Tài liệu Neural Networks and Neural-Fuzzy Approaches in an In-Process Surface Roughness Recognition System for End Milling Operations pptx

Chen, Joseph C. "Neural Networks and Neural-Fuzzy Approaches in an In-Process Surface
Roughness Recognition System for End Milling Operations"
Computational Intelligence in Manufacturing Handbook
Edited by Jun Wang et al
Boca Raton: CRC Press LLC,2001


16
Neural Networks and
Neural-Fuzzy
Approaches in an
In-Process Surface
Roughness Recognition
System for End
Milling Operations
16.1
16.2
16.3
16.4

Joseph C. Chen
Iowa State University

Introduction
Methodologies
Experimental Setup and Design
The In-Process Surface Roughness Recognition
Systems
16.5 Testing Results and Conclusions

16.1 Introduction

Different machining processes produce different products with varying qualities. When evaluating the
quality of a finished piece, surface roughness is the most important result of the machining process to
consider, because many product attributes can be determined by how well the surface finish is produced.
The quality of the surface finish, or surface roughness, affects several functional attributes of parts, such
as surface friction, wear, reflectivity, heat transmission, porosity, coating adherence, and fatigue resistance.
The desired surface roughness value is usually specified for individual parts, and a particular process is
selected in order to achieve the specified roughness.
Typically, surface roughness measurement has been carried out by manually inspecting machined
surfaces at fixed intervals. A surface profilometer containing a contact stylus is used in the manual
inspection procedure. This procedure is both time-consuming and labor-intensive. In addition, a number of defective parts could be produced during the time needed to complete an off-line surface
inspection, thereby creating additional production costs. Another disadvantage of using surface profilometers is that they register the serious interference of extraneous vibration generated in the surrounding
environment. This extraneous vibration might significantly influence the accuracy of surface measurements. For these reasons, researchers are seeking solutions to model the surface roughness in an online or in-process fashion.

©2001 CRC Press LLC


The studies of Martellotti [1941, 1945] are among the earliest that represent a major contribution to
the understanding of kinematics and the mechanism of surface generation in milling processes. Martellotti developed parametric equations to describe the trochoidal path that the tool follows. These studies
also provide approximate analytical expressions for the ideal peak-to-valley roughness generated in upand down-slab milling, and face milling.
Numerous other studies have explored the topography of milled surfaces. Many of these focused on
predicting the two- or three-dimensional shape of a milled surface under ideal and non-ideal conditions.
Kline et. al. [1982] demonstrated the effects of cutter runout on surface errors, and surface errors or
dimensional inaccuracies were predicted using the cantilever beam theory for cutter runout. Another
study by Babin et al. [1985] applied the cantilever beam theory to predict the topography of wall surfaces
produced by end milling. Armarego and Deshpande [1989] presented one more milling process geometry
model that incorporates cutter runout to predict cutting forces.
Sutherland and Babin [1985] demonstrated a two-dimensional worst-case analysis of the slot floor
surface. However, the model for the slot floor surface significantly underpredicted surface roughness
values. Research by Kolarits and DeVries [1989] extended the previous model to account for varying cut
geometries and feed rates. This extended floor surface generation model improved prediction capabilities

considerably. However, the roughness parameter predictions for some of the tests were found to deviate
greatly from measured values.
You and Ehmann [1989] developed a comprehensive model to predict the three-dimensional surface
texture generated by ball end mills. They also presented an algorithm for three-dimensional representations of the machined surface; however, the effect of flexibility of the cutter-workpiece system was not
considered in this model. Montgomery and Altintas [1991] presented the effects of the cutter-workpiece
system flexibility in their force and surface prediction model in order to analyze the surface generation
mechanism in peripheral milling under dynamic cutting conditions.
All models previously discussed represent only deterministic cutting models, but most machined
surfaces exhibit interrelated characteristics of both random and deterministic components. Zhang and
Kapoor [1991] demonstrated the effect of random vibrations on surface roughness in the turning process.
These vibrations were shown to occur due to random variations in the microhardness of the workpiece
material. Ismail and others presented a surface generation model in milling that included both cutter
vibrations and the effects of tool wear [Ismail et al., 1993]. Melkote and Thangaraj [1994] presented
another enhanced end milling surface texture model including the effects of radial rake and primary
relief angles. These three models, limited to laboratory usage or based on theoretical analysis, could not
be implemented as an in-process monitoring system.
The findings of this literature review, in addition to communication with leading private industrial
research and development laboratories in the state of Iowa (including Winnebago Co. in Forest City;
Delavan Inc. in Des Moines; Sauer-Sundstrand Inc. in Ames), point to the feasibility of in-process surface
roughness recognition (ISRR) systems for implementation in the newer generation of milling machines.
The successful implementation of this surface roughness recognition system will enable metal cutting
industries to reduce manufacturing costs by eliminating the relatively inefficient off-line quality control
aspect of surface roughness inspection. Therefore, reductions in manufacturing costs will increase competitiveness in worldwide markets. This implication supports the development of an effective and inexpensive ISRR system. The development of this system will enable implementation of adaptive control in
modern manufacturing environments.

16.2 Methodologies
In order to provide an adaptive control mechanism, ISRR systems require two major components:
(i) sensors, which receive the dynamics signal from the machining cutting processes; and (ii) an intelligent
technique able to learn the dynamics of the machining system while allowing for control features to be
built in. The research described in this chapter employed an accelerometer to detect the dynamics

mechanism of the tool and material interface. This study also used two major intelligent learning
©2001 CRC Press LLC


methodologies to incorporate data about the machining process through actual cuts. These methodologies were also employed to construct a control system that predicts surface roughness during the
execution of the machining process. These two learning methodologies are artificial neural networks
(ANN) and fuzzy neural (FN) systems. An overview of these two approaches follows in the next section.

16.2.1 Neural Networks Model
Several learning methods have been developed for ANNs. Many of these learning methods are closely
connected with a certain network topology, with the main categorization method distinguished by
supervised vs. unsupervised learning. Backpropagation was chosen from among various learning methods
already existing in this field. This approach was adopted into this research for two reasons: primarily, it
is the most representative and commonly used algorithm, in addition to being relatively easy to apply;
additionally, it has been consistently successful when used in practical applications [Das et al., 1996;
Huang and Chiou, 1996].
The backpropagation algorithm can be divided into two main processes, the process of learning and
the process of recalling.
16.2.1.1 The Learning Process
Step 1. Given network parameters:
Set all the necessary parameters, such as the number of input neurons (i), the number of
hidden layers and the number of neurons included in each hidden layer (h), the number of
output neurons (j), etc.
Step 2: Initialize the beginning weights and biases:
Set all the initial weights and biases values randomly.
Step 3: Load the input vector X and the target output vector T of a training example.
Step 4: Calculate and infer the actual output vector Y.
(a) Calculate the output vector H of hidden layers.

net h =

∑W _ xhih • Xi – θ _ h h

Equation (16.1)

i

(

)

H h = f net h =

1

Equation (16.2)

1 + exp – net h

(b) Infer the actual output vector Y.

net j = h W _ hy hj 〈 H h − θ _ y j

( )

Y j = f net j =

Equation (16.3)

1

1 + exp

Equation (16.4)

− net j

Step 5: Calculate the error term.
(a) The error term of the output layer: δ j = Y j 1 – Y j

(

(b) The error term of the hidden layer: δ h

)( T – Y )
= H (1 H )
W _ hy ã
h

j

h

hi

j

â2001 CRC Press LLC

j

j

Equation (16.5)
Equation (16.6)


Step 6: Calculate the revised weight of the weight matrix and the revised bias of the bias vector.
Equation (16.7)
(a) For the output layer: ∆W _ hy hj = ηδ j H h , ∆θ _ y j = –ηδ j
(b) For the hidden layer: ∆W _ xhih = ηδ h X i , ∆θ _ h h = – ηδ h

Equation (16.8)

Step 7: Adjust and renew the weight matrix and the bias vector.
(a) For the output layer:

W_hyhj = W_hyhj + ∆W_hyhj , θ _y j = θ_y j + ∆θ_ y j

Equation (16.9)

(b) For the hidden layer:

W_xhih = W_xhih + ∆W_xhih , θ _hh = θ _hh + ∆θ_hh

Equation (16.10)

Step 8: Repeat steps 3 through 7, until the energy function has converged or the specified learning
cycles are completely executed.
16.2.1.2 The Recalling Process
Step 1: Set all the network parameters.

Step 2: Read the weight matrix W_xh and W_hy, and the bias vector θ_h and θ_y.
Step 3: Load the input vector X of a testing example.
Step 4: Calculate and infer the actual output Y.
(a) Calculate the output vector H of hidden layers.

net h =

∑W _ xhih • Xi – θ _ h h

Equation (16.11)

i

(

)

H h = f net h =

1

Equation (16.12)

1 + exp – net h

(b) Infer the actual output vector Y.

net j =

∑W _ hy hj • H h – θ _ y j

Equation (16.13)

h

(

)

Y j = f net j =

1
1 + exp

Equation (16.14)

– net j

16.2.2 Fuzzy-Nets Modeling
The proposed fuzzy-nets system was developed by fuzzy rules generated from sampled input–output
pairs. This model is built in five steps.
16.2.2.1 Step 1: Divide the Input and Output Spaces into Fuzzy Regions

[

]

Assume that the domain intervals of input variable xi are x i– , x i+ , and that the domain intervals of
[y –, y+].

Each domain interval can be divided into 2N + 1 regions. The value of
output variable y are
N is dynamic for different variables, and the lengths of each region can be equal or unequal. Each
region is denoted by

©2001 CRC Press LLC


SN (Small N), S(N-1) (Small N-1), …, MD (Medium), … , LN (Large N),

Equation (16.15)

and then assigned a fuzzy membership function. The divisions of the input and output spaces are shown
in Figure 16.1, where N is 2 for x1, and 3 for x2 and y. The width for each variable is the same.

µ( x1 )
S2

S1

MD

L1

L2

0

x1−

x1+

d

µ( x2 )
S3

S2

S1

MD L1 L2

L3

1

0
−
x2

+
x2

d

µ(y)
S3

S2

S1

MD L1 L2

L3

1

0

y−

y+

d

FIGURE 16.1 The domain intervals of the input–output variables and triangular membership function.

In this study, the input variables are spindle speed (S), feed rate (F), depth of cut (D), and vibration
average per revolution (V). The output variable is the surface roughness average value, Ra. A triangular
membership function specified by three parameters {a, b, c} is employed as follows:

 0
x –a

b–a
triangle x ; a ,b ,c = 
c–x
 c –b

 0


(

©2001 CRC Press LLC

)

x ≤a
a ≤x ≤b
Equation (16.16)
b≤x ≤c
c≤x


The spread of the input feature shown in Figure 16.1 is defined as

d=

x i+ – x i–
(i = 1,2, ..., k),
2N

Equation (16.17)

where x i– and x i+ are the domain intervals of variable xi, xi ∈ Xi. There are 2N + 1 fuzzy regions
quantifying the universe of discourse Xi.
The center points of each linguistic variable are

(x

–
i

(

)

)

, x i– + d ,… , x i– + 2 N – 1 d , x i+ .

Equation (16.18)

Equations 16.17 and 16.18 are also used for the output variable y.
16.2.2.2 Step 2. Generate Fuzzy Rules from Given Data Pairs through Experimentation
Three steps are used for generating fuzzy rules:
1. Determine the degree of input–output data obtained from the successful experiment.
2. Assign the input–output pairs to the region with the maximum degree.
3. Obtain one rule from one pair of designated input–output data.
In this study, the experimental input–output pairs were

[S ,F ,D ,V ,R ],
i

i

i

i

i
a

Equation (16.19)

where i denotes the number of input–output pairs.
1. A human expert examined these rules to ensure their usefulness and correctness.
2. The degrees of each data pair were determined by the function

 xi – xc
1 –
, xi ∈ xc , xc + x s
xs

 x –x
c
i

, xi ∈ xc – x s , xc .
µ x i = 1 –
xs

otherwise
0,




[

( )

]

[

]

Equation (16.20)

where xc is the center of the linguistic level x, and xs is the width of the linguistic level x, equal to d.
3. After all of the input and output elements were determined, each element was assigned to the
region with the maximum degree.
4. One rule from one pair of the desired input–output pair [S1,F1,D1,V1,Ra1], was assigned. For
example, the degree of one input–output pair: was determined by Equation 16.20 as

©2001 CRC Press LLC


}
( ) {
µ ( F ) = {0.9 ∈ L 3 , 0.1 ∈ L 2 }
µ ( D ) = {0.8 ∈ L 2 ,0.2 ∈ L1}
µ (V ) = {0.2 ∈MD, 0.8 ∈S1}
µ ( R ) = {0.3 ∈S2, 0.7 ∈S1} .

µ S 1 = 0.7 ∈MD, 0.3 ∈S1
1

1

Equation (16.21)

1

1
a

The region of each datum with a maximum degree was assigned as follows:

S 1 ∈ MD , F 1 ∈ L 3 , D 1 ∈ L 2 ,V 1 ∈S 1 , R1 ∈S 1.
a

Equation (16.22)

Rule one was obtained by

IF S1 is MD and F1 is L3 and D1 is L2 and V1 is S1 THEN Ra1 is S1,

Equation (16.23)

where AND indicates that the conditions of the IF statement must all be met simultaneously in order
for the result of the THEN statement to be true.
16.2.2.3 Step 3: Assign a Degree to Each Rule and Resolve the Conflicting Rules
If two or more rules generated in step 2 have the same IF command but a different THEN command,
then the rules conflict. To resolve conflicts between the two data sets, a degree must be assigned to each
rule, generated from the data pairs as

( )

( )( )( )( )( )( )

i
d Ri = µ S i µ F i µ D i µ V i µ Ra µ E i

where µ(Si)
µ(Fi)
µ(Di)
µ(Vi)
µ(Rai)
µ(Ei)

=
=
=
=
=
=

Equation (16.24)

the degree of the spindle speed variable,
the degree of the feed rate variable,
the degree of the depth of cut variable,
the degree of the vibration variable,
the degree of the surface roughness variable,
the degree assigned by the human expert to determine the importance of this rule.

The following function resolved the conflict between rules:

( ) ( )

d Rk – d Rl > ε

Equation (16.25)

where Rk and Rl are two conflicting rules, d(Rk) and d(Rl) are the degree of rules, Rk and Rl, and ε is the
user-defined parameter 0 < ε < 0.05. Next, the rule with the maximum degree is selected. If the above
function cannot resolve the conflict, ε may be decreased, or two more regions to one feature of the input
vector may be increased and the input–output data pairs retrained. If these rules still conflict, the region
number of the next input feature is extended to two more regions and then retrained. These procedures
are repeated until all of the conflicting problems are resolved.

©2001 CRC Press LLC


L2

S

L1

L1
MD
S1
S2

S1

S2 S1 MD L1 L2

F
FIGURE 16.2 Illustration of a combined fuzzy rule base.

16.2.2.4 Step 4. Create a Combined Rule Base
A combined rule base consists of two kinds of rules: rules generated from numerical data by means of
steps 1 through 3, and linguistic rules determined by a human expert. As shown in Figure 16.2, the
combined rule base includes two fuzzy rules:

IF S is L1 and F is L1 THEN Ra is L1;
IF S is MD and F is S1 THEN Ra is S1.

Equation (16.26)

16.2.2.5 Step 5. Determine a Mapping Based on the Combined Fuzzy Rule Base
A defuzzification strategy is used to determine the output control y for any given input datum. There
are many methods for defuzzification. In this study, the following centroid of area method was applied:

y=

{ ( ) ( ) ( ) ( )}

where µ io = min µ S i , µ F i , µ D i , µ V i

∑ µ 0i y i
∑ µ 0i

Equation (16.27)

, yi = the center value of the region, and y = the output

for a given input datum. This is the most widely adopted defuzzification strategy today, and it is
reminiscent of the calculation of expected values of probability distributions.

16.3 Experimental Setup and Design
Figure 16.3 shows the complete experimental setup in this research. A computer numerical control (CNC)
program was written to perform the end milling cutting processes. The electromagnetic proximity sensor
was fixed at a close distance to the spindle, as shown in Figure 16.4, and the accelerometer sensor was
mounted on the vise beneath the workpiece (Figure 16.5).
Rotation and vibration data were collected simultaneously by the proximity sensor and the accelerometer sensor, respectively, when the cutter had cut the workpiece at a distance of 0.35 in. The main
concern was to avoid the impact of initially unstable or significant vibration. Figure 16.6 illustrates the
two types of signals (i.e., rotation data and vibration data). These two types of data were connected to
an analog-to-digital (A/D) board (the vibration data from the accelerometer sensor had to be amplified
by the PCB battery power unit beforehand) and then transmitted to a 486 personal computer for further
data recording, processing, and analysis.

©2001 CRC Press LLC


CNC
Machine
Center
Vise
Accelerometer
Sensor

486 Personal Computer

VM

DESIGNER:
Wei-Liang

Proximity Sensor

PCB Battery
Power Unit

Workpiece

A/D Board

FIGURE 16.3 Experimental setup.

A computer program was written in C language that allowed the collection of the two kinds of data
transformed from analog to digital signals by an A/D converter. The collection time for each run was
about 0.54 s, and the runs contained 6000 rotation or vibration data from the proximity sensors or
accelerometer, respectively.
The cutting parameters (spindle speed, feed rate, and depth of cut) were changed manually according
to different cutting conditions for each run. Also, after each specimen was cut, the cutting tool was
cleaned to avoid chip formation or a built-up edge (BUE), which would affect the surface roughness of
the following cut. The tool condition was also checked to ensure that it was free from defects.
All specimens in this experiment were conducted under dry cutting conditions without coolants.
Coolants are not generally used in order to reduce costs and prevent tool breakage due to thermal shock.
Moreover, the decision to use dry cutting conditions was based on the need to isolate the correlation
between cutting vibrations and surface roughness in end milling.
With the experimental setup complete, the next step was to develop the ISRR-ANN and -FN models.
In this study, identifying the parameters of the training and testing data sets was a very important factor

in establishing the experimental runs. These runs could not exceed the suggested cutting parameters,
which were based on machine capabilities and the nature of the material composition of both the
workpiece and end mill.

©2001 CRC Press LLC


D (less than 5 mm)

Connected
to A/D board

Spindle
Proximity
Sensor

FIGURE 16.4 Top view of the proximity sensor and spindle.

WORKPIECE
Accelerometer sensor

VISE

To PCB battery power
CNC TABLE

FIGURE 16.5 The accelerometer setup.

16.3.1 Training and Testing Experiments
A total of 48 specimens were cut based on the following combination of cutting parameters: four levels

of spindle speed (S = 750, 1000, 1250, and 1500 rpm); four levels of feed rate (F = 6, 12, 18, and 24
in./min); and three levels of depth of cut (D = 0.01, 0.03, and 0.05 in.).
After these cuts were made, all specimens were measured off-line with a stylus profilometer to obtain
their roughness average Ra values. The average Ra value shown in Table 16.1 is the average of three Ra
measurements of each specimen.
During the machining process, the accelerometer sensor produces vibration data. Each cut produces
13 or more revolutions of vibration data. The statistical average of the vibration voltages (V), transformed
from analog to digital data, is based on one revolution of the spindle. This vibration average served as
the fourth independent variable for the ISRR systems. Three statistical averages of vibration voltage (V)
data were collected per specimen to provide a better empirical representation of the training data set.
Therefore, a total of 492 data sets were available for training.
In this research, 92 data sets were selected randomly from the 492 training data. In addition, 36
experimental cuts under different cutting conditions than those in the training data set were used for a

©2001 CRC Press LLC


Voltage
6
5
4
3
2
1

Data points
p

807

776

745

714

683

652

621

590

559

528

497

466

435

404

373

342

311

280

249

218

187

156

94

125

63

1
-1

32

0

Vibration data
Revolution data

FIGURE 16.6 Vibration and rotation signals.

flexible testing data set (Table 16.1). Therefore, a total of 128 pieces of data were used to evaluate the
accuracy of ISRR systems. The evaluation criteria are summarized in the next section.

16.3.2 Test Criteria
Criteria used in this experiment to judge the predictive capabilities of the ISRR–ANN and FN systems
included the percentage deviation (φi) of each testing sample, given as:

φi =

Ra i′ – Ra i~M
,

× 100%,

Ra i′

Equation (16.28)

where φi = percentage deviation of single sample data

Ra i′ = actual Ra measured by a profilometer
Ra i~M = predicted Ra generated by the ISRR systems; M indicates the predicted value using ISRR,
ANN or ISRR-FN models, respectively.
Additionally, the overall average percentage deviation (φ) of all 128 samples is given as
m

φi =

∑φ

i

i =1

m

Equation (16.29)

–
where φ = average percentage deviation of all sample data
m = the size of testing samples; in this study m = 128.

16.4 The In-Process Surface Roughness Recognition Systems
In this research, two ISRR systems were developed and tested. Their training and testing processes are
presented in the following sections.

©2001 CRC Press LLC


TABLE 16.1 Experimental Results
Sample No.

Spindle Speed
(rpm)

Feed Rate (ipm)

Depth of Cut (in.)

Roughness

(Ra: µin.)

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28

29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48

750
750
750
750
750
750
750
750
750

750
750
750
1000
1000
1000
1000
1000
1000
1000
1000
1000
1000
1000
1000
1250
1250
1250
1250
1250
1250
1250
1250
1250
1250
1250
1250
1500
1500
1500

1500
1500
1500
1500
1500
1500
1500
1500
1500

6
6
6
12
12
12
18
18
18
24
24
24
6
6
6
12
12
12
18
18

18
24
24
24
6
6
6
12
12
12
18
18
18
24
24
24
6
6
6
12
12
12
18
18
18
24
24
24

0.01

0.03
0.05
0.01
0.03
0.05
0.01
0.03
0.05
0.01
0.03
0.05
0.01
0.03
0.05
0.01
0.03
0.05
0.01
0.03
0.05
0.01
0.03
0.05
0.01
0.03
0.05
0.01
0.03
0.05
0.01

0.03
0.05
0.01
0.03
0.05
0.01
0.03
0.05
0.01
0.03
0.05
0.01
0.03
0.05
0.01
0.03
0.05

65.40
62.75
72.40
143.85
101.90
94.05
184.80
146.60
121.05
186.55
170.40
172.40

58.40
78.30
62.20
129.90
83.60
92.05
137.50
124.15
85.75
163.15
153.30
142.30
49.95
63.30
70.85
101.30
98.75
84.95
115.00
92.25
94.70
155.45
108.85
120.65
36.55
55.70
55.65
87.55
81.65
94.05

119.45
86.50
104.25
119.20
103.30
109.40

16.4.1 ISRR-ANN Model
Figure 16.7 shows the structure of the ISRR-ANN system. The input variables were spindle speed (S), feed
rate (F), depth of cut (D), and the statistical average of the vibration voltages (V) based on one revolution.
In their original form, the four independent variables (S, F, D, and V) for both data sets (training and
testing) could not be trained by neural networks due to the wide range of values distributed among them.
©2001 CRC Press LLC


Workpiece
Vibration

Accelerometer
Sensor

ISRR-ANN

Machining
Process
Spindle
Rotation
Machining
Parameters

Ra

Proximity
Sensor
Spindle Speed
Feed Rate
Depth of Cut

Input

Output

FIGURE 16.7 Structure of the ISRR-ANN.

In order to make them feasible as input neurons, all values in the first four columns (i.e., input neurons)
were preprocessed by normalizing, and then transformed into a value between zero and one.
Based on the RMS error of the training examples and testing examples, it is clear that the 4-5-1
structure had the lowest RMS-error among all the structures with one hidden layer, and 4-7-7-1 was
better than the other structures with two hidden layers. In the final step of development for this ANN
model, two programs were written in C language to predict the Ra values for one or two hidden layers
by retrieving the weighted files that resulted from the training and testing.

16.4.2 ISRR-FN System
The structure of the ISRR-FN, as shown in Figure 16.8, consisted of the sensing system, machining
parameters, and ISRR-FN. In this sensing system, an accelerometer sensor was used to measure the realtime vibration of the workpiece, a proximity sensor was used to measure the real-time rotation of the
spindle of the CNC machine center, and an A/D board and interface program were applied for analogto-digital conversion with 12-bit resolution. Machining parameters, such as spindle speed, feed rate, and
depth of cut, were transmitted to the ISRR-FN before or during machining.
The primary objective was to train the fuzzy system by generating fuzzy rules from input–output pairs,
and combining these generated and linguistic rules into a common fuzzy rule base. After input vectors
were fuzzified by the fuzzification interface, the fuzzy inference engine generated output values by

inferencing these fuzzified input values based on the fuzzy rule bank. Finally, the defuzzification interface
determined the final prediction of Ra values based on input variables. Through training, the ISRR-FN
learned to detect different conditions for individual machines, build the fuzzy rule base, and infer the
surface roughness, Ra. All of these processes were based on the experimental data.
Before the fuzzy-nets training takes up the experimental data, some parameters need to be preset based
on the limitations of the machine and the machining processes. They are summarized as follows:
1. The domain intervals of the input–output data pairs were assigned as follows:
• Spindle speed: [500, 2000] rpm
• Feed rate: [6, 42] inches per minute
• Depth of cut: [0.01, 0.07] inches
• Vibration: [780, 2460] microvolts
• Ra : [38, 168] micro inches
©2001 CRC Press LLC


Workpiece
Vibration

Accelerometer
Sensor

Fuzzy Rule Bank

Machining
Process
Spindle
Rotation
Machining
Parameters

Proximity
Sensor

Fuzzifier

Deffuzzifier

Ra

Spindle Speed
Feed Rate

Fuzzy Inference
Engine

Depth of Cut

ISRR-FN
FIGURE 16.8 Structure of the ISRR-FN.

2. Each domain interval was divided into 2N + 1 regions. To increase the accuracy of prediction and
properly decrease the number of fuzzy rules, N was selected as follows:
• N = 3 for input variables: spindle speed, feed rate, depth of cut, and vibration
• N = 5 for the output variable Ra
3. The ε, as shown in Equation 14.10, was assigned the value of 0.01. The degree assigned by a human
expert, µ(E')is 1, since all the experiments assume the tool to be in good condition, free from
chatter or tool-wear conditions during data collection.

16.5 Testing Results and Conclusions
16.5.1 Testing Results

After the experiment was conducted and both the ANN and FN systems were developed, testing experiments were designed to test the predictive ability of this model. The two sets of testing data mentioned
in Section 3.1 were used for testing the accuracy of these two systems.
Using the same four variables (spindle speed, feed rate, depth of cut, and the VAPR) as independent
variables or input neurons, predicted roughness Ra values (response variable or output neuron) were
generated for either the ANN or FN model. Tables 16.2 and 16.3 contain the measured Ra values and
their predicted Ra values, as well as the percentage deviation of both ISRR systems with the 92-item
testing set and the 36-item flexible testing set, respectively. Considering the percentage deviation, both
models performed well in their ability to predict the roughness of machined surfaces. In summary, the
prediction accuracy of ISRR-FN, ISRR-ANN 4-5-1, and ISRR-ANN 4-7-7-1 models are 95.78%, 95.87%,
and 99.27%, respectively.

16.5.2 Conclusions
The fuzzy logic and neural-networks-based ISRR models demonstrated that learning and reasoning
capabilities could be used for an in-process surface roughness recognition system. With better than 95%

©2001 CRC Press LLC


TABLE 16.2 Testing Data Set—92 Samples
Spindle
Speed

Feed Rate

Depth of cut

Vibration

Ra

Ra
ANN 4-5-1

Ra
ANN 4-7-7-1

750
750
1500
1000
750
1250
1250
1000
1500
750
1500
1000
1000
1250
1250
750
1500
1000
1500
1250
1500
1000
750
1250

1500
1250
1500
1000
1250
1500
1000
1250
1250
1500
1000
1500
750
750
750
1250
1250
750
1000
1500
1500
1000
1500
1250
1250
1000
1500
1250
1250
1500

1500

24
18
12
12
24
6
24
18
12
12
12
24
6
6
12
6
18
18
12
6
12
18
18
18
12
18
6
12

18
24
6
12
18
24
18
12
18
6
6
24
12
24
24
24
12
18
18
18
24
24
6
24
12
12
18

1
3

3
3
3
5
5
5
5
3
1
5
3
1
1
3
5
5
5
5
3
3
5
1
1
1
3
3
5
3
5
5

3
1
3
1
5
1
1
5
3
5
1
5
3
5
5
5
3
5
5
1
5
5
1

0.1123
0.1488
0.1162
0.1330
0.1659
0.0977

0.1764
0.1674
0.1225
0.1509
0.1249
0.1859
0.1013
0.8855
0.1418
.01045
0.1509
0.1647
0.1180
0.0974
0.1232
0.1691
0.1656
0.1252
0.1203
0.1266
0.1119
0.1391
0.2122
0.2164
0.0989
0.1253
0.1700
0.1372
0.1592
0.1151

0.1699
0.0899
0.1003
0.1740
0.1575
0.1705
0.1202
0.1900
0.1053
0.1610
0.1410
0.2160
0.2196
0.1932
0.0895
0.1249
0.1331
0.1155
0.1674

187
147
82
84
171
71
121
86
94
102

88
142
78
50
100
63
104
86
94
71
82
124
121
115
88
115
56
84
95
103
62
85
92
120
124
88
121
66
66
121

99
172
163
110
82
86
104
95
109
142
56
156
85
94
120

190
155
84
91
177
67
121
105
91
104
83
146
73
44

104
68
102
105
92
67
85
108
128
116
84
117
57
91
102
106
69
93
100
125
109
85
126
63
62
121
91
168
166
107

83
105
102
102
118
144
62
155
93
92
129

186
147
82
84
171
67
121
86
94
100
87
142
76
49
101
64
104
86

94
71
82
124
121
116
84
116
55
84
94
103
63
85
92
118
124
84
121
64
63
121
98
172
162
109
82
86
104
94

108
143
56
155
85
94
119

©2001 CRC Press LLC


TABLE 16.2 (continued) Testing Data Set—92 Samples
Spindle
Speed

Feed Rate

Depth of cut

Vibration

Ra

Ra
ANN 4-5-1

Ra
ANN 4-7-7-1

1500

1500
1250
1500
1250
1500
1500
1000
1500
750
1500
1500
1250
1000
1500
1500
1000
750
1250
1000
1250
750
1500
1500
1250
750
1000
1000
750
1250
1500

1250
1000
1500
750
1000
1500

12
18
12
6
12
6
18
24
12
6
24
12
24
12
24
24
18
18
6
24
24
12
18

18
24
6
24
18
18
6
18
24
12
18
18
18
12

3
5
1
3
5
1
3
1
3
3
1
3
1
5
1

1
1
3
1
3
3
5
1
5
5
5
5
1
3
3
5
1
5
3
5
1
5

0.1101
0.1500
0.1288
0.1084
0.1253
0.0647
0.1359

0.1187
0.1151
0.1028
0.1394
0.1233
0.1316
0.1832
0.1339
0.1374
0.1011
0.1734
0.0759
0.1723
0.2145
0.1331
0.1300
0.1513
0.1857
0.0916
0.2035
0.0979
0.1446
0.1060
0.1516
0.1216
0.1599
0.1536
0.1622
0.0979
0.1183

82
104
100
56
85
37
87
163
82
63
120
82
156
92
120
120
138
147
50
153
109
94
120
104
121
72
142
138
147

63
104
156
92
87
121
138
94

84
102
106
57
93
39
96
166
84
68
125
85
153
92
126
125
139
145
46
157
120

101
118
102
119
69
141
140
155
67
102
155
92
98
129
140
92

82
105
101
55
85
36
87
163
82
65
118
82
155

92
118
118
137
147
49
153
108
96
119
105
121
70
143
137
147
64
105
155
91
87
129
137
94

accuracy of prediction, this system could be implemented as an in-process surface roughness prediction
system; however, some directions for further research could make future implementation more effective:
1. The cutting tool used in the study was a four-flute, high-speed steel cutter. Use of the model with
more diverse cutter materials or tools with a different number of flutes will benefit from further
investigation.

2. The workpiece material used in the study was Aluminum 6061 T6. Different workpiece materials
widely used in industry, such as carbon steel, aluminum alloys 380 and 390, or alloy steel, merit
further exploration in order to build an overall ISRR system capable of extensive application in
actual manufacturing environments.
3. Feedback control is necessary for automated production systems. The ISRR should be a closedloop system, ensuring that output of the ISRR feeds back to the CNC machine center, causing the
machine to adapt to the predicted real-time surface roughness value by adjusting the feed rate of
the machine table. Thus, a study of interface techniques between the ISRR system and the CNC
machining center is also necessary.

©2001 CRC Press LLC


TABLE 16.3 Flexible Testing Data — 36 Samples

Sample No.

Spindle
Speed (rpm)

Feed Rate
(ipm)

Depth of Cut
(in.)

1
2
3
4
5

6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35

36

1500
1500
1500
1500
1500
1500
1500
1500
1500
1250
1250
1250
1250
1250
1250
1250
1250
1250
1000
1000
1000
1000
1000
1000
1000
1000
1000
750

750
750
750
750
750
750
750
750

9
9
9
15
15
15
21
21
21
9
9
9
15
15
15
21
21
21
9
9
9

15
15
15
21
21
21
9
9
9
15
15
15
21
21
21

0.01
0.03
0.05
0.01
0.03
0.05
0.01
0.03
0.05
0.01
0.03
0.05
0.01
0.03

0.05
0.01
0.03
0.05
0.01
0.03
0.05
0.01
0.03
0.05
0.01
0.03
0.05
0.01
0.03
0.05
0.01
0.03
0.05
0.01
0.03
0.05

VAPR

Measured Ra
(µin.)
ISRR-FN

Ra ISRR-FN

0.0883
0.1110
0.1056
0.1464
0.1256
0.1638
0.1473
0.1787
0.1980
0.1197
0.1381
0.1202
0.1338
0.1521
0.1444
0.1300
0.1726
0.1846
0.0911
0.1226
0.1426
0.1001
0.1487
0.1598
0.1034
0.1680
0.1687
0.0931
0.1255

0.1171
0.0950
0.1514
0.1530
0.1135
0.1624
0.1659

53
74
70
110
84
99
119
102
113
80
82
92
107
97
87
129
99
105
95
97
102
129

108
95
149
145
112
109
99
95
126
122
104
178
163
150

53
74
70
110
84
99
119
102
113
80
82
92
107
97
87

129
98
105
92
96
102
129
108
92
149
145
112
109
99
95
125
122
104
178
163
150

4. The ISRR systems used in this study were controlled by a personal computer system. To increase
operation speed, minimize size, reduce costs, and enhance efficiency, the ISRR could be created
in an electronic model operated by a microprocessor, fuzzy chip, memory chip, and other related
circuits. These kinds of techniques are important and may be required for further development
of ISRR technology for the next century.

References
Armarego, E. J. A. and Deshpande, N. P., 1989, Computerized predictive cutting models for forces in

end-milling including eccentricity effects, Annals of the CIRP, 38(1), pp. 45-49.
Babin, T. S., Lee, J. M., Sutherland, J. M., and Kapoor, S. G., 1985, A model for end milling surface
topography, Proc. of 13th North American Metalworking Research Conference, pp. 362-368.
Das, S., Roy, R., and Chaptopadhyay, A. B., 1996. Evaluation of wear of turning carbide inserts using
neural networks, Int. J. Mach. Tools Manuf., 36(7), pp. 789-797.

©2001 CRC Press LLC


Huang, S. J. and Chiou, K. C., 1996. The application of neural networks in self-tuning constant force
control, Int. J. Mach. Tools Manuf., 36, pp. 17-31.
Ismail, F., Elbestawi, M. A., Du, R., and Urbasik, K., 1993, Generation of milled surfaces including tool
dynamics and wear, ASME J. Eng. Ind., 115, pp. 245-252.
Kline, W. A., DeVor, R. E., and Shareef, I. A., 1982, The prediction of surface accuracy in end milling,
ASME J. Eng. Ind., 104, pp. 272-278.
Kolarits, F. M. and DeVries, W., 1989, A model of the geometry of the surface generated in end milling
with variable process inputs, in Mechanics of Deburring and Surface Finishing Processes, J. R. Stango,
and P. R. Fitzpatrick, Eds., ASME, PED, vol. 38, pp. 63-78.
Martellotti, M. E., 1941, An analysis of the milling process, Trans. ASME. 12:677-700.
Martellotti, M. E., 1945, An analysis of the milling process, part II — Down milling, Trans. ASME, 67,
pp. 233-251.
Melkote, S. N. and Thangaraj, A. R., 1994, An enhanced end milling surface texture model including the
effects of radial rake and primary relief angles, ASME J. Eng. Ind., 116, pp. 166-174.
Montgomery, D. and Altintas, Y., 1991, Machanism of cutting force and surface generation in dynamic
milling, ASME J. Eng. Ind., 113 (1), pp. 160-168.
Sutherland and Babin 1985,
You, S. J. and Ehmann, K. F., 1989, Scallop removal in die milling by tertiary cutter motion, ASME J.
Eng. Ind., 111, pp. 213-219.
Zhang, G. M. and Kapoor, S. G., 1991, Dynamic generation of machined surface, part 1: Description of
a random excitation system, ASME J. Eng. Ind., 113(3), pp. 137-144.

Zhang, G. M. and Kapoor, S. G., 1991b, Dynamic generation of machined surface, part 2: Construction
of surface topography, ASME J. Eng. Ind., 113(3), pp. 145-153.

©2001 CRC Press LLC