From press emporium to validation: BMW Group Plant Munich builds on synthetic comprehension and intelligent use of data

Munich. BMW Group Plant Munich is creation increasing
use of applications with synthetic comprehension (AI). AI is fast,
arguable and easy to confederate into a several prolongation processes
and, joined with intelligent information analytics and cutting-edge measurement
technologies, it opens adult new opportunities for some-more fit automobile production.

Robert Engelhorn, Director of BMW Group Plant Munich, is operative to
allege a concentration of these technologies: “At Plant Munich, it
takes about 30 hours to make a vehicle. During that time, each
automobile we make generates large amounts of data. With a assistance of
synthetic comprehension and intelligent information analytics, we can use this data
to conduct and analyse a prolongation intelligently. AI is assisting us
to streamline a prolongation even serve and safeguard premium
peculiarity for any customer. It also saves a employees from carrying to
do monotonous, repeated tasks.” As with any innovation, a key
cause is effectiveness: “Our group in prolongation are highly
gifted specialists, so they are a best judges of possibly an AI
concentration can boost peculiarity and potency during any given theatre of
production,” says Robert Engelhorn.

The options for regulating AI and intelligent information analytics are now being
tested in several areas of BMW Group Plant Munich. In some areas, the
technologies are already in use in array production, such as the
press emporium and duty validation.


Smart Data and AI in a press shop

The press emporium during a BMW Group’s home plant in Munich turns more
than 30,000 blanks a day into automobile physique parts. Since 2019 each
vacant has been given a laser formula during a start of prolongation so the
physique partial can be clearly identified throughout. This formula is picked up
by a iQ Press system, that annals element and routine parameters
– such as a density of a steel and oil layer, and the
heat and speed of a presses. The parameters are afterwards related
to a peculiarity of a tools produced.

Uploaded to a cloud in genuine time, a information is immediately available
in a entirety for a prolongation group to benefit a clearer design of
a prolongation process. iQ Press information is an critical apparatus for
them, as it eliminates a need for any physique partial to be checked in
notation detail, in peculiarity control for example, and picks out only
irregularities that need action.

Artificial comprehension also offers intensity to brand recurring
patterns in a process, formed on a information collected, to support
continual optimisation. So, as good as improving a potency of
prolongation systems, iQ Press helps to serve boost hourly output
from a press shop.


Predictive upkeep in a physique shop

Body emporium robots are propitious with a sum total of over 600 welding
tongs. If a tongs ever need replacing unexpectedly, it costs
poignant time and money. Moreover, many of a robots are difficult
to access, so dismantling and replacing their tongs can take hours.

Until now, a condition of tongs has been monitored by eye, by a
member of a prolongation team. But in new months, a maintenance
specialists during Plant Munich have been wise sensors to all a tongs
to magnitude attrition levels 3 times per change and news any
abnormalities. The information they furnish is constantly evaluated by
software, permitting intensity appurtenance failures to be predicted. Martin
Hilt, Innovation and Digitalisation Officer during Plant Munich, explains:
“Because we have a sensors and collect their information in a cloud, we can
now guard round-the-clock possibly any upkeep work is needed.
So, we can devise any replacements improved and potentially report them
for a prolongation break.”


Dust molecule research in a paint shop

Despite extensive clarification systems, automobile bodies can collect up
dirt particles as they make their approach to a paint line. Though
invisible to a tellurian eye, a particles can impact a peculiarity of
a finish. Until now, intensity defects have left undiscovered after
a portrayal process, suggested usually by a involuntary surface
inspection. They afterwards had to be reworked, or a bodies repainted completely.

Now, however, any paint emporium complement incorporates sensors that
magnitude dirt levels and concede a peculiarity of paintwork to be
predicted. “We can now tell fast if a environmental parameters
are not utterly right during some point, possibly within a paint emporium or in
one of a aegis areas. It takes a lot of information to do this, that we
collect via a process, weigh historically and analyse in
genuine time,” explains Martin Hilt.

Over a final few months, a serve special sensor grown by Plant
Munich has been measuring dirt levels on physique tools during a beginning
of a portrayal process, before and after a emu plume rollers. In
a future, when dirt levels are too high, automobile bodies will pass
by a paint emporium untreated and be sent for serve cleansing.


AI-based picture approval in assembly

AI projects in public especially concentration on programmed picture recognition.
Here, a record is used to weigh images of a member and
review them in milliseconds with hundreds of other images from the
same sequence. The complement afterwards identifies any deviations from the
norm, such as tools that are wrongly positioned or fitted, or absent.

At Plant Munich, programmed picture approval allows a production
group to brand possibly a jeopardy warning triangle, wiper caps and
doorway sills have all been rightly propitious to any car. Previously,
tiny froth in a foil cover of a doorway sill were mostly adequate to
forestall a compulsory camera gates from saying if a trademark on the
doorway sill was correct. But now an associate photographs any part
endangered in spin and can even use a mobile apparatus to check parts
that are some-more formidable to access. Distance, angle and light hardly
have any outcome on AI evaluations, that exhibit within fractions of a
second possibly all is in place or not.

The AI complement is lerned by associates. They start by photographing
a member endangered from several perspectives and marking
intensity deviations on a images. This allows them to rise an
picture database that can be used to build adult a neural network for
evaluating a images. Evaluations are carried out fully
automatically, and a appurtenance decides by itself possibly or not a part
meets all a specifications.


RFID identifies components in a vehicle

Radio Frequency Identification (RFID) allows components to be
identified automatically and contactlessly via a value chain.
“Our idea is to save prolongation workers from carrying to indicate components
manually, and concurrently to streamline prolongation even further
by ensuring a right components are propitious to a right vehicles,”
says Martin Hilt. RFID is now being used in chair prolongation at
Plant Munich though will shortly be used via a automobile public as well.

Smart RFID labels compulsory for a complement are practical before the
member leaves a supplier. They sojourn in place throughout
production, permitting line-side antennae to collect adult any labelled
member within any automobile as it passes.


Function validation with a Comfort Access robot

Comfort Access was initial introduced in a stream BMW 3 Series. A
tiny group from Electrics/Electronics Validation in Munich has now
grown a special drudge to countenance a integration.

Vehicles with Comfort Access use 3 extraneous antennae to generate
a three-dimensional electromagnetic margin around a car. When the
motorist enters a field, a complement recognises a automobile key. At about 3
metres from a car, it switches a Welcome Light on to illuminate
a area outward a driver’s door. At about 1.5 m, a doors clear –
and relock automatically if a motorist walks away.

Until now, this special underline has been certified manually, with
parameterisation in growth alone holding dual days per vehicle. The
Comfort Access zones and a change of prolongation processes on them
are afterwards checked manually again in a plant, before production
begins, holding into comment a several country-specific requirements
and apparatus features, such as trailer couplings. All in all, it is a
extensive routine and not always wholly accurate, given a multitude
of opposite functions.

To solve a problem, a BMW Group and a University of Applied
Sciences (HTW) in Dresden have grown a dimensions drudge that
autonomously circles a automobile several times in a pre-defined
settlement to establish a strength of a captivating margin during various
compulsory points. Attached to a drudge is a box containing a car
key. The box can be set during opposite heights to simulate a different
ways a motorist competence lift it: in their hand, their sports bag or a
breast pocket, perhaps. As shortly as a drudge detects a vehicle
wiring locking or unlocking a doors, a inbuilt Lidar scanner
measures a stretch between a pivotal and a vehicle, and surveys the
vehicle’s surroundings. The information that is generates goes true to a
executive computer, where it is portrayed as a graphic.

The advantages of a complement are obvious: “This drudge is not only
many faster, it’s also some-more precise. The formula we obtain are highly
minute and, many importantly, objective. So we can even start
validating a duty before a automobile has a initial test-drive,”
explains Martin Hilt.


Vehicle plcae in a prolongation system

The specialists obliged for product formation during Plant Munich
safeguard fast processes to broach defect-free vehicles throughout.
They are also obliged for integrating pre-series vehicles into
prolongation to concede a well-spoken central prolongation launch with series
peculiarity right from a start.

“Since a start of this year, a specialists have been regulating a new
app that notifies them as shortly as a pre-series automobile they are
tracking reaches a specified indicate in assembly. It allows them to
locate any automobile they wish – so they can check, say, a particular
multiple of apparatus features,” says Martin Hilt.

The new app not usually replaces a primer routine though also improves
validation. In a future, it can also be used on array vehicles.