by Godfrey Hands, Vianen, The Netherlands
About the author
Automation of NDT has both advantages and disadvantages. This discussion article tries to highlight these, and allows NDT users to decide for themselves if they should automate or not. The two main advantages of automation are reproducibility and man-power reduction; the disadvantages are the investment needed to implement this and the complexity of an automated system.
Once the test system is set up, it will (or at least should) produce
reproducible results, 60
seconds every minute, 60 minutes every hour, 24 hours every day. The overall
sensitivity of
the system may be a little less than that of a vigilant operator (but
with signal
processing, could provide significantly superior performance), but the
reproducibility will be
almost 100%. An automated system's performance is, however, so reproducible that if the
operator does not
adjust the machine correctly, it will test all parts incorrectly. For
security, a well
documented set-up procedure should be specified for the machine, to reduce
the risk of this
happening.
The manual operator has very poor performance reproducibility. In many industries,
detection of 80% of
observable defects is
considered the normal performance of a human inspector (whether the system is Ultrasonic,
Eddy Current,
Visual, Magnetic Particle, Penetrant, or any other.)
Let us look at a manual ultrasonic operator testing components for defects.
Under ideal
conditions, such as when he is alert, just starting his shift and has just
calibrated his
instrument and probe, perhaps he can find defects 3mm long and larger.
Assume also that
an automated system can only detect defects of 5mm and longer, and that 3mm
is the
acceptance limit. Which is to be preferred? The more sensitive manual
operator or the
automated system that cannot meet the requirements of the specification
?
Because the manual operator can miss one defect in every 5 (but is less
likely to miss larger
ones than smaller ones), he could potentially accept one component with a
life-endangering
defect for perhaps every 10,000 pieces he tests. If the component that he is
testing is a car
road wheel (for instance), then because each car has 4 wheels in use, one
car in 2 500 has a
life-endangering defect.
What about the automatic system ? In the example given here, it will never miss serious defects, but
cannot find any
of the defects that are between 3mm and 5mm long (example!). How serious is that? The
manufacturer will have a few cases with reduced component life, but
the reduction
will probably only be noticeable in the long term, and when statistically
analyzed. In many
cases, the consumer will attribute the failure of the component to it having
reached the end
of its life. The effect on the manufacturer's sales due to dissatisfied
customers will be
negligible, while the manufacturer using manual operators will not be
in business due to
liability claims.
This is an extreme case, but demonstrates the principles.
For safety-critical parts, where it is not possible to automate the
inspection process, it is
common practice to "double inspect" the parts. This means that an operator
will inspect the
production and reject 80% of the defective parts in the batch. Another
operator will then
reinspect all the components that the first operator accepted.
find 80% of the
remaining defects (leaving about 4% of the defects instead of 20%). An
additional (third)
inspection will reduce this to 0.8% of the defects.
Automation can therefore bring a possible throughput increase into the
manufacturing (or
inspecting) process of up to maybe 50 times, but anyway should not reduce it.
If the manufacturer has large enough production amounts to warrant it, he
will be able to
inspect his parts with less personnel. If he only manufactures small
quantities, and set-up
and calibration times of an automatic system are high, then he is unlikely
to have any
significant benefit in this area.
Let us assume that the manufacturer can reduce his manpower for inspection
by only 50%
through automation. If he only has two operators, he cannot continue
with one tenth of
an operator unless the operator is also employed in other areas of the
factory. Thus, the cost
of one man is (an assumed) $30 000 per year - the limit of the
manufacturer's
direct financial advantage.
If he can reduce his inspection manpower by 5 or even 10 men for the
investment of one
machine, his savings become very significant.
The ideal system is one that is user friendly and not difficult to adjust
for component
changes. It can ideally be computer controlled, with the computer
controlling all necessary
adjustments, based purely on an operator's simple input to the system. This
can be through a
keyboard if he is computer literate, but if the level of personnel to be
employed on such
systems have no potential for computer literacy, then alternative
arrangements have to be
made.
It may not take a scientist to operate the system, but it could possibly
need one to adjust the
system in the first place for different component types. If the system is
user friendly enough,
it may be possible to remotely (or in advance) program the machine so that
the operator
only has to press a button with his name on it to change from the production
types of the last
shift to those that he has to test on his shift.
Lets look at the pipeline inspection situation mentioned above. Supposing
you are presently
manually testing 15 welds per man per day, and you employ 3 Level 2
operators to make the
tests (45 welds per day). An automated inspection system may take only 5
minutes to test
each weld, but 15 minutes to accurately set up. You will still have the
potential to test 25
welds per day per operator, but you may be able to use 2 Level 1 operators
to perform the
testing with two systems. One Level 2 or Level 3 will also be needed to
interpret the results
generated by the automated systems, so you return to about the same manpower
costs, and
the same throughput. Where is the advantage here? It lies in the
reproducibility, plus you have
the potential to store the test results digitally for future reference, which is not
possible when the
operator is relied upon to interpret the screen and translate this into
"accept" or "reject" in real
time.
Automated NDT
Let's look at a definition of Automation, then at some of the points mentioned above, and try
to analyze them.
An automated NDT station can be as simple as a scanning guide combined with a monitor and audible alarm to
warn of a defect to help the
ultrasonic
operator position his probe, (
a semi-automated system), or as complex as a multi-channel million dollar computer
controlled machine
that can fill a living room. Anything that mechanically
or
electronically assists or replaces the manual operator can be considered
as some form of
automation. Reproducibility
An automated NDT system, be it Ultrasonic, Eddy Current or any other type,
almost always
has reproducible performance. It is not significantly dependent on how
much beer the
operator drank last night (or at lunch time), temperature or humidity,
how tired the
operator is, how many hours of overtime he has worked this week, or how
attentive the
operator is to his work.
To answer this question, we have to look at the distribution of defects
in components,
and at the effects of defects of different sizes on the product. Assume
that the
components have many very small defects (from less than 0.1mm long) and very
few large
defects (several centimeters long), and that a 3mm long defect could just
reduce the life of
the component, whilst a 3cm long defect will cause a catastrophic failure,
and 5cm long may
endanger life. The 0.1mm long defects occur at a rate of 1 defect in every
10 pieces, 1mm
long at a rate of 1 in every 100 pieces, 10mm long at a rate of 1 in every 1
000 pieces etc. in
a logarithmic progression. This means that a 5cm long defect could occur
once in every 5
000 pieces.
Manpower Reduction
In many cases, automated systems will operate at the maximum possible
throughput, limited in speed only by the laws of physics. It only takes one pulse on an
ultrasonic instrument
(operating with a Pulse Repetition Frequency or PRF of 2000Hz) to
trigger an alarm,
while it will need probably 100 pulses (or 0.05 seconds) for the alert
manual operator to
observe an echo exceeding his threshold. This means that the automated scan
can
potentially move at up to 200 times the speed of the equivalent manual scan
without missing
defects. The automated system can also operate with an optimal probe
movement,
eliminating double scanning, or missed areas when manually inspecting.
Investment
An automated NDT station will not be cheap. Depending on the complexity of
the system, the
price could range from $10,000 to $10,000,000, but let us assume that a
machine costs
about $100,000. In pure financial terms, if it can replace three operators,
it may come close
to paying for itself in just over a year. One has to consider, however,
the interest on
capital borrowed to purchase the machine (or at least loss of interest by
not having the
capital to invest), running costs etc., as well as possibly extra training
for the operators in the
new technology. Consider also the situation that you could be in if the
machine breaks down,
so you have to have a back-up of some type, or a significant investment in
spare parts and a
good TPM (Total Preventative Maintenance) scheme.
For smaller companies, more complex test systems may prove to be too high an
investment
to make economic sense.
Discussion Point
Taking mass production to its extreme, we need to consider if NDT should be
employed in
the production process at all.
If your process is producing defective parts, then the process is
potentially costing you
money in the form of reduced throughput (due to parts rejected), as well as
inspection costs.
If the cause of the defects can be determined, and the process so controlled
that these
defects do not occur, then the need for NDT becomes questionable.
References
The author
Godfrey Hands started his career in NDT in 1968. His specialties are
Ultrasonics and
Resonant Inspection, and his NDT experience has encompassed many different
industrial
sectors from offshore industry and training to research. His employers have ranged
from construction
and the NDT service industry, through sales to his present position within a
multi-national
mass-production organisation.
E-Mail: Godfrey@hands-ndt.co.uk
Homepage: www.hands-ndt.co.uk
| Frontpage | |Top to this
page|
Rolf Diederichs 1.March.1996, info@ndt.net
