KR20210061249A - Apparatus and method for scheduling process based on machine learning - Google Patents

Abstract

A machine learning based process scheduling device and a method thereof are disclosed. According to an embodiment of the present invention, the machine learning based process scheduling device comprises: a reference information generating unit for generating reference information including a process lead type and facility preference for products to be produced based on machine learning; a congestion facility selection unit for allocating the products to a plurality of facilities for producing the products by using the reference information and selecting a facility with the highest product production quantity among the plurality of facilities as a congestion facility by a date; and a scheduling unit that calculates difficulty rates of the products allocated to congestion facilities selected by a date and determines a production schedule of the products based on the difficulty rates.

Description

Machine learning-based process scheduling device and method {APPARATUS AND METHOD FOR SCHEDULING PROCESS BASED ON MACHINE LEARNING}

The present invention relates to a process management technology, and more particularly, to a process scheduling technology based on machine learning.

Process management refers to the activity of comprehensively managing all factory activities in order to most efficiently produce products of a certain quality, quantity, and price within a certain period of time in a production factory.

In general, process management means a production process, and the production process is an example of a series of processes performed in the manufacturing process from raw materials or materials to the completion of a product. Planning processes such as design and technical education, labor, machinery, materials, and products It means the manufacturing, work process, such as completion.

However, since the conventional process management system and method establishes process plans and manages performance through human mental activity, it was difficult not only to establish process plans that meet efficiency or process standards, but also process information quickly Since it is not reflected in the product, there is a disadvantage in that rapid work cannot be performed when changes in facilities, manpower, and materials occur, and production scheduling is required.

Production scheduling requires that scheduling rules be adjusted according to the situation according to the status of production facilities and the distribution of orders for each product, so that good quality scheduling results can be expected.

Meanwhile, Korean Patent No. 10-2064490 “Manufacturing and Production Process Management System” is a system that manages parts production and workers so that data related to warehousing, work orders, and process inspections can be checked in real time by computerizing parts production process management. Is being disclosed.

An object of the present invention is to effectively perform process management according to the state of production facilities and distribution of orders received.

In addition, it is an object of the present invention to provide a high quality process scheduling suitable for production equipment.

The machine learning-based process scheduling apparatus according to an embodiment of the present invention for achieving the above object generates reference information for generating reference information including process lead types and facility preferences for products to be produced based on machine learning. part; A congestion facility selection unit that allocates to a plurality of facilities for producing the products using the reference information, and selects a facility with the highest product production amount among the plurality of facilities as congestion facilities by date, and congestion selected by the date. And a scheduling unit that calculates a difficulty rate of products allocated to facilities and determines a production schedule of the products based on the difficulty rate.

In this case, the congested facility selection unit may allocate the products for each of the orders to the plurality of facilities based on an order quantity for each order of the products and the facility preference.

In this case, the congestion facility selection unit may arrange orders for non-input products of the products in order of delivery date, and allocate non-input product orders for the products to the plurality of facilities in the order of delivery date.

At this time, the congested facility selection unit, when there is a residual order that has not been allocated on the first date by exceeding the available order quota of the plurality of facilities, the second date after the first date by using the reference information The remaining order may be preferentially allocated to the most preferred facility among the plurality of facilities.

At this time, the scheduling unit calculates the ratio of the total available time of the congested facility to the product of the product of the order quantity of the products and the unit production time of each product of the congested facility as the difficulty rate, and in the order of the difficulty rate The order of production of these products can be determined.

In this case, the scheduling unit may determine the production order in the order in which the order quantity is large among product orders of the same product among products whose production order is determined in the order in which the difficulty rate is large.

In this case, the scheduling unit may further include, in the unit production time of the first product, a job replacement time that occurs when a production job is replaced from a first product to a second product in the congested facility.

At this time, the scheduling unit is preferred after the most preferred facility if there is an unallocated order that cannot be allocated to the congested facility after the production order of the products is determined according to the difficulty rate in the total available time of the congested facility. The unallocated order can be assigned to the next preferred facility.

In this case, the scheduling unit may determine a production order of the unallocated order by calculating the difficulty rate of products allocated to the sub-preferred facility and the difficulty rate of the unallocated order.

In addition, the machine learning-based process scheduling method according to an embodiment of the present invention for achieving the above object is a machine learning-based process scheduling method of a machine learning-based process scheduling device, Generating reference information including a related process lead type and facility preference; Allocating to a plurality of facilities for producing the products using the reference information, and selecting a facility with the largest production amount among the plurality of facilities by date as a congestion facility; And calculating a difficulty rate of products allocated to the congested facilities selected for each date, and determining a production schedule of the products based on the difficulty rate.

The present invention can effectively perform process management according to the state of production facilities and distribution of orders received.

In addition, the present invention can provide excellent quality process scheduling suitable for production equipment.

1 is a block diagram showing a machine learning-based process scheduling apparatus according to an embodiment of the present invention.
2 is an operation flow diagram showing a process scheduling method based on machine learning according to an embodiment of the present invention.
3 is an operation flow diagram showing in detail an example of the step of generating reference information shown in FIG. 2.
FIG. 4 is a detailed operation flow diagram illustrating an example of a congestion facility selection step shown in FIG. 2.
5 is an operation flow diagram showing in detail an example of the order scheduling step shown in FIG. 2.
6 is a diagram illustrating a process lead time calculation process based on machine learning according to an embodiment of the present invention.
7 is a diagram illustrating a process of calculating a manufacturing cycle time according to an embodiment of the present invention.
8 is a diagram illustrating a process of calculating machine learning-based facility preferences according to an embodiment of the present invention.
9 is a diagram illustrating a process of calculating equipment preference based on past production results according to an embodiment of the present invention.
10 is a diagram illustrating a process of selecting a congestion facility according to an embodiment of the present invention.
11 is a diagram showing a difficult order priority scheduling process according to an embodiment of the present invention.
12 is a diagram illustrating a process scheduling process based on machine learning according to an embodiment of the present invention.
13 is a block diagram showing a computer system according to an embodiment of the present invention.

The present invention will be described in detail with reference to the accompanying drawings as follows. Here, repeated descriptions, well-known functions that may unnecessarily obscure the subject matter of the present invention, and detailed descriptions of configurations are omitted. Embodiments of the present invention are provided to more completely explain the present invention to those with average knowledge in the art. Accordingly, the shapes and sizes of elements in the drawings may be exaggerated for clearer explanation.

Throughout the specification, when a part "includes" a certain component, it means that other components may be further included rather than excluding other components unless specifically stated to the contrary.

Hereinafter, preferred embodiments according to the present invention will be described in detail with reference to the accompanying drawings.

1 is a block diagram showing a machine learning-based process scheduling apparatus according to an embodiment of the present invention.

Referring to FIG. 1, a machine learning-based process scheduling apparatus according to an embodiment of the present invention includes a reference information generation unit 110, a congestion facility selection unit 120, and a scheduling unit 130.

The reference information generation unit 110 may generate reference information including a process lead type and facility preference for products to be produced based on machine learning.

In this case, the reference information generation unit 110 may learn a product-specific process lead time (LEAD TIME, L/D) based on machine learning and calculate a process lead time.

At this time, the reference information generation unit 110 learns the order quantity for each product, the amount of work in progress (WIP) for each product, and the production performance for each product based on the previously stored past production history, and leads the process for each product. You can learn the time and calculate the process lead time.

In this case, the reference information generation unit 110 may calculate the process lead time by using the predefined theoretical lead time when the previously stored production history does not exist.

In this case, the reference information generation unit 110 may correct the process lead time by further considering the manufacturing cycle time (CYCLE TIME, C/T) for each product according to the amount provided.

For example, the reference information generation unit 110 determines the difference between the order quantity up to the first time T1, which is the first process lead time for the first process, and the production quantity up to the first time T1, of the first process. It can be calculated by serving.

At this time, the reference information generation unit 110 determines the difference between the order quantity up to the second time T2, which is the second process lead time for the second process, and the production quantity up to the second time T2, the amount provided by the second process. Can be calculated as

In addition, the reference information generation unit 110 may calculate facility preferences for each product based on machine learning.

In this case, the reference information generation unit 110 may learn facility preferences for each product by learning the amount of work in progress (WIP) and production performance for each product, and calculate facility preferences for each product.

For example, the reference information generation unit 110 has a past production performance of 2,000 for a first facility of a first product, a past production performance of 500 for a second facility, and a past production performance of 0 for the third facility. If the first facility is the most preferred facility (the first preferred facility), the second facility is the second preferred facility (the second preferred facility), and the third facility is the non-preferred facility (third preferred facility). can do.

At this time, the reference information generation unit 110 has a past production performance of 1,000 for the first facility of the second product, 1,500 for a past production performance of the second facility, and 500 for the third facility. In this case, the facility preference is learned and calculated as the first facility as the second preferred facility (the second preferred facility), the second facility as the most preferred facility (the first preferred facility), and the third facility as the second most preferred facility (the third preferred facility) can do.

In this case, the reference information generation unit 110 may learn and calculate the facility preference in the order of the shortest manufacturing cycle time for each product calculated above, if the product does not have a past production record.

The congestion facility selection unit 120 may use the reference information to allocate to a plurality of facilities for producing the products, and select a facility with the largest production amount among the plurality of facilities by date as the congestion facility.

In this case, the congestion facility selection unit 120 may allocate the products for each of the orders to the plurality of facilities based on the order quantity for each order of the products and the facility preference.

In this case, the congested facility selection unit 120 may arrange orders for non-input products of the products in order of delivery date, and allocate non-input product orders for the products to the plurality of facilities in the order of delivery date.

In this case, when there is a residual order that has not been allocated on the first date by exceeding the available order quota of the plurality of facilities, the congestion facility selection unit 120 uses the reference information to On a date, the remaining order may be preferentially allocated to the most preferred facility among the plurality of facilities.

In this case, the congestion facility selection unit 120 may select a facility with the largest order quantity among the facilities to which products are allocated as the congestion facility.

The scheduling unit 130 includes a scheduling unit that calculates the difficulty rate of products allocated to the congested facilities selected for each date and determines a production schedule of the products based on the difficulty rate.

At this time, the scheduling unit 130 calculates the ratio of the total available time of the congested facility to the product of the product of the order quantity of the products and the unit production time of each product of the congested facility as the difficulty rate, as shown in Equation 1, and , It is possible to determine the order of production of the products in the order of the greatest difficulty.

In this case, the scheduling unit 130 may determine the production order in the order in which the order quantity is large among product orders of the same product among products whose production order is determined in the order in which the difficulty rate is large.

In this case, the scheduling unit 130 may further include, in the unit production time of the first product, a job replacement time that occurs when a production job is replaced from a first product to a second product in the congested facility.

At this time, the scheduling unit 130 determines the order of production of the products according to the difficulty rate in the total available time of the congested facility, and if there is an unallocated order that cannot be allocated to the congested facility, the most preferred facility Next, the unallocated order can be assigned to the preferred sub-preferred facility.

In this case, the scheduling unit 130 may determine the order of production of the unallocated order by calculating the difficulty rate of products allocated to the sub-preferred facility and the difficulty rate of the unallocated order.

In addition, the scheduling unit 130 may check whether all order schedules of products allocated on the day of the day have been completed.

At this time, the scheduling unit 130 may request the congestion facility selection unit 220 to recalculate the congestion degree and reselect the congested facility when there is an order for which the production order has not been determined.

At this time, when all order schedules are completed, the scheduling unit 130 checks whether an order exceeding the available quota of the facility exists, and if there is an order that exceeds the available quota without being allocated to the facility, the facility will be installed on the next day. It is possible to request the congestion facility selection unit 220 to set an assignment priority so that it can be assigned first.

2 is an operation flow diagram showing a process scheduling method based on machine learning according to an embodiment of the present invention. 3 is an operation flow diagram showing in detail an example of the step of generating reference information shown in FIG. 2. FIG. 4 is a detailed operation flow diagram illustrating an example of a congestion facility selection step shown in FIG. 2. 5 is an operation flow diagram showing in detail an example of the order scheduling step shown in FIG. 2.

Referring to FIG. 2, the machine learning-based process scheduling method according to an embodiment of the present invention may first generate reference information (S210).

That is, in step S210, reference information including a process lead type and facility preference for products to be produced may be generated based on machine learning.

Referring to FIG. 3, in step S210, a process lead time may be learned (S211 ).

That is, in step S211, a process lead time (LEAD TIME, L/D) for each product may be learned based on machine learning, and a process lead time may be calculated.

At this time, step (S211) learns the process lead time for each product by learning the order quantity for each product, the amount of work in progress (WIP) for each product, and the production performance for each product based on the previously stored past production history. And the process lead time can be calculated.

In this case, in step S211, if there is no past production history previously stored, the process lead time may be calculated using a predefined theoretical lead time.

In this case, in step S211, the process lead time may be corrected by further considering the manufacturing cycle time (CYCLE TIME, C/T) for each product according to the amount provided.

For example, in step S211, the difference between the order amount up to the first time T1, which is the first process lead time for the first process, and the production amount up to the first time T1, is calculated as the amount provided for the first process. I can.

At this time, in step S211, the difference between the order amount up to the second time T2, which is the second process lead time for the second process, and the production amount up to the second time T2, can be calculated as the amount provided for the second process. have.

In addition, in step S210, facility preferences may be learned (S212).

That is, in step S212, facility preferences for each product may be calculated based on machine learning.

At this time, in step S212, by learning the amount of work in progress (WIP) and production performance for each product, the facility preference for each product may be learned, and the facility preference for each product may be calculated.

For example, in step S212, when the past production performance for the first facility of the first product is 2,000, the past production performance for the second facility is 500, and the past production performance for the third facility is 0, The facility preference can be learned and calculated as the first facility as the most preferred facility (first preferred facility), the second facility as the second preferred facility (second preferred facility), and the third facility as non-preferred facility (third preferred facility). .

At this time, in step S212, if the past production performance for the first facility of the second product is 1,000, the past production performance for the second facility is 1,500, and the past production performance for the third facility is 500, Facility preferences can be learned and calculated as 1 facility as the next preferred facility (the second preferred facility), the second facility as the most preferred facility (the first preferred facility), and the third facility as the second most preferred facility (the third preferred facility) .

In this case, in step S212, in the case of a product for which no past production performance exists, the facility preference may be learned and calculated in the order of the shortest manufacturing cycle time for each product calculated above.

In addition, the machine learning-based process scheduling method according to an embodiment of the present invention may select a congested facility (S220).

That is, in step S220, a plurality of facilities for producing the products may be allocated using the reference information, and a facility with the largest production amount among the plurality of facilities may be selected as a congestion facility by date.

Referring to FIG. 4, in step S220, orders may be sorted in order of delivery (S221).

That is, in step S221, orders for non-input products of the products may be arranged in order of delivery date, and non-input product orders for the products may be allocated to the plurality of facilities in the order of delivery date.

In addition, step S220 may allocate an order to a preferred facility (S222).

That is, the step S222 may allocate the products for each of the orders to the plurality of facilities based on the order quantity for each order of the products and the facility preference.

At this time, step (S222), if there is a residual order that has not been allocated on the first date by exceeding the available order quota of the plurality of facilities, the reference information is used on the second date following the first date. The remaining order may be preferentially allocated to the most preferred facility among the plurality of facilities.

In addition, in step S220, congestion facilities may be selected (S223).

That is, in step S223, a facility with the largest order quantity among the facilities to which products are allocated may be selected as the congestion facility.

In addition, the machine learning-based process scheduling method according to an embodiment of the present invention may perform scheduling (S230).

That is, step S230 includes a scheduling unit that calculates the difficulty rate of products allocated to the congested facilities selected for each date and determines a production schedule of the products based on the difficulty rate.

Referring to FIG. 5, in step S230, the difficulty rate may be first calculated (S231).

That is, in step S231, as shown in Equation 1, the ratio of the total available time of the congested facility to the product of the product of the order quantity of the products and the unit production time of each product of the congested facility may be calculated as the difficulty rate.

In this case, step S233 may further include, in the unit production time of the first product, a work replacement time that occurs when the production work is replaced from the first product to the second product in the congested facility.

In addition, step S230 may select a difficult order (S232).

That is, in step S232, an order having the greatest difficulty rate may be selected as the difficulty order.

In addition, in step S230, a production schedule may be determined (S233).

That is, in step S233, the order of production of the products may be determined in the order in which the difficulty rate is large.

In this case, in step S233, the order of production may be determined in the order of the order quantity of products of the same product among the products for which the production order is determined in the order of the increase in the difficulty rate.

At this time, in step S233, after the production order of the products is determined according to the difficulty rate in the total available time of the congested facility, if there is an unallocated order that cannot be allocated to the congested facility, the most preferred facility is followed by As a result, the unallocated order can be assigned to the preferred sub-preferred facility.

In this case, step S233 may determine the order of production of the unallocated order by calculating the difficulty rate of the products allocated to the sub-preferred facility and the difficulty rate of the unallocated order.

In addition, in the machine learning-based process scheduling method according to an embodiment of the present invention, it is possible to check whether all order schedules of products allocated on the day of the day have been completed (S240).

In this case, step S240 may return to step S220 in order to recalculate the congestion level and reselect congestion equipment when there is an order for which the production order has not been determined.

At this time, in step S240, when all order schedules are completed, it is checked whether there is an order that exceeds the available quota of the facility, and if there is an order that exceeds the available quota without being allocated to the facility, the facility is sent to the facility on the next day. Allocation priority can be set so that priority can be assigned.

6 is a diagram illustrating a process lead time calculation process based on machine learning according to an embodiment of the present invention. 7 is a diagram illustrating a process of calculating a manufacturing cycle time according to an embodiment of the present invention.

6, the machine learning-based process scheduling apparatus according to an embodiment of the present invention learns a process lead time (LEAD TIME, L/D) for each product based on machine learning, and calculates a process lead time. have.

At this time, the machine learning-based process scheduling device learns the order quantity for each product, the work in progress (WIP) amount for each product, and the production performance for each product based on the previously stored production history, and the process lead time for each product. Can be learned and the process lead time can be calculated.

In this case, the machine learning-based process scheduling apparatus may calculate a process lead time by using a predefined theoretical lead time when there is no past production history previously stored.

Referring to FIG. 7, the machine learning-based process scheduling apparatus may correct the process lead time by further considering the manufacturing cycle time (CYCLE TIME, C/T) for each product according to the amount provided.

For example, the machine learning-based process scheduling device determines the difference between the order quantity up to the first time T1, which is the first process lead time for the first process, and the production quantity up to the first time T1, the amount provided by the first process. Can be calculated as

At this time, the machine learning-based process scheduling apparatus uses the difference between the order quantity up to the second time T2, which is the second process lead time for the second process, and the production quantity up to the second time T2, as the amount provided for the second process. Can be calculated.

8 is a diagram illustrating a process of calculating machine learning-based facility preferences according to an embodiment of the present invention. 9 is a diagram illustrating a process of calculating equipment preference based on past production results according to an embodiment of the present invention.

Referring to FIG. 8, the machine learning-based process scheduling apparatus according to an embodiment of the present invention may calculate a facility preference for each product based on machine learning.

At this time, the machine learning-based process scheduling device may learn facility preferences for each product by learning the amount of work in progress (WIP) and production performance for each product, and calculate the facility preferences for each product.

9, in the machine learning-based process scheduling device, the past production performance for the first facility of the first product is 2,000, the past production performance for the second facility is 500, and the past production performance for the third facility is In case of 0, the first facility is the most preferred facility (the first preferred facility), the second facility is the second preferred facility (the second preferred facility), and the third facility is the non-preferred facility (the third preferred facility). Can be calculated.

In this case, the machine learning-based process scheduling device has a past production performance of 1,000 for the first facility of the second product, 1,500 for the second facility, and 500 for a third facility. , The first facility is the second preferred facility (the second preferred facility), the second facility is the most preferred facility (the first preferred facility), and the third facility is the second most preferred facility (the third preferred facility). I can.

In this case, the machine learning-based process scheduling apparatus may learn and calculate facility preferences in the order of the shortest manufacturing cycle time for each product calculated above, in the case of a product that does not have a past production record.

10 is a diagram illustrating a process of selecting a congestion facility according to an embodiment of the present invention.

Referring to FIG. 10, the machine learning-based process scheduling apparatus according to an embodiment of the present invention first allocates an order quantity of 500 products A included in WIP on the date of July 1 to the first facility, which is the most preferred facility, and It can be seen that the order quantity 200 of product B included in is allocated to the second facility, and the order quantity 200 of product C included in the WIP is allocated to the second facility.

At this time, since the order quantity of the first facility on July 1 is 500 and the order quantity of the second facility is 400, it can be seen that the first facility with the largest load peak on the date of July 1 is selected as the congested facility. .

In addition, the machine learning-based process scheduling apparatus according to an embodiment of the present invention divides orders with a delivery date on July 4 and orders with a delivery date on July 5 and sorts them in the order of delivery date, respectively. It can be seen that orders are assigned to the facilities on the 3rd production day of the month.

In addition, the machine learning-based process scheduling device allocates 900 orders for Product A, which is the delivery date of July 2 to July 4, to the first facility, which is the most preferred facility, and 1000 for Product B, which is the delivery date for July 4 It can be seen that it is assigned to the second facility, which is the most preferred facility.

At this time, the machine learning-based process scheduling device further allocates 500 of the order quantity of Product C, which is the delivery date on July 4, to the second facility on July 2, taking into account the process lead time for each product of the second facility. If all of the products C are not produced, it is determined that the available quota has been exceeded, and it can be seen that product C is preferentially assigned to the second facility on July 3 instead of July 2nd.

At this time, since the order quantity of the first facility on July 2 is 900 and the order quantity of the second facility is 1000, it can be seen that the second facility with the largest load peak on the date of July 2 is selected as the congested facility. .

In addition, the machine learning-based process scheduling device allocates the order amount of Product A, which is the delivery date of July 3 to July 5, to the first facility, the most preferred facility, and assigns the order amount of Product C, which is the delivery date of July 4, to 500. It can be seen that the order quantity of Product B, which is the delivery date on July 5, is allocated 900, and the order quantity of Product D, which is the delivery date of July 5, is allotted 800.

At this time, the machine learning-based process scheduling device further allocates 500 of the order quantity of Product C, which is the delivery date on July 5, to the second facility on July 3, taking into account the process lead time for each product of the second facility. If not all of product C is produced, it is judged that the available quota has been exceeded, and it can be seen that product C is assigned preferentially to the second facility on July 4 instead of July 3.

At this time, since the order quantity of the first facility on July 3 is 1500 and the order quantity of the second facility is 1400, it can be seen that the first facility with the largest load peak on the date of July 3 is selected as the congested facility. .

In addition, it can be seen that the machine learning-based process scheduling apparatus allocates the order quantity of product C, which is the delivery date from July 4 to July 5, to the second facility, which is the most preferred facility.

11 is a diagram showing a difficult order priority scheduling process according to an embodiment of the present invention.

Referring to FIG. 11, the machine learning-based process scheduling apparatus according to an embodiment of the present invention relates to an order amount of 100 of a first order of a product A selected as a congestion facility, an order amount of 50 of a second order, and an order amount of 100 of a third order. You can calculate the difficulty rate.

In this case, the machine learning-based process scheduling apparatus may calculate the difficulty rate according to Equation 1 described above, and select an order having the largest difficulty rate as the difficulty order.

In this case, the machine learning-based process scheduling apparatus may allocate the difficulty order to the earliest available time of the first facility.

12 is a diagram illustrating a process scheduling process based on machine learning according to an embodiment of the present invention.

Referring to FIG. 12, in the process scheduling process based on machine learning according to an embodiment of the present invention, product orders may be first allocated to preferred facilities, and congested facilities may be selected.

At this time, in the machine learning-based process scheduling process, when the congested facility is selected as the first facility, the difficulty rate for products A to D is calculated, and the order of production of product orders is determined preferentially to the congested facility in the order of the highest difficulty rate. I can.

In this case, the machine learning-based process scheduling process may calculate the difficulty rate by further considering the job replacement time when the production job is replaced from product A to product B.

In this case, the machine learning-based process scheduling process may first determine an order with a large order quantity from the same product as a production order.

At this time, in the machine learning-based process scheduling process, when the scheduling on July 1 is completed, it checks whether there are any remaining orders that have not been completed on the same day, and if there are remaining orders, the order is assigned to the facility on the next day to determine the congestion level of the remaining facilities Can be recalculated.

13 is a diagram showing a computer system according to an embodiment of the present invention.

Referring to FIG. 13, the machine learning-based process scheduling apparatus according to an embodiment of the present invention may be implemented in a computer system 1100 such as a computer-readable recording medium. As shown in FIG. 13, the computer system 1100 includes one or more processors 1110, memory 1130, a user interface input device 1140, and a user interface output device 1150 communicating with each other through a bus 1120. And a storage 1160. Further, the computer system 1100 may further include a network interface 1170 connected to the network 1180. The processor 1110 may be a central processing unit or a semiconductor device that executes processing instructions stored in the memory 1130 or the storage 1160. The memory 1130 and the storage 1160 may be various types of volatile or nonvolatile storage media. For example, the memory may include a ROM 1131 or a RAM 1132.

Machine learning-based process scheduling apparatus according to an embodiment of the present invention includes one or more processors 1110; And an execution memory 1130 storing at least one program executed by the one or more processors 1110, wherein the at least one program is a process lead type and facility for products to be produced based on machine learning. Generate reference information including preferences, use the reference information to allocate to a plurality of facilities for producing the products, select a facility with the highest product production amount among the plurality of facilities by date as a congestion facility, and Then, the difficulty rate of products allocated to the congested facilities selected for each date is calculated, and the production schedule of the products is determined based on the difficulty rate.

In this case, the at least one program may allocate the products for each of the orders to the plurality of facilities based on an order quantity for each order of the products and the facility preference.

In this case, the at least one program may arrange orders for non-delivery products of the products in order of delivery date, and allocate non-delivery product orders for the products to the plurality of facilities in the order of delivery date.

In this case, when there is a residual order that has not been allocated on the first date by exceeding the available order quota of the plurality of facilities, the at least one program uses the reference information to the second date following the first date. The remaining order may be preferentially allocated to the most preferred facility among the plurality of facilities.

In this case, the at least one program calculates a ratio of the total available time of the congested facility to the product of the product of the order quantity of the products and the unit production time of each product of the congested facility as the difficulty rate, and the difficulty rate is It is possible to determine the order of production of the products in large order.

In this case, the at least one program may determine the order of production in the order of the order quantity of products of the same product among the products for which the production order is determined in the order of the highest difficulty rate.

In this case, the at least one program may further include, in the unit production time of the first product, a job replacement time that occurs when a production job is replaced from a first product to a second product in the congested facility.

At this time, the at least one program is the most preferred facility when there is an unallocated order that cannot be allocated to the congested facility after the production order of the products is determined according to the difficulty rate in the total available time of the congested facility. Next, the unallocated order can be assigned to the preferred sub-preferred facility.

In this case, the at least one program may determine the order of production of the unallocated order by calculating the difficulty rate of products allocated to the sub-preferred facility and the difficulty rate of the unallocated order.

As described above, the machine learning-based process scheduling apparatus and method according to an embodiment of the present invention is not limitedly applicable to the configuration and method of the embodiments described above, but various modifications may be made to the embodiments. All or part of each of the embodiments may be selectively combined to be configured.

110: reference information generation unit 120: congestion facility selection unit
130: scheduling unit
1100: computer system 1110: processor
1120: bus 1130: memory
1131: ROM 1132: RAM
1140: user interface input device
1150: user interface output device
1160: storage 1170: network interface
1180: network

Claims (10)

A reference information generator for generating reference information including process lead types and facility preferences for products to be produced based on machine learning;
A congestion facility selection unit that allocates to a plurality of facilities for producing the products by using the reference information, and selects, as a congestion facility, a facility with the largest production amount among the plurality of facilities by date; And
A scheduling unit that calculates a difficulty rate of products allocated to the congested facilities selected for each date and determines a production schedule of the products based on the difficulty rate;
Machine learning-based process scheduling apparatus comprising a. The method according to claim 1,
The congestion facility selection unit
The machine learning-based process scheduling apparatus, characterized in that, based on the order quantity for each order of the products and the equipment preference, allocating the products to the most preferred equipment for each order to the plurality of equipments. The method according to claim 2,
The congestion facility selection unit
Machine learning-based process scheduling apparatus, comprising: sorting the non-delivery product orders of the products in order of delivery date, and allocating the non-delivery product orders of the products to the plurality of facilities in the order of delivery date. The method of claim 3,
The congestion facility selection unit
When there is a residual order that has not been allocated on the first date by exceeding the available order quota of the plurality of facilities, the remaining order is placed on the second date following the first date using the reference information. Machine learning-based process scheduling device, characterized in that prioritized allocation to the most preferred equipment among the. The method of claim 4,
The scheduling unit
The ratio of the total available time of the congested facility to the product of the order quantity of the products and the unit production time of each product of the congested facility is calculated as the difficulty rate, and the order of production of the products in the order of the greatest difficulty rate is calculated. Machine learning-based process scheduling device, characterized in that to determine. The method of claim 5,
The scheduling unit
The machine learning-based process scheduling apparatus, characterized in that the order of production is determined in the order of a large order quantity among product orders of the same product among the products whose production order is determined in the order of the highest difficulty rate. The method of claim 6,
The scheduling unit
The machine learning-based process scheduling apparatus, further comprising, in the unit production time of the first product, a job replacement time generated when a production job is replaced from a first product to a second product in the congested facility. The method of claim 7,
The scheduling unit
After the order of production of the products is determined according to the difficulty rate in the total available time of the congested facility, if there is an unallocated order that has not been allocated to the congested facility, the second preferred facility after the most preferred facility is selected. Machine learning-based process scheduling device, characterized in that allocating unallocated orders. The method of claim 8,
The scheduling unit
The machine learning-based process scheduling apparatus, characterized in that the order of production of the unallocated order is determined by calculating the difficulty rate of products allocated to the sub-preferred facility and the difficulty rate of the unallocated order. In a machine learning-based process scheduling method of a machine learning-based process scheduling device,
A reference information generator for generating reference information including process lead types and facility preferences for products to be produced based on machine learning;
A congestion facility selection unit that allocates to a plurality of facilities for producing the products by using the reference information, and selects, as a congestion facility, a facility with the largest production amount among the plurality of facilities by date; And
A scheduling unit that calculates a difficulty rate of products allocated to the congested facilities selected for each date and determines a production schedule of the products based on the difficulty rate;
Machine learning-based process scheduling method comprising a.

Images