1994
DOI: 10.1109/66.286857
|View full text |Cite
|
Sign up to set email alerts
|

Diagnosis of semiconductor manufacturing equipment and processes

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
1
1
1
1

Citation Types

0
7
0

Year Published

1994
1994
2005
2005

Publication Types

Select...
5
1

Relationship

0
6

Authors

Journals

citations
Cited by 8 publications
(7 citation statements)
references
References 15 publications
0
7
0
Order By: Relevance
“…Extracted from the equipment models and measurement values, it corresponds to how well the predicted output(s), assuming a hypothetical input change, matches the measurements. It is similar to the algorithms used by May [8], and Saxena and Unruh [10], but instead of assigning the probability directly to a faulty input, our algorithm assigns the probability to the evidence that a faulty input could have caused the problem. Then, if multiple inputs could each have caused the problem, the diagnostic system will calculate the probability of each one, depending on their past frequency (5).…”
Section: Number Of Faulty Measurements Total Number Of Measurementsmentioning
confidence: 96%
See 1 more Smart Citation
“…Extracted from the equipment models and measurement values, it corresponds to how well the predicted output(s), assuming a hypothetical input change, matches the measurements. It is similar to the algorithms used by May [8], and Saxena and Unruh [10], but instead of assigning the probability directly to a faulty input, our algorithm assigns the probability to the evidence that a faulty input could have caused the problem. Then, if multiple inputs could each have caused the problem, the diagnostic system will calculate the probability of each one, depending on their past frequency (5).…”
Section: Number Of Faulty Measurements Total Number Of Measurementsmentioning
confidence: 96%
“…It is very powerful when these models are physically based and well known, because deep level diagnostic systems can find the root cause of the problem by deriving it from the theory of the domain [5], [12], [15]. If the domain model is empirical, this method is not as powerful because the evidence must lie within the experimental range of the model [8]- [10]. In either case, however, a deep level diagnostic system is still very desirable because it can find the proper solutions in unanticipated situations.…”
Section: A Deep Versus Shallow Knowledge Base Approachesmentioning
confidence: 99%
“…The Js determined by (5) and (6) appear in the last row in Table 4. In Table 4, the J = 1 for the first test pattern means that the fault is mainly attributed to a variation in the source power, which is clearly represented in the corresponding actual pattern (0.3, 0, 0, 0) shown in the second row.…”
Section: Fault Sensitivitymentioning
confidence: 99%
“…Wafer measurements such as an etch rate or a deposition rate, meanwhile, have been used to identify process faults [1][2][3][4][5]. Diagnostic systems relying upon ex situ measurements are somewhat limited in that they can be applied only to in-line fault detection.…”
Section: Introductionmentioning
confidence: 99%
“…During the MMST demonstration, the process control system was responsible for invoking the equipment-level diagnosis techniques described in [9]. Although the diagnosis system was designed to be an independent system that can operate independently if necessary, it was designed with the same automatic component activation framework used for process control.…”
Section: Selecting Process Control Modelsmentioning
confidence: 99%