Information Technology Reference
In-Depth Information
4.2 Materials for Fatigue Life Assessment of Multivariable
and Multiaxial Loadings
The investigated materials were multidirectional laminates of E-glass/polyester
(Vassilopoulos and Philippidis
2002
) and E-glass fabrics/epoxy (Mandell and
Samborsky
2010
), typical materials used in wind turbine blade applications. The
corresponding lay-ups were [0/(
±
45)
2
/0]
T
and [
±
45/0
4
/
±
45/], respectively. The
materials were cut by diamond saw wheel at on-axis (0
) and off-axis orientations.
For E-glass/polyester material, the corresponding off-axis orientations were 15
°
°
,
30
(Vassilopoulos and Philippidis
2002
), while for E-glass
fabrics/epoxy material, the only off-axis orientation was 90
°
,45
°
,60
°
,75
°
and 90
°
°
(Mandell and Sam-
borsky
2010
).
In addition, the corresponding database containing fatigue data of various stress
ratio values and the corresponding on-axis/off-axis orientations of R = 0.1:
ʸ
=0
°
,
15
°
,45
°
,75
°
and 90
°
; R = 0.5:
ʸ
=0
°
and 45
°
; R =
−
1:
ʸ
=0
°
,30
°
,45
°
,60
°
and
90
°
; and R = 10:
ʸ
=0
°
,30
°
,45
°
,60
°
and 90
°
for E-glass/polyester, and of R = 0.1:
ʸ
=0
°
and 90
°
; R = 0.5:
ʸ
=0
°
and 90
°
; R =
−
0.5:
ʸ
=0
°
and 90
°
; R =
−
1:
ʸ
=0
°
and 90
for E-glass fabrics/
epoxy. The database comprised, respectively, 85 and 96 fatigue data, making the
database suitable for the study purpose. Note that number of stress levels in each
stress ratio value employed were 5 and 8 for E-glass/polyester and E-glass fabrics/
epoxy, respectively.
From the fatigue data, stress ratio (R), on-axis/off-axis orientation (
°
; R =
−
2:
ʸ
=0
°
and 90
°
; and R = 10:
ʸ
=0
°
and 90
°
) and
maximum stress (S
max
) values were used as input set and the output was the
corresponding fatigue cycles (log N) for the input set. For each particular R value,
mean fatigue life values were used. As in the previous section, all the data were also
normalized into the range of
ʸ
−
1 to 1 using Eq. (
9
).
Table
2
summarizes the materials examined together with the training and
testing sets employed. Note that for the assessment task, stress ratio values-R were
arranged in CCW direction according to the CLD, moving across from tensile-
tensile sector to compressive-compressive sector. In addition, fatigue data as
training set of R = 0.1 and 10 were chosen because the best relative positions of the
R values in the CLD (Hidayat and Melor
2009
; Hidayat et al.
2011
; Hidayat and
Berata
2011
). The corresponding
.
With the training and testing data, the NN model will develop multivariable and
multiaxial fatigue life assessment analysis.
ʸ
value chosen for both the stress ratios was 0
°
4.3 Methods
In the present study, the training algorithm of Levenberg-Marquardt was chosen
and utilized to result in fast and ef
cient NN model (Nocedal and Wright
2006
).
The use of Levenberg-Marquardt algorithm for fast and ef
cient NN modeling has
Search WWH ::
Custom Search