With
the
development
of
smart
manufacturing,
there
are
many
studies
have
been
devoted
to
machining
processes
monitoring,
predicting
the
failures
of
mechanical
system
or
tool
life.
In
order
to
get
the
real-time
information
from
the
system
relies
on
the
assistance
of
sensors.
Furthermore,
the
sensor
is
usually
installed
at
a
location
where
the
interference
to
the
cutting
process
is
minimum.
However,
the
sensor
location
is
usually
fixed
so
the
distance
between
the
sensor
and
the
tool
tip
varies
all
the
time.
This
might
lead
to
errors
in
decision
making
based
on
vibration
signals.
In
this
study,
the
effect
of
sensor
location
on
detecting
size
effect
with
vibration
signals
in
slot
milling
was
investigated.
Two
different
methods
were
compared.
The
first
method
(method
A)
was
to
fix
the
location
of
the
accelerometer.
The
other
method
(method
B)
was
to
change
the
sensor
position
before
every
cutting
test
so
the
sensor
remained
the
same
position
relative
to
the
cutting
tool.
The
features
of
vibration
signals
in
method
A
and
method
B
were
compared.
Analysis
shows
that
the
characteristic
frequency
of
the
size
effect
mostly
falls
on
the
tool
passing
frequency,
and
the
features
of
signal
were
different
between
these
two
methods.
It
suggested
that
the
relative
positions
between
the
sensor
and
cutting
zone
should
be
taken
into
consideration
in
machining
process
monitoring,
otherwise,
it
would
lead
to
errors
in
decision
making
based
on
vibration
signals.
Taking
the
feed
direction
as
an
example,
the
accuracy
of
the
accelerometer
in
method
A
is
reduced
to
91.63%
after
changing
the
feed
position
sequence.
As
a
result,
a
recommended
solution
is
proposed
in
this
study,
that
is,
use
energy
normalization
process.
The
results
show
that
it
is
possible
to
increase
the
accuracy
to
95.8%,
and
then
establish
a
neural
network
model
with
feature
fusion
method,
which
can
effectively
improve
the
average
accuracy
rates
of
98.1%.
Keywords:
Sensor
location;
Size
effect;
Class
mean
scatter
criteria;
Fisher’s
linear
discriminant
analysis;
Neural
network;
Normalization.