Experimental Investigations of Impact Damage Influence on Behavior of Thin-Walled Composite Beam Subjected to Pure Bending

The paper deals with buckling, postbuckling, and failure of pre-damaged channel section beam subjected to pure bending. The channel section beams made of eight-layered GFRP laminate with different symmetrical layups have been considered. The specimens with initially pre-damaged web or flange were investigated to access the influence of impact damage on work of thin-walled structure in the full range of load till failure. The bending tests of initially pre-damage beams have been performed on a universal tensile machine with especially designed grips. The digital image correlation system allowing to follow the beam deflection have been employed. The experimentally obtained results are presented in graphs presenting load-deflection or load vs. angle of rotation relations and in photos presenting impact damages areas before and after bending test. The results show that the impact pre-damages have no significant influence on the work of channel section beams.


Introduction
Modern composites are light and durable, their properties can be adjusted by fitting them to the shape and manner of loading the designed structure. The disadvantage of laminates based on fiber-reinforced resins is their manufacturing cost, rather fragile behavior, the inability to work at high temperatures, unknown behavior after minor operational damage. Despite that, a different type of laminates (i.e., FRP, FML) manufactured in a different way (i.e., autoclaving or pultruded method) and different types of sandwich or FGM structures are widely used. They are more and more popular in many branches of industry, i.e., aircraft industry, aerospace engineering, automotive industry or civil engineering and many others.
Nowadays, in worldwide literature it is quite easy to find papers dealing with thin-walled laminates-structures made of composites. Taking into account such structures, the following two aspects should be mentioned: (i) buckling, postbuckling, and failure of thin-walled structures, (ii) not obvious work and failure of composites. As it is well known, buckling and postbuckling behavior of thin-walled structures are very important, even more than their strength. The same range of importance for thin-walled composite (especially laminate) structures seems to be initial damages i.e., operating damages (barely visible impact damage).
The number of papers devoted to buckling and postbuckling behavior of thin-walled structures made of composite materials in the latest years has grown rapidly, but in author's opinion they still need further investigations. It is well-known that papers dealing with above-mentioned issue started to appear in the 1990s. Modern composite beams are made using an autoclaving or pultrusion method. The pultruded beams are generally used in civil engineering. For example, a paper written by more durable and safe structures. The above will be possible if more results for structures with operating damages (impact damages) and their influence on postbuckling and failure are presented.

Object of Investigations
The analysis of the influence of impact pre-damage on buckling, postbuckling, and failure of thin-walled beam was carried out. As an object of analysis, the thin-walled channel section beams ( Figure 1) with cross-section dimension of a × b × t = 84 mm × 40 mm × 2.1 mm (a-a web width, ba flange width and t-a wall thickness) were considered. The beams for tests were produced using an autoclaving technique to ensure the best quality so that possible production imperfections (inclusions, voids) did not affect the test result. The unidirectional glass fiber reinforced polymer (GFRP) pre-impregnate produced by GURIT have been chosen. The eight layers laminate with following lay-ups were produced and then tested: properties (E1, E2-Young moduli in fiber direction and transverse to fiber direction, respectively, G12-Kirchhoff modulus, 12-Poisson ratio, ultimate stresses: in fiber direction in case of tension X T and compression X C , in transverse to fiber direction in case of tension Y T and compression Y C , Sultimate shear stress) of unidirectional plies are presented in Table 1 [30].   [30]. Pre-damaged channel section beams were taken for experimental tests. The low-velocity impact damage (pre-damage) was introduced in mid-span of tested beam and two different places, i.e., midwidth of the web and mid-width of the flange. Thus, the determined series of data were denoted in the following form Cbx-W or Cbx-F where x refers to the analyzed arrangement of layers and F or W concerns the considered element (F-flange, W-web) of the channel section profile.
The impact damages were introduced with the energy of impact equal to 20 J using an impactor with the hemispherical shape of the tip. A detailed description of the stand to introduce the initial damage can be found in the authors' previous papers [24,25,28]-only for clarification, the schemes of the test stands for impact damage introduced in web and flange are presented in Figure 2.  Pre-damaged channel section beams were taken for experimental tests. The low-velocity impact damage (pre-damage) was introduced in mid-span of tested beam and two different places, i.e., mid-width of the web and mid-width of the flange. Thus, the determined series of data were denoted in the following form Cbx-W or Cbx-F where x refers to the analyzed arrangement of layers and F or W concerns the considered element (F-flange, W-web) of the channel section profile.
The impact damages were introduced with the energy of impact equal to 20 J using an impactor with the hemispherical shape of the tip. A detailed description of the stand to introduce the initial damage can be found in the authors' previous papers [24,25,28]-only for clarification, the schemes of the test stands for impact damage introduced in web and flange are presented in Figure 2.

Experimental Test Stand
The initially damaged beams were subjected to pure bending in the plane of the flange (plane bending due to the symmetry of cross-section). The four-point bending test was performed employing the test stand [30]. The dimensions describing the span of considered beam and the span of the load are presented in the scheme of the test stand ( Figure 3).  The bending after impact tests of the considered profile were conducted on a universal testing machine (UTM) produced by Instron (Norwood, MA, USA) and modernized by Zwick-Roel (Ulm, Germany). The tests were performed at a constant velocity of the cross-bar equal to 1.5 mm/min. The loading force and the displacement in the points where the load was applied were obtained directly from the machine sensors. To determine the deflection of the beam in the whole range of load, a digital image correlation system Aramis produced by GOM (Braunschweig, Germany) was employed. The data from the DIC were captured with a frequency of 1 Hz together with load transferred directly from the testing machine. The whole test stand is presented in Figure 4.

Experimental Test Stand
The initially damaged beams were subjected to pure bending in the plane of the flange (plane bending due to the symmetry of cross-section). The four-point bending test was performed employing the test stand [30]. The dimensions describing the span of considered beam and the span of the load are presented in the scheme of the test stand ( Figure 3).

Experimental Test Stand
The initially damaged beams were subjected to pure bending in the plane of the flange (plane bending due to the symmetry of cross-section). The four-point bending test was performed employing the test stand [30]. The dimensions describing the span of considered beam and the span of the load are presented in the scheme of the test stand ( Figure 3).  The bending after impact tests of the considered profile were conducted on a universal testing machine (UTM) produced by Instron (Norwood, MA, USA) and modernized by Zwick-Roel (Ulm, Germany). The tests were performed at a constant velocity of the cross-bar equal to 1.5 mm/min. The loading force and the displacement in the points where the load was applied were obtained directly from the machine sensors. To determine the deflection of the beam in the whole range of load, a digital image correlation system Aramis produced by GOM (Braunschweig, Germany) was employed. The data from the DIC were captured with a frequency of 1 Hz together with load transferred directly from the testing machine. The whole test stand is presented in Figure 4.  The bending after impact tests of the considered profile were conducted on a universal testing machine (UTM) produced by Instron (Norwood, MA, USA) and modernized by Zwick-Roel (Ulm, Germany). The tests were performed at a constant velocity of the cross-bar equal to 1.5 mm/min. The loading force and the displacement in the points where the load was applied were obtained directly from the machine sensors. To determine the deflection of the beam in the whole range of load, a digital image correlation system Aramis produced by GOM (Braunschweig, Germany) was employed. The data from the DIC were captured with a frequency of 1 Hz together with load transferred directly from the testing machine. The whole test stand is presented in Figure 4.

Experimental Test Stand
The initially damaged beams were subjected to pure bending in the plane of the flange (plane bending due to the symmetry of cross-section). The four-point bending test was performed employing the test stand [30]. The dimensions describing the span of considered beam and the span of the load are presented in the scheme of the test stand ( Figure 3).  The bending after impact tests of the considered profile were conducted on a universal testing machine (UTM) produced by Instron (Norwood, MA, USA) and modernized by Zwick-Roel (Ulm, Germany). The tests were performed at a constant velocity of the cross-bar equal to 1.5 mm/min. The loading force and the displacement in the points where the load was applied were obtained directly from the machine sensors. To determine the deflection of the beam in the whole range of load, a digital image correlation system Aramis produced by GOM (Braunschweig, Germany) was employed. The data from the DIC were captured with a frequency of 1 Hz together with load transferred directly from the testing machine. The whole test stand is presented in Figure 4.   The collected data allow to determine bending moment vs. angle of rotation or bending moment vs. deflection plots. The obtained results allow to determine buckling loads and modes, analyze postbuckling behaviors and determine the failure loads with corresponding failure mechanisms.

Results of Experimental Tests
The results of the experimental tests are presented and discussed in this section. They are presented in the form of (i) pictures showing impact damage before and after bending test, buckling mode and failure mechanism, (ii) graphs presenting the relation between load and deflection or load vs. angle of rotations.
Impact pre-damages for all considered cases of layups and place of impact are presented in Table 2. In most cases that were analyzed, the size of damage, as well as the character of damage, do not change after bending tests, which have been performed till failure of the beam.  It also should be mentioned that damage due to the bending load have not been started at place of the initial pre-damages Even in the cases denoted as Cb3-F and Cb1-W (see Table 2) the impact damage is very close to damages due to the bending load, and the impact pre-damage has not initiated the failure in the bending test. It should be noted that in case Cb2-F (cf. Figure 5) the failure due to bending goes through the initial impact pre-damage.
Concluding all test observations of which results are presented in Table 2 and Figure 5, it can be  It also should be mentioned that damage due to the bending load have not been started at place of the initial pre-damages Even in the cases denoted as Cb3-F and Cb1-W (see Table 2) the impact damage is very close to damages due to the bending load, and the impact pre-damage has not initiated the failure in the bending test. It should be noted that in case Cb2-F (cf. Figure 5) the failure due to bending goes through the initial impact pre-damage.
Concluding all test observations of which results are presented in Table 2 and Figure 5, it can be  It also should be mentioned that damage due to the bending load have not been started at place of the initial pre-damages Even in the cases denoted as Cb3-F and Cb1-W (see Table 2) the impact damage is very close to damages due to the bending load, and the impact pre-damage has not initiated the failure in the bending test. It should be noted that in case Cb2-F (cf. Figure 5) the failure due to bending goes through the initial impact pre-damage.
Concluding all test observations of which results are presented in Table 2 and Figure 5, it can be  It also should be mentioned that damage due to the bending load have not been started at place of the initial pre-damages Even in the cases denoted as Cb3-F and Cb1-W (see Table 2) the impact damage is very close to damages due to the bending load, and the impact pre-damage has not initiated the failure in the bending test. It should be noted that in case Cb2-F (cf. Figure 5) the failure due to bending goes through the initial impact pre-damage.
Concluding all test observations of which results are presented in Table 2 and Figure 5, it can be  It also should be mentioned that damage due to the bending load have not been started at place of the initial pre-damages Even in the cases denoted as Cb3-F and Cb1-W (see Table 2) the impact damage is very close to damages due to the bending load, and the impact pre-damage has not initiated the failure in the bending test. It should be noted that in case Cb2-F (cf. Figure 5) the failure due to bending goes through the initial impact pre-damage.
Concluding all test observations of which results are presented in Table 2 and Figure 5, it can be  It also should be mentioned that damage due to the bending load have not been started at place of the initial pre-damages Even in the cases denoted as Cb3-F and Cb1-W (see Table 2) the impact damage is very close to damages due to the bending load, and the impact pre-damage has not initiated the failure in the bending test. It should be noted that in case Cb2-F (cf. Figure 5) the failure due to bending goes through the initial impact pre-damage.
Concluding all test observations of which results are presented in Table 2 and Figure 5, it can be  It also should be mentioned that damage due to the bending load have not been started at place of the initial pre-damages Even in the cases denoted as Cb3-F and Cb1-W (see Table 2) the impact damage is very close to damages due to the bending load, and the impact pre-damage has not initiated the failure in the bending test. It should be noted that in case Cb2-F (cf. Figure 5) the failure due to bending goes through the initial impact pre-damage.
Concluding all test observations of which results are presented in Table 2 and Figure 5, it can be  It also should be mentioned that damage due to the bending load have not been started at place of the initial pre-damages Even in the cases denoted as Cb3-F and Cb1-W (see Table 2) the impact damage is very close to damages due to the bending load, and the impact pre-damage has not initiated the failure in the bending test. It should be noted that in case Cb2-F (cf. Figure 5) the failure due to bending goes through the initial impact pre-damage.
Concluding all test observations of which results are presented in Table 2 and Figure 5, it can be  It also should be mentioned that damage due to the bending load have not been started at place of the initial pre-damages Even in the cases denoted as Cb3-F and Cb1-W (see Table 2) the impact damage is very close to damages due to the bending load, and the impact pre-damage has not initiated the failure in the bending test. It should be noted that in case Cb2-F (cf. Figure 5) the failure due to bending goes through the initial impact pre-damage.
Concluding all test observations of which results are presented in Table 2 and Figure 5, it can be  It also should be mentioned that damage due to the bending load have not been started at place of the initial pre-damages Even in the cases denoted as Cb3-F and Cb1-W (see Table 2) the impact damage is very close to damages due to the bending load, and the impact pre-damage has not initiated the failure in the bending test. It should be noted that in case Cb2-F (cf. Figure 5) the failure due to bending goes through the initial impact pre-damage.
Concluding all test observations of which results are presented in Table 2 and Figure 5, it can be  It also should be mentioned that damage due to the bending load have not been started at place of the initial pre-damages Even in the cases denoted as Cb3-F and Cb1-W (see Table 2) the impact damage is very close to damages due to the bending load, and the impact pre-damage has not initiated the failure in the bending test. It should be noted that in case Cb2-F (cf. Figure 5) the failure due to bending goes through the initial impact pre-damage.
Concluding all test observations of which results are presented in Table 2 and Figure 5, it can be  It also should be mentioned that damage due to the bending load have not been started at place of the initial pre-damages Even in the cases denoted as Cb3-F and Cb1-W (see Table 2) the impact damage is very close to damages due to the bending load, and the impact pre-damage has not initiated the failure in the bending test. It should be noted that in case Cb2-F (cf. Figure 5) the failure due to bending goes through the initial impact pre-damage.
Concluding all test observations of which results are presented in Table 2 and Figure 5, it can be  It also should be mentioned that damage due to the bending load have not been started at place of the initial pre-damages Even in the cases denoted as Cb3-F and Cb1-W (see Table 2) the impact damage is very close to damages due to the bending load, and the impact pre-damage has not initiated the failure in the bending test. It should be noted that in case Cb2-F (cf. Figure 5) the failure due to bending goes through the initial impact pre-damage.
Concluding all test observations of which results are presented in Table 2 and Figure 5, it can be  It also should be mentioned that damage due to the bending load have not been started at place of the initial pre-damages Even in the cases denoted as Cb3-F and Cb1-W (see Table 2) the impact damage is very close to damages due to the bending load, and the impact pre-damage has not initiated the failure in the bending test. It should be noted that in case Cb2-F (cf. Figure 5) the failure due to bending goes through the initial impact pre-damage.
Concluding all test observations of which results are presented in Table 2 and Figure 5, it can be  It also should be mentioned that damage due to the bending load have not been started at place of the initial pre-damages Even in the cases denoted as Cb3-F and Cb1-W (see Table 2) the impact damage is very close to damages due to the bending load, and the impact pre-damage has not initiated the failure in the bending test. It should be noted that in case Cb2-F (cf. Figure 5) the failure due to bending goes through the initial impact pre-damage.
Concluding all test observations of which results are presented in Table 2 and Figure 5, it can be  It also should be mentioned that damage due to the bending load have not been started at place of the initial pre-damages Even in the cases denoted as Cb3-F and Cb1-W (see Table 2) the impact damage is very close to damages due to the bending load, and the impact pre-damage has not initiated the failure in the bending test. It should be noted that in case Cb2-F (cf. Figure 5) the failure due to bending goes through the initial impact pre-damage.
Concluding all test observations of which results are presented in Table 2 and Figure 5, it can be  It also should be mentioned that damage due to the bending load have not been started at place of the initial pre-damages Even in the cases denoted as Cb3-F and Cb1-W (see Table 2) the impact damage is very close to damages due to the bending load, and the impact pre-damage has not initiated the failure in the bending test. It should be noted that in case Cb2-F (cf. Figure 5) the failure due to bending goes through the initial impact pre-damage.
Concluding all test observations of which results are presented in Table 2 and Figure 5, it can be  It also should be mentioned that damage due to the bending load have not been started at place of the initial pre-damages Even in the cases denoted as Cb3-F and Cb1-W (see Table 2) the impact damage is very close to damages due to the bending load, and the impact pre-damage has not initiated the failure in the bending test. It should be noted that in case Cb2-F (cf. Figure 5) the failure due to bending goes through the initial impact pre-damage.
Concluding all test observations of which results are presented in Table 2 and Figure 5, it can be  It also should be mentioned that damage due to the bending load have not been started at place of the initial pre-damages Even in the cases denoted as Cb3-F and Cb1-W (see Table 2) the impact damage is very close to damages due to the bending load, and the impact pre-damage has not initiated the failure in the bending test. It should be noted that in case Cb2-F (cf. Figure 5) the failure due to bending goes through the initial impact pre-damage.
Concluding all test observations of which results are presented in Table 2 and Figure 5, it can be  It also should be mentioned that damage due to the bending load have not been started at place of the initial pre-damages Even in the cases denoted as Cb3-F and Cb1-W (see Table 2) the impact damage is very close to damages due to the bending load, and the impact pre-damage has not initiated the failure in the bending test. It should be noted that in case Cb2-F (cf. Figure 5) the failure due to bending goes through the initial impact pre-damage.
Concluding all test observations of which results are presented in Table 2 and Figure 5, it can be It also should be mentioned that damage due to the bending load have not been started at place of the initial pre-damages Even in the cases denoted as Cb3-F and Cb1-W (see Table 2) the impact damage is very close to damages due to the bending load, and the impact pre-damage has not initiated the failure in the bending test. It should be noted that in case Cb2-F (cf. Figure 5) the failure due to bending goes through the initial impact pre-damage. It also should be mentioned that damage due to the bending load have not been started at place of the initial pre-damages Even in the cases denoted as Cb3-F and Cb1-W (see Table 2) the impact damage is very close to damages due to the bending load, and the impact pre-damage has not initiated the failure in the bending test. It should be noted that in case Cb2-F (cf. Figure 5) the failure due to bending goes through the initial impact pre-damage.
Concluding all test observations of which results are presented in Table 2 and Figure 5, it can be said that failure due to bending starts from matrix and fiber cracking on the flange-web edge (c.f. Figure 5c) and then propagates along the flange to the free edge and along the web. The direction of damage propagation corresponds to layer arrangements. For cases denoted as Cb5 where the fibers in relation to the axis of the beam longitudinal edges are located at an angle of ±45, the plastic behavior is observed-the beams did not break and after unloading permanent deflection was visible (see Figure 6). The beam's behavior under pure bending till failure can be analyzed on the basis of the bending moment vs. angle of rotations at support curves ( Figure 7) as well as the bending moment-relative deflection curves (Figure 8). The relative deflection is calculated in chosen cross-section as a difference between mid-with of the web deflection and web-flange edge deflection. Observing the course of the curves in Figures 7 and 8, it can be said that the place of initial pre-damage has no significant influence on beam behavior, which is also proved by the further presented results.
Analyzing the course of load vs. angle of rotation curves (Figure 7) it is visible that beams with the most cases of layer arrangement have the brittle behavior-almost straight line till failure load. Only in the case denoted as Cb5 the plastic behavior is observed-the bending moment vs. angle of rotation at support have a nonlinear relation, which was proved by permanent deflection after unloading, visible in Figure 6. Such behavior can also be observed on load-relative deflection curve (Figure 8e), where for high bending moment (>450 Nm) the relative deflections are constant, which means that there is no relation between the deflection of mid-width of the web and web-flange edge in the chosen cross-section. Concluding all test observations of which results are presented in Table 2 and Figure 5, it can be said that failure due to bending starts from matrix and fiber cracking on the flange-web edge (c.f. Figure 5c) and then propagates along the flange to the free edge and along the web. The direction of damage propagation corresponds to layer arrangements.
For cases denoted as Cb5 where the fibers in relation to the axis of the beam longitudinal edges are located at an angle of ±45, the plastic behavior is observed-the beams did not break and after unloading permanent deflection was visible (see Figure 6). It also should be mentioned that damage due to the bending load have not been started at place of the initial pre-damages Even in the cases denoted as Cb3-F and Cb1-W (see Table 2) the impact damage is very close to damages due to the bending load, and the impact pre-damage has not initiated the failure in the bending test. It should be noted that in case Cb2-F (cf. Figure 5) the failure due to bending goes through the initial impact pre-damage.
Concluding all test observations of which results are presented in Table 2 and Figure 5, it can be said that failure due to bending starts from matrix and fiber cracking on the flange-web edge (c.f. Figure 5c) and then propagates along the flange to the free edge and along the web. The direction of damage propagation corresponds to layer arrangements. For cases denoted as Cb5 where the fibers in relation to the axis of the beam longitudinal edges are located at an angle of ±45, the plastic behavior is observed-the beams did not break and after unloading permanent deflection was visible (see Figure 6). The beam's behavior under pure bending till failure can be analyzed on the basis of the bending moment vs. angle of rotations at support curves (Figure 7) as well as the bending moment-relative deflection curves (Figure 8). The relative deflection is calculated in chosen cross-section as a difference between mid-with of the web deflection and web-flange edge deflection. Observing the course of the curves in Figures 7 and 8, it can be said that the place of initial pre-damage has no significant influence on beam behavior, which is also proved by the further presented results.
Analyzing the course of load vs. angle of rotation curves (Figure 7) it is visible that beams with the most cases of layer arrangement have the brittle behavior-almost straight line till failure load. Only in the case denoted as Cb5 the plastic behavior is observed-the bending moment vs. angle of rotation at support have a nonlinear relation, which was proved by permanent deflection after unloading, visible in Figure 6. Such behavior can also be observed on load-relative deflection curve (Figure 8e), where for high bending moment (>450 Nm) the relative deflections are constant, which means that there is no relation between the deflection of mid-width of the web and web-flange edge in the chosen cross-section. The beam's behavior under pure bending till failure can be analyzed on the basis of the bending moment vs. angle of rotations at support curves (Figure 7) as well as the bending moment-relative deflection curves (Figure 8). The relative deflection is calculated in chosen cross-section as a difference between mid-with of the web deflection and web-flange edge deflection. Observing the course of the curves in Figures 7 and 8, it can be said that the place of initial pre-damage has no significant influence on beam behavior, which is also proved by the further presented results.  The results of the experimental test also allow to determine the buckling load (Table 3) with a corresponding buckling mode (Table 4).
It is well known that especially in the case of experimental tests, the structure subjected to bending the bifurcation load does not exist. A different approximation method like for example "top of the knee", P-w 2 Table 4. Taking into account the similarity of load-deflection and load-angle of rotations curves, as well as the buckling modes for Analyzing the course of load vs. angle of rotation curves (Figure 7) it is visible that beams with the most cases of layer arrangement have the brittle behavior-almost straight line till failure load. Only in the case denoted as Cb5 the plastic behavior is observed-the bending moment vs. angle of rotation at support have a nonlinear relation, which was proved by permanent deflection after unloading, visible in Figure 6. Such behavior can also be observed on load-relative deflection curve (Figure 8e), where for high bending moment (>450 Nm) the relative deflections are constant, which The results of the experimental test also allow to determine the buckling load (Table 3) with a corresponding buckling mode (Table 4).  Table 4. Deflection corresponding to buckling mode.

Cb5
As it can be observed on pictures presented in Table 4, the buckling modes differ only due to layer arrangement, the places of impact pre-damage have no significant influence on the buckling  Table 4. Deflection corresponding to buckling mode.

Cb5
As it can be observed on pictures presented in Table 4, the buckling modes differ only due to layer arrangement, the places of impact pre-damage have no significant influence on the buckling  Table 4. Deflection corresponding to buckling mode.

Cb5
As it can be observed on pictures presented in Table 4, the buckling modes differ only due to layer arrangement, the places of impact pre-damage have no significant influence on the buckling mode as well as the buckling load. Only in the case Cb4, the change of buckling mode depends on  Table 4. Deflection corresponding to buckling mode.

Cb5
As it can be observed on pictures presented in Table 4, the buckling modes differ only due to layer arrangement, the places of impact pre-damage have no significant influence on the buckling mode as well as the buckling load. Only in the case Cb4, the change of buckling mode depends on  Table 4. Deflection corresponding to buckling mode.

Cb5
As it can be observed on pictures presented in Table 4, the buckling modes differ only due to layer arrangement, the places of impact pre-damage have no significant influence on the buckling mode as well as the buckling load. Only in the case Cb4, the change of buckling mode depends on  Table 4. Deflection corresponding to buckling mode.

Cb5
As it can be observed on pictures presented in Table 4, the buckling modes differ only due to layer arrangement, the places of impact pre-damage have no significant influence on the buckling mode as well as the buckling load. Only in the case Cb4, the change of buckling mode depends on  Table 4. Deflection corresponding to buckling mode.

Cb5
As it can be observed on pictures presented in Table 4, the buckling modes differ only due to layer arrangement, the places of impact pre-damage have no significant influence on the buckling mode as well as the buckling load. Only in the case Cb4, the change of buckling mode depends on the place of impact pre-damage-the impact introduced on the web causes the creation of the initial  Table 4. Deflection corresponding to buckling mode.

Cb5
As it can be observed on pictures presented in Table 4, the buckling modes differ only due to layer arrangement, the places of impact pre-damage have no significant influence on the buckling mode as well as the buckling load. Only in the case Cb4, the change of buckling mode depends on the place of impact pre-damage-the impact introduced on the web causes the creation of the initial  Table 4. Deflection corresponding to buckling mode.

Cb5
As it can be observed on pictures presented in Table 4, the buckling modes differ only due to layer arrangement, the places of impact pre-damage have no significant influence on the buckling mode as well as the buckling load. Only in the case Cb4, the change of buckling mode depends on the place of impact pre-damage-the impact introduced on the web causes the creation of the initial  Table 4. Deflection corresponding to buckling mode.

Cb5
As it can be observed on pictures presented in Table 4, the buckling modes differ only due to layer arrangement, the places of impact pre-damage have no significant influence on the buckling mode as well as the buckling load. Only in the case Cb4, the change of buckling mode depends on the place of impact pre-damage-the impact introduced on the web causes the creation of the initial It is well known that especially in the case of experimental tests, the structure subjected to bending the bifurcation load does not exist. A different approximation method like for example "top of the knee", P-w 2 [31] or proposed by Teter and Kolakowski [32] allows to estimate the buckling load. The estimated buckling loads based on load vs. relative deflection curves (Figure 8)-M b1 after employing "top of the knee" method [31] and on the basis of load vs. angle of rotation curves (Figure 7)-M b2 employing Teter-Kolakowski method [32] are presented in Table 4. Taking into account the similarity of load-deflection and load-angle of rotations curves, as well as the buckling modes for given layer arrangement and both places of pre-damages, it can be noted that better results of the buckling load have been obtained based on load-angle of rotation curves [32]-the results are closer to each other.
As it can be observed on pictures presented in Table 4, the buckling modes differ only due to layer arrangement, the places of impact pre-damage have no significant influence on the buckling mode as well as the buckling load. Only in the case Cb4, the change of buckling mode depends on the place of impact pre-damage-the impact introduced on the web causes the creation of the initial deflection, which translates into the deflection corresponding to the buckling half-wave.
The failure loads for the cases from Cb1 to Cb4 and maximum load collected during the test for case Cb5 are presented in Table 5. Due to no damage and plastic behavior of beam with layer arrangement denoted as Cb5, only the maximal load with the corresponding angle of rotation at support are presented-those results cannot be compared with others. It only can be said that such layer arrangement causes the beam to be very flexible-the beam subjected to similar or lower bending moment which in other cases leads to damage in this one leads to rotate at the support by more than twice the angle measured in other cases. Table 5. Failure moments with corresponding angle of rotation.

Failure Moment M bf (Nm)/Corresponding Angle of Rotation (deg) Lay-Up Web Pre-Damaged Flange Pre-Damaged
Comparing the failure bending moments for beams with pre-damaged webs and flange with no influence of place of pre-damage on the failure due to the bending moment, the results listed in Table 5 appear surprising. In relation to the previously described results, they should have similar values, however, beams with pre-damaged flanges carry a larger bending moment than those with a pre-damaged web. The differences are from approx. 3-5% for beams Cb1, Cb2 and Cb4 to c.a. 20% for case denoted as Cb2. After all, slightly higher load-bearing values for beams with pre-damaged flanges than beams with pre-damaged webs should be considered as an accident. Differences in load capacities of up to 5% can be considered with a high probability of spreading associated with the measurement accuracy and accuracy of beam manufacturing. The 20% difference for the Cb2 case is simply due to the place where the damage due to bending occurred-it is not related to the place of initial damage in any way. The Cb2-W beam was destroyed in the place where the beam was fixed in the grip, and the Cb2-F beam in the mid-span (see Table 2). Therefore, the lower failure bending moment of 732 Nm for beam with pre-damaged web in comparison to failure bending moment equal to 890 Nm for beam with pre-damaged web could be a coincidental result of the way of mounting the beam in the aluminum grip (Figure 3b). The above results show no differences in load-deflection and load-angle of beam rotation curves as well as no correlation between pre-damages and failure due to bending.
It is worth to mention that observations based on obtained results are similar to the other authors' investigations performed on columns subjected to uniform compression [28,29].

Conclusions
In this paper, buckling, postbuckling and failure of pre-damaged channel section beam subjected to pure bending have been investigated. The channel section beams made of 8-layered GFRP laminate with two different symmetrical layups was used. The specimens with impact pre-damages on web or flange have been investigated to access the influence of initial impact damage on work of a thin-walled structure in the full range of load till failure.
Analyzing the results obtained from experimental research, it can be concluded that introduced impact damages do not have an influence on the buckling load as well as the failure load of channel section beams under consideration. Analysis of the influence of layer arrangement on sensitivity to size and shape of impact damages, and then on the operation of a pre-damaged beam subjected to pure bending also show that such relations have not been observed. However, in case of beams with layups where the outer layers have fiber inclined to the edges at an angle of 45 • (Cb-1 and Cb-5) the size of delamination after impact in flange is higher than in other cases. Despite the above, for these cases, a decrease of buckling or failure loads has not been observed.
It is worth to mention that the introduced damages were small in comparison to channel web or flange dimensions (i.e., their length and width), what could be the main reason of the observed behavior (no influence of initial impact damage). The reason for this can be probably not only the size of pre-damages but also the places of damage introduction are not sensitive enough to affect the beam behavior.
The obtained results for the considered cases suggest performing further investigations, in which the stiffest part of a beam (i.e., web-flange junction) will be initially damaged-then the influence of pre-damage on beam behavior could be visible.