Upload
tobias-schroeder
View
14
Download
3
Embed Size (px)
DESCRIPTION
A MULTIFACTOR EXPERIMENT IN AN INDUSTRIAL ENVIROMENT: SOME PECULIAR PROBLEMS. Ing. Marco Testa – Pianelli&Traversa S.p.A. Prof. Grazia Vicario - Dipartimento di Matematica Prof. Raffaello Levi - Dip. di Sistemi di Produzione ed Economia dell’Azienda Politecnico di Torino. Agenda. - PowerPoint PPT Presentation
Citation preview
Cagliari, May, 31, 2003 1
A MULTIFACTOR EXPERIMENT IN AN INDUSTRIAL ENVIROMENT: SOME
PECULIAR PROBLEMS
Ing. Marco Testa – Pianelli&Traversa S.p.A.
Prof. Grazia Vicario - Dipartimento di Matematica
Prof. Raffaello Levi - Dip. di Sistemi di Produzione ed Economia dell’Azienda
Politecnico di Torino
Cagliari, May, 31, 2003 2
Agenda
Introduction and background
Objects of the experiment
Factors, Responses and Design
Statistical analysis and modeling
Final remarks and discussion
Cagliari, May, 31, 2003 3
Introduction and background (1)
Production process is under control, product
quality meets requirements; yet management
feel there’s definitely room for improvement, in
terms of production rate, costs, and product
performance.
O.V.a.T. experiments (One Variable at a
Time) were unable to deliver meaningful
information, so the proposal of trying a new tack
was accepted.
Cagliari, May, 31, 2003 4
Introduction and background (2)
Experiments were performed on motor car
subassemblies, manufactured as production items
and tested to destruction. Some constraints led to
an odd plan, that is namely a replicated fractional
factorial. Sizable interactions, peculiar failure
modes and scatter pattern suggested critical
analysis of results.
Cagliari, May, 31, 2003 5
Objects of the experiment (1)
Metal and glass components are bonded by a
plastic interface, cured under controlled
temperature and pressure in an automated
robotic cell.
Proprietary end effectors handle components.
A heating unit and a pneumatic actuator act under robot control in a tight sequence.
Cagliari, May, 31, 2003 6
Objects of the experiment (2)
Storage affects to some extent bond resistance, as evaluated by peel and torsion tests.
Tensile load increases up to failure at constant rate at room temperature.
As a rule failure occurs in glass component, as both metal part and bond exhibit superior resistance.
Cagliari, May, 31, 2003 7
Typical Initial Problems ….
Fear of “new methods”, especially if rather different from traditional ones Lack of awareness of importance of interactions among concerned people Managerial resistance to shift from traditional O.V.a.T. strategy to D.o.E. Blind faith in former beliefs not supported by experimental evidence Lack of specific know-how – expert procedure
had to be explained in detail, step by step
Cagliari, May, 31, 2003 8
… and Steps to Overcome Them
Gather a solid knowledge of process, best with the help of an insider
Track down and collect systematically all relevant information
Stimulate key player’s curiosity and interest about experimental approach
Group meetings proved helpful, a blend of technical knowledge and intuition.
Cagliari, May, 31, 2003 9
Factors, Responses and Design (1)
Quantitative Factors
Plastic film temperature at assembly
Conditions of fitting plastic film to metal component
Preheating temperature of glass component prior to assembly
Pressure applied to components during assembly
Duration of application of pressure during assembly
Plastic film thickness
Heat treatment parameters
Glass component shape
Drying cycle parameters
Cagliari, May, 31, 2003 10
Factors, Responses and Design (2)
Qualitative factors
Surface conditioning of metal component
Shape of metal component
Surface treatment of metal component
Type of plastic film
Presence/absence of enamel on glass
Percentage humidity on stored plastic film
Cagliari, May, 31, 2003 11
Factors, Responses and Design (3)
Factors considered in the experiment Plastic film temperature at assembly: Substantial effects expected, no affect cycle time, lower limit getting film “sticky”, upper limit melting temperature
Force applied to components during assembly: upper limit resistance of glass, loading rate constant over all tests
Duration of application of force during assembly: covers bonding phase, for tightly scheduling production Surface conditioning of metal component: comparison with standard procedure in terms of cost and performance
Surface treatment of metal component: for solving problems with unprotected metal surfaces (possible trouble)
Cagliari, May, 31, 2003 12
Factors, Responses and Design (4)
List of responses
Time to failure under constant load at elevated temperature: evaluation in creep tests, either short term (LSL one hour) or long
term (LSL three days) Static tension tests: repeatable thanks to self aligning features, may fail to evaluate ultimate bond resistance, retained for checking purposes only
N.B. Resistance in torsion was not considered among responses analysed owing to poor repeatibility, traceble to excessive sensitivity to minor misalignment of test fixture
Cagliari, May, 31, 2003 13
Factors, Responses and Design (5)
Possible initial model
Educated guesses could be obtained about some main effects;
a priori estimates of existence, let alone magnitude, of
interactions were not available, some suspects being however
entertained by experienced foremen.
A simple model was therefore initially considered, catering for
linear combinations of single effects and two factor interactions only.
Cagliari, May, 31, 2003 14
Factors, Responses and Design (6)
Initial design:
251 fractional design
Resolution IV (I = ABCDE)
two replications
Cagliari, May, 31, 2003 15
Problems with Communication
To exploit results people must grasp their meaning – and if they are
to understand facts must be explained appropriately, that is in a
crystal clear way. Easier said than done, especially on production floor.
All sorts of graphs, charts, diagrams are helpful – if only because
most people try hard to understand what they feel you do your best
to explain.
On the other hand, to expose any audience to statistical treatment
beyond their grasp is self defeating, as it seldom fails to originate
antagonistic attitudes.
Cagliari, May, 31, 2003 16
Some final Remarks
Results were accepted with positive remarks by managers concerned. That such an amount of meaningful
information could be inferred from such a small set of
data came as a pleasant surprise to plant management,
and the reward came as a request to carry on with
further investigations.