ShowDaily Day3 - Flipbook - Page 15
sponsored by
To reiterate, an IIoT platform with
AI in mind enables the contextualization and traceability of production
data. A concrete example of context
and traceability at work is attaching
quality, metrology, and process data
to the material flow. It should be
emphasized that this kind of orchestration can happen at the die, wafer,
or cassette level.
Once a foundry is data-ready,
phase two of next-level production can commence. It consists
of deploying AI for continual
improvement.
The AI deployment that best
affects holistic process improvement
in complex manufacturing is unsupervised Deep Learning, with
humans-in-loop. Advanced Deep
Learning algorithms do not merely
identify or predict production
anomalies. This AI ingests all
upstream and downstream interdependent variables in a production
process—systematically. In this way,
it adjusts for momentum effects and
can prevent production loss from
occurring in the first place with
expert modelling and deployment.
www.semiconductordigest.com
How is this possible?
An unsupervised Deep Learning
algorithm discovers “Best of Best”
(BoB) batches from a selected
period of historical production.
This AI-driven process can compartmentalize the manufacturing
conditions that will lead to poorer
or higher-quality wafers. Using
learned relationships concerning the
historical BoB regions, the Deep
Learning model delivers prioritized prescriptions to operators
(FIGURE 2).
Enacting the prescriptions locks in
an optimal plant state preemptively.
Furthermore, it efficiently
moves each subsequent production
run closer to the BoB state over
time. The result is a continual
improvement — as the AI’s
knowledge about the process’s inner
workings accumulates with each
iteration of it.
For example, a Deep Learning
algorithm can investigate how
the parameters in the lithography
and etch and strip processes
work together — in tandem with
the quality variables of yield,
throughput, and
metrology data outputs.
To recap, next-level
production is a two-phase
digital transformation
journey in which data is
consciously orchestrated
for advanced AI-readiness. AI-as-a-Service is
the short answer to the
second question of how
to transform production
quickly.
Humans-in-the-loop
can seamlessly combine
a flexible IIoT platform
with a proactive
AI-driven solution that
can achieve ROI in under a year.
Assuming ongoing and complete
post-installation support, additional
data sources and full model maintenance — AI-as-a-Service is proven
to eliminate the risks of adopting
AI. In doing so, it significantly
reduces digital transformation costs.
Best practice AI-as-a-Service, as
discussed above, also works with
established production constraints:
maintaining critical tolerances and
relaxing non-critical tolerances to
optimize production holistically.
This practice ensures that the current
manufacturing regime is not destabilized or forestalled.
With the right technology and
stewardship, semiconductor professionals can quickly accomplish
data readiness and AI-guided digital
transformation. Only integrated solutions will meet the industry’s present
production challenges and prepare
for the fabs of tomorrow. AI-for-manufacturing is not about piecemeal,
or incremental operational improvements — it is a long-term strategy
in the journey towards autonomous
manufacturing.
Thursday, December 9 | 15
