CN114723286A

CN114723286A - Learning type dispatching system and dispatching method

Info

Publication number: CN114723286A
Application number: CN202210364080.3A
Authority: CN
Inventors: 陈仕涵
Original assignee: Nanjing Dinghua Intelligent System Co ltd
Current assignee: Nanjing Dinghua Intelligent System Co ltd
Priority date: 2022-04-08
Filing date: 2022-04-08
Publication date: 2022-07-08
Also published as: TW202341028A

Abstract

The invention provides a learning type dispatching system and a dispatching method. The learning type dispatching system comprises a storage device and a processor. The storage device stores an advanced planning and scheduling system, a manufacturing execution system and an association reasoning module. The processor is coupled to the storage device and executes the advanced planning and scheduling system, the manufacturing execution system and the association reasoning module. The processor inputs the input data generated by the advanced planning and scheduling system and the manufacturing execution system into the correlation inference module, so that the correlation inference module generates output data. The processor controls the work station to dispatch the work in process according to the output data. The output data comprises dispatching time information and suggested machine information. The learning type dispatching system and the dispatching method can automatically generate the corresponding dispatching time information and the suggested machine information.

Description

Learning type dispatching system and dispatching method

Technical Field

The invention relates to a learning type dispatching system and a dispatching method.

Background

Currently, in the Process of Work In Process (WIP), since the processes of Work in Process processed by each workstation may be different, the Work In Process (WIP) of the conventional Work in Process needs to perform a lot of labor to evaluate the status of each workstation and each machine thereof in the field for dispatching. In contrast, the conventional dispatching operation is not only inefficient, but also fails to consider the conditions of the physical production field and the real-time status of the production flow of the product, so that the conventional dispatching operation cannot effectively improve the production efficiency of the workstation.

Disclosure of Invention

The invention is directed to a learning-based dispatching system and a dispatching method, which can automatically generate corresponding dispatching time information and suggested machine information according to input data generated by an advanced planning and scheduling system and a manufacturing execution system.

According to an embodiment of the present invention, the learning-based dispatching system includes a storage device and a processor. The storage device stores an advanced planning and scheduling system, a manufacturing execution system and an association reasoning module. The processor is coupled to the storage device and executes the advanced planning and scheduling system, the manufacturing execution system and the relevance inference module. The processor inputs the input data generated by the advanced planning and scheduling system and the manufacturing execution system into the correlation inference module, so that the correlation inference module generates output data. The processor controls the work station to carry out work dispatching operation of work-in-process according to the output data. The output data comprises dispatching time information and suggested machine information.

According to the embodiment of the invention, the dispatching method comprises the following steps: generating input data through an advanced planning and scheduling system and a manufacturing execution system; inputting the input data to a relevance reasoning module; generating output data through a relevance reasoning module; and controlling the work station to dispatch the work in process according to the output data.

Based on the above, the learning-based dispatching system and the dispatching method of the invention can utilize the relevance reasoning module to analyze the input data provided by the advanced planning and scheduling system and the manufacturing execution system, and automatically generate effective dispatching time information and suggested machine information, so as to effectively improve the production efficiency.

In order to make the aforementioned and other features and advantages of the invention more comprehensible, embodiments accompanied with figures are described in detail below.

Drawings

FIG. 1 is a diagram of a learning-based dispatching system according to an embodiment of the present invention;

FIG. 2 is a flow chart of a dispatching method according to an embodiment of the present invention;

FIG. 3 is a flow diagram of a training relevance inference module of an embodiment of the invention.

Description of the reference numerals

100: a learning type dispatch system;

110: a processor;

120: a storage device;

121: an advanced planning and scheduling system;

122: a manufacturing execution system;

123: a relevance reasoning module;

s210 to S240, S310 to S380: and (5) carrying out the following steps.

Detailed Description

Reference will now be made in detail to exemplary embodiments of the invention, examples of which are illustrated in the accompanying drawings. Wherever possible, the same reference numbers will be used throughout the drawings and the description to refer to the same or like parts.

Fig. 1 is a schematic diagram of a learning-based dispatching system according to an embodiment of the invention. Referring to fig. 1, the learning-based dispatching system 100 includes a processor 110 and a storage device 120. The processor 110 is coupled to the storage device 120. The storage device 120 is used for storing an Advanced Planning and Scheduling (APS) 121, a Manufacturing Execution System (MES) 122 and a relational reasoning module (relational forecasting module) 123. In the present embodiment, the ADPS 121 and the MES 122 may provide dispatch services (or other Business services) through a specific Application Programming Interface (API). In other words, the user can operate the application program interfaces of the advanced planning and scheduling system 121 and the manufacturing execution system 122 to generate the input data, and the learning type dispatching system 100 can automatically execute the relevance inference module 123 according to the input data to generate the relevant dispatching recommendation information.

In the present embodiment, the learning dispatching system 100 can be, for example, disposed in a computing device of a workstation for a user to operate. The workstations may be used to control multiple process tools and multiple workstations may achieve a single production goal. Alternatively, the learning-oriented dispatching system 100 may be installed in a cloud server for user operation and controlling the workstation in parallel. In the present embodiment, the Processor 110 may include, for example, a Central Processing Unit (CPU), or other Programmable general purpose or special purpose Microprocessor (Microprocessor), Digital Signal Processor (DSP), Application Specific Integrated Circuit (ASIC), Programmable Logic Device (PLD), other similar Processing Circuits, or a combination thereof. The storage device 120 may include a Memory (Memory) and/or a database (database), wherein the Memory may be, for example, a Non-Volatile Memory (NVM). The storage module 120 may store relevant programs, modules, systems or algorithms for implementing the embodiments of the present invention, so as to be accessed and executed by the processor 110 to implement the relevant functions and operations described in the embodiments of the present invention. In the present embodiment, the advanced planning and scheduling system 121 and the manufacturing execution system 122 can be implemented by a program Language such as json (javascript Object notification), Extensible Markup Language (XML), YAML, etc., but the invention is not limited thereto.

In this embodiment, the processor 110 may execute the advanced planning and scheduling system 121 and the manufacturing execution system 122 to generate input data for work in process dispatch, and the processor 110 may execute the relevance inference module 123 to input the input data to the relevance inference module 123. Relevance inference module 123 can include machine-learned algorithms and can be used to generate a relevance function between input data and output data. In this way, the relevance reasoning module 123 can automatically generate output data according to the input data and the relevance function. The output data may include dispatch time information and suggested tool information. The processor 110 may automatically control a plurality of Process tools of the Work Station (WS) according to the dispatch time information and the recommended tool information to perform Work In Process (WIP) manufacturing operations.

FIG. 2 is a flow chart of a dispatching method according to an embodiment of the invention. Referring to fig. 1 and 2, the learning-based dispatching system 100 of the present embodiment can perform the following steps S210 to S240. In the present embodiment, the user may operate the application program interfaces of the ADM 121 and the MES 122. In step S210, the processor 110 executes the apc system 121 and the mes system 122 to generate input data through the apc system 121 and the mes system 122. In this embodiment, both the production planning data and the Equipment (EQ) data can be input data, and the input data can be vector feature data or tensor feature data. Notably, the production planning data may include at least one of a plurality of vector feature data as shown in tables 1 and 2 below. The plurality of vector feature data shown in tables 1 and 2 include feature values, actual values, and vector values. The plurality of vector eigen data may constitute tensor eigen data. The production Planning data may be provided by an Advanced Planning and Scheduling (APS) System, and the tool data may be provided by a Manufacturing Execution System (MES), for example.

TABLE 1

TABLE 2

At step S220, the processor 110 may execute the relevance reasoning module 123 and input the input data to the relevance reasoning module 123. In step S230, the relevance reasoning module 123 may generate output data after performing relevance reasoning operation. In this embodiment, the relevance reasoning module 123 can perform relevance reasoning according to the characteristic data as described above to find out the relevance between the production strategy and the site dispatch. The output data may include dispatch time information and suggested tool information. In this regard, the output data may include, for example, the data content shown in table 3 below.

Suggested Time (Time)	0
		Equipment (EQ)	21LAMBLK002

TABLE 3

In step S240, the processor 110 may control the workstation to perform a manual or automatic dispatching operation of work in process according to the output data, so as to effectively optimize the production efficiency of the workstation. Therefore, the learning-based dispatching system 100 and the dispatching method of the present embodiment can effectively generate the corresponding dispatching suggestion information according to the input data provided by the ADM 121 and the MES 122.

FIG. 3 is a flow diagram of a training relevance inference module of an embodiment of the invention. Refer to fig. 1 and 3. In the embodiment, the relevance inference module 123 may be a Neural Network (Neural Network) module constructed by a self-attention mechanism (SAM) model, or other similar Neural Network modules, and may implement a Deep enhanced machine Learning (DRL) function. The relevance inference module 123 can, for example, utilize a multi-layer neural network. Moreover, the relevance inference module 123 can be trained to learn and infer the dispatching data of the ADP 121 and the MES 122, so as to find the relevance between the corresponding tool and the on-site dispatching parameters, time, material number, order delivery time, production line, etc. from the big data, and generate the corresponding dispatching relevance function. In this way, the dispatching system configured with the relevance inference module 123 is implemented as a learning-based dispatching system. In the present embodiment, the learning-based dispatching system 100 can execute the following steps S310 to S380 to train the relevance inference module 123 in advance. At step S310, the processor 110 may obtain reference input data. In this embodiment, the input data may include a reference production planning feature vector, a reference reporting data feature vector, field resource information, and field material resource information. It is noted that the reference production plan feature vector may be provided by an advanced planning and scheduling system, and the reference report data feature vector may be provided by a manufacturing execution system, for example. In step S320, the processor 110 may initialize reference input data to generate training data. The initialization may be, for example, converting the input data into vector eigen data or tensor eigen data.

At step S330, the processor 110 may train the relevance reasoning module 123 according to the training data. In this embodiment, the number of training data may be multiple strokes. During the training process, the relevance inference module 123 may generate corresponding output data for each piece of training data. In step S340, the processor 110 may control the workstation according to each output data, and the workstation may generate the feedback data accordingly. It should be noted that the feedback data may include a plurality of device information and a plurality of weight values. The feedback data may include data content as shown in table 4 below.

Equipment (number)	Correlation weight
		21LAMBLK002	0.86
21LAS003	0.07
		21LDR000001	0.07

TABLE 4

In step S350, the processor 110 may obtain feedback data. In step S360, the processor 110 may determine whether the feedback data meets the expectation. In this embodiment, the processor 110 compares the feedback data with the scheduling data generated by the APC system to determine whether to modify the relevance formula module 123. In other words, the relevance reasoning module 123 can be trained to generate dispatch suggestions through more efficient and accurate relevance reasoning instead of at least the scheduling suggestion provided by the advanced planning and scheduling system. In this embodiment, the processor 110 may compare the compliance with the advanced scheduling plan, wherein the compliance refers to the actual dispatch condition. For example, the processor 110 may compare the differences according to a numerical value such as a production line, production equipment, or expected time error. In this regard, the processor 110 may determine that the value error between the feedback data and the scheduling data is less than or equal to the threshold. If yes, the feedback data is in accordance with the expectation. The processor 110 may execute step S340 to continuously (recursively) train the relevance inference module 123 according to the next reference output data. If not, the feedback data is not in accordance with the expectation. The processor 110 may execute step S370 to modify the decision factor (auto-correction) in the relevance reasoning module 123. Decision factors may include, for example, determining whether a route is correct, whether a job is correct, and/or whether equipment is correct, etc. The decision factor can be, for example, the factors of the sequence of the dispatched batches, the delivery period, the material number characteristics, etc., and can relate to whether the overall result is in accordance with the reality. If the decision factor has an error, the processor 110 may correct the error decision factor automatically or by receiving an external correction instruction (provided by a user). At step S380, the processor 110 may re-control the workstation to re-dispatch the job, and then proceed to step S330 to continue (recursively) train the relevance inference module 123 according to the next reference output data. In this way, the relevance inference module 123 of the present embodiment can generate the dispatching recommendation after training, so that the dispatching result realized by the workstation can conform to the related planning of the production schedule and the field dispatching factor can be considered at the same time, so as to provide an accurate dispatching recommendation or automatic dispatching operation.

In summary, the learning-based dispatching system and the dispatching method of the present invention enable a user to generate input data by operating the application program interfaces of the advanced planning and scheduling system and the manufacturing execution system, and automatically execute the relevance inference module according to the input data to generate relevant dispatching suggestion information. Therefore, the learning type dispatching system can carry out good dispatching operation on the workstation according to the relevant dispatching suggestion information, and can effectively improve the production efficiency of the workstation.

Finally, it should be noted that: the above embodiments are only used to illustrate the technical solution of the present invention, and not to limit the same; while the invention has been described in detail and with reference to the foregoing embodiments, it will be understood by those skilled in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some or all of the technical features may be equivalently replaced; and the modifications or the substitutions do not make the essence of the corresponding technical solutions depart from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A learning-based dispatching system, comprising:

the storage device is used for storing the advanced planning and scheduling system, the manufacturing execution system and the relevance reasoning module; and

a processor coupled to the storage device and executing the advanced planning and scheduling system, the manufacturing execution system, and the correlation inference module,

the processor inputs the input data generated by the advanced planning and scheduling system and the manufacturing execution system to the relevance reasoning module so as to enable the relevance reasoning module to generate output data, and controls a work station to perform work-in-process dispatching operation according to the output data, wherein the output data comprises dispatching time information and suggested machine information.

2. The system of claim 1, wherein production planning data and tool data are used as the input data.

3. The system of claim 2, wherein the input data is vector feature data or tensor feature data.

4. The system of claim 1, wherein the processor retrieves reference input data and initializes the reference input data to generate training data, the processor pre-trains the relevance inference module based on the training data,

the input data comprises a reference production planning feature vector, a reference reporting data feature vector, field resource information and field material resource information.

5. The system of claim 4, wherein the processor controls the workstation according to reference output data generated by training the relevance inference module and retrieves feedback data generated by the workstation,

the processor compares the feedback data with the scheduling data generated by the advanced planning and scheduling system to determine whether to modify the relevance inference module.

6. The system of claim 5, wherein the processor continuously controls the workstation according to the reference output data when the processor determines that the numerical error between the feedback data and the scheduling data is less than or equal to a threshold value; and

and when the processor judges that the numerical error between the feedback data and the scheduling data is larger than the threshold value, the processor corrects a decision factor in the relevance reasoning module.

7. The system of claim 1, wherein the relevance inference module is a neural network module constructed by a self-attention mechanism model.

8. A method for dispatching is characterized by comprising the following steps:

generating input data by an advanced planning and scheduling system and the manufacturing execution system;

inputting the input data to a relevance reasoning module;

generating output data by the relevance reasoning module; and

and controlling the work station to dispatch the work in process according to the output data.

9. The method of claim 8, wherein production planning data and tool data are used as the input data.

10. The method of dispatching according to claim 9, wherein the input data is vector eigen data or tensor eigen data.

11. The method of dispatching according to claim 8, further comprising:

obtaining reference input data;

initializing reference input data to generate training data; and

pre-training the relevance reasoning module according to the training data,

the input data comprise reference production planning feature vectors, reference reporting data feature vectors, field resource information and field material resource information.

12. The method of dispatching according to claim 11, further comprising:

controlling the workstation in accordance with reference output data generated via training the relevance reasoning module;

obtaining feedback data generated by the workstation; and

and comparing the feedback data with the scheduling data generated by the advanced planning and scheduling system to determine whether to correct the relevance reasoning module.

13. The method of dispatching according to claim 12, further comprising:

when the numerical error between the feedback data and the scheduling data is judged to be less than or equal to a threshold value, the workstation is continuously controlled according to the reference output data; and

and when the numerical error between the feedback data and the scheduling data is judged to be larger than the threshold value, correcting the decision factor in the relevance reasoning module.

14. The method of dispatching as claimed in claim 8, wherein the relevance reasoning module is a neural network module constructed by a self-attention mechanism model.