Development of An Intelligent System For Tool Wear Monitoring Applying Neural Networks
Development of An Intelligent System For Tool Wear Monitoring Applying Neural Networks
Development of An Intelligent System For Tool Wear Monitoring Applying Neural Networks
VOLUME 14
ISSUE 1-2
January-February
2006
1.
Introduction
1. Introduction
Modern procedures for part manufacturing impose cost
reductions that can be realized in the following ways: increasing
turning regime, reducing manufacturing time and number of
rejects. In order to accomplish that in various processing
procedures, extreme efforts of tools and machines are required.
Processing tool condition is very influential on reducing rejects
146
Research paper
1.1.
Tool
wear
monitoring
1.1. Tool
wear
monitoring
Systems for tool wear monitoring, both old and new
generation, as its measuring value utilizes process parameters that
are indirectly linked to tool wear, those being: force or vibration,
Acoustic Emissions (AE) etc. The process is also influenced by
conditions under which the processing is taking place, like tool
geometry, tool material and product, etc. For modelling non-linear
dependencies that are separated from the measuring signal,
processing conditions, tool wear or tool breakage, neural
networks, fuzzy logic systems or the combination of both methods
are used. Balazinski et al [2] state that intelligent neural networks
and neural fuzzy techniques are intensely studied and they present
the most selected intelligence neural network methods for
merging monitoring properties. However, with commercially
available systems, the approach one sensor/one tool per process
is dominant and the application of AI method can rarely be found.
In his survey paper, Siek [3] established that, in the previous
period, most researchers elaborated on the tasks of classifying
wear or breakage. Tool wear is a term not uniformly defined and
it has to be defined clearly before stating the monitoring task.
Tool breakage is always defined and classified by two states,
broken or not broken. Tool wear classification has to use more
than two tool states, that is, it should be continual evaluation of
wear condition [4, 5].
Parameters defining wear are average and maximal width of
flank wear, as well as depth, length and widths of crater wear.
Criteria that should define wear as uniform, needs to be fixed in
order to present the state of tool wear. If wear is defined in two
groups (wear width), it becomes quite wide and one can recognize
only new and significantly worn tools. To monitor wear in
practice, it is necessary to establish several wear groups, which
practically presents very promising monitoring strategy. It can be
said that wear is a continual and monotonously increasing
process; therefore, continual evaluation would most appropriately
suit the physical processing.
Last years have seen the intensive work to apply artificial
intelligence (AI) method for monitoring tool wear. Thus
Balazinski et al [2] and [6] compare the application of three AI
methods: a feed forward back-propagation (FF-BP) neural
network, a fuzzy decision support system (FDSS) and an artificial
2. Proposal
Proposal for
system
2.
fora amonitoring
monitoring
system
model
model
The proposed tool monitoring system model can basically be
observed through four segments united in a whole that is, using
back propagation, connected with the machine-managing unit, as
shown in (Fig. 1).
Special segments of the system are:
sensor part
part for data acquiring, processing and analyzing
part for training neural network
part for presenting results.
Sensor part of the machine tool is made of measuring bearing
placed in the front bed of the machine tool main spindle. Besides
the measuring bearing, there is also another sensor working on the
principle of measuring strain gauge, placed on the processing tool
holder and designed specially for this case for measuring cutting
forces appearing on the tool itself [10, 11, 12, 13].
Part for data acquiring, processing and analyzing contains
standard A/D card ED 300, which receives input data from the
existing sensors, converts them to digital information, and sends
them to computer database. For information flow, the composite
software named ED LINK is responsible, allowing the possibility
for programming conditioning speed and type of input data [10, 11].
Neural network built into the system is a multi-layer
perception network with signal spreading in one direction (feedforward topology), and one of the best-known types of feed
forward neural networks. The network has three layers: the input
layer contains three neurons, intermediate hidden layer contains m
neurons, and the output layer has one neuron.
Software system is designed to acquire and process
information in on-line work regime and to manage the work of
hardware components, so it can be based on set limitations for
monitor processing and tool wear. To establish the degree of tool
wear, we can utilize comparative analysis of wear curves obtained
by system training using neural networks. Determining the
leftover tool duration is set on the basis of wear trend gained by
comparative analysis with the wear curve and real condition.
147
unit value. For i-th value of the input vector from the variable
registered by the measuring bearing FRprom, normalization
formula can be written in the following way:
p i
where
p sr
p i p sr
.
sp
1
N
(1)
p.
i 1
(2)
sp
3. Neural
network
for tool
3.
Neural
network
forwear
tool wear
monitoring
monitoring
3.1.
Pre-processing
and
training
3.1. Pre-processing
and
training
setset
As already stated, neural network has three inputs to which
the force values from the sensor on the tool holder, measured
force from measuring bearing, and cutting speed are directed.
Using these three values, the neural network at its output
evaluates the values of flank wear VB in the same time moment.
For the need of training process, a set was formed containing
30,900 input vectors and the same number of precise values of
output variable. In creating the set, special attention was given to
data representatives, that is, data was selected to cover all the
interval of possible values of input variables and to be adequate to
real change conditions. The training set formed in such a manner
ensured that neural network correctly approximated the
dependence of input values and output variable on the whole
range of input sizes.
In order to have efficient training, all the values in the training
set were previously normalized. Normalization was performed in
a way that every input and output size in the training set had the
average value equal to zero, and standard deviation reduced to
148
Research paper
1 N
( p i p sr ) 2
N 1 i 1
(3)
y (net i )
2
1.
1 e neti
(4)
where
M
net i w ji x j bi ,
(5)
j 1
4. Experiment
setup
4.
Experiment
setup
Machine parameters were selected in order to respond to
industrial application in real manufacturing. Machine conditions
for every experiment are presented in Table 1. The experiments
were repeated under the same conditions for the possibilities of
training, verifying and testing neural network model. Basically,
cutting speed and training number varied. There were ten
experiments in total, all of them with the same basic
configuration. However, some of the experimental conditions
were being changed to isolate disturbances and identify the
properties of adequate monitoring signal. Special focus was on
ensuring that all experimental conditions remained the same,
except for the parameters that were changed under control. The
basic configuration of the experimental measuring setup is shown
in (Fig. 4), and it contains CNC lathe equipped with sensors for
measuring cutting force, those being: promess sensor and
specially designed sensor with measuring strain gauge placed on
the tool holder.
Development of an intelligent system for tool wear monitoring applying neural networks
149
Table 1.
Experiment parameters
Machine
Holder
Insert
Cutting depth [mm]
Speed [m/min]
Material workpiece
Number of passing
Total time [min]
Diameter [mm]
Passing length [mm]
Exp. 1
110
36
60
10
Exp. 2
...
Exp. 10
INDEX GU 600
PTGNL 25x25
TNMG 220408
1 mm
200
.4730
132
...
112
43
...
37
60
...
60
10
...
15
300 m
a)
300 m
b)
5.
Results
5. Results
The selected appropriate model was established to be
relatively reliable method for monitoring tool wear during hard
turning. During the research, several different network
configurations were used and studied for their application in tool
wear monitoring during hard turning.
It is known that static cutting forces are good tool wear indicators;
however, adequate dynamic analysis of cutting forces can also give
satisfactory properties for wear monitoring. (Fig. 5) presents cutting
force components measured during tool wear monitoring.
300 m
c)
40
35
F(x,z) force
30
25
20
15
10
5
0
time s
150
Research paper
VB [mm]
measured values
measured values
measured values
measured values
measured values
0.16
0.14
0.12
0.1
0.08
0.06
0.04
0.02
0
References
References
[1]
[2]
time [min]
10
11
[3]
[4]
[5]
[6]
[7]
[8]
6. Future
work
6.
Future
work
The lack of neural networks (same as many other
experimental models) requires long-term training with data
normalization with the values expected to be working in the real
conditions. The network cannot work without previous training.
To expand ones work, it is necessary to utilize both numerical
and experimental methods. Considering the fact that the network
should be re-trained from time to time, training period can be
considered as a major drawback in the application of neural
networks in real manufacturing. However, future research could
include the integration of the existing system into CNC-machine,
instead of currently separated device; this would ensure that
monitoring system and machine could react synchronically, i.e.
the machine could react by stopping once the over-worn tool is
detected. More precisely, dynamic neural networks to ensure
additional correction of the trained static network in on-line work
regime could also enlarge the existing model.
[9]
[10]
[11]
[12]
[13]
7. Conclusions
7.
Conclusions
The paper presented that neural networks (NN) can be used for
efficient wear monitoring during hard turning, with the listed
limitations. After considering many possible setups for wear
monitoring model using different configuration types of neural
networks, and based on input and output parameters, the one
selected performed with optimal results for the selected number of
[14]
[15]
Development of an intelligent system for tool wear monitoring applying neural networks
151