- Introduction/Abstract
Surface Mount Technology, SMT, is increasingly the dominant methodology
in the assembly of sophisticated electronic devices. It involves bonding electronic
packages to substrates via solder paste pads. These solder deposits are accurately
placed on the substrate by screen printing. Faults in the printing process
are a major source of board failure. However, understanding these failures
is a challenging problem as the printing process has a large number of non
linearly dependent variables such as factors relating to paste (formulation,
viscosity), the environment (temperature, humidity) and printing machine technology
(alignment, pressure and speed of squeegee, blade ha,rdness etc). Nevertheless,
if solder paste printing can be better understood and monitored then there
is considerable scope for adaptive process control which would produce economic
benefits by leading to enhanced product yield.
The ASPMIC project intends to integrate process monitoring and inspection
techniques with research on solder printing process models to provide information
for a fuzzy rule based control system that will adaptively control the printing
process. It aims to demonstrate the feasibility of using adaptive control
for the improvement of the solder paste printing process. In order to achieve
these aims, significant advances will have to be made in terms of both science
and technology. These will include development of: a prototype range sensor
suitable for real-time inspection of solder pad boards, algorithms for analysis
of solder pad shape, novels methods for monitoring of printing process variables,
enhancement of models of the solder paste printing process and finally, a
fuzzy rule based adaptive control system. Such wide reaching aims necessitate
the participation of the diverse specialist teams that form the ASPMIC consortium.
-
Scientific/Technological Relevance
It is estimated that in 1995 Surface Mount Technology accounts for about
75% of the market in PCBs and by 1997 its worldwide use will grow to 800 billion
pieces per year. Although SMT is an established manufacturing method, the pressure
to increase component densities and decrease lead pitches means that even though
the solder joint failure rate is typically only 100 ppm, the board failure rate
may approach 50 %. Studies have shown that over 63% of defects identified after
reflow originate during the solder paste printing step and his (together with
the ease and low cost of reworking bare solder boards) suggests that improvement
of the solder paste printing process is one of the most important ways of increasing
product yield.
Improving solder paste printing yield requires three interdependent components.
Firstly, enhancement of fundamental understanding of the printing of solder
paste and development of better mathematical process models. Secondly, faster,
more accurate and more consistent ways of measuring both the process and its
end product. Thirdly, development of flexible adaptive control algorithms
that can exploit improvements in the first two items but are robust in the
sense that they gracefully degrade as a function of the inevitable incompleteness
and imprecision of that knowledge. These three components match naturally
with the specialist interests of the three academic partners within ASPMIC.
University of Salford is interested
in solder process modelling, University
of Surrey are developing advanced inspection techniques and Nottingham
Trent University work in the area of adaptive control. In the sub-sections
below each of the three areas are considered in turn and the state of the
art is discussed to identify the critical issues that the ASPMIC consortium
hopes to address.
Solder Paste Print Modelling
For many industrial practitioners the printing of solder paste for reflow
soldering of SMT assemblies has been regarded as a black art. However, recent
research has cast considerable new light on the process. Internationally,
there is significant work in both the CATERU group at University College Galway
in Ireland and the Universal Group at Binghamton in the States. Within the
U.K. significant results have been achieved (under SERC grant GR/H 32939)
by the University of Salford group.
They have advanced understanding considerably and have produced quantitative
models of many of the component processes. This has resulted in an on-line
Intelligent Advisor system which can perform functions such as predicting
defects for specific printer settings, diagnosing of printer faults and provision
of guidance on solder paste selection and printer set-up. The system has attracted
much industrial interest and is currently being validated. However, the models
used within the system are derived from manually acquired data and experiments
which isolate particular. aspects of a much richer process. Inevitably, the
current models neglect some higher order interactions and would benefit from
better data to instantiate them. Enhanced process models can be best developed
within a more fully automated printing system of the type that the APSMIC
project envisages.
Product Inspection and Process Monitoring
There is much interest in industrial inspection and monitoring for defect
recognition. Less work has been devoted to inspection for process control.
In the ASPMIC project a major area will be the geometric inspection of solder
pads using range data. This is not entirely novel but the few commercial systems
which exist are expensive, slow and, most importantly for process control,
limited in the information which they provide. For example, the SVS system
uses a laser spot scanner and although it costs nearly 150,000 can only achieve
a maximum speed of 30 pads per second. The information that it provides is
volume, area and height of solder pad. This is sufficient to detect some fault
classes but inadequate for process control purposes. Important shape information
relating to aspects such as scooping and skipping of solder paste or angle
of solder is not extracted. The ASPMIC project aims to develop advanced analysis
algorithms which will provide much richer information for process control
and detailed quantitative understanding of the solder paste printing process.
It will also provide a route for the routine collection of the large amounts
of'data necessary to develop and initialise improved mathematical models.
In addition to their general expertise in the inspection field, Surrey
University are well placed to advance this work as they have just completed
an EPSRC grant concerned with inspection of component mounted PCB boards using
range data.
Adaptive Process Control
The solder printi g process has a large number of inter-dependent variables
including factors which relate to paste (formulation, viscosity), the environment
(ambient temperature and humidity) and the printing machine technology (alignment,
pressure and speed of squeegee, blade hardness etc). Some of these are directly
controllable but the efect of others has to be compensated indirectly. Although
adaptive control is a subject with a long and rich history, the majority of
successful solutions involve systems where there is complete and accurate
system knowledge and an approximately linear response between control variables
and process response. In the solder paste printing process both of these assumptions
are unrealistic and therefore conventional control methods are inappropriate.
It is therefore proposed to investigate a fuzzy logic approach to control.
The essence of the approach is for a collection of reasonable control decisions
to be derived from incomplete knowledge of the system and for the weighted
combination of these results to be used to determine control outputs. The
approach has proven efective in a range of dificult control problems which
share the characteristics of the solder printing process. The Nottingham
Trent group are particularly well placed to explore this methodology as
they have been using it in the very closely related problems of machine control.
Initially, it is anticipated that a limited fixed rule-set will give reasonable
performance but further improvements may be investigated by adopting a hybrid
neural network/fuzzy approach.
Relevance to Beneficiaries
The consortium includes both academic and industrial collaborators. The industrial
partners are interested in the research as they can see that it relates to one
of the dominant technologies in the electronic manufacturing area and to a difficult
but important problem whose solution would quickly realise commercial benefits.
The participating companies will benefit through having first exposure to the
reseach and access to prototypes developed. Specifically, 3D Scanners intend
to contribute greatly to sensor development work and use it as the basis for
a generic low cost, high speed scanner for high resolution work in the electronic
manufacturing industry. DEK Printing Machines would like to augment existing
machine with more sophisticated monitoring and control capabilities. The two
user companies, Multitone and D2D, form potential customers for developed products
and hope to gain an increased understanding of their manufacturing process so
as to produce higher product yields. More generally, both the UK electronics
industry and its academic community will benefit as the work will be published
in the open literature, subsequent to steps to protect IPR and commercial interests.
Dissemination and Exploitation
The work will be disseminated via standard academic and commercial routes
i.e. trade journals, archival academic journals, conference and exhibition presentations.
Where necessary, work will be protected by commercial licensing agreements or
via patenting. Individual IPR agreements will be negotiated as appropriate for
pieces of work related to each partners activities.
The Programme
Objectives
The aims of the project are to improve fundamental understanding of the
solder paste printing process and to apply this new knowledge in an effective
way for real-time, adaptive feedback control of the process. To achieve this
requires work in several areas including:
- study of the interactions between process variables, process parameters
and printing defects to enable enhancement of existing process models,
- development of automated inspection techniques and on-line sensing
of process variables for effective real-time monitoring of the process,
- development of advanced data analysis algorithms for converting
raw inspection data into useful features for control,
- research into algorithms for interpreting the sensed data over time
and integrating it into the process models in the most effective way,
- development of control strategies for real-time updating of the
printing process.
Members of the consortia cover the full range of skills required to address
the aims of the project and in several cases have existing collaborative work
together (USUR/3DS, USAL/DEK/D2D, DEK/MLT). The University of Surrey
and 3D Scanners have been heavily involved in range image sensing and
analysis (which is the primary inspection technique proposed),
Universitu of Salford and DEK have extensive experience in solder paste
printer modelling and Nottingham Trent
has been conducting industrially sponsored research on understanding and
adaptively controlling printing machines. Both D2D and Multitone have significant
manufacturing capabilities and will play an essential role in providing access
to realistic, large scale systems for validating the research.
Workpackages
The research programme involves three academic sites and they are requesting
8 man-years of effort spread over a two year period. The choice of a two year
project is motivated by the fact that the academic teams already have considerable
experience to enable a prompt start to the work and also there is a desire
to provide commercially exploitable results within the relatively short timescales
characteristic of the electronics industry marketplace.
The division of work among partners follows their areas of established
expertise. University of Salford
will be responsible for process model development while the University
of Surrey will develop the hardware and advanced analysis algorithms
for real-time solder paste inspection. Nottingham
Trent University will integrate these elements into a process control
system. There will be significant interaction between the groups and industrialists
as each must relate their work to the findings of others and integrate them
into the ma,nufacturing process.
Printing Process Modelling (USAL):
The relationship and interactions between printing process variables,
parameters and their efect on final product will be studied. As mentioned
previously, up to now this has been possible only by laborious and manually
intensive experiments. The automation of the measurement and inspection
processes will enable more consistent data to be collected and therefore
process models can be enhanced. It will then be possible to look for higher
order effects and to consider studying how faults/defects propagate through
the various printing sub-processes which form the full process model.
Process Monitoring and Inspection (USUR):
This workpackage is concerned with measuring process variables and the
finished solder paste product. As many of the process variables are sensed
on existing printing machines, the major area of work will be development
of a low cost, high accuracy range scanner for solder pad inspection. This
is based on a novel design that exploits recent CCD array technology. It
will be coupled to a specialised DSP based processing architecture to provide
the computational power to meet the real-time requirements of the inspection
and control task. The specification and design will be heavily induenced
by 3D Scanners considerable expertise in this area. NTU
will perform some work on instrumenting a printer to measure squeegee
pressure as in their experience this is an important factor in the printing
process.
Advanced Data Analysis Algorithms (USUR):
Although range data provides a rich source of geometric information,
thus far only simple measures such as solder area and volume have been used
in defect classification. In this workpackage more advanced shape features
such as moments, morphological shape descriptors or edge and curvature based
features will be extracted and assessed for use in defect classification
and process measurement. In addition, work will be carried out to determine
the best measurements and features which can be used for feedback control
purposes. This will require more subtle discrimination than simple defect
detection.
Adaptive Process Control ( NTU ):
Key process parameters will be measured and the relationship with the
process determined. The results from the printing process model work package
will be analysed and used as the basic data to be encoded in a fuzzy rule
base enabling online control of the influential process parameters. This
will be further enhanced by the inclusion of primary results from the process
monitoring and inspection package, which provides the linkage of process
variables with output characteristics. The fuzzy control will select the
best set of feasible adjustments that bring the process to its optimal performance.
System Integration and Validation (USAL/USUR/
NTU ):
A significant activity will involve aspects of integrating the hardware,
algorithms and studies developed within the project. Component processes
will be developed to facilitate easy adoption of the work within the industrial
environment. For example, careful attention will be given to development
of interfaces and visualisation software which aids interpretation by trained
human operators. Extensive work will also be required to carry out field
trials in collaboration with the industrial partners.
Industrial Contributions
Each of the industrial partners has an essential role to play in the project.
3D Scanners will be highly involved in sensor design and will be responsible
for supplying range data capture facilities prior to the commissioning of
the prototype sensor. DEK will provide expertise on printing equipment, parameter
control and technical advice on solder paste printing. Both D2D and Multitone
will provide samples of solder paste boards for test purposes and access to
equipment and manufacturing lines for field testing. The DEK site at Weymouth
provides a focus and demonstration facilities for testing the adaptive control.
DEK, D2D and Multitone will all provide expertise on current problems and
trends in the SMT market and advise how the research is addressing them.
Management and Resources
Each of the sites has a Technical Manager nominated as follows: Dr Ndy Ekere
(USAL), Dr John
Illingworth (USUR), Mr
Martin Howarth ( NTU ), Mr S Crampton
(3DS), Mr Alan Harling (DEK), Mr Mike Waters (Multitone) and Mr Philip Hamilton
(D2D). The role of these managers is to oversee the smooth running of the work
at each of their sites and to attend regular management meetings where progress
will be reviewed and future planning will be discussed. Regular meetings will
take place every 4 months with partners providing a short written report of
progress at their site. These meetings may occupy only half a day but will be
accompanied by complementary sessions involving detailed technical presentations
and discussions of work in progress. Other technical meetings will take place
at more frequent intervals between partners who are pursuing related pieces
of work. In the second year integration is likely to necessitate longer visit
to the integration site by Research Fellows.
Staff:
The University of Salford has one
research Associate for two years to work on printing process model enhancement.
Nottingham Trent has one researcher
working on monitoring and control. University
of Surrey has two Research Associates. The first will primarily be responsible
for design and development of the prototype sensor and associated processor
architecture while the second will be concerned with analysis and interpretation
of the range data. University of Surrey
also requested technical staff to assist with sensor development. Each
of the industrial partners will provide several months of manpower.
Equipment:
At University of Salford an upgraded
solder paste printing facility has been installed so that experiments can
be performed for enhancing process models. Nottingham
Trent are improving instrumentation on their printing machine and implementing
PLC control for the final system. A high specification PC is to be assigned
to run design and analysis software. Surrey
University are to build the customised prototype laser range scanner and
associated dedicated processor architecture. This prototype is derived from
a feasibility study conducted in collaboration with 3D Scanners.
The industrial partners will be providing capital equipment of considerable
value to the project. 3D Scanners will contribute an existing scanner which
can be used at the early stages of the project for gathering test data.
Initial experiments show that it provides sufficient resolution for this
purpose but will not be suitable for the detailed work required towards
the end of the project. In addition, 3D Scanners will provide a FPGA development
kit and associated PCB design and layout software. DEK and Multitone will
provide and refurbish/upgrade a manual printing machine suitable for initiating
work on the project.
Download a preprint copy of our book in postscript format: