PRODUCTION SYSTEM

• US 20170336775A1
• Filed: 11/20/2016
• Published: 11/23/2017
• Est. Priority Date: 05/23/2016
• Status: Active Application

• Current Assignee: Renesas Electronics Corporation
• Original Assignee: Renesas Electronics Corporation
• Inventors: TOKOROZUKI, Kazuyuki, NAKAJIMA, Toshihiro
• Granted Patent: US 10,261,496 B2
• Time in Patent Office: Days
• CPC Class Codes: G05B 19/406 characterised by monitoring or safety G05B19/19 takes precedence
• CPC Class Codes: G05B 19/4184 characterised by fault tolerance, reliability of production system
• CPC Class Codes: G05B 2219/33322 Failure driven learning
• CPC Class Codes: G05B 2219/45031 Manufacturing semiconductor wafers
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: Y02P 90/02 Total factory control, e.g. smart factories, flexible manufacturing systems [FMS] or integrated manufacturing systems [IMS]

Abstract

A sensor monitors a treatment status of a predetermined manufacturing device, and an abnormality detection device detects an abnormality of a sensor signal that is a monitoring result of the sensor. The sensor signal is a digital data group obtained by sampling an analog waveform at a predetermined sampling period. A management apparatus learns characteristics of a plurality of digital data groups accumulated in past times through use of artificial intelligence to generate a learned model. An abnormality detection device holds the learned model and determines whether an abnormality is present in the digital data group of a current processing target by using the learned model.

Claims

1. A production system for manufacturing a product by using a plurality of manufacturing processes, the production system comprising:

a plurality of edge devices performing treatment associated with the plurality of manufacturing processes;

a master device managing the entire production system;

a sensor provided corresponding to a predetermined edge device among the plurality of edge devices and monitoring a treatment status of the predetermined edge device;

an abnormality detection device provided corresponding to the predetermined edge device and detecting an abnormality of a sensor signal that is a monitoring result of the sensor; and

a communication network connecting the plurality of edge devices, the master device, and the abnormality detection device,

wherein the sensor signal is a digital data group obtained by sampling an analog waveform at a predetermined sampling period,

wherein the master device learns characteristics of a plurality of the digital data groups accumulated in past times through use of artificial intelligence to generate a learned model for determining whether an abnormality is present in the digital data group, and

wherein the abnormality detection device holds the generated learned model and determines whether an abnormality is present in the digital data group which is a current processing target by using the held learned model.

2. The production system according to claim 1,

wherein the master device classifies the plurality of digital data groups into a plurality of categories and learns to generate a plurality of the learned models corresponding to the plurality of categories, respectively, and

wherein the abnormality detection device recognizes a category of the digital data group which is the current processing target and determines whether an abnormality is present by using the learned model corresponding to the recognized category.

3. The production system according to claim 2,

wherein the master device determines a classification method of the plurality of categories based on an instruction of an engineer, and registers the determined classification method in the abnormality detection device, and

wherein the abnormality detection device recognizes the category of the digital data group which is the current processing target based on the registered classification method.

4. The production system according to claim 1,

wherein the abnormality detection device includes

a memory circuit holding the learned model, and

a processor circuit determining whether an abnormality is present by using the held learned model.

5. The production system according to claim 1,

wherein the plurality of manufacturing processes are preceding processes of a semiconductor product.

6. A production system for manufacturing a plurality of types of semiconductor products by using a plurality of manufacturing processes, the production system comprising:

a plurality of manufacturing devices performing processing treatment associated with the plurality of manufacturing processes by using a predetermined process recipe depending on a type of the semiconductor product;

a management apparatus managing the entire production system;

a first sensor and a second sensor each provided corresponding to a predetermined manufacturing device among the plurality of manufacturing devices and each monitoring a status of the processing treatment of the predetermined manufacturing device;

an abnormality detection device provided corresponding to the predetermined manufacturing device and detecting an abnormality of a first sensor signal that is a monitoring result of the first sensor and an abnormality of a second sensor signal that is a monitoring result of the second sensor; and

a communication network connecting the plurality of manufacturing devices, the management apparatus, and the abnormality detection device,

wherein the first sensor signal is a first digital data group obtained by sampling an analog waveform at a first sampling period,

wherein the second sensor signal is a second digital data group obtained by sampling an analog waveform at a second sampling period,

wherein the management apparatus learns characteristics of a plurality of the first digital data groups accumulated in past times through use of artificial intelligence to generate a first learned model for determining whether an abnormality is present in the first digital data group, and learns characteristics of a plurality of the second digital data groups accumulated in past times through use of artificial intelligence to generate a second learned model for determining whether an abnormality is present in the second digital data group, and

wherein the abnormality detection device holds the generated first learned model and the generated second learned model and determines whether an abnormality is present in the first digital data group which is a current processing target by using the held first learned model and determines whether an abnormality is present in the second digital data group which is a current processing target by using the held second learned model.

7. The production system according to claim 6,

wherein the management apparatus classifies the plurality of first digital data groups into a plurality of categories based on a first classification method and learns to generate a plurality of the first learned models corresponding to the plurality of categories, respectively, and classifies the plurality of second digital data groups into a plurality of categories based on a second classification method and learns to generate a plurality of the second learned models corresponding to the plurality of categories, respectively, and

wherein the abnormality detection device recognizes a category of the first digital data group which is the current processing target and determines whether an abnormality is present by using the first learned model corresponding to the recognized category, and recognizes a category of the second digital data group which is the current processing target and determines whether an abnormality is present by using the second learned model corresponding to the recognized category.

8. The production system according to claim 7,

wherein the management apparatus determines the first classification method and the second classification method based on an instruction of an engineer, and registers the determined first classification method and the determined second classification method in the abnormality detection device, and

wherein the abnormality detection device recognizes the category of the first digital data group which is the current processing target based on the registered first classification method and recognizes the category of the second digital data group which is the current processing target based on the registered second classification method.

9. The production system according to claim 7,

wherein each of the first classification method and the second classification method is determined by using a single identifier or a plurality of identifiers among a plurality of production identifiers narrowing down types of the semiconductor products in a stepwise manner, a process identifier identifying the plurality of manufacturing processes, a plurality of device identifiers narrowing down types of the plurality of manufacturing devices in a stepwise manner, and a recipe identifier identifying the predetermined process recipe.

10. The production system according to claim 9,

wherein the single identifier or the plurality of identifiers to be used in the first classification method differs from the single identifier or the plurality of identifiers to be used in the second classification method.

11. The production system according to claim 6,

wherein the abnormality detection device includes

a memory circuit holding the first learned model and the second learned model, and

a processor circuit determining whether an abnormality is present by using the held first learned model and the held second learned model.

12. The production system according to claim 6,

wherein at least one of the first sampling period and the second sampling period is equal to or less than 100 ms.

13. A production system for inspecting a product through use of a plurality of inspection processes, the production system comprising:

a plurality of edge devices performing measurement associated with the plurality of inspection processes and generating measurement data;

a master device managing the entire production system;

an abnormality detection device provided corresponding to a predetermined edge device among the plurality of edge devices and detecting an abnormality of the measurement data from the predetermined edge device; and

a communication network connecting the plurality of edge devices, the master device, and the abnormality detection device,

wherein the measurement data is a digital data group forming an analog waveform,

wherein the master device learns characteristics of a plurality of the digital data groups accumulated in past times through use of artificial intelligence to generate a learned model for determining whether an abnormality is present in the digital data group, and

wherein the abnormality detection device holds the generated learned model and determines whether an abnormality is present in the digital data group which is a current processing target by using the held learned model.

14. The production system according to claim 13,

wherein the master device classifies the plurality of digital data groups into a plurality of categories and learns to generate a plurality of the learned models corresponding to the plurality of categories, respectively, and

wherein the abnormality detection device recognizes a category of the digital data group which is the current processing target and determines whether an abnormality is present by using the learned model corresponding to the recognized category.

15. The production system according to claim 14,

wherein the master device determines a classification method of the plurality of categories based on an instruction of an engineer and registers the determined classification method in the abnormality detection device, and

wherein the abnormality detection device recognizes the category of the digital data group which is the current processing target based on the registered classification method.

16. The production system according to claim 13,

wherein the abnormality detection device includes

a memory circuit holding the learned model, and

a processor circuit determining whether an abnormality is present by using the held learned model.