JPH11227931A - Conveying order deciding method and its device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To enable the conveying order of articles to be decided in a short time regardless of the number of the articles in the conveying order to be generated. SOLUTION: When the conveying order is decided in a manufacturing device having a conveying means and a working means, a means 21 to input a known conveying order to a parallel computer to carry out a genetic algorithm as one of initial solution, and a program to carry out the genetic algorithm on the parallel computer, stores the initial conveying order into a conveying order preserver 24, and the result of a crossover operation carried out in a crossover operation part 25 by making the initial conveying order as an input is stored again in the conveying order preserver 24, so as to carry out the totalization and the selection of the preserved conveying order in a selection and totalization execution part 26. In this case, the selection and totalization operations of a crossover operation by plural conveying orders, and a crossover operation by only a single conveying order, are carried out at a time and in parallel in the parallel computer, so as to reduce the deciding time.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a scheduling method and apparatus for determining an optimal transfer order in a manufacturing apparatus by a genetic algorithm using a parallel computer.

[0002]

2. Description of the Related Art Conventionally, when a transfer order and the like are determined using a genetic algorithm, a product is shipped through various processing steps such as a fine processing step and an inspection step for forming an LSI on a silicon wafer. In a semiconductor production line, the in-process products for each lot are transported to and stored in a storage shelf such as a cassette so that a plurality of processing process facilities can operate efficiently without waste according to the production plan, and then stored in the next processing process. An optimal solution for scheduling the wafer transfer order in the case where the transfer is performed again is obtained by using a genetic algorithm. As a specific procedure of the genetic algorithm, a plurality of genes generated by replacing a gene composed of a bit string of [0, 1] with a model of a “problem” such as a transfer order so that the genetic algorithm can solve the problem. A parent gene is selected by selecting and selecting two genes. Based on the determined crossover position on the parent gene, the two parent genes are each divided at the same crossover position, and a part of the gene is given to the child gene, and the crossover position is given. In this method, two parent genes are crossed to generate a new child gene by, for example, changing the child giving the boundary to a new child gene. The computer that actually executes the genetic algorithm determines the initial transfer order, and sequentially performs the crossover operation on the repeatedly executed transfer order.

[0003]

However, in the above-mentioned conventional example, since the crossover operation is performed sequentially,
In general, in the case of a sequential operation, the transfer order determination time increases in proportion to the square of the number of individuals in the generated transfer order, and there is a problem that the reliability decreases due to an increase in the operation time. there were. Also, other means from the same sense, for example, operator input based on a partially decentralized rule of thumb,
Alternatively, the total processing time according to the transfer order determined by referring to the actual past history may be better than the total processing time of the transfer order determined using the computer in all cases. There was a problem. Therefore, the present invention provides a transfer order determining method capable of generating a transfer order equivalent to or better than the transfer order input as a reference initial value and determining the transfer order in a time independent of the number of individuals in the transfer order. It is intended to provide.

[0004]

In order to achieve the above-mentioned object, the invention according to claim 1 is a method for controlling a product transfer order in a manufacturing apparatus having a product transfer means and a plurality of product processing means by using a genetic algorithm. A method of determining a transfer order determined by a parallel computer to be executed, the method including storing the initial transfer order, performing a crossover operation using the stored initial transfer order as an input, storing the crossover operation result again, and storing the stored crossover operation result. In the transfer order determining method for selecting and totaling the transfer order, a known transfer order is input to the parallel computer as one of the initial solutions. According to the above configuration, a known reliable transport order is input as an initial value to the parallel computer as an initial transport order, and the crossover operation is performed so that the total processing time is minimized based on the transport order. Is the same as the total processing time according to the initial transport order,
Alternatively, the transfer order that becomes shorter will be saved as the calculation progresses, and the total processing time according to the finally obtained transfer order will be a somewhat reliable initial value transferred to the computer. Is it the same as the total machining time according to the order,
Or it can be shorter. According to a second aspect of the present invention, in the method of determining a transport order according to the first aspect,
It is characterized in that the calculation of the plurality of transfer orders by the crossover operation and the calculation of the only transfer order by the selection / aggregation are performed in parallel on the parallel computer. According to a third aspect of the present invention, there is provided an apparatus for realizing the method according to the second aspect, wherein a genetic algorithm is used to determine a product transfer order in a manufacturing apparatus having a product transfer means and a plurality of product processing means. In the transfer order determination device determined by the parallel computer, the parallel computer may include a transfer order storage unit that stores an initial transfer order and a result of a cross operation, and a transfer order that receives the initial transfer order stored in the transfer order storage unit as an input. A crossover operation means for executing the crossover operation and storing the transfer order as a result of the crossover operation again in the transfer order storage section, and selecting and removing the transfer order stored in the transfer order storage section.
Selection / aggregation execution means for counting, and calculating the plurality of transport orders by the crossover operation means;
It is characterized in that the calculation of the sole transfer order by the tallying execution means is executed simultaneously and in parallel. According to the above configuration, since the plurality of crossover operations are all executed simultaneously and in parallel on the parallel computer, the transfer order determination time is independent of the number of transfer orders generated by the computer, and the operation time of each genetic operation is Can be related only to the number of repetitions.

[0005]

Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is a configuration diagram of a semiconductor manufacturing system in which a method of determining a transfer order according to an embodiment of the present invention is demonstrated. FIG. 2 is a conceptual block diagram of the transport order determining method in FIG. FIG. 3 is a flowchart of the process of the initial transfer order input / command unit shown in FIG. FIG. 4 is a flowchart of the process of the crossover execution unit shown in FIG. FIG. 5 is a flowchart of the process of the selection / aggregation execution unit shown in FIG. FIG. 1 shows an example of a semiconductor manufacturing system as an example of a manufacturing apparatus. However, of course, the invention is not limited to semiconductor manufacturing. This semiconductor manufacturing system includes a semiconductor manufacturing apparatus 11, a controller 12 for controlling the semiconductor manufacturing apparatus 11, and a parallel computer 13 for executing a genetic algorithm. Further, the semiconductor manufacturing apparatus 11 includes a transfer robot 14
, A cassette 15 for storing semiconductors, and three semiconductor processing devices 16, 17, 18. First, the parallel computer 13 determines the transfer order, outputs the transfer order to the controller 12, and the controller 12 controls the semiconductor manufacturing apparatus 11 including the transfer robot 14 in accordance with the commanded transfer order. doing. The transfer order in this case refers to the semiconductor transfer order of the transfer robot 14 inside the semiconductor manufacturing apparatus 11. Each semiconductor is taken out of the cassette 15 and is firstly conveyed to the processing device 16 and then to the processing device 17 or the processing device 18. After the processing is completed, the semiconductor is returned to the cassette 15 and stored. The movement of the transfer robot 14 in this case is (1) transfer from the cassette 15 to the processing device 16. (2) Transport from the processing device 16 to the processing device 17 or the processing device 18. (3) The cassette 1 from the processing device 17 or the processing device 18
Transport to 5. Can be considered. Each semiconductor (wafer group for each lot as a work in process) must basically be processed in the order of transport (1), (2), (3), but of course transport between different semiconductors There is no such order restriction. The most typical case is the case of an ASIC manufacturing line or the like. However, it is possible to select whether to execute the transfer (3) for a certain semiconductor A or to execute the transfer (1) for a different semiconductor B. . This selection, that is, the transfer order affects the total processing time, and is reflected in the operation efficiency of the production line and the product cost. Therefore, the method of determining the transfer order according to the present embodiment is for scheduling of the transfer order so as to reduce the total processing time. Next, in FIG. 2, a block 21 is an initial transfer order input / command unit for receiving an input of a known initial transfer order and instructing the block 22 to generate the initial transfer order. The block 22 is a semiconductor manufacturing apparatus model section, which is a part that models a semiconductor manufacturing apparatus to be scheduled, and includes robots, processing apparatuses, cassettes, and semiconductor objects, like the actual semiconductor manufacturing apparatus. , And is controlled from the initial transfer order input / command unit. That is, based on data such as whether a production plan is a single-product large-quantity LSI or a multi-product, small-quantity LSI by ASIC, the quantity of semiconductors and the configuration of the processing equipment are input and modeled, and the semiconductor (here, silicon The achievable transfer order of the wafer (wafer) is replaced with, for example, a gene of a genetic algorithm composed of an 8-bit string such as [1101O110] (other conversion methods may be used), and generated as an initial transfer order. In the parallel computer 13, various operations such as a crossover operation are executed based on the transfer order (gene) or the like replaced by the bit string. Block 23 is a genetic algorithm execution unit, which is a main operation part of the parallel computer 13. Block 24 is a transfer order storage unit, which is a memory portion for storing the initial transfer order and the transfer order generated by the crossover operation. A block 25 is a crossover operation execution unit, which inputs the transfer order stored in the transfer order storage unit 24, executes the crossover operation on the transfer order, and outputs the generated transfer order to the transfer order storage unit 24 again. And save. 26 is selection
The aggregation execution unit evaluates the priority of the transfer order stored in the transfer order storage unit 24, deletes the transfer order that is inferior in evaluation, and saves the transfer order that is inferior, thereby generating a huge amount of data. Selection / selection of transport order
It is a selection department. Here, the transfer order that is superior to each other at the time of each evaluation is stored as a provisional best transfer order, and is output in response to the output request of the best solution.

Next, the operation will be described with reference to the flowcharts of FIGS. 3, 4 and 5. First, the initial transfer order input / command unit 21 inputs and accepts a known transfer order having a certain degree of reliability, and refers to the generation of the initial transfer order. As shown in FIG. 3, a storage area is newly secured in the transfer order storage unit 24 (S30), a request is made to the semiconductor manufacturing apparatus model unit 22 to generate an initial transfer order, and the generated initial transfer order is transferred to the transfer order storage unit. 24 (S31), and the storage area of the stored transport order is write-protected (S32). By repeating this (S33) to (S36), is the creation of the entire initial transport sequence completed? Is confirmed (S37), and the input process of the initial transport order is terminated. Next, from the transfer order stored in the transfer order storage unit 24, the plurality of crossing operation executing units 25 immediately execute the calculation and generate a new transfer order. As a specific procedure for this, as shown in FIG. 4, two storage areas for crossover operation are secured in the transfer order storage unit 24 (S4).
0), two previously stored transfer orders are selected and read from the transfer order storage unit 24 (S41). A crossover operation is executed by the crossover execution unit 25 using the crossover position preset for the two read transfer orders (S42). The transfer order of the child newly generated by the crossover operation is written in the storage area of the transfer order storage unit 24 (S43). In parallel with the operation of the crossover operation execution unit 25, the selection / aggregation execution unit 26 also starts execution and is performed as shown in FIG. First, the transport order storage unit 2
The transfer order stored in No. 4 is selected and counted based on the set value (S50). One transfer order is read from the transfer order storage unit 24 (S51), and is it better than the stored transfer order? ? Is determined (S52). If the judgment result is good, it is stored as the provisional best conveyance order (S53),
The old provisional best conveyance order is discarded (S54). On the other hand, S
If the determination result in 52 is bad, the read transport order is deleted from the original book of the transport order storage unit 24, and the next transport order is read (S55). This flow ends when there is no more transfer order to be read in the transfer order storage unit 24. As described above, according to the present embodiment, in determining the optimum transfer order using the genetic algorithm, a known transfer order having a certain degree of reliability is set as an initial value, and the total processing time is set accordingly. Is performed, the useless calculation portion when all calculations are performed by the computer can be considerably reduced, and there is no danger that an unexpected calculation result will appear. In addition, the generation of individuals such as the transfer order in the genetic algorithm is also configured to be executed in parallel on a parallel computer, so the decision time can be greatly reduced, and the total processing time is shortened accordingly. System efficiency and product yield are improved. Although the example of the transfer scheduling of the semiconductor production line including the LSI fine process processing steps such as pattern formation, etching, thin film formation, and impurity diffusion has been described above, the transfer order determining method of the present invention is not limited to this. However, it is not limited to the above, and optimal scheduling, such as when production lines other than semiconductors or multiple production bases higher than the production line unit are distributed and cooperatively, or logistics and operation schedules The present invention can be applied to, for example, juling.

[0007]

As described above, according to the present invention,
The parallel computer sets the initial transfer order based on a known transfer order that is somewhat reliable, and performs calculations to minimize the total machining time. The processing time can be equal to or shorter than the total processing time according to the initial transport order, which is the initial value. Further, when the total processing time is the same, the present system can be used as an evaluation means for the input transfer order. Further, since the transfer order determination time is independent of the number of individuals in the transfer order to be created, by generating more transfer orders, a transfer order that is closer to the overall minimum value of the total processing time is determined. There is an effect that can be.

[Brief description of the drawings]

FIG. 1 is a configuration diagram of a semiconductor manufacturing system in which a transfer order determining method according to an embodiment of the present invention is demonstrated.

FIG. 2 is a conceptual block diagram of a transfer order determining method in FIG.

FIG. 3 is a flowchart of a process of an initial transfer order input / command unit shown in FIG. 2;

FIG. 4 is a flowchart of a process of a cross operation execution unit shown in FIG. 2;

FIG. 5 is a flowchart of a process of a selection / aggregation execution unit shown in FIG. 2;

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 11 Semiconductor manufacturing apparatus 12 Controller 13 Parallel computer 14 Transfer robot 15 Cassette 16, 17, 18 Processing apparatus 21 Initial transfer order input / command part 22 Semiconductor manufacturing equipment model part 23 Genetic algorithm execution part 24 Transfer order storage part 25 Crossover operation execution Part 26 Selection and tabulation execution part

Claims (3)

[Claims] 1. A transfer order determining method for determining a transfer order of a product in a manufacturing apparatus having a product transfer unit and a plurality of product processing units by a parallel computer that executes a genetic algorithm, wherein the initial transfer order is Remember
A crossover operation is performed using the stored initial transfer order as an input, the crossover operation result is stored again, and the stored transfer order is selected and counted. Is input as one of the initial solutions. 2. The method for determining a transfer order, wherein the plurality of transfer order operations by the crossover operation and the sole transfer order operation by the selection / aggregation are simultaneously executed in parallel on the parallel computer. 2. The method according to claim 1, further comprising the steps of: 3. A transfer order determining apparatus for determining a transfer order of a product inside a manufacturing apparatus having a product transfer unit and a plurality of product processing units by a parallel computer executing a genetic algorithm, wherein the parallel computer comprises: A transfer order storage unit that stores a transfer order and a crossover operation result, and performs a crossover operation of the transfer order with the initial transfer order stored in the transfer order storage unit as an input, and again executes the transfer order that is the crossover operation result. Cross-operating means for storing in the transfer-order storing unit, and selection and totaling execution means for selecting and counting the transfer order stored in the transfer-order storing unit; and An apparatus for determining a transfer order, wherein the calculation and the calculation of the sole transfer order by the selection / aggregation execution means are performed simultaneously in parallel.