Global predictive monitoring system for a manufacturing facility

• US 20080082197A1
• Filed: 08/10/2007
• Published: 04/03/2008
• Est. Priority Date: 09/22/2006
• Status: Active Grant

• Current Assignee: PDF Solutions
• Original Assignee: PDF Solutions
• Inventors: Lacaille, Jerome Henri Noel
• Granted Patent: US 7,894,926 B2
• Time in Patent Office: Days
• US Class Current: 700/121
• CPC Class Codes: G05B 23/0232 based on qualitative trend analysis, e.g. system evolution
• CPC Class Codes: G06N 3/126 Evolutionary algorithms, e.g. genetic algorithms or genetic programming
• CPC Class Codes: G06N 5/043 Distributed expert systems; Blackboards

Abstract

A global predictive monitoring system for a manufacturing facility. The system may be employed in an integrated circuit (IC) device fabrication facility to monitor processing of semiconductor wafers. The system may include deployment of a swarm of individually separate agents running in computers in the facility. Each agent may comprise a genetic algorithm and use several neural networks for computation. Each agent may be configured to receive a limited set of inputs, such as defectivity data and WIP information, and calculate a risk from the inputs. A risk may be a value indicative of a production yield. Each agent may also generate a quality value indicative of a reliability of the risk value. New agents may be generated from the initial population of agents. Outputs from the agents may be collected and used to calculate projections indicative of a trend of the production yield.

Claims

1. A global predictive monitoring system for a manufacturing facility, the system comprising:

a plurality of individually separate and independent agents distributed on different computers in the manufacturing facility, each of the agents comprising a genetic algorithm and being configured to receive a set of inputs for calculating and outputting a risk value indicative of production yield; and

an agent response collector configured to receive outputs of the plurality of agents to generate a projection indicative of a health of production in the manufacturing facility.

2. The system of claim 1 wherein each of the agents comprises a plurality of neural networks.

3. The system of claim 1 wherein each of the agents calculates and outputs a quality value indicative of a reliability of the risk value.

4. The system of claim 1 wherein the manufacturing facility comprises an integrated circuit (IC) device fabrication facility and the plurality of agents is configured to receive defectivity data from wafers processed in the fabrication facility.

5. The system of claim 4 wherein the plurality of agents is configured to collect defectivity data from the wafers along route of the wafers in the fabrication facility.

6. The system of claim 4 wherein the plurality of agents is configured to collect information from wafers-in-process.

7. The system of claim 6 wherein the information from wafers-in-process includes equipment ID.

8. A computer-implemented method of monitoring an IC device fabrication facility, the method comprising:

creating an initial population of agents in the fabrication facility, each of the agents being configured to receive at least one input comprising data collected on a route of a wafer being processed in the fabrication facility, each of the agents generating a risk value indicative of production yield;

creating new agents from agents in the initial population of agents to create a new population of agents; and

generating a projection indicative of a trend of a production yield in the fabrication facility based on outputs from agents in the new population of agents.

9. The method of claim 8 wherein each of the agents in the initial population of agents comprises a genetic algorithm.

10. The method of claim 8 wherein the new agents are generated using a random agent generation process.

11. The method of claim 8 wherein the new agents are generated from a pair of agents randomly selected from the initial population of agents.

12. The method of claim 8 wherein each agent in the initial and new population of agents is configured to generate a quality value indicative of a reliability of a corresponding risk value generated by the agent.

13. The method of claim 8 wherein each agent in the initial and new population of agents is configured to collect defectivity data from wafers-in-process in the fabrication facility.

14. The method of claim 8 each agent in the initial and new populations of agents comprises a plurality of neural networks.

15. A computer-implemented method of monitoring an IC device fabrication facility, the method comprising:

running a plurality of individually separate and independent agents in computers in the fabrication facility, each agent in the plurality of agents being configured to receive a limited set of input data from the central database and calculate a first value indicative of a yield of a production process in the fabrication facility; and

collecting outputs from the plurality of agents to generate a collected output; and

performing statistical analysis on the collected output to determine a trend of the yield of the production process.

16. The method of claim 15 wherein each agent comprises a genetic algorithm.

17. The method of claim 15 wherein each agent comprises a plurality of neural networks.

18. The method of claim 15 further comprising:

creating a new population of agents from the plurality of agents.

19. The method of claim 16 wherein the new population of agents is created from randomly selected pairs of agents in the plurality of agents.

20. The method of claim 15 wherein the limited set of input data includes defectivity data.