3.1. Modelling of Cracks: An Overview of the Discrete Spring Model
The identification of the location and depth of a crack in beam-type structures is an important topic in structural health monitoring and has been the subject of a significant amount of research. The literature on crack modelling is very rich, and various models have been proposed in the technical literature, such as local stiffness reduction, discrete spring models and complex models in two or three dimensions. In this topic, the state of the art can be found in a review works by Alijani et al. [
17] and by Palmeri and Cicirello [
29]. In both papers, the authors provide a coherent but concise review of as many publications as possible, and the main topics covered are modelling and simulation of the static behaviour of cracked beams.
According to the classification by Friswell and Penny [
30], the proposed approach falls into the broad category of “discrete spring models”, being equivalent to an internal hinge coupled with a linear elastic spring. This model is widely adopted when structural analysis focuses on the overall performance of frames and trusses rather than on crack initiation and propagation phenomena. Although very simple, the discrete spring model proves to be very efficient for static problems. The idea of treating cracked beams with linear springs equivalent in crack location is based on the division of each element into undamaged pieces between two consecutive cracks. One of the main advantages of this model is the effective representation of the crack in terms of position and gravity. Several attempts to provide values for the rotational stiffness of the spring using cracking parameters such as depth or geometry can be found in the literature. Among these, Palmeri and Cicirello [
29] studied the static analysis of multi-cracked Euler–Bernoulli and Timoshenko beams using this relationship. Subsequently, Okamura et al. [
31] showed the buckling behaviour of rectangular-section cracked columns. Furthermore, Ricci and Viola [
32] extended the method of Kienzler and Herrmann [
33] to evaluate the stress intensity factors of cracked beams and bars and derived two relationships between the stress intensity factor and the rotational stiffness of the spring. In the following analysis, two relationships between the stress intensity factor and the rotational stiffness of the spring were considered, as introduced by Alijani et al. in [
17].
Figure 2 shows a beam with a crack in the non-dimensional position
and depth
a. In addition, it is assumed that the shear effect as well as axial and torsional loads are neglected. According to this assumption, the stress intensity factors may be evaluated through the following equations:
where
is given from the following geometric function
Finally, the following relationship is defined:
which is used to determine a rotational spring stiffness factor equivalently in terms of the geometrical and material parameters of the crack and where
in Equation (
27) is equal to
and
is the stiffness of the i-th equivalent spring.
In the present paper, the rotational spring stiffness factor is assumed to be different for each beam. In particular, and denote the rotational spring stiffness factor for the upper and lower beams, respectively.
3.2. Method of Solution: Cell Discretisation Method (CDM)
The Cell Discretisation Method (CDM) is an efficient numerical method for solving linear partial differential equations. It has become an important tool in the field of structural engineering due to its approximation capabilities and ease of implementation. The advent of sophisticated and fully generalised discretisation tools, such as the finite element method and the Boundary Element Method (BEM), has made it possible to simulate the behaviour of structures by taking into account more variables due to the elimination of simplified assumptions, but, on the other hand, such procedures can lead to losing the physical sense of the real behaviour of structures that should always be the basis of engineering studies. In this sense, the CDM can be considered a technique capable of tackling such problems. Since the beginning of the 20th century, this method has found various applications, for example, the dynamics and stability of arches, masonry arches, the static and dynamic analysis of Euler–Bernoulli beams under different load and boundary conditions, the static and dynamic analysis of Timoshenko and Rayleigh beams and the static analysis of plates under different load and boundary conditions [
23,
24,
25,
26,
27,
28]. More recently, some of the present authors have applied the method to the dynamic analysis of single- and double-walled carbon nanotubes, taking into account non-local effects [
23,
26], and have obtained results showing that the method is able to describe the behaviour of the nanostructure satisfactorily with little computational effort. In several articles, the procedure proves to be very versatile and able to work on a finite number of Lagrangian parameters in each case, bringing the solution within the scope of the usual numerical analysis methods.
In the present paper, the two beams are reduced to a set of
t rigid bars with the same length
, connected by n = t + 1 elastic cells (see
Figure 3). The moment of inertia
, with j = 1, 2, will be evaluated on the abscissa of the cells, resulting in the concentrated stiffness
and
for the upper and lower beam, respectively. Both quantities can be organised into the so-called unassembled stiffness diagonal matrix
with dimension (n × n), j = 1, 2 for each of the two beams.
If the cross-section is not uniform, the average inertia across the elastic cell can be considered for each rigid section of length .
In this way, the structure is reduced to a classical holonomic system, with 2
n degrees of freedom; in particular,
n vertical displacements
for the upper beam and
n vertical displacements
for the lower beam. The abscissa of the cells will be conveniently taken as Lagrangian coordinates and will be organised in the 2n-dimensional vector v. Moreover, for the upper and lower beams, the
n − 1 rotations of the rigid bars can be calculated as a function of the Lagrangian coordinates as follows:
or, in matrix form,
V and
V, where
is a rectangular transfer matrix with
n− 1 rows and n columns.
The relative rotations between the two faces of the elastic cells are given by:
or, in matrix form,
=
for upper rigid bar and
=
for lower rigid bar, where
is another rectangular transfer matrix with n rows and n−1 columns.
The strain energies
, with j = 1, 2, (the first two terms of Equation (
1)) are given by:
and they are concentrated at the cells of the upper and lower beams, respectively.
The strain energies should be expressed as functions of the Lagrangian coordinates as follows:
so that the total strain energy can be expressed as:
where
=
and
=
. The global assembled stiffness matrix
K, with 2
n rows and 2
n columns, assumes the following form:
The potential energy, as a function of the Lagrangian coordinates, is given by:
Being the double-beam system subjected to a uniformly distributed load
q
with
the global assembled load matrix
Q with 2
n rows and 2
n columns assumes the following form:
The strain energy due to the elastic medium, Equation (
2), can be expressed as:
The terms of matrix
are given by:
Matrix
, with
n rows and
n columns and half-band widths equal to 2, takes the following form:
Finally, the governing static equation can be written as:
where
is the global assembled stiffness matrix:
Boundary Conditions in Presence of a Crack
Finally, from the strain energy terms of the flexible constraints at the ends in Equiation (1), the assembled stiffness matrix
K must be modified as follows:
The rotational stiffnesses of the constraints of each beam can be taken into account by adding the corresponding flexibilities of the rigid bars and obtaining:
These terms will be organised in two matrices,
and
, given in (40). In presence of a crack on the abscissa of cell
, the local stiffness of the upper beam is given by:
whereas at the
cell abscissa relative to the lower beam, the local stiffness is:
if the height of the crack is different in the two beams. The values
and
are derived from Equation (
34).
3.3. Method of Solution: Finite Element Method (FEM)
In order to highlight the efficiency of the proposed computational model (CDM), the authors considered the derivation of a finite beam element for the static analysis of a cracked double-beam system. Specifically, a finite element of a cracked double-beam system was developed based on a variational approach, and shape functions for rotational and translational displacements were used to develop the stiffness matrix in the presence and absence of cracks.
For the structure under consideration (see
Figure 4), the total potential energy is given by:
Using cubic polynomial functions for transverse displacements, the following shape functions are derived:
where
i = 1, 2 refers to the upper and lower beams, respectively. The transverse displacements are then given as:
where
denotes the vector of nodal displacements and is given by:
By deriving the nodal displacements with respect to the spatial coordinate z, the following expressions are obtained:
which, substituted for Equation (
54), leads to:
By applying the principle of stationary potential energy, one obtains:
Then, rearranging the terms in
and
, one obtains:
or
being
Equation (
62) can be rewritten as:
From the second Equation (
64), we obtain
, which, substituted in the first equation, leads to:
or
In the presence of a crack, the stiffness matrix of finite element with a crack must be modified. In the presence of a crack, the stiffness matrix of the cracked beam element must be modified. Several investigators have derived the stiffness matrix for a cracked beam element and studied the effect of the crack on the stiffness matrix (see, e.g., [
34,
35]). In the present analysis, the structure under investigation was analysed using finite beam elements that did not take into account the effect of shear forces. Therefore, the stiffness matrix, which represents the relationship between the vector of nodal forces and the vector of nodal displacements, can be expressed as:
so as deduced in [
35] (see Formula (22)), where
,
is given by:
and where
This matrix will be translated into at the appropriate abscissa.