FATIGUE LIFE CHARACTERISATION 1 Data analysis

Load and displacement data was acquired every 100th cycle and stored for analysis. This data was used to investigate any stiffness degradation over the specimen life. Fitting a straight line to the hysteresis loop of the load-deflection curve and obtaining the gradient provided the stiffness of the joint. The results are plotted against number of cycles. In addition, the area inside the hysteresis loop is calculated. A change in this area with number of cycles indicates a release of energy due to damage within the structure [77] assuming no rig damage. More detail with respect to the data analysis can be found in Chapter 4.

6.2.2.2 Compression-compression (17-34 percent UCSnom, specimen 12a)

Specimen 12a was tested in a compression-compression cycle between approximately 17 and 34 percent of its UCSnom, corresponding to -17.2 kN to -34.4 kN. The test was carried

out at a frequency of 1 Hz to ensure that the correct load levels were being reached and to limit the effect of thermal stress caused by rapid deformation. After 106 cycles the test was stopped. Figure 6.5 shows that there is very little variation in the stiffness of the specimen or energy released by the system. In addition, after 106 cycles there was no visual indication of damage. Based on these results it could be assumed that there is no damage within the structure and so its residual strength should be similar to the static compressive strength.

6.2.2.3 Compression-compression (34-69 percent UCSnom, specimen 12b)

In an attempt to induce failure of the hybrid joint before 106 _{cycles the mean load was}

increased. Specimen 12b was tested in a compression-compression cycle at a mean load level of approximately 52 percent of UCSnom with a load amplitude of approximately 17

percent at a fatigue rate of 2Hz. The frequency was increased from 1Hz to 2Hz as the actuator had no difficulty in reaching the prescribed loads in time and no change in temperature was observed in the specimen during cycling. This specimen showed no visible signs of damage after nearly 1.3x106 cycles. The acquired data from the test was analysed for specimen stiffness degradation and energy release. Figure 6.6 shows the results of the analysis. The two peaks in the stiffness plot are caused by tightening the rig

during the test when the rig bolts began to work themselves loose. The stiffness plot indicates that there is very little if any loss in stiffness of the specimen. Based on the results obtained from the first two fatigue tests (specimens 12a and 12b) it could be assumed that if there is no loss in stiffness or increase in energy release, both indicative of damage, then it can be assumed that there will be little drop in residual strength.

6.2.2.4 Zero-compression (0-86 percent UCSnom, specimen 13a)

The fatigue tests carried out on Specimen 12b (mean load equal to 52 percent of UCSnom

and a load amplitude equal to 17 percent of UCSnom) did not produce failure therefore it was

decided to test a specimen with the load amplitude increased to 43 percent of UCSnom and

the mean load reduced slightly. The results were again analysed for loss in stiffness and plotted in Figure 6.7. It is apparent from this plot that there is a loss in stiffness and a corresponding increase in energy release. Both of these results indicate damage within the joint, which is verified by failure occurring after 1700 cycles.

6.2.2.5 Zero-compression (0-69 percent UCSnom, specimen 14a)

Specimen 14a was tested at a load range of between 0 kN and –68.8 kN (or 0-69 percent UCSnom). Complete failure occurred after 80264 cycles. The results were again analysed

for loss in stiffness and energy dissipation in the system. It was observed that at approximately 50,000 cycles that a small area of debonding occurred at the interface of the steel and GRP on the tapered side of the specimen. The crack grew progressively with increasing numbers of cycles. Figure 6.8 indicates a loss in stiffness just below 50,000 cycles corresponding to the appearance of the debond. Unusually the stiffness of the specimen appears to increase between the point of initial failure and the next observed debond growth at 63,000 cycles. However from this point to failure there is a steady loss in stiffness as the crack grew. Similarly, there is a corresponding increase in energy dissipation as energy is used to propagate the crack towards failure.

6.2.2.6 Zero-compression (0-60 percent UCSnom, specimen 14b, 15a)

Two specimens were tested at cycling load levels from 0 kN to -60.2 kN. Specimen 14b failed after 187153 cycles but on removal of the specimen from the test rig, one of the four mounting bolts on the composite end of the specimen was found to have failed. This failure

premature failure of the hybrid joint due to increased peel stresses at the interface between the steel and GRP on the flat side. A second test was therefore carried out on specimen 15a, the results of which are shown in Figure 6.9. Specimen 15a failed after 561700 cycles.

6.2.2.7 Zero-compression (0-51 percent UCSnom, specimen 13b)

Specimen 13b was tested at a load level cycling from 0 kN to -51.6 kN. On visual inspection, after 3548500 cycles the specimen appeared to be virtually undamaged, with perhaps the slightest of matrix cracking in the overlaminate radius on the tapered side and slight whitening apparent on the flat side of the specimen between the GRP skin and the steel (indicative of debonding). The stiffness degradation and energy release plots are given in Figure 6.10 and reflect the fact that whilst the specimen appears to be relatively undamaged, there are progressive failure mechanisms at work. Despite this it seems that with these prescribed boundary conditions and load levels, the specimen is operating at, or close to, the endurance limit.

6.2.2.8 Zero-compression (0-77 percent UCSnom, specimen 15b, 16a)

Two specimens were tested at cycling load levels from 0 kN to –77.4 kN. Specimen 15b failed after 30880 cycles and specimen 16a after 11018 cycles, the results of which are shown in Figures 6.11 and 6.12 respectively. The failure of both specimens was a result of complete debonding of the GRP from the steel on the flat side of the specimen.

6.2.2.9 Fatigue life curve

The tests performed to produce the fatigue life curve are shown in Table 6.1. This test matrix became evolutionary in nature as it became apparent that some of the tests were not suitable to this joint configuration. For example, the performance of the joint at low load amplitudes (below 25 percent UCSnom) and relatively high mean load has been shown to be

operating under the endurance limit and hence low load amplitude tests were replaced. Figure 6.13 shows the fatigue life curve of the hybrid joint in terms of load amplitude (Pa)

Table 6.1 Fatigue life test matrix

Test Result Comment

17-34% UCSnom N/a No Failure, increase mean load

34-69% UCSnom N/a No Failure, increase load amp

0-86% UCSnom 1700 Failure

0-77% UCSnom 20949 Failure

0-69% UCSnom 101432 Failure

0-60% UCSnom 374427 Failure

0-51% UCSnom 3548500 No Failure, approaching endurance limit

A representation of the whole life of the joint may be described by a number of discrete functions. Figure 6.13 shows a discrete logarithmic function representing the experimental data without consideration of the static test case. The level of correlation between the chosen logarithmic function and the experimental data is high, with a variance accounted for value, VAC or R2, equal to 0.95. A more appropriate and useful function to represent the data would be continuous in nature. A Weibull cumulative density function, W(N), commonly used to describe the fatigue behaviour of composite structures [85, 86] is therefore also presented in Figure 6.13. The Weibull function is defined from the UCS and is fitted using the shape, β, and scale, η, parameters as described in Equation 6.1. These parameters are selected so as to provide a minimum value of error through the least squares approach. The correlation between the experimental data and the function used to fit the data is very good with an R2 value also equal to 0.95, validating the use of the Weibull function to describe the as-determined fatigue life of the hybrid joint. It must be noted that the function is based upon a fatigue load endurance limit of 25 percent of UCSnom at this

load level the specimen exceeded 3.5 million cycles but did not fail. A refinement of the endurance limit may produce a slightly different curve.

⎟⎟ ⎟ ⎠ ⎞ ⎜⎜ ⎜ ⎝ ⎛ ⋅ − = ⎟⎟⎠ ⎞ ⎜⎜ ⎝ ⎛ − β η N e UCS UCS N W( ) (6.1)

The curve obtained from the present study is typical of those obtained from fatigue characterisation of composite materials [85]. The goodness of fit of the Weibull cumulative density function provides confidence in predicting the joint life and the corresponding probability density function can predict the probability of failure for any given load amplitude under the prescribed boundary conditions.