IEEE Computational Intelligence Magazine - February 2023 - 10

Industrialand
Governmental
Activities
Piero P. Bonissone
Piero P. Bonissone Analytics, USA
Alessandro Liani
Video Systems, srl, ITALY
Industrial AI
I
n a previous communication (Bonissone
and Fogel, 2022), we highlighted
the inaugural 2023 IEEE
Conference on AI (IEEE CAI 2023) -
https://cai.ieee.org/2023/ - our first
conference with an industry focus. IEEE
CAI 2023 is structured along six verticals,
covering Industrial AI, AI in Healthcare/
Life Science, Transportation/Aerospace,
Energy, Earth System Decision Support, and
Social Implications ofAI/Privacy (Fig. 1).
In this column, we will focus on the
first vertical, Industrial AI. Today's
industrial machines are hyper-connected
to the Industrial Internet of
Things (IIoT) and have highly capable
Edge computing. As a result, they constantly
generate a wealth of diverse data
that needs to be analyzed and consumed
by other machines or human operators.
This requires advanced descriptive, predictive,
and prescriptive analytics, often
combining physics-based and datadriven
models. Industrial AI covers the
use of these models from production to
manufacturing, quality control, operations,
and maintenance. Given our
space limitation, we will limit our
discussion to Quality Control (QC)
and Prognostics and Health Management
(PHM)for maintenance.
Quality Control (QC)
In manufacturing production processes,
one of the most important goals is to
Digital Object Identifier 10.1109/MCI.2022.3223460
Date ofcurrent version: 13January 2023
guarantee the quality of the produced
goods that are placed on the market.
Quality control processes have evolved
dramatically from operator-based QC
to AI-based Non-Destructive Inspection
(NDI) automatic solutions.
Non-Destructive Inspection is a
broad, interdisciplinary field that plays a
critical role in assuring the reliability of
products and components. NDI technicians
and engineers define and implement
tests to locate and characterize
material, product conditions, and flaws
that might otherwise cause planes to
crash, reactors to fail, trains to derail,
pipelines to burst, or other catastrophic
events. NDI allows parts and materials
to be inspected and measured without
damaging them. In NDI portfolio techniques,
we find Visual and optical testing
(VT), X-Ray (RT), Magnetic
particle testing (MT), Ultrasonic testing
(UT), electromagnetic testing (ET), and
more.
Over the last 10 years, AI has automated
many NDI processes and product
quality tasks previously delegated to
operators. Enabled by rapid growth in
hardware technology, such as GPUs,
and impressive investments in the
development of new AI algorithms,
today AI represents one of the most
critical technologies in the Industry 4.0
paradigm.
AI is applied in production lines for
product quality control. Together with
other technologies like machine vision,
robotics, and big data, AI solves many
10 IEEE COMPUTATIONAL INTELLIGENCE MAGAZINE | FEBRUARY 2023
quality control tasks that always
required operator supervision. Nowadays,
AI reduces the configuration time
of automatic quality control systems
and adapts machine vision applications
to be more robust to environmental
changes than traditional solutions. For
example, machine vision applications
developed with traditional algorithms
required a laborious parameter settings
process to manage environment lighting
changes. Now, AI-based quality
control approaches provide more resilient
solutions to external effects.
AI technologies have been successfully
implemented in a large number of
Horizon 2020 projects on the Zero
Defect Manufacturing paradigm
(ZDM). For example, in the ZDMP
project [3], in which more than 30 partners
developed a Docker-based cloud
platform for process control and NDI
control solutions, the contribution
made by AI technologies was essential
to solve advanced parts inspection like
automotive electronics car displays.
Similarly, in the InterQ project [4], in
which the team dealt with advanced
surface quality needs for avionics parts
and automotive components, a quality
control solution integrated robotics,
machine vision, and AI. The excellent
results achieved would not have been
possible without the use of deep-learning
engines for the identification of
cracks and other surface defects. By
leveraging this technology, many stressing
control processes are delegated to
https://orcid.org/0000-0001-7051-2501 https://cai.ieee.org/2023/
IEEE Computational Intelligence Magazine - February 2023

Table of Contents for the Digital Edition of IEEE Computational Intelligence Magazine - February 2023
