The Impact of AI on the automotive sector

AI is having a large impact upon the
automotive sector with great attention on self-driving cars.
However our focus is on applying AI to the production of vehicles. Here we see
AI as part of Industry 4.0 initiatives driving up efficiencies in manufacturing
plants by improving overall equipment effectiveness, reducing defects and
improving automation on the line. It is important to understand the different
use cases in value-added AI projects and provide to the line. Automating visual
inspection helps reduce human error in the process and improve traceability
however it does not reduce the chance of the defect occurring only the chance of
one being shipped. Predictive maintenance can help overall efficiencies in the
plant, however we see the greatest immediate value add in production is in
the optimization of the control parameters by actively optimizing
control parameters with AI, production can expect a reduction in defects that occur,
reducing the cost of non quality. AI is a value add to data, in order to leverage
AI it needs to be enabled by data. This means that the manufacturer needs to
have a good data environment or a route to a good environment. Practically what we
emphasize is that most of the data collection hardware is already installed
capital equipment that was installed in the last 20 years will have a good set
of sensors on them, however, the collection of that data is
important especially from a holistic point of view. We work with many of our
clients to improve their data environment to reach a state where they
can leverage AI and put a lot of emphasis on an AI for Industry 3.0
rather than Industry 4.0 This speaks to working with the current
set of sensors on the line. We’re creating value from the data that they
currently produce. It is often not necessary to add more data streams
before the existing ones are ordered and the value is created from them. One
further important point we like emphasizing is that real-time data is
often too late. Real-time data is only valuable if you can respond in real-time.
In practice, a good AI solution should be able to act in advance of real time. We
provide two solutions DataProphet PRESCRIBE an AI for optimizing control
parameters. Here we have helped customers reduce costing defects significantly, in
some cases reducing shipped defects to zero percent. Records of up to three
months in reducing robotic weld defects by 50 to 70 percent in body shops. Our
second solution DataProphet INSPECT is an AI for automated visual inspection.
Here the solution helps customers automate the visual inspection of the
surface of items. We also support the two solutions through the digitization
services we offer where we work with our customers to help them achieve the data
environment to realize the value from their data. I
think much of AI in manufacturing has been limited to predictive maintenance
and in the next five years there will be a much wider set of well understood use
cases of AI in manufacturing as their value is better and better understood. I
have two key takeaways I hope attendees leave with, One: their current data streams
are more than sufficient to start extracting value from AI. What we see
across most industries are strategies to aggregate data and this should really be
done with the value adds upon the data in mind. Two: real-time is often too late.
In the process control environment it is important that information is received
well before action is taken upon it.