Download as pdf or txt
Download as pdf or txt
You are on page 1of 12

Mechanics of Materials 154 (2021) 103715

Contents lists available at ScienceDirect

Mechanics of Materials
journal homepage: www.elsevier.com/locate/mecmat

Research paper

Improvement and validation of residual stress measurement in composite


laminates using the incremental hole-drilling method
Xiaodong Liu, Xiaodong Wang ∗, Zhidong Guan ∗, Ting Jiang, Kunhao Geng, Zengshan Li
School of Aeronautic Science and Engineering, Beihang University, Beijing 100191, PR China

ARTICLE INFO ABSTRACT

Keywords: In this paper, the incremental hole-drilling (IHD) method has been improved by attaching strain gauge rosettes
Composite laminates on both top and bottom surfaces, and changing the calculation algorithm for solving the ill-conditioned
Residual stress problem. In order to assess the feasibility and accuracy of the improved IHD method, an experiment of
Incremental hole-drilling method
measuring the in-plane stress distribution in a composite laminate under bending was proposed. The composite
Finite element method
laminate, which is bended by an external pure bending moment and fixed in a stable state, contains an in-plane
stress distribution with obvious gradients in thickness direction, which is essentially similar to residual stresses.
A plain woven T800H/803 specimen was tested in the proposed experiment with the improved IHD method.
The results calculated using both top-surface and bottom-surface measured strains, and by iteration algorithm,
showed good agreement with the finite element method (FEM) simulation and high calculation efficiency.

1. Introduction method to the measurement of non-uniform residual stresses. The incre-


mental hole-drilling (IHD) method, which involves the drilling in a se-
Residual stresses resulted by the curing process are present in all ries of small steps, is able to determine non-uniform stress profiles and
composites, including ceramic, metal and organic composites (Low gradients. It uses the calibration coefficient matrix to relate the mea-
et al., 1995). In composite materials, mismatches between the thermal sured strain relaxation field with the residual stresses originally existing
expansion coefficients of fibers and matrix, contraction and expansion inside composites. The determination of the calibration coefficients is
of different layers due to differences in their material directions, and based on the principle of superposition by FEM (Akbari et al., 2014).
temperature gradients in the manufacturing process cause residual The FEM model is established based on the specimen in the experiment
stresses (Ghasemi et al., 2014b). The negative effects caused by residual with the same strain gauge rosette arrangement. Usually the FEM model
stresses reduce the stiffness, strength and life of the composites and is set as a multi-layer structure and the thickness of each layer equals
can lead to undesirable defects such as matrix microcracking, interface to the stepping depth. In recent years, many researchers (Ghasemi and
debonding and warping (Ersoy and Vardar, 2000). Measuring these Mohammadi, 2016; Smit and Reid, 2018; Amir-Ahmadi et al., 2019)
stresses is very important in order to evaluate their effects during design successfully measured the non-uniform residual stresses in composites
step (Low et al., 1995). via the IHD method.
Non-destructive residual stress measurement methods such as X-ray However, the IHD method needs further improvement. Zuccarello
and neutron diffraction cannot be applied, or have highly restricted (1999) investigated the effects of errors in measured strains on the
application in composite laminates (Schajer, 2010). Measurement of calculated residual stresses. The author showed that in each hole-
the residual stresses in composite laminates usually demands the use drilling step, the error sensitivity depends on the hole depth and the
of relaxation methods which involve the measurement of deformations stepping depth. So most existing works utilized the IHD method on thin
arising from the removal of stressed material (Schajer, 2001). In early or simple lay-up composites. The number of drilling steps is usually
years, the through-hole drilling method was commonly used. However, less than 10. For thick and complicated lay-up composite laminates,
it is not appropriate for laminated composites. Because the elastic mate- the non-uniform residual stress distributions have rarely been studied
rial properties change abruptly at layers’ boundaries and, consequently, using the IHD method. One reason is that based on Saint-Venant’s
residual stresses are always non-uniform in thickness direction (Akbari principle, the response of strains on the top surface is insensitive to the
et al., 2014). Then many researchers extended the through-hole drilling deep internal stress release (Schajer, 1988). On the other hand, small

∗ Corresponding authors.
E-mail addresses: wangxd819@163.com (X. Wang), d5062010@163.com (Z. Guan).

https://doi.org/10.1016/j.mechmat.2020.103715
Received 1 May 2020; Received in revised form 10 December 2020; Accepted 11 December 2020
Available online 14 December 2020
0167-6636/© 2020 Elsevier Ltd. All rights reserved.
X. Liu et al. Mechanics of Materials 154 (2021) 103715

Fig. 1. Strain gauge rosette placement in the traditional IHD method.

errors in measured strains may result in much larger errors in the final
calculated stress distribution because of the ill conditioning of the cal-
ibration coefficient matrix. So the existing IHD method cannot obtain
accurate estimations of residual stresses in thick composite laminates.
Meanwhile, the validation of the feasibility and accuracy of the residual
stress measurement method is also worth studying. Actually residual
stresses in composites cannot be observed directly. Although some
researchers have proposed analytical methods (Ghasemi and Moham-
madi, 2016) or numerical methods by simulating the manufacturing
process (Gower et al., 2016) to predict residual stresses in composites,
these methods depend on many assumptions and complicated constants
of properties of materials. So the results of these prediction methods
can be used as references, but they are not so convincing to support the
validation. So there needs a more standardized and convincing method. Fig. 2. Major procedure of the traditional IHD method.
In this paper, the IHD method has been improved in both exper-
imental aspect and calculation algorithm. The strain gauge rosettes
were attached on both top and bottom surfaces of a specimen to direction. 𝐶 𝑝𝑞 (𝑧, ℎ𝑖 ) is the coefficient describes the measured strain
enlarge the thickness range of measurement and improve the accuracy of strain gauge 𝑝 caused by a unit stress in 𝑞 direction at depth 𝑧
of stress distribution calculation. Three calculation methods, including within the hole depth ℎ𝑖 . 𝜎 𝑞 (𝑧) is the residual stress at depth 𝑧 in 𝑞
least squares method, Tikhonov regularization method and iteration direction. In order to solve Eq. (1), an initial residual stress distribution
algorithm, for solving the ill-conditioned linear equation system were in thickness must be assumed. The stress distribution can be assumed as
analyzed to select the most suitable algorithm. Then to assess the a continuous polynomial with unknown coefficients (Ghaedamini et al.,
feasibility and accuracy of the improved IHD method, an experiment 2018):
of measuring the in-plane stress distribution in a composite laminate ∑ 𝑞 𝑞
𝜎 𝑞 (𝑧) = 𝐴𝑗 𝜙𝑗 (𝑧) (2)
under bending was proposed and accomplished using a plain woven 𝑗
composite specimen. The results calculated by various calculation al-
gorithms, based on different kinds of inputs of measured strains, were where 𝜙𝑞𝑗 (𝑧) are the base polynomials and 𝐴𝑞𝑗 is the domain coefficient
compared with the results simulated by FEM. for each 𝜙𝑞𝑗 (𝑧). Substituting Eq. (2) into Eq. (1) results in Eq. (3):
3 [
∑∑ ]
2. Improvement of the incremental hole-drilling method 𝜀𝑝 (ℎ𝑖 ) = 𝐶𝑗𝑝𝑞 (ℎ𝑖 )𝐴𝑞𝑗 (3)
𝑗 𝑞=1

2.1. Traditional incremental hole-drilling method where


ℎ𝑖
In the traditional IHD method, one strain gauge rosette is placed on 𝐶𝑗𝑝𝑞 (ℎ𝑖 ) = 𝐶 𝑝𝑞 (𝑧, ℎ𝑖 )𝜙𝑞𝑗 (𝑧)𝑑𝑧 (4)
∫0
the top surface of a specimen containing residual stresses, as shown in
The most commonly used base polynomials type for composites
Fig. 1. The major procedure is shown in Fig. 2.
is the pulse function (Ghasemi and Mohammadi, 2016; Ghaedamini
In the IHD method, the relation between the measured strains and
et al., 2018). In this type, the entire range of stress distribution in
the residual stresses has an integral form (Schajer and Prime, 2006).
thickness direction is divided into a number of small subranges, and
And the measured strains and the residual stresses both have three
the stresses in each subrange (usually one-layer thickness) are assumed
components corresponding to three in-plane directions. The equation
to be uniform. The pulse function is described as follow:
is shown as follow: {
1 ℎ𝑗−1 ≤ 𝑧 ≤ ℎ𝑗
ℎ𝑖 ∑
3
[ ] 𝜙𝑞𝑗 (𝑧) = ,1 ≤ 𝑗 ≤ 𝑖 (5)
𝜀𝑝 (ℎ𝑖 ) = 𝐶 𝑝𝑞 (𝑧, ℎ𝑖 )𝜎 𝑞 (𝑧) 𝑑𝑧 1 ≤ 𝑖 ≤ 𝑛, 𝑝 = 1, 2, 3 (1) 0 𝑧 < ℎ𝑗−1 , 𝑧 > ℎ𝑗
∫0
𝑞=1
Then Eq. (3) is transformed as follow:
where 𝜀𝑝 (ℎ𝑖 ) is the measured strain of strain gauge 𝑝 when the hole [ ]

𝑖 ∑
3 ℎ𝑗 ∑ 3 (
𝑖 ∑ )
depth is ℎ𝑖 . 𝑖 refers to various drilling steps. 𝑛 is the number of total
𝜀𝑝 (ℎ𝑖 ) = 𝜎𝑗𝑞 𝐶 𝑝𝑞 (𝑧, ℎ𝑖 )𝑑𝑧 = 𝜎𝑗𝑞 𝐶𝑖𝑗𝑝𝑞 (6)
drilling steps. 𝑝 means the gauge number. 𝑞 means the in-plane stress ∫ℎ𝑗−1
𝑗=1 𝑞=1 𝑗=1 𝑞=1

2
X. Liu et al. Mechanics of Materials 154 (2021) 103715

⎤⎡ ⎤ ⎡
11 12 13
⎡ ⎡ 𝜀11 ⎤ ⎡ 𝐶11 𝐶11 𝐶11 ⎤ ⎡ 𝜎1𝑥 ⎤ ⎤
⎢ ⎢ 2 ⎥ ⎥ ⎢⎢ ⎢ 21 22 23 ⎥
⎥ ⎢
⎥ ⎢
⎢ 𝑥𝑦 ⎥ ⎥
⎢ ⎢ 𝜀1 ⎥ ⎥⎢ ⎢ 𝐶11 𝐶11 𝐶11 ⎥ 0 ⋯ 0 ⎢ 𝜏1 ⎥ ⎥
⎢ ⎢ 3 ⎥ ⎥⎢ ⎢ 31 32 33 ⎥
⎥ ⎢ ⎢ 𝑦 ⎥ ⎥
⎢ ⎣ 𝜀1 ⎦ ⎥⎢ ⎣ 𝐶11 𝐶11 𝐶11 ⎦ ⎥ ⎢ ⎣ 𝜎1 ⎦ ⎥
⎢ ⎥⎢ 11 12 13 11 12 13
⎥ ⎢ ⎥
⎢ ⎡ 𝜀12 ⎤ ⎥⎢ ⎡ 𝐶21 𝐶21 𝐶21 ⎤ ⎡ 𝐶22 𝐶22 𝐶22 ⎤ ⎥ ⎢ ⎡ 𝜎2𝑥 ⎤ ⎥
⎢ ⎢ 2 ⎥ ⎥⎢ ⎢ 21 22 23 ⎥ ⎢ 21 22 23 ⎥ ⎥ ⎢ ⎢ 𝑥𝑦 ⎥ ⎥
⎢ ⎢ 𝜀2 ⎥ ⎥⎢ ⎢ 𝐶21 𝐶21 𝐶21 ⎥ ⎢ 𝐶22 𝐶22 𝐶22 ⎥ ⋯ 0 ⎥=⎢ ⎢ 𝜏2 ⎥ ⎥ (7)
⎢ ⎢ 3 ⎥ ⎥⎢ ⎢ 31 32 33 ⎥ ⎢ 31 32 33 ⎥ ⎥ ⎢ ⎢ 𝑦 ⎥ ⎥
⎣ 𝜀2 ⎦ ⎣ 𝐶21 𝐶21 𝐶21 ⎦ ⎣ 𝐶22 𝐶22 𝐶22 ⎦ ⎥ ⎢ ⎣ 𝜎2 ⎦
⎢ ⎥⎢ ⎥ ⎢ ⎥
⎢ ⋮ ⎥⎢ ⋮ ⋮ ⋱ ⋮ ⋮ ⎥
⎢ ⎥⎢ 11 12 13 11 12 13 ⎥ ⎢ ⎥
⎡ 𝜀1𝑛 ⎤ ⎡ 𝐶𝑛1 𝐶𝑛1 𝐶𝑛1 ⎤ ⎡ 𝐶𝑛2 𝐶𝑛2 𝐶𝑛2 ⎤ ⎡ 𝐶𝑛𝑛
11 12
𝐶𝑛𝑛 13 ⎤
𝐶𝑛𝑛 ⎥ ⎢ ⎡ 𝜎𝑛𝑥 ⎤
⎢ ⎢ 2 ⎥ ⎥⎢ ⎢ 21 ⎥
⎢ ⎥⎢ 22 23 ⎥ ⎢ 21 22 23 ⎥ ⎢ 21 22 23 ⎥ ⎥ ⎢ ⎢ 𝑥𝑦 ⎥

⎢ 𝜀𝑛 ⎥ ⎢ 𝐶𝑛1 𝐶𝑛1 𝐶𝑛1 ⎥ ⎢ 𝐶𝑛2 𝐶𝑛2 𝐶𝑛2 ⎥ ⋯ ⎢ 𝐶𝑛𝑛 𝐶𝑛𝑛 𝐶𝑛𝑛 ⎥ ⎥ ⎢ ⎢ 𝜏𝑛 ⎥
⎢ ⎢ 3 ⎥ ⎥⎢ ⎢ 31 33 ⎥ ⎢ 31 33 ⎥ ⎢ 31 33 ⎥ ⎥ ⎣ ⎢ 𝑦 ⎥ ⎥
⎣ ⎣ 𝜀𝑛 ⎦ ⎦⎣ ⎣ 𝐶𝑛1 32
𝐶𝑛1 𝐶𝑛1 ⎦ ⎣ 𝐶𝑛2 32
𝐶𝑛2 𝐶𝑛2 ⎦ ⎣ 𝐶𝑛𝑛 32
𝐶𝑛𝑛 𝐶𝑛𝑛 ⎦ ⎦ ⎣ 𝜎𝑛 ⎦ ⎦

Box I.

where 𝜎𝑗𝑞 means the value of the residual stress component in 𝑞 di- Table 1
rection at the depth range of ℎ𝑗−1 ≤ 𝑧 ≤ ℎ𝑗 . 𝐶𝑖𝑗𝑝𝑞 equals to the strain Equivalent radial and shear stress distributions of three load cases.

measured by strain gauge 𝑝 as a unit stress in 𝑞 direction is applied to Load case Cartesian coordinate system Polar coordinate system
{ 𝑥 { 𝜌
the hole wall at the depth range of ℎ𝑗−1 ≤ 𝑧 ≤ ℎ𝑗 , when the hole depth 𝜎 =1 𝜎 = 𝑐𝑜𝑠2 𝜑
Load case 1
is ℎ𝑖 after the 𝑖𝑡ℎ drilling. Eq. (6) can be expressed as a matrix form as 𝜏 𝑥𝑦 = 𝜎 𝑦 = 0 𝜏 𝜌𝜑 = −𝑐𝑜𝑠𝜑𝑠𝑖𝑛𝜑
{ {
given in Box I. 𝜏 𝑥𝑦
=1 𝜎 𝜌 = 𝑠𝑖𝑛2𝜑
For stating simply, Eq. (7) is expressed as follow: Load case 2
𝑥 𝑦
𝜎 =𝜎 =0 𝜏 𝜌𝜑 = 𝑐𝑜𝑠2𝜑
{ {
⎡ 𝜀1 ⎤⎡ 𝐶11 0 0 ⋯ 0 ⎤ ⎡ 𝜎1 ⎤ 𝜎 =1𝑦
𝜎 𝜌 = 𝑠𝑖𝑛2 𝜑
⎢ 𝜀2 ⎥⎢ 𝐶21 𝐶22 0 ⋯ 0 ⎥ ⎢ 𝜎2 ⎥ Load case 3
𝑥 𝑥𝑦
⎢ ⎥⎢ ⎥ ⎢ ⎥ 𝜎 =𝜏 =0 𝜏 𝜌𝜑 = 𝑐𝑜𝑠𝜑𝑠𝑖𝑛𝜑
⎢ 𝜀3 ⎥⎢ 𝐶31 𝐶32 𝐶33 ⋯ 0 ⎥=⎢ 𝜎3 ⎥ (8)
⎢ ⋮ ⎥⎢ ⋮ ⋮ ⋮ ⋱ ⋮ ⎥ ⎢ ⋮ ⎥
⎢ ⎥⎢ ⎥ ⎢ ⎥
⎣ 𝜀𝑛 ⎦⎣ 𝐶𝑛1 𝐶𝑛2 𝐶𝑛3 ⋯ 𝐶𝑛𝑛 ⎦ ⎣ 𝜎𝑛 ⎦
In the IHD method, the released strains recorded after a drilling step
are affected by not only the material removal in this step, but also the
material removals in previous steps. The previous removed material can
result in additional released strains as the hole gets deeper (Ghasemi
et al., 2014a). In Eq. (8), matrix [𝐶] is lower triangular. 𝐶𝑖𝑗 (𝑖 = 𝑗)
represents the influence of the material removal in this step. 𝐶𝑖𝑗 (𝑖 > 𝑗)
represents the influence of material removals in previous steps.
In the calculation of the calibration coefficient matrix using FEM,
the single layer is assumed as elastic, orthotropic material, and the
stress component perpendicular to the surface is negligible (Smit and
Reid, 2018). It is also assumed that the material removal does not in- Fig. 3. Three load cases applied to the hole wall in FEM model.

duce significant additional residual stresses in the remaining laminate,


which has been experimentally validated by Nobre et al. (2011). In
Eq. (8), each sub-matrix [𝐶𝑖𝑗 ] has nine elements which reflect the one- the hole is going through the entire laminate thickness, another rosette
to-one relations between three different unit in-plane stress components can be placed right opposite the top-surface rosette on the bottom
(𝜎𝑗𝑥 = 1, 𝜏𝑗𝑥𝑦 = 1, 𝜎𝑗𝑦 = 1) and three different strains measured by surface, as shown in Fig. 5. Both rosettes record the released strains
the rosette, respectively. Each of the nine components is obtained by which are the inputs to the calculation of residual stresses. Thereby the
the corresponding load case and strain record. In each load case, the
improved IHD method can ascertain the residual stresses in composites
equivalent radial and hoop stresses are applied simultaneously to the
with the thickness less than 2𝐷.
hole wall, which is the same as work in (Akbari et al., 2014). Equivalent
radial and shear stress distributions for three load cases are shown in Two rosettes are used to record the released strains in the first and
Table 1 and Fig. 3. The meanings of the components in the calibration last halves of the hole-drilling process, respectively. [𝜀] and [𝐶] in the
coefficient matrix are shown in Fig. 4. Finally, the in-plane residual improved IHD method are derived as follow:
stress distribution in thickness direction of the composite laminate can 𝑡𝑜𝑝
be calculated by Eq. (8). [𝜀] = [ 𝜀1 ⋯ 𝜀𝑡𝑜𝑝
⌈𝑛∕2⌉
𝜀𝑏𝑜𝑡𝑡𝑜𝑚
⌈𝑛∕2⌉+1
⋯ 𝜀𝑏𝑜𝑡𝑡𝑜𝑚
𝑛 ]⊤ (10)
{ 𝜌
𝜎 = 𝜎 𝑥 𝑐𝑜𝑠2 𝜑 + 𝜎 𝑦 𝑠𝑖𝑛2 𝜑 + 𝜏 𝑥𝑦 𝑠𝑖𝑛2𝜑
(9)
𝜏 𝜌𝜑 = (𝜎 𝑦 − 𝜎 𝑥 )𝑐𝑜𝑠𝜑𝑠𝑖𝑛𝜑 + 𝜏 𝑥𝑦 𝑐𝑜𝑠2𝜑 ⎡ 𝐶11𝑡𝑜𝑝
0 0 ⋯ 0 ⎤
⎢ ⋮ ⋮ ⋮ ⋱ ⋮ ⎥
⎢ 𝑡𝑜𝑝 𝑡𝑜𝑝 𝑡𝑜𝑝 ⎥
2.2. Improvement in experiment ⎢ 𝐶⌈𝑛∕2⌉ 1 𝐶⌈𝑛∕2⌉ 2
𝐶⌈𝑛∕2⌉ 3
⋯ 0 ⎥
[𝐶] = ⎢ 𝑏𝑜𝑡𝑡𝑜𝑚 𝑏𝑜𝑡𝑡𝑜𝑚 𝑏𝑜𝑡𝑡𝑜𝑚 ⎥ (11)
⎢ 𝐶⌈𝑛∕2⌉+1 1 𝐶⌈𝑛∕2⌉+1 2
𝐶⌈𝑛∕2⌉+1 3
⋯ 0 ⎥
In the traditional IHD method, the strain gauge rosette is only placed ⎢ ⋮ ⋮ ⋮ ⋱ ⋮ ⎥
on the top surface. Akbari et al. (2014) found that the measured strains ⎢ 𝑏𝑜𝑡𝑡𝑜𝑚 𝑏𝑜𝑡𝑡𝑜𝑚 𝑏𝑜𝑡𝑡𝑜𝑚 𝑏𝑜𝑡𝑡𝑜𝑚 ⎥
⎣ 𝐶𝑛1 𝐶𝑛2 𝐶𝑛3 ⋯ 𝐶𝑛𝑛 ⎦
are sensitive only to the released residual stresses within 𝑧 ≤ 𝐷 of the
hole depth. 𝐷 is the diameter of the hole. In this paper, considering that where the superscripts, 𝑡𝑜𝑝 and 𝑏𝑜𝑡𝑡𝑜𝑚, refer to the locations of rosettes.

3
X. Liu et al. Mechanics of Materials 154 (2021) 103715

Fig. 4. Meanings of the components in the calibration coefficient matrix.

Fig. 5. Placement of strain gauge rosettes in the improved IHD method.

2.3. Improvement in calculation algorithm where [𝐷] is the matrix to define the penalization type. The identity
matrix [𝐼] is usually chosen. 𝛼 is the parameter that controls the level of
According to Eq. (8), the unknown residual stresses, [𝜎𝑗 ], can be regularization. A large magnitude of 𝛼 gives a rise to the misfit and the
calculated by solving this linear equation system. Some researchers intensification of the model error. While a small magnitude of 𝛼 leads
(Ghaedamini et al., 2018; Ghasemi and Mohammadi, 2016) used the to an amplification of the measured data error. Therefore, 𝛼 needs to
conventional method, which means the direct calculation of [𝐶]−1 . be optimized to minimize both errors (Salehi and Shokrieh, 2019).
However, this method makes the strain fit exactly match the measured Iteration algorithm was proposed by Deng et al. (2015) for solving
data, which causes the solutions to be quite sensitive to measure- the ill-conditioned linear equation system based on Cholesky decom-
ment errors for the ill-conditioned [𝐶]. To improve the stabilization of position and Tikhonov regularization method. The iterative equation is
the calculated results, three methods, including least squares method, shown as follow:
Tikhonov regularization method and iteration algorithm, are usually ([𝐶]⊤ [𝐶] + 𝛼𝐼)[𝜎](𝑘+1) = [𝐶]⊤ [𝜀] + 𝛼[𝜎]𝑘 (13)
considered in the calculation algorithm.
Least squares method can be used when the number of measured The calculation steps are shown in Appendix. To solve Eq. (13)
strains is greater than the number of unknown stresses (Prime and and avoid inverting ([𝐶]⊤ [𝐶] + 𝛼𝐼) matrix, Cholesky decomposition is
Hill, 2005). However, in the IHD method, if the stepping depth is not adopted. The initial value of [𝜎] is [0]. So actually, Tikhonov regu-
less than one-layer thickness, this method may omit the stress changes larization method is a particular case of iteration algorithm when the
between the adjacent layers in composites (Salehi and Shokrieh, 2019). number of iterations is only 1. The details of the relationship between
If the stepping depth is less than one-layer thickness, the implement these two methods will be discussed in Section 4.4. As the iterations
of the experiment will be harder, and may cause inaccurate measured continue, the results of [𝜎] approach to the results of the conventional
strains. So least squares method is not so suitable here. method. So the number of iterations affects the final results. In this
Tikhonov regularization method (Tikhonov et al., 2013) is often paper, the termination criterion is established as all the values of strain
used to smooth the results by allowing a misfit between the calculated misfit components, as depicted in Eq. (14), are restricted to 1% of the
strains and the measured strains. It is accomplished by adding a penalty corresponding measured strains. Considering that some measured strain
function to Eq. (8): components may be very small and the 1% of their values are less
than 1𝜇𝜀. For these components, the corresponding misfit values are
[𝐶]⊤ [𝜀] = ([𝐶]⊤ [𝐶] + 𝛼[𝐷]⊤ [𝐷])[𝜎] (12) restricted to 1𝜇𝜀. This criterion ensures the results not deviate from the

4
X. Liu et al. Mechanics of Materials 154 (2021) 103715

Table 2 3.3. Experiment


Mechanical properties of the plain woven T800H/803
composite.
Actually, the bended specimen may originally contain residual
𝐸1 (GPa) 𝐸2 (GPa) 𝜇12 𝐺12 (GPa)
stresses, which means the measured strains of Rosettes A are affected
80.1 80.1 0.18 4.5
by both original residual stresses and additional bending stresses. For
measuring the pure bending stresses, three procedures were needed to
eliminate the effects of original residual stresses. Firstly, the IHD pro-
measured strains severely. Here this criterion is also used for Tikhonov cess was completed using Rosettes B when the specimen was under no
regularization method to determine the value of 𝛼. bending. Secondly, the specimen was bended with a specific deflection
by the designed fixture. Then another IHD process was completed using
[𝜀𝑚𝑖𝑠𝑓 𝑖𝑡 ] = [𝐶][𝜎] − [𝜀] (14) Rosettes A. Thirdly, the differences between two recorded strains at
each drilling step in the above two procedures were regarded as the
𝜀𝑚𝑖𝑠𝑓 𝑖𝑡,𝑖 < 0.01𝜀𝑖 𝑜𝑟 𝜀𝑚𝑖𝑠𝑓 𝑖𝑡,𝑖 < 1 (15) released strains on account of the pure bending stresses.
The experimental setup is shown in Fig. 7. Therein CNC drilling
machine with a high-speed turbine and a flat-bottomed carbide drilling
3. Validation experiment of residual stress measurement methods blade was used to drilling holes. The radius of the drilling blade
was 2 mm. The spindle speed was 24000 rpm and the feed rate
3.1. Theory was 0.001 mm/s. It has been demonstrated that this combination can
create enough qualified holes in carbon/epoxy composites (Flaman and
Essentially, residual stresses in a composite laminate usually have Herring, 1985; Sicot et al., 2004; Ghaedamini et al., 2018; Ghasemi
several features: stresses change along thickness direction, the stresses and Mohammadi, 2016; Akbari et al., 2014; Sicot et al., 2003; Nobre
components in the out-of-plane direction is negligible, and stresses are et al., 2011). The designed fixture, which was used to apply bending
balanced in forces and moments. Considering that residual stresses forces to the specimen and also fasten it, was mounted to the dovetail
cannot be observed directly, introducing a known stress distribution groove which was a part of the CNC machine. The designed fixture
which have the same features as residual stresses to a composite has two fixed indenters and two removable indenters. The removable
laminate can be a good way of ’prefabrication residual stresses’. Then indenters are controlled by the adjustable bolt to apply four-point
the measurement of the ’prefabricated residual stresses’ can be used bending forces to the specimen. In the experiment, the displacement
of the removable indenters was set as 2.5 mm. The perpendicularity
to assess the feasibility and accuracy of the measurement methods.
and coaxiality between the hole and the rosettes must be ensured to
This introduction can be accomplished by applying external constant
obtain accurate released strains. Here the bottom-surface rosette was
forces to a composite laminate, such as stretch, compression or bending.
accurately placed opposite the top-surface rosette. The perpendicularity
Therein bending can result in more obvious gradients along thickness
was ensured by the precise arrangement of the designed fixture on
direction in the stress distribution. If a laminate is bended and fixed in
dovetail groove via bolt fastening. The coaxiality was ensured by trial
a stable state, the stress distribution has the same features as residual
drillings with a 0.001 mm stepping depth on the top-surface rosette
stresses except for the moment equilibrium, which does not affect
and adjusting the location of the blade, until the relative locations were
the measurement. And the stress distribution caused by bending can
coaxial. The rosettes after the experiment on both surfaces are shown
be accurately calculated via FEM. So the measurement of the stress in Fig. 6, which can prove the accuracy of the drilling hole’s location.
distribution in a composite laminate under bending can be used to In addition, it is worth mentioning that the pressure force applied
evaluate the improved IHD method. by the translational movement of the blade during the hole drilling
may cause the specimen to deform. So the stepping depth should be
3.2. Specimen introduction slightly larger than one-layer thickness to ensure that the actual drilling
depth precisely equals to one-layer thickness. The magnitude of the
The composite specimen used in this paper was cut from a plain additional stepping depth, which depends on the material properties of
woven T800H/803 composite plate. The lay-up was [(0∕90)∕(+45∕ − the specimen, needs to be obtained via tests. In this paper, the stepping
depth was chosen as 0.217 mm, which can be used to exactly drill
45)∕(0∕90)∕(+45∕ − 45)2 ∕(0∕90)∕(+45∕ − 45)∕(0∕90)]𝑠 . The thickness was
through the specimen, after the tests with stepping depths varying from
3.42 mm and in-plane dimensions were 200 mm×30 mm. It has been
0.214 mm to 0.216 mm, which all resulted in blind holes after drilling.
mentioned in (Akbari et al., 2014) that with the dimensions of approx-
There were totally 16 drilling steps corresponding to 16 layers of the
imately 15 times bigger than the diameter of the hole, which is 2 mm
specimen. In each step, the rotating drilling blade reached the specified
in this paper, the edge effects can be neglected. Tests for determining
location to machine away the material in the specified layer, and then
the in-plane mechanical properties of the specimen containing only
retracted out of the specimen. Strains were recorded steadily for 180s
(0∕90) plies were carried out using ASTM D3039 (ASTM, 2017) and
in each step to allow all thermal effects to dissipate.
ASTM D3518 (ASTM, 2018) standards. The results are presented in
Table 2. Two pairs of BX120-1CG commercial strain gauge rosettes, 3.4. Calculation of the calibration coefficient matrix
marked as Rosettes A and Rosettes B, respectively, were attached to
both top and bottom surfaces of the specimen at specified positions as In this paper, ABAQUS commercial software was used for obtaining
shown in Fig. 6. Rosettes A were at the center of the specimen, and the calibration coefficient matrix. A three dimensional model was
Rosettes B were set 40 mm away from the center to avoid affecting created using SC8R continuum shell elements, as shown in Fig. 8. The
the measurement of Rosettes A. The ratio between the drilling hole mesh of the model was similar to other researchers’ works (Akbari
radius and the inner radius of the rosettes was 0.5, which was in et al., 2014; Amir-Ahmadi et al., 2019). The accuracy of this model
accordance with the recommendations in (Sicot et al., 2004). Before was also verified by a convergence analysis. The model had 16 solid
the attachment, the surfaces of the specimen were polished with sand layers which coincided with the specimen. The minimum mesh size
paper, and alcohol was used to clean superfluous particles. Then the was about 0.0025 mm. As mentioned before, the calculation of the
rosettes were installed on the specimen’s surfaces precisely. Therein the calibration coefficient matrix is a repetitive work. Totally 3*16 calcula-
strain gauge No.1 and No.3 were located in the length (0◦ ) and width tion processes were accomplished corresponding to three kinds of load
(90◦ ) directions, respectively. cases, as shown in Fig. 3, applied to the hole walls of 16 different layers

5
X. Liu et al. Mechanics of Materials 154 (2021) 103715

Fig. 6. Specimen and strain gauge rosettes.

Fig. 7. Experimental setup of the stresses measurement of the specimen under bending.

the indenters were the same as those in the experiment. Considering


that there may be gaps among different parts of the designed fixture,
the displacement of removable indenters was set by ensuring that the
strains at the center of the bended model were the same as the strains
measured by Rosettes A after the specimen was bended in experiment.
There were no residual stresses inside the model before bending. So the
stress distribution was caused only by the bending deformation. The
in-plane stresses at the center of each layer were recorded.

4. Results and discussion

4.1. Experimental results

The measured strains of Rosettes B with the specimen under no


bending after the IHD process were nearly all zero, which means
Fig. 8. FEM model for calculating the calibration coefficient matrix. that there were nearly no original in-plane residual stresses inside the
specimen. Fig. 10 shows measured strains of Rosettes A in the whole
IHD process. The results had six components corresponding to three
individually, as shown in Fig. 4. After each load case was applied, the in-plane directions (0◦ , 45◦ , 90◦ ) and two surfaces. At the start of each
strain at each gauge location was determined by nodal displacements drilling step, all the strains decreased due to the deformation of the
at gauge boundaries as well as the gauge length in the model (Akbari specimen, which was caused by the pressure force of the blade. The
et al., 2014). coefficients of thermal expansion mismatch between strain gauges and
the specimen also resulted in additional negative strains as the local
3.5. FEM simulation of the stress distribution temperature arisen caused by the friction effect. Meanwhile the strains
were unstable because of the vibration. After the blade retracting out of
To examine the results measured by the improved IHD method, the the specimen, the thermal effect dissipated and the specimen resumed
in-plane stress distribution of the specimen under bending was calcu- to a steady state gradually. Then the measured strains stabilized as
lated via FEM, as shown in Fig. 9. The model was three dimensional and time increased. All the stabilized strains at the end of each step (cor-
created using SC8R elements. Two fixed indenters and two removable responding to each layer) were extracted to be the inputs to the stress
indenters were also created as rigid bodies. The relative locations of distribution calculation, as shown in Fig. 11. Fig. 11 clarifies that the

6
X. Liu et al. Mechanics of Materials 154 (2021) 103715

Fig. 9. Stress distribution of the specimen under bending calculated by FEM.

Fig. 10. Measured strains of Rosettes A in the whole IHD process.


Fig. 11. Extracted measured strains of Rosettes A after each step (corresponding to
each layer).

measured strains of the rosettes on the top and bottom surfaces changed
obviously in the first three and the last three steps, respectively, and where 𝑥 and 𝑦 are two vectors or matrices which have the same dimen-
nearly kept constants in other steps. It is worth mentioning that a sion. EDS can be unitized to give more intuitional results regardless of
spike of the last step of 0◦ strain on the bottom surface occurred the dimensions, as shown in follow:
because of the bottom layer’s geometry change and the big released
0◦ stress inside the bottom layer. Before the removal of the material 1
𝑈 𝐸𝐷𝑆(𝑥, 𝑦) = √ (17)
at the bottom layer center, the strains of the bottom-surface rosette 1+ (𝛴(𝑥𝑖 − 𝑦𝑖 )2 ∕𝑛)
changed little for the intact bottom layer structure. After the removal,
Then similarities between the results of all various calculation meth-
the big compressive 0◦ stress inside the bottom layer released, and a
ods and FEM are calculated and shown in Figs. 12–14. Therein the
hole occurred. These two factors resulted in the spike, and the final
magnitude of the bottom-surface 0◦ strain was similar to the top-surface UEDS of all results means that all results’ components (𝜎 𝑥 , 𝜏 𝑥𝑦 , 𝜎 𝑦 ) are
one. put into Eq. (17), which reveals an overall similarity. The UEDS of a
specified stress component means that only the specified component of
4.2. Calculation only using top-surface measured strains results is put into Eq. (17), which reveals the similarity of this compo-
nent. It is worth mentioning that the value of UEDS shows exponential
Figs. 12–14 illustrate the stresses calculated by conventional decline as differences increase. So the absolute value of UEDS is quite
method, Tikhonov regularization method and iteration algorithm using smaller than 1. But the comparison of the UEDS of various methods
top-surface measured strains. These figures are plotted corresponding can obviously reveal the accuracy differences. As shown in Figs. 12–14,
to three stress components (𝜎 𝑥 , 𝜏 𝑥𝑦 , 𝜎 𝑦 ). Layer number 1 refers to the top for the stress distributions calculated only using top-surface measured
layer, and layer number 16 refers to the bottom layer. FEM simulation strains, Tikhonov regularization method and iteration algorithm gen-
results are all fluctuant along thickness direction due to the existing of erate nearly the same results and show relatively good accuracy. The
(+45/−45) layers. Therein 𝜎 𝑥 changes from 146 MPa to 0 MPa, and to
values of 𝛼 used in these two methods and the reason of the nearly
−146 MPa as the layer number increases from 1 to 8, and to 16. 𝜏 𝑥𝑦
same results will be discussed later. The conventional method generates
nearly remains 0 MPa in all layers. 𝜎 𝑦 changes also around 0 MPa, but it
several unreasonable stress values and shows the worst accuracy. It can
has a bigger magnitude than 𝜏 𝑥𝑦 for Poisson effect. In this paper, united
Euclidean distance similarity (UEDS) is utilized to quantificationally be observed that all methods obtain relatively accurate results at the
compare the accuracy of various calculation methods. Euclidean dis- first 8 layers, especially the first 3 layers. The unreasonable stress values
tance similarity (EDS) (Elmore and Richman, 2001) is a simple and of the conventional method verify that the ill-conditioned problem is
intelligible method to evaluate the similarity between two vectors or nonnegligible. In summary, the IHD method only using top-surface
matrices. The formula of EDS is: measured strains can estimate the stress distribution in the first half
1 layers, especially accurately in the first 3 layers. But the estimation
𝐸𝐷𝑆(𝑥, 𝑦) = √ (16)
1 + (𝛴(𝑥𝑖 − 𝑦𝑖 )2 ) accuracy in the last half layers is not acceptable.

7
X. Liu et al. Mechanics of Materials 154 (2021) 103715

Fig. 12. 𝜎 𝑥 calculated by various calculation methods only using top-surface measured strains and FEM simulated results.

Fig. 13. 𝜏 𝑥𝑦 calculated by various calculation methods only using top-surface measured strains and FEM simulated results.

4.3. Calculation using both top-surface and bottom-surface measured strains the estimation accuracy. But the estimation of stress distributions in
middle thickness is still less precise than the estimations in the fields
Figs. 15–17 illustrate the stresses calculated by the three methods near top and bottom surfaces, which is the intrinsic limitation of the
using both top-surface and bottom-surface measured strains. Compar- IHD method.
ing to the calculated results only using top-surface measured strains,
the introduction of bottom-surface measured strains improves the es- 4.4. Discussion about Tikhonov regularization method and iteration algo-
timation accuracy in the last half layers apparently, which results in rithm
higher UEDS of all results. Therein Tikhonov regularization method and
iteration algorithm still generate nearly the same results and fit well As mentioned above, actually Tikhonov regularization method is a
to the FEM simulated results. The conventional method still generates particular case of iteration algorithm when the number of iterations is
several unreasonable stress values, which means the ill-conditioned only 1. Fig. 18 shows the UEDS of all results in Tikhonov regularization
problem is nonnegligible yet despite the introduction of bottom-surface method and iteration algorithm with different 𝛼. When 𝛼 ≤1E-3, the
measured strains. For Tikhonov regularization method and iteration results of two methods are the same because the number of iterations
algorithm, the bottom-surface measured strains improve the accuracy is 1 for iteration algorithm. And by this time, the results are all satisfied
in both the first half and the last half layers. In summary, the use of both with Eqs. (14) and (15) because of the small 𝛼. The smaller 𝛼, the
top-surface and bottom-surface measured strains apparently improves worse results are. With 𝛼 continually increases from 1E-3, the results

8
X. Liu et al. Mechanics of Materials 154 (2021) 103715

Fig. 14. 𝜎 𝑦 calculated by various calculation methods only using top-surface measured strains and FEM simulated results.

Fig. 15. 𝜎 𝑥 calculated by various calculation methods using both top-surface and bottom-surface measured strains and FEM simulated results.

calculated by Tikhonov regularization method are not satisfied with as long as 𝛼 is big enough. In conclusion, iteration algorithm can
Eqs. (14) and (15). So these results cannot be used in practice although obtain the most proper results more easily than Tikhonov regularization
the UEDS of results may be bigger. Because in the actual experiment, method with the same accuracy.
the real residual stress distribution is unknown, which means the only
criterion to decide the results are reasonable or not is from Eqs. (14)
5. Conclusion
and (15), but not UEDS. That is to say, it needs several trial calculations
with different values of 𝛼 in Tikhonov regularization method to find the
most proper results, and the choice of 𝛼 should be from a big value to In this paper, the IHD method is improved in both experiment
a small value gradually. However, for iteration algorithm, when 𝛼 is aspect and calculation algorithm. An experiment method of measuring
bigger than 1E−3, it can always obtain the most proper results that are the in-plane stress distribution in a composite laminate under bending
satisfied with Eqs. (14) and (15) by iterative calculations. The bigger is proposed and accomplished to assess the feasibility and accuracy
value of 𝛼, the bigger number of iterations is. That means iteration of the improved IHD method. The results calculated from different
algorithm can always approach to the most proper results in only kinds of measured strains inputs and by various calculation methods,
one calculation process without needing additional trial calculations, including the conventional method, Tikhonov regularization method

9
X. Liu et al. Mechanics of Materials 154 (2021) 103715

Fig. 16. 𝜏 𝑥𝑦 calculated by various calculation methods using both top-surface and bottom-surface measured strains and FEM simulated results.

Fig. 17. 𝜎 𝑦 calculated by various calculation methods using both top-surface and bottom-surface measured strains and FEM simulated results.

and iteration algorithm, are compared to the FEM simulated results. CRediT authorship contribution statement
Some conclusions can be extracted:
Xiaodong Liu: Methodology, Software, Investigation, Writing -
(1) The IHD method only using top-surface measured strains can original draft. Xiaodong Wang: Conceptualization, Methodology, Su-
estimate the stress distribution in the first half layers. But the pervision, Writing - review & editing. Zhidong Guan: Conceptualiza-
estimation accuracy in the last half layers is not acceptable. The tion, Supervision. Ting Jiang: Investigation, Visualization. Kunhao
introduction of bottom-surface measured strains can improve the Geng: Investigation, Validation. Zengshan Li: Methodology, Supervi-
estimation accuracy in both the first and the last half layers. sion.
(2) The ill-conditioned problem is nonnegligible in the IHD method.
Iteration algorithm can get stable and accurate results with high Declaration of competing interest
calculation efficiency.
(3) The stress measurement of a composite laminate under bending The authors declare that they have no known competing finan-
is suitable to assess the feasibility and accuracy of the residual cial interests or personal relationships that could have appeared to
stress measurement method. influence the work reported in this paper.

10
X. Liu et al. Mechanics of Materials 154 (2021) 103715

Fig. 18. UEDS of Tikhonov regularization method and iteration algorithm with different 𝛼.

Appendix. The calculation steps of iteration algorithm equation [𝑁][𝜎] = [𝑊 ].

⎧ 𝜎(𝑖) = 𝛽(𝑖)∕𝑈 (𝑖, 𝑖), 𝑖 𝑖𝑠 𝑓 𝑟𝑜𝑚 1 𝑡𝑜 𝑛



(1) 𝛼 is assigned with a big enough value. It can be the similar ⎨ ∑
𝑖+1

⎪ 𝜎(𝑖) = 𝜎(𝑖) − [𝑈 (𝑖, 𝑗)∕𝑈 (𝑖, 𝑖) × 𝜎(𝑗)], 𝑖 𝑖𝑠 𝑓 𝑟𝑜𝑚 𝑛 − 1 𝑡𝑜 𝑛


magnitude as residual stresses. [𝜎] is assigned as [0]. ⎩ 𝑗=𝑛
(2) Setting [𝑁][𝜎] = [𝑊 ]. Assuming [𝑁], [𝑊 ] are:
{ (A.6)
[𝑁] = [𝐶]⊤ [𝐶] + 𝛼𝐼
(A.1) (7) Stopping the iteration when the termination criterion is satisfied.
[𝑊 ] = [𝐶]⊤ [𝜀] + 𝛼[𝜎]𝑘

[𝑁] can be decomposed into the product of a lower triangular References


matrix [𝐴] and its conjugate transpose [𝐴]⊤ , i.e., [𝑁] = [𝐴][𝐴]⊤ .
Akbari, S., Taheri-Behrooz, F., Shokrieh, M.M., 2014. Characterization of residual
This will yield an equation ([𝐴][𝐴]⊤ )[𝜎] = [𝑊 ]. stresses in a thin-walled filament wound carbon/epoxy ring using incremental
(3) Assuming the initial value of [𝐴] is: hole drilling method. Compos. Sci. Technol. 94, 8–15. http://dx.doi.org/10.1016/
j.compscitech.2014.01.008.
[𝐴] = [0] (A.2) Amir-Ahmadi, S., Ghasemi, A.R., Mohammadi, M., 2019. Evaluation of thermal residual
𝑛×𝑛 stresses of thin-walled laminated composite pipes to characterize the effects of
where [0] is zero matrix; 𝑛 × 𝑛 is the size of [𝑁]. mandrel materials and addition MWCNTs. Mech. Mater. 136, 103083. http://dx.
doi.org/10.1016/j.mechmat.2019.103083.
(4) According to Cholesky decomposition, computing the elements ASTM, 2017. D3039/D3039M-17 standard test method for tensile properties of polymer
of matrix [𝐴] by following equations: matrix composite materials. http://dx.doi.org/10.1520/D3039_D3039M-17.
{ √ ASTM, 2018. D3518/D3518M-18 standard test method for in-plane shear response
𝐴(1, 1) = 𝑁(1, 1) of polymer matrix composite materials by tensile test of a ±45◦ laminate. http:
, 𝑖 𝑖𝑠 𝑓 𝑟𝑜𝑚 2 𝑡𝑜 𝑛 (A.3) //dx.doi.org/10.1520/D3518_D3518M-18.
𝐴(𝑖, 1) = 𝑁(𝑖, 1)∕𝐴(1, 1) Deng, X., Yin, L., Peng, S., Ding, M., 2015. An iterative algorithm for solving ill-
conditioned linear least squares problems. Geod. Geodyn. 6 (6), 453–459. http:
⎧ √ //dx.doi.org/10.1016/j.geog.2015.06.004.

⎪ √ ∑
𝑗−1 Elmore, K.L., Richman, M.B., 2001. Euclidean distance as a similarity metric for
⎪ 𝐴(𝑗, 𝑗) = √𝑁(𝑗, 𝑗) − 𝐴(𝑗, 𝑘)2 principal component analysis. Mon. Weather Rev. 129 (3), 540–549. http://dx.
⎪ 𝑘=1 doi.org/10.1175/1520-0493(2001)129<0540:EDAASM>2.0.CO;2.
⎨ [ ] , Ersoy, N., Vardar, O., 2000. Measurement of residual stresses in layered composites
⎪ ∑
𝑗−1
by compliance method. J. Compos. Mater. 34 (7), 575–598. http://dx.doi.org/10.
⎪ 𝐴(𝑖, 𝑗) 𝑁(𝑖, 𝑗) − [𝐴(𝑖, 𝑘) × 𝐴(𝑗, 𝑘)] ∕𝐴(𝑗, 𝑗) 1177/002199830003400703.
⎪ 𝑘=1
⎩ Flaman, M.T., Herring, J.A., 1985. Comparison of four hole-producing techniques for
the center-hole residual-stress measurement method. Exp. Tech. 9 (8), 30–32.
𝑗 𝑖𝑠 𝑓 𝑟𝑜𝑚 2 𝑡𝑜 𝑛, 𝑖 𝑖𝑠 𝑓 𝑟𝑜𝑚 𝑗 + 1 𝑡𝑜 𝑛 (A.4) http://dx.doi.org/10.1111/j.1747-1567.1985.tb02036.x.
Ghaedamini, R., Ghassemi, A., Atrian, A., 2018. A comparative experimental study
(5) Letting [𝛽] = [𝐴]⊤ [𝜎]. Then solving [𝛽] by the equation [𝐴][𝛽] = for determination of residual stress in laminated composites using ring core,
[𝑊 ]: incremental hole drilling, and slitting methods. Mater. Res. Express 6 (2), 025205.
http://dx.doi.org/10.1088/2053-1591/aaee46.
⎧ 𝛽(𝑖) = 𝑊 (𝑖)∕𝐴(𝑖, 𝑖), 𝑖 𝑖𝑠 𝑓 𝑟𝑜𝑚 1 𝑡𝑜 𝑛 Ghasemi, A.R., Mohammadi, M.M., 2016. Residual stress measurement of fiber metal
⎪ laminates using incremental hole-drilling technique in consideration of the integral
⎨ ∑
𝑖−1
(A.5) method. Int. J. Mech. Sci. 114, 246–256. http://dx.doi.org/10.1016/j.ijmecsci.
⎪ 𝛽(𝑖) = 𝛽(𝑖) − [𝐴(𝑖, 𝑗)∕𝐴(𝑖, 𝑖) × 𝛽(𝑗)], 𝑖 𝑖𝑠 𝑓 𝑟𝑜𝑚 2 𝑡𝑜 𝑛
2016.05.025.
⎩ 𝑗=1
Ghasemi, A.R., Taheri-Behrooz, F., Shokrieh, M.M., 2014a. Determination of non-
uniform residual stresses in laminated composites using integral hole drilling
(6) Letting [𝑈 ] = [𝐴]⊤ , taking the values of [𝛽] and solving [𝜎] by method: experimental evaluation. J. Compos. Mater. 48 (4), 415–425. http://dx.
the equation [𝑈 ][𝜎] = [𝛽]. This will give the solution [𝜎] in the doi.org/10.1177/0021998312473858.

11
X. Liu et al. Mechanics of Materials 154 (2021) 103715

Ghasemi, A.R., Taheri-Behrooz, F., Shokrieh, M.M., 2014b. Measuring residual stresses Schajer, G.S., 1988. Measurement of non-uniform residual stresses using the hole-
in composite materials using the simulated hole-drilling method. Residual Stress. drilling method. Part I—Stress calculation procedures. J. Eng. Mater. Technol. 110
Compos. Mater. 76–120. http://dx.doi.org/10.1533/9780857098597.1.76. (4), 338–343. http://dx.doi.org/10.1115/1.3226059.
Gower, M.R., Shaw, R.M., Wright, L., Urquhart, J., Hughes, J., Gnaniah, S., Morrell, R., Schajer, G.S., 2001. Residual stresses: measurement by destructive methods. In: Section
Garstka, T., 2016. Determination of ply level residual stresses in a laminated carbon 5a in Encyclopedia of Materials: Science and Technology. Elsevier, Oxford.
fibre-reinforced epoxy composite using constant, linear and quadratic variations of Schajer, G.S., 2010. Hole-drilling residual stress measurements at 75: origins, advances,
the incremental slitting method. Composites A 90, 441–450. http://dx.doi.org/10. opportunities. Exp. Mech. 50 (2), 245–253. http://dx.doi.org/10.1007/s11340-009-
1016/j.compositesa.2016.08.005. 9285-y.
Low, B.Y., Gardner, S.D., Pittman Jr, C.U., Hackett, R.M., 1995. A micromechanical Schajer, G.S., Prime, M.B., 2006. Use of inverse solutions for residual stress measure-
characterization of residual thermal stresses in carbon fiber/epoxy composites ments. J. Eng. Mater. Technol. 128 (3), 375–382. http://dx.doi.org/10.1115/1.
containing a non-uniform interphase region. Compos. Eng. 5 (4), 375–396. http: 2204952.
//dx.doi.org/10.1016/0961-9526(94)00104-H. Sicot, O., Gong, X.L., Cherouat, A., Lu, J., 2003. Determination of residual stress in
Nobre, J.P., Stiffel, J.H., Van Paepegem, W., Nau, A., Batista, A.C., Marques, M.J., composite laminates using the incremental hole-drilling method. J. Compos. Mater.
Scholtes, B., 2011. Quantifying the drilling effect during the application of 37 (9), 831–844. http://dx.doi.org/10.1177/002199803031057.
incremental hole-drilling technique in laminate composites. Mater. Sci. Forum 681, Sicot, O., Gong, X.L., Cherouat, A., Lu, J., 2004. Influence of experimental parameters
510–515. http://dx.doi.org/10.4028/www.scientific.net/MSF.681.510. on determination of residual stress using the incremental hole-drilling method.
Prime, M.B., Hill, M.R., 2005. Uncertainty, model error, and order selection for Compos. Sci. Technol. 64 (2), 171–180. http://dx.doi.org/10.1016/S0266-3538(03)
series-expanded, residual-stress inverse solutions. J. Eng. Mater. Technol. 128 (2), 00278-1.
175–185. http://dx.doi.org/10.1115/1.2172278. Smit, T.C., Reid, R.G., 2018. Residual stress measurement in composite laminates
Salehi, S.D., Shokrieh, M.M., 2019. Residual stress measurement using the slitting using incremental hole-drilling with power series. Exp. Mech. 58 (8), 1221–1235.
method via a combination of eigenstrain, regularization and series truncation http://dx.doi.org/10.1007/s11340-018-0403-6.
techniques. Int. J. Mech. Sci. 152, 558–567. http://dx.doi.org/10.1016/j.ijmecsci. Tikhonov, A.N., Goncharsky, A., Stepanov, V., Yagola, A.G., 2013. Numerical Methods
2019.01.011. for the Solution of Ill-Posed Problems, Vol. 328. Springer Science & Business Media.
Zuccarello, B., 1999. Optimal calculation steps for the evaluation of residual stress by
the incremental hole-drilling method. Exp. Mech. 39 (2), 117–124. http://dx.doi.
org/10.1007/BF02331114.

12

You might also like