selection, tournament selection, rank method or the like is used. Next, the partial gene exchange between the selected parents is performed to generate two children (step S33). This crossover includes one-point crossover, two-point crossover, uniform crossover, and the like. Subsequently, a mutation process is performed to randomly mutate a gene at an arbitrary position with respect to two children obtained by crossover (step S34). This mutation is provided to randomly change the locus of an individual and prevent it from falling into a local solution. As shown in FIG. 9, the next generation gene pool 81 is generated by mutation. Then, an end condition check is performed (step S35), and the process ends when the planned generation has ended or when no improvement in solution has been observed for a certain period of time. If the termination condition is not satisfied, the new gene pool 81 is updated and the process is repeated from step S32.

  In the present invention, feeder sequences are expressed as genetic codes (individuals), and generational change operations are performed on them using GA to try to improve the solution.

  The gene coding method is as follows. First, a unique number is assigned to each feeder in the feeder arrangement in the optimization result of the empirical method, and the number of the feeder is stored as a gene code. This state is shown in the upper part of FIG. The permutation of the sequence storing this number represents various feeder arrangements. This genetic code is interpreted as follows.

  First, the sequence storing the gene code is referenced from the top, and the feeders are arranged in the feeder bank (23 in FIG. 1) according to the appearance order. When all the genetic codes are stored, the interpretation of the genetic codes is complete. Thereby, the arrangement of feeder positions (feeder arrangement) as shown in FIG. 1 or FIG. 6 is obtained.

  In the gene evaluation method, an adsorption pattern matrix as shown in FIG. 6 is created from the gene code of FIG. 10, and the total number of adsorptions is calculated and evaluated based on this matrix. The feeder rearrangement process is performed by GA so as to minimize the evaluation value (total number of adsorption times). By generating another genetic code by the process of FIG. 8, for example, if a genetic code as shown in the lower side of FIG. 10 is generated, the feeder arrangement is as shown in FIG. The total number of adsorption is improved from 32 times to 23 times.

  In this way, various gene codes are generated by a genetic algorithm, and each gene code is evaluated for each, and the feeders are rearranged based on the gene codes that have obtained the optimum evaluation value (the total number of adsorptions is small). And optimize feeder placement.

If the above gene coding is used and crossover or mutation is performed easily, two or more of the same numbers will appear, resulting in a state that is not established as a feeder placement solution (lethal gene). Therefore, the above coding is further converted, and the following encoding is performed so that a lethal gene does not appear even by crossover or mutation. This is illustrated in FIG.
(A) In addition to the gene code to be encoded, a reference list is prepared. The genetic code is C = {2, 7, 0, 1, 3, 6, 4, 5}, the reference list is S = {0, 1, 2, 3, 4, 5, 6, 7}, and after encoding Let the list be E = {}. Encoding is performed with information indicating the number in the reference list of the element having the same number as the first element of C, and stored in E.
(B) Since the top element of C is 2, it is located at the second (0th start) of S. Therefore, E = {2}. At the same time, remove the element from C and S.
(C) Repeat (2) with C = {7,0,1,3,6,4,5} and S = {0,1,3,4,5,6,7}.

In the above example, as shown in the lowermost stage of FIG. 11, the final encoding is {2, 6, 0, 0, 0, 2, 0, 0}. This encoded code is used for crossover and mutation. Also for decoding, C can be created from E based on the reference list S. This is illustrated in FIG.
(D) Generate C, which is the original gene code, from the gene code E to be decoded and the reference list S. Since the information in E is position information of the reference list S, the element of S at the position of the head element of E is stored in C. The encoded gene code is E = {2,6,0,0,0,2,0,0}, the reference list is S = {0,1,2,3,4,5,6,7}, The decoded list is C = {}.
(E) Since the top element of E is 2, the second (0th start) of S is stored in C. Therefore, C = {2}. At the same time, remove the element from E and S.
(F) Repeat (2) with E = {6,0,0,0,2,0,0} and S = {0,1,3,4,5,6,7}.

  By the above method, the original gene code is restored. Since the feeder arrangement is thus determined, it is possible to evaluate the solution for each gene code.

It is the top view which showed schematic structure of the component mounting machine. It is the block diagram which showed the control structure of the component mounting machine. It is the block diagram which showed the structure of the suction head provided with the several suction nozzle. It is the flowchart which showed the flow of the whole process flow and the flow which rearranges the feeder by GA. It is a table | surface figure which shows the example of a suction pattern. It is a table | surface figure which shows an adsorption pattern matrix. It is a table | surface figure which shows the adsorption | suction pattern matrix after replacing a feeder position and trying improvement. It is a flowchart which shows the general flow of a process of a genetic algorithm. It is explanatory drawing which shows the detailed flow in the process of a genetic algorithm. It is explanatory drawing which showed rearrangement of the gene code. It is a table | surface figure which shows the encoding of a gene. It is a table | surface figure which shows the decoding of a gene. It is a flowchart which shows the conventional optimization method. It is the table | surface which showed the calculation time by the conventional algorithm.

Explanation of symbols

20, 25 Substrate 23 Feeder bank 23a Feeder 26 Host computer 30 Electronic component mounting device 70 Suction head 70a-70d Suction nozzle 71a-71d Electronic component

Claims (2)

A feeder placement optimization method for a component mounting machine that includes a suction head equipped with a plurality of suction nozzles, and sucks the components supplied from the feeder with the suction nozzles and mounts them on a circuit board.
After the feeder arrangement is determined according to a predetermined algorithm, the order of component suction mounting is determined with respect to the arrangement, and each suction nozzle of the suction head is determined at the time of component suction with respect to the feeder arrangement in which the component suction mounting order is determined. Find the suction pattern that shows the position of the feeder that supplies the parts to be sucked
A matrix column is formed according to the type of the suction pattern, and the feeder positions are arranged in a matrix according to the type of the suction pattern, with the feeder positions being matrix rows.
When replacing the matrix rows and calculating the number of suctions by each suction pattern for each replacement, and finding the optimal feeder placement based on the results ,
The number of times of adsorption by each adsorption pattern is obtained from the matrix, the matrix rows are replaced so that the sum is minimized, and the feeder arrangement corresponding to the feeder position of the matrix row is set as the optimum feeder arrangement. Feeder layout optimization method for component mounters. 2. The feeder placement optimization method for a component mounter according to claim 1, wherein a genetic algorithm is used for replacing the matrix rows.

Images