The ASSET Project

First informal meeting: Introduction and Aim of the Project

Brussels, May 20–21 2010

Thursday, May 20, 2010

 Agenda of the Meeting

✤ Thursday 20
✤ 13.30: Welcome to the partners and project introduction (UniTN)
✤ 14.00: Partners presentations:
✤ DMG Electronics
✤ IBM + Robo-team
✤ NEC Lab
✤ CERTH
✤ UniReading
✤ NationalInstruments
✤ 17.45: Office is closed (they will ban us!)
✤ 19.00: Dinner

Thursday, May 20, 2010

 Agenda of the Meeting

✤ Thursday 20
✤ 8.00: Partners presentations
✤ D.Electron
✤ TWT
✤ Zimroy
✤ Shadow Robot
✤ 10.00: Definition of WP details
✤ Sub-projects and tasks assignment
✤ 14.00: End of meeting

Thursday, May 20, 2010

 Outline

✤ The Machining Process:

✤ As it is
✤ As we want to improve it

✤ What we have to do: ASSET subprojects

✤ Process
✤ Perception
✤ CogSys
✤ Cloud

Thursday, May 20, 2010

 The Machining process
As it is, and as we want it

Thursday, May 20, 2010

 The Machining Process

CAM Operator

Thursday, May 20, 2010

 paranoid) confidentiality level would mean minimal (even none) advantage from
cloud manufacturing.

Toolpath optimal planning and control

The milling process involves a rotating tool that moves inside the raw piece thus
removing the excess material with respect to the final geometry that we want to

The Machining Process

obtain.
The first step in designing a milling process is the choice of the right tool on the
basis (mainly) of the workpiece material and of the type of operation (e.g. face
milling, drilling, threading, contour shaping, and so on).
The second step is the computation of the toolpath. In this process, the most
significant free parameters are:
• Tool rotational speed
• Tool feed rate
• Toolpath strategy (e.g. raster or offset, see Figure 2)
• Toolpath parameters (e.g. stepover, see Figure 2)
Designing the manufacturing process actually means choosing the right
combination of these parameters that provides a process yield (however defined) as
large as possible.

CAM
Figure 2 – Optimization parameters: green lines represent different possible toolpaths for
finishing the pocket surface

The optimal toolpath is the one that provides the best yield, given a particular
metric that could be different from component to component, but also in different
parts of the same piece. For example, the optimum metric can be either minimum
roughness, minimum machining time, maximum tool life, minimum cost per piece,
minimum shape errors, minimum energy consumption, and so on. Moreover, there
could be an overall optimum metric (e.g. minimum cost) and one or more detail
CAM Operator optimum metrics (e.g. minimum roughness on the pocket surface).
It is important to note that today these optima are pursued on the basis of a trial
and error procedure that involves the designer of the toolpaths (that works on the so
called Computer Aided Manufacturing/CAM software), the machine tool operator,

CONFIDENTIAL 4/11

Thursday, May 20, 2010

 paranoid) confidentiality level would mean minimal (even none) advantage from
cloud manufacturing.

Toolpath optimal planning and control

The milling process involves a rotating tool that moves inside the raw piece thus
removing the excess material with respect to the final geometry that we want to

The Machining Process

obtain.
The first step in designing a milling process is the choice of the right tool on the
basis (mainly) of the workpiece material and of the type of operation (e.g. face
milling, drilling, threading, contour shaping, and so on).
The second step is the computation of the toolpath. In this process, the most
significant free parameters are:
• Tool rotational speed
• Tool feed rate
• Toolpath strategy (e.g. raster or offset, see Figure 2)
• Toolpath parameters (e.g. stepover, see Figure 2)
Designing the manufacturing process actually means choosing the right
combination of these parameters that provides a process yield (however defined) as
large as possible.

CAM
G-Code
Figure 2 – Optimization parameters: green lines represent different possible toolpaths for
finishing the pocket surface

The optimal toolpath is the one that provides the best yield, given a particular
metric that could be different from component to component, but also in different
parts of the same piece. For example, the optimum metric can be either minimum
roughness, minimum machining time, maximum tool life, minimum cost per piece,
minimum shape errors, minimum energy consumption, and so on. Moreover, there
could be an overall optimum metric (e.g. minimum cost) and one or more detail
CAM Operator optimum metrics (e.g. minimum roughness on the pocket surface).
It is important to note that today these optima are pursued on the basis of a trial
and error procedure that involves the designer of the toolpaths (that works on the so
called Computer Aided Manufacturing/CAM software), the machine tool operator,

CONFIDENTIAL 4/11

Thursday, May 20, 2010

 paranoid) confidentiality level would mean minimal (even none) advantage from
cloud manufacturing.

Toolpath optimal planning and control

The milling process involves a rotating tool that moves inside the raw piece thus
removing the excess material with respect to the final geometry that we want to

The Machining Process

obtain.
The first step in designing a milling process is the choice of the right tool on the
basis (mainly) of the workpiece material and of the type of operation (e.g. face
milling, drilling, threading, contour shaping, and so on).
The second step is the computation of the toolpath. In this process, the most
significant free parameters are:
• Tool rotational speed
• Tool feed rate
• Toolpath strategy (e.g. raster or offset, see Figure 2)
• Toolpath parameters (e.g. stepover, see Figure 2)
Designing the manufacturing process actually means choosing the right
combination of these parameters that provides a process yield (however defined) as
large as possible.

CAM
G-Code
Figure 2 – Optimization parameters: green lines represent different possible toolpaths for
finishing the pocket surface

The optimal toolpath is the one that provides the best yield, given a particular
metric that could be different from component to component, but also in different
parts of the same piece. For example, the optimum metric can be either minimum
roughness, minimum machining time, maximum tool life, minimum cost per piece,
minimum shape errors, minimum energy consumption, and so on. Moreover, there
could be an overall optimum metric (e.g. minimum cost) and one or more detail
CAM Operator optimum metrics (e.g. minimum roughness on the pocket surface).
It is important to note that today these optima are pursued on the basis of a trial
and error procedure that involves the designer of the toolpaths (that works on the so
called Computer Aided Manufacturing/CAM software), the machine tool operator,

CONFIDENTIAL 4/11

MT Setup + test runs

Thursday, May 20, 2010
 paranoid) confidentiality level would mean minimal (even none) advantage from
cloud manufacturing.

Toolpath optimal planning and control

The milling process involves a rotating tool that moves inside the raw piece thus
removing the excess material with respect to the final geometry that we want to

The Machining Process

obtain.
The first step in designing a milling process is the choice of the right tool on the
basis (mainly) of the workpiece material and of the type of operation (e.g. face
milling, drilling, threading, contour shaping, and so on).
The second step is the computation of the toolpath. In this process, the most
significant free parameters are:
• Tool rotational speed
• Tool feed rate
• Toolpath strategy (e.g. raster or offset, see Figure 2)
• Toolpath parameters (e.g. stepover, see Figure 2)
Designing the manufacturing process actually means choosing the right
combination of these parameters that provides a process yield (however defined) as
large as possible.

CAM
G-Code
Figure 2 – Optimization parameters: green lines represent different possible toolpaths for
finishing the pocket surface

The optimal toolpath is the one that provides the best yield, given a particular
metric that could be different from component to component, but also in different
parts of the same piece. For example, the optimum metric can be either minimum
roughness, minimum machining time, maximum tool life, minimum cost per piece,
minimum shape errors, minimum energy consumption, and so on. Moreover, there
could be an overall optimum metric (e.g. minimum cost) and one or more detail
CAM Operator optimum metrics (e.g. minimum roughness on the pocket surface).
It is important to note that today these optima are pursued on the basis of a trial
and error procedure that involves the designer of the toolpaths (that works on the so
called Computer Aided Manufacturing/CAM software), the machine tool operator,

CONFIDENTIAL 4/11

MT Setup + test runs

Inspection
Thursday, May 20, 2010
 paranoid) confidentiality level would mean minimal (even none) advantage from
cloud manufacturing.

Toolpath optimal planning and control

The milling process involves a rotating tool that moves inside the raw piece thus
removing the excess material with respect to the final geometry that we want to

The Machining Process

obtain.
The first step in designing a milling process is the choice of the right tool on the
basis (mainly) of the workpiece material and of the type of operation (e.g. face
milling, drilling, threading, contour shaping, and so on).
The second step is the computation of the toolpath. In this process, the most
significant free parameters are:
• Tool rotational speed
• Tool feed rate
• Toolpath strategy (e.g. raster or offset, see Figure 2)
• Toolpath parameters (e.g. stepover, see Figure 2)
Designing the manufacturing process actually means choosing the right
combination of these parameters that provides a process yield (however defined) as
large as possible.

CAM
G-Code
Figure 2 – Optimization parameters: green lines represent different possible toolpaths for
finishing the pocket surface

The optimal toolpath is the one that provides the best yield, given a particular
metric that could be different from component to component, but also in different
parts of the same piece. For example, the optimum metric can be either minimum
roughness, minimum machining time, maximum tool life, minimum cost per piece,
minimum shape errors, minimum energy consumption, and so on. Moreover, there
could be an overall optimum metric (e.g. minimum cost) and one or more detail
CAM Operator optimum metrics (e.g. minimum roughness on the pocket surface).
It is important to note that today these optima are pursued on the basis of a trial
and error procedure that involves the designer of the toolpaths (that works on the so
called Computer Aided Manufacturing/CAM software), the machine tool operator,

CONFIDENTIAL 4/11

MT Setup + test runs

Inspection
Thursday, May 20, 2010
 paranoid) confidentiality level would mean minimal (even none) advantage from
cloud manufacturing.

Toolpath optimal planning and control

The milling process involves a rotating tool that moves inside the raw piece thus
removing the excess material with respect to the final geometry that we want to

The Machining Process

obtain.
The first step in designing a milling process is the choice of the right tool on the
basis (mainly) of the workpiece material and of the type of operation (e.g. face
milling, drilling, threading, contour shaping, and so on).
The second step is the computation of the toolpath. In this process, the most
significant free parameters are:
• Tool rotational speed
• Tool feed rate
• Toolpath strategy (e.g. raster or offset, see Figure 2)
• Toolpath parameters (e.g. stepover, see Figure 2)
Designing the manufacturing process actually means choosing the right
combination of these parameters that provides a process yield (however defined) as
large as possible.

CAM
G-Code
Figure 2 – Optimization parameters: green lines represent different possible toolpaths for
finishing the pocket surface

W !
The optimal toolpath is the one that provides the best yield, given a particular

SLO
metric that could be different from component to component, but also in different
parts of the same piece. For example, the optimum metric can be either minimum
roughness, minimum machining time, maximum tool life, minimum cost per piece,
minimum shape errors, minimum energy consumption, and so on. Moreover, there
could be an overall optimum metric (e.g. minimum cost) and one or more detail
CAM Operator optimum metrics (e.g. minimum roughness on the pocket surface).
It is important to note that today these optima are pursued on the basis of a trial
and error procedure that involves the designer of the toolpaths (that works on the so
called Computer Aided Manufacturing/CAM software), the machine tool operator,

CONFIDENTIAL 4/11

MT Setup + test runs

Inspection
Thursday, May 20, 2010
 paranoid) confidentiality level would mean minimal (even none) advantage from
cloud manufacturing.

Toolpath optimal planning and control

The milling process involves a rotating tool that moves inside the raw piece thus
removing the excess material with respect to the final geometry that we want to

The Machining Process

obtain.
The first step in designing a milling process is the choice of the right tool on the
basis (mainly) of the workpiece material and of the type of operation (e.g. face
milling, drilling, threading, contour shaping, and so on).
The second step is the computation of the toolpath. In this process, the most
significant free parameters are:
• Tool rotational speed
• Tool feed rate
• Toolpath strategy (e.g. raster or offset, see Figure 2)
• Toolpath parameters (e.g. stepover, see Figure 2)
Designing the manufacturing process actually means choosing the right
combination of these parameters that provides a process yield (however defined) as
large as possible.

CAM
H u m a n Figure 2 – Optimization parameters: green lines represent different possible toolpaths for
G-Code

e !
finishing the pocket surface

Ex pe r tis LO W !
The optimal toolpath is the one that provides the best yield, given a particular
metric that could be different from component to component, but also in different

CAM Operator
S parts of the same piece. For example, the optimum metric can be either minimum
roughness, minimum machining time, maximum tool life, minimum cost per piece,
minimum shape errors, minimum energy consumption, and so on. Moreover, there
could be an overall optimum metric (e.g. minimum cost) and one or more detail
optimum metrics (e.g. minimum roughness on the pocket surface).
It is important to note that today these optima are pursued on the basis of a trial
and error procedure that involves the designer of the toolpaths (that works on the so
called Computer Aided Manufacturing/CAM software), the machine tool operator,

CONFIDENTIAL 4/11

MT Setup + test runs

Inspection
Thursday, May 20, 2010
 The Machining Process
Roughing Finishing
Path generation
Simulation
Kinematic

Thursday, May 20, 2010

 The Machining Process
Roughing Finishing
Simulation
Kinematic

✤ Only “nominal” optimization

✤ Trial-and-error approach:
✤ short loop (within CAM system)
✤ long loop (with shopfloor)

Thursday, May 20, 2010

 The Machining Process

Factors (noise affected)

Yield
Parameters Process or
State variables

Thursday, May 20, 2010

 The Machining Process

✤ Factors influencing process:

✤ work material properties

✤ machine tool dynamics

✤ nominal tool shape and material

✤ tool wear

Thursday, May 20, 2010

 The Machining Process

✤ Process parameters:

✤ cutting speed
(~ tool rotational speed)

✤ chip cross-section geometry

(~ feedrate and depth-of-cut)

✤ lubrication/
lubrorefrigeration

Thursday, May 20, 2010

 The Machining Process

✤ Process state variables:

✤ spindle torque
✤ absorbed axes power
✤ vibration magnitude and spectra
✤ material removal rate (MRR)
✤ working volume occupancy
✤ surface quality
✤ geometrical tolerances

Thursday, May 20, 2010

 The ASSET Machining Process

Process
Targets
model
CAD model++

State
Optimiz’n
perception

Path Autonom.
generation learning

MT Setup
Thursday, May 20, 2010
 The ASSET Machining Process

Design of process first “tentative solution”

Autonomous learning
Four layers action

In-cycle continuous optimization

Trans-cycle optimization

Surveillance and monitoring

Thursday, May 20, 2010

 The ASSET Machining Process

Design of process first “tentative solution”

Autonomous learning
Thursday, May 20, 2010
 The ASSET Machining Process

In-cycle continuous optimization

Autonomous learning
Spindle
torque

Model 12 Nm
Reduce
feedrate
Actual 15 Nm

Thursday, May 20, 2010

 The ASSET Machining Process

Trans-cycle optimization

Autonomous learning
Inappropriate
measured
crest height Modify
toolpaths

Thursday, May 20, 2010

 The ASSET Machining Process

Surveillance and monitoring

Autonomous learning
✤ Tool breakage ✤ Safe escape

✤ Excess vibration ✤ Path recalculation

or chatter
✤ Approach and
✤ Risk of collision restart

Thursday, May 20, 2010

 ASSET M.T. Architecture

Cloud
Knowledge CNC/PLC

Process Machine
STEP-NC tool
controller

S.Fusion Sensors
controller

Thursday, May 20, 2010

 ASSET Subprojects

Thursday, May 20, 2010

 Autonomous Process Planning
AP
P

✤ CAD Model and metadata import

✤ Process model and optimal toolpath planning

✤ In-process qualification of model parameters

✤ MT/CNC interfaces, toolpath modification tactics

Thursday, May 20, 2010

 Perception System
PS

✤ Sensor fusion algorithms and strategies

✤ HW architecture

✤ Sensor collection

✤ Platform for rapid SF tailoring on different MTs

Thursday, May 20, 2010

 Cognitive System
CS

✤ Classification/regression (reinforcement learning)

✤ Compensation of missing data

✤ Targets

Thursday, May 20, 2010

 Cloud Manufacturing
CM

✤ Design of cloud infrastructure (paranoid environment!)

✤ Definition of protocols and interfaces

✤ Abstraction and support libraries

✤ Application to simpler designs (standard tool/material tables)

Thursday, May 20, 2010

 Workpackages
Just a rough subdivision

Thursday, May 20, 2010

 Workpackages

1) Management ? 7) Cloud manufacturing CM

2) Req’ments & Specs. AL

L 8) Data import/processing PS

(STEP-NC)
3) System and process simul’nAP
P
9) CNC and MT interfaces AP
P

4) Process modelling & Optimal

path planning APP 10) Prototypes AP
P PS CM

5) Perception subsys PS 11) Dissemination AL

L

6) Cognitive subsys CS

Thursday, May 20, 2010

 Technology Demonstrators
First NC prototype, Parsons and MIT labs, 1952

Thursday, May 20, 2010

 Technology Demonstrators

I. Lathe (Gildemeister)

II. 5-axes milling centre (Deckel-Maho)

III. Reduced cloud-connected machine tool (e.g. laser cutting)

IV. Cloud infrastructure

Thursday, May 20, 2010

 A note about the name

Advanced
Safety and Driver
Support for
Essential Road
Transport

It is on another call.
Should we change the name/acronym?

Thursday, May 20, 2010

 Comments by missing partners

✤ DMG Electronics:

✤ Available as WP/SP coordinator, not to lead consortium

✤ Have willingness to participate in almost every WP
✤ Are open towards opening consortium to other partners

✤ manufacturers of large machine tools

✤ manufacturers of laser cutting machines

✤ tool manufacturers

Thursday, May 20, 2010

 Comments by missing partners

✤ Delcam:

✤ says «interesting consortium», but advises about opportunity of having more

partners:
✤ “machining partners” (e.g. cutting tools)
✤ real, standard end user like Hide Group (more mass-production oriented than Shadow
Robot)

✤ warns about difficulties of «Flowcharts/methodologies/ideas transforming

in working software»
✤ wants not to «take a dominant role», however
✤ is willing to have a pro-active role, since it looks like toolpath generation
plays a central role

Thursday, May 20, 2010

 What’s next

✤ During this meeting:

✤ Partners present themselves (in ASSET-oriented way!)
✤ Discussion about proposed WP/SP structure
✤ Candidates as SPs coordinators
✤ Before the call:
✤ Each partner states his commitment in the project
✤ Coordinating partners develop WBS and define budgets for each SP
✤ After the call:
✤ Main coordinator writes the proposal, by merging contributes of partners

Thursday, May 20, 2010

 UniTN expertises
✤ Group of Mechatronics (3 active FP7):

✤ Process modeling and simulation

✤ Symbolic and numerical modeling of multibody/multiphysics systems
✤ Identification of dynamic systems
✤ Parameter optimization
✤ Formulation and solution of optimal control problems for mechanical systems
(also for real-time applications)
✤ High-performance custom optimization algorithms (C++)
✤ Sensor-fusion, SLAM, multistereo image analysis

✤ ELEDIA Group (ELEctromagnetic DIAgnostic Laboratory)

Thursday, May 20, 2010
 UniTN expertises

✤ ELEDIA Group (ELEctromagnetic DIAgnostic Laboratory) (2 active FP7)

✤ Optimization (local and global iterative methods)

✤ Learning-by-example (SVM, NN, Gaussian Processes)
✤ Cognitive Control (Reinforcement Learning, Particle Swarm Opt’n-based
Control)

✤ Within ASSET:

✤ Analysis and Understanding of Sensed Data:

classification, regression, and compensation of missing data
✤ Cognitive Control Algorithms:
non supervised experience exploitation and exploration

Thursday, May 20, 2010