-Phase 1 Ideation Submission Requirements

aspiringtokΤεχνίτη Νοημοσύνη και Ρομποτική

15 Οκτ 2013 (πριν από 3 χρόνια και 2 μήνες)

65 εμφανίσεις

-
Phase

1



Ideation

Submission

Requirements

A.


COVER PAGE (1 PAGE)





Program Name: MTConnect Challenge





Title of Idea:

Agile Product

and Process

Development using MTConnect





Respondent Name:

Arvind Rangarajan






Organization (if applicable):

PARC





Phone Number:

650 812 4720





Email Address:

arangara
@parc.com





Date of Proposal:
5
/1
6
/2013



B.


SECTION I: ABSTRACT
(200 WORDS MAXIMUM),

PROPOSED IDEA:

A knowledge
-
centric enterprise needs rapid communication
between engineering, planning and
supply chain

focused around
specific
product
s
.
This is typically accomplished by
an
organizational structure dedicated to
managing
new product
s through their life
cycle
. This
paradigm doesn’t extend
well
to high
-
mix low volume or agile production facilities.
We propose
a solution for operators in the shop floor
interfacing seamlessly with MTConnect enabled
machines
to provide feedback to designers and manufacturi
ng
planners with minimal effort. This
will
enable rapid process improvements/redesigns

and build an enterprise knowledge base for
defects and best practice identification
.

Our
approach utilizes MTConnect to establish
the
context
of
the
defect
,

process improvement
,

or
design suggestion
.
In software parlance, “a stack trace” would be automatically created for
operator feedback
.

This can be
package
d

and emailed/made available for retrieval based on a
workflow established in the PLM system.
Over ti
me

machine specific

models

created
through
supervised learning will
automate
this function

resulting in a self
-
reporting intelligent machine
tool
.


The primary benefits of the proposed system would be to accelerate
product yield
learning curve
of enterpri
se
in real
-
time.

This is accomplished by creating a h
igher level of supply chain
involvement
for
product refinement

including cost reduction

and improving manufacturing
efficiency
.


C.


SECTION II: PROPOSED

IDEA:

The goal is to bring “
situational
intelligence” to machines
;

using MTConnect to provide the
context
,

with intelligence provided from multiple sources including the human operator,

simulations u
tilized

during planning, and

algorithms supported by a library of signatures of
faulty behavior
.
This idea is best illustrated with some uses cases of this solution.

Scenario 1


Operator hears a chatter and
uses speech to

report that he heard chatter.
Speech
-
to
-
text

software can create a label to
classify

the defect.
Now, the solution

developed

with

this
approach
would pull the MTConnect data associated with location during program execution,
status of various components, assets used including specific tool, current life of the tool, part
number, lot sizes,
and results

from previous 5 parts at the
sa
me location, any reports on same part
on other machine at the same location
and email it to the manufacturing process engineer.

Scenario 2


Operator applies feed rate override. This can be identified without reporting and
packaged along with component an
d location in tool path to manufacturing process planner to
revise the program supported by simulations to provide feedback to shop floor operator.

This
would be connected to a machine learning system to learn the model for shop floor operation.

There are other situations involving alternate tool recommendations, lack of chip clearance,
excess power drawn, small clearance for part, worn tools,
incorrect coolant
, rapid tool wear
where this information can be used to re
fine manufacturing process pla
n.
T
he operator could
also
provide feedback on excessive burring for certain features, poor finish observed on certain
location to enable designers consider redesign alte
rnatives.

This will enable introducing agile development concepts commonly employed in

software
development and system level problem solving to product design and supply chain enabled by
MTConnect.

D.


SECTION III: TECHNIC
AL REQUIREMENTS

This
solution will be implemented in 2 phases:

Phase 1


Model building

In the first phase, information to supplement observed defect or suggested process improvement
could be a verbal description of the condition from the operator

through speech or selection
from
a set of labels
.
A standard set of or custom defined manufacturing stack trace
MTConnect data
would be collected and bundled along with the operator description of the condition.
This will be
routed to the appropriate manufacturing/design engineer bas
ed on label classification.
This
information would be
used as a training set for
a machine learning framework that analyzes the
MTConnect data to automatically detect signatures that are associated with “labels”

identifying
the issue/recommendation
.
Additional information can be obtained from simulations to predict
expected values of parameters monitored with MTConnect

to train a better model
. This builds
model specific to the machine
representing enterprise knowledge
in a supervised learning
approach

while providing immediate pay back for the factory. At the end of phase1, there would
be a library of models that are machine specific for various common defects encountered during
Figure
1

Agile Product and Process Development using MTConnect

operation. This information can also be mined in other forms, where the la
bels can be used to
provide immediate

feedback
on
prior solutions applied for similar solutions.


Phase 2


Intelligent Machine Tool

An automated diagnosis from machine specific models is utilized and validated by operator
when probability of detection i
s below a threshold.
In the second phase of this implementation,
the role of human operator would be to provide secondary feedback. The knowledge would be
supplemented with relevant MTConnect information from part genealogy and machine history to
create an

intelligent machine that would self
-
report process improvement/defect prevention
opportunities.

An analogy of the first phase of proposed idea would be a program crash reporting tool for
computer aided design (CAD) platform that bundles the component, st
age of operation,
necessary diagnostic information and an automated workflow to direct these reports for bug
fixing and product improvements. The second phase would be similar to automated development
of software design patterns.

E.


SECTION IV: BENEF
ITS

Shop operator utilization
:

Seventy
-
four percent of the survey respondents reported that workforce shortages or skills
deficiencies in production roles are having a significant impact on their ability to expand
operations or improve productivity [1].

Proposed approach will improve the utilization of
skilled
operators in shop floor
, increasing availability of operator while expanding their sphere of
influence within the enterprise
.

The knowledge worker in a factory would now be able to
participate in a

larger segment of product development cycle improving their impact.

Accelerating learning curve
:

Learning curves can vary significantly in a given industry or even within a given organization
across different facilities. Studies by Lapre[2] on organizatio
nal learning cites case studies in
manufacturing where organizational learning curve with in an industry can vary by a factor of 2.
Epple et al. [4] study showed that labor productivity stayed the same for first shift when switched
from operating one shif
t to two shifts at roughly the same rate as the plant did before the switch,
whereas almost no learning occurred on the night shift. Similar studies exist in semiconductor
manufacturing where the starting yields could be off by a factor of 8 and significan
tly different
rates of learning [5]. MTConnect provides a significant opportunity to facilitate learning in the
shop and democratize operator contributions by combining user input with a supervised learning
model.
This could lead up to a 2 fold increase i
n rate of learning.


There is also a direct impact of learning in the shop floor that facilitates acceleration of learning
curve.
A second study by Lapre[3] states that the operationally (shop floor) validated theories
accelerate the learning curve by 3.2
%
, obtaining reduced costs earlier in product life cycle
.




Rapid inclusion of feedback from shop floor operators
using MTConnect
in manufacturing
planning and product redesign would accelerate

and reduce

the
variation in
learning curve
across
operators and shops
by providing necessary information for decision makers by integrating this
to PLM systems and creation of new enterprise knowledge by development of machine models
and fault signatures based on MTConnect data/intelligence.

REFER
ENCES
:

[1]
http://www.themanufacturinginstitute.org/Research/Skills
-
Gap
-
in
-
Manufacturing/2011
-
Skills
-
Gap
-
Report
/2011
-
Skills
-
Gap
-
Report.aspx

[2]
Lapré, Michael A. "Inside the organizational learning curve: Understanding the
organizational learning process."

Foundations and Trends® in Technology, Information and
Operations Management

4.1 (2010): 1
-
103.

[3] Lapré,
Michael A., and Luk N. Van Wassenhove. "Managing learning curves in factories by
creating and transferring knowledge."

California Management Review

46.1 (2003): 53
-
71.

[4]

D. Epple, L. Argote, and K. Murphy, “An empirical investigation of the

microstructure of
knowledge acquisition and transfer through learning by doing,”
Operations Research
, vol. 44, no.
1, pp. 77

86, 1996.

[5] N. W. Hatch and J. H. Dyer, “Human capital and learning as a source of sustainable
competitive advantage,”
Strategic
Management Journal
, vol. 25, no. 12, pp. 1155

1178, 2004.