calender
Date & Time
Search
Datum
{{range.dates[index].day}}
{{range.dates[index].date}}
Time
Mornings Noon Afternoons Evenings
  • from
  • to
  • o'clock
Topic
Event location
Event
Properties
{{item.name}}
{{item.name}}
Exhibition venue

(please choose the desired areas)

Lecture language
Format

Events calendar 2019

Information about forums, live demonstrations, Accelerating Talents and panel discussion. The entire forum program is available in the event database.

Back to the EventList

Digital Twin / Ion Source Prediction

NOV
12
2019
12. NOV 2019

Presentation Hall ICM - Internationales Congress Center München SEMICON EUROPA > Fab Management Forum > Session 3: Poster session - Shop Floor Innovations with Immediate Payback

15:45-15:50 h | ICM - Internationales Congress Center München ICM Room 14c, 1st Floor

Subjects: SEMICON EUROPA

Type: Presentation

Speech: English

The ion source is a major concern in utilising implant tools to their full potential. The issue is mostly related to the thermionic filament which either exhibits beam imaging problems or sudden fusing. This work is examining both failure modes by implementing different prediction models. It can be demonstrated that the combination of different prediction models can enhance the overall accuracy of the prediction. The best prediction accuracy is achieved by combining linear regression methods with random survival forest as well as exponentially gradient-boosted trees.Moreover a fusing event can be predicted within 24 hours before occurrence in more than 75% of breakdowns and uniformity issues are forecast with a minimum accuracy of greater than 50% of the cases.The achieved accuracy has enabled the introduction of the predictive maintenance strategy for implant tools at Globalfoundries which has led to an improved utilisation and a reduction of 12.7% of total maintenance costs. Taking into account additional influences such as cycle time, mean time to failure or time to repair it can be shown that with a balanced approach additional 4.7% of cost saving is achieved.Based on this outcome the general condition for successfully applying predictive maintenance is discussed. In particular the process of identifying potential opportunities is highlighted. Dependencies in scaling this method effectively are studied with regards to: a) available feature information, b) the opportunity to generate run to fail data and c) easy access to failing components.

Informations

Dr. Ramin Madani

Dr. Ramin Madani studied Physics and worked at the Max-Planck-Institute for Plasma Physics in Greifswald until 2004 before joining Infineon/Qimonda at the 300 mm site in Dresden. He joined Odersun AG, a German manufacturer of flexible CIS solar film, in 2009 and was responsible for backend manufacturing. In 2012 he started at Globalfoundries where he initially led projects in the area of process control and successfully reduced tool variability to pave the way for a steady increase in 28nm LPQ prime yield - a major success for Fab1. He later headed a multi-year program to drive manufacturing excellence with sophisticated tool data processing and best in class automation solutions. In line with Globalfoundries’ smart manufacturing initiative, he launched in 2016 the equipment and process analytics program which is currently driving data intelligence and ML use cases in Fab1 manufacturing.

Dr. Ramin Madani
Sr Staff Project Manager

Location

Eingang
Nord-West
ICM
Eingang
Nord
Eingang
West
Atrium
Eingang
Nord-Ost
Eingang
Ost
Conference
Center Nord
Freigelände
C1
C2
C3
C4
C5
C6
B0
B1
B2
B3
B4
B5
B6
A1
A2
A3
A4
A5
A6

More Events