CADCAM/SUPPLEMENTCADCAM/PRODUCTION IT SUPPLEMENT
ENTER ARTIFICIAL INTELLIGENCE
machine, workholding, workpiece and the
tooling. And much of that standard data is
published, by DMG Mori and Sandvik
Coromant, for example. We collect that data
and assemble it into the software and then
that information, along with the AI engine, is
used to automate a lot of the tasks.
“So, from the CAD you have the part
geometry, commonly today you also have the
product manufacturing information (PMI),
CAD feature tree and original design intent,
with this all brought into our system. The AI
engine will analyse that and categorise it
into the things that it knows how to make,
using a rules-based AI engine that we
originally patented about 10 years ago.”
This is basically an expert system
approach, he adds. “Esprit basically learns
from the user, applying any result as the
next preference when that situation arises.
So, for the actual material cutting data, you
take that from the suggested data from a
cutting tool supplier. Then, if the customer
uses different speeds and feeds, the system
will learn the new preference.” But that is
just one of the learning elements it takes
from the user, he adds.
“In creating toolpaths, roughly speaking
there are 25 to 50 properties that control
the behaviour of the toolpath and CAM
software, depending upon what cutting cycle
in which CAM software you’re talking about.
So, you know, there’s the radial depth of
cut, width of cut, depth of cut, cutting speed,
and cutting feed, but there are also things
like how do you enter the material, what kind
of clearance do you want, how do you exit
the material, what kind of axial engagement
do you want to maintain. And for rotary axes,
there are also acceleration limits required,
so that you don’t mark the part – if you
overdrive axes and acceleration, especially
rotary access, you can start to get chatter on
the part. So, it’s much more than speeds
and feeds. It learns, that’s the key, it learns
and so the user is putting preferences in
order, taking the baseline and accepting it or
the baseline is not my preference and is
changed via input screens and those values
are then memorised and used next time the
same situation comes up.”
In addition to supporting toolpath
creation, Matthews explains that AI is also
used to support process control. “As an
example, take a multi-tasking machine
where we have some toolpaths to cut the
part. The AI will take the toolpaths and the
digital denition of the machine, digital twin,
and it will create what we call a machine
process. And so an example might be that I
have a two-spindle machine and an upper
and lower head turret. I’m not just going to
use things one at a time, I’m going to
simultaneously cut workpieces on the
primary spindle, on the front side of the
workpiece, and at the same time on the
secondary spindle I will cut the rear side of
the workpiece. From a machine point of
view, there are two workpieces being cut
simultaneously, but from a part point of
view, the process is still linear – step one,
two, three, four etc. And so, at the click of a
button, we can take a process that is
sequential – a lathe with one spindle one
turret – swap out the machine for one that
has two spindles and two turrets and
optimise for that, such that Esprit will
actually create a new machine program.
“It’s re-sequencing the toolpath and
synchronising it, because at some point
there needs to be a synchronisation for
potential collisions that could occur and, of
course, there’s the process of part transfer
– you cut the front half, you pick the pick the
part up and then you can cut the back half,
so there’s a part transfer that has to be
synchronised. The AI engine does that
automatically. So, we use AI in in toolpath
creation, but we also use AI in process
control.”
There’s more, though, Matthews
continues: “Let’s say you have a milling
workpiece and let’s say you have 10 steps
to cut it. You’re going to face it and then do
some drilling etc. So, 10 different steps
using 12 or using 10 tools or eight tools.
You have that part process, and let’s say
you’re running it on a vertical machining
centre. So, we have the part process, the
digital machine, and we have a part program
produced by our AI engine. Let’s assume
that production volumes go up. Now you’re
going to use a pallet system and a
tombstone on a horizontal machining centre.
You swap the machine out, you dene your
new machine, now it’s a horizontal not a
vertical; it’s a 4-axis with a tombstone
xture, with six parts on the xture
“We take the same process, we add the
new xturing and the software will
automatically replicate the process for you
and optimise it, because now the sequence
of machining is different. Also the coordinate
systems are different. So, it will recongure
the coordinate systems for you – work
offsets are a big deal for horizontal
machines and we can automatically
recongure the coordinate systems. Our AI
engine is process aware, or what we call
machine aware. This is a big time-saver for
users that run into these situations.”
Esprit takes accurate part and
machining environment geometry, and
applies verifi ed machining data to it
to create programs that do not require
debugging. In the UK, Esprit is available
from the company’s subsidiary,
https://is.gd/calepa
62 November 2019 www.machinery.co.uk @MachineryTweets
/calepa
/www.machinery.co.uk