To be sure of the presence of cracks or not , try to apply a liquid penetrant test (a NDT method) to examine if the wind turbine blade  has cracks or not.   The test is cheap & easy to use.   A liquid penetrant is normally applied on  a machine component for visual examinations of the component.   Then a fluorescent light is used to illuminite the component to detect any visual cracks or defects.   

can u share the graph please.. 

The change of the natural frequency is caused by the growth of cracks in your blade, which changes the blade's stiffness.

How the stiffness will change will presumably depend not only on the size of the cracks, but also their location (both within the laminate and within the structure). To make your analysis more effective you would have to relate the stiffness (and thus, by extension, the natural frequency) to both the size and the location of the cracks.

Muhammad Amjad, the plot is attached. 

Another figure for testing the lade in another direction 

Experimentally or using software you can analyze frequency of blade having no crack and blade having crack at known location. There is going to change in natural frequency of blade with crack propagation and will be function of location in blade from support end to open end being function of stiffness and mass.

A good way of analyzing such is to intentionally introduce crack at a known location and then vary the crack location on subsequent samples. If the natural frequency changes with the variation in crack location you can be assure of the crack effect.

Thereafter you can test the turbine blade (with and without crack) and possible use your model to accertain the crack location if there is, other NDT techniques can possible be use to confirm your findings wrt to the crack location. 

I have worked on the effect of laser welding parameters on the flexural characteristics of plates,  natural frequency of the plates was our indicator of the effect.  

You also need to pay adequate attention to the sensitivity your data acquisition (DAQ) system, it plays an important role in the accuracy of your results 

Since the difference are relatively small and the tendency is not evident (e.g. initially decreased and then increased stiffness after fatigue), I would be worried that this has more to do with scatter and uncontrolled test parameters than real deterministic behaviour due to fatigue damage. Also, were these values the results of a single test at each number of cycles or the average of multiple tests? I think that without information on the inherent variability of the natural frequency measurement, very little can be said.

You also need to be aware that with polymer composites, temperatuer may have a strong influence on the stiffness and damping properties of the structure, so you need to make sure that the test environment is quite steady between trials. You would have to at least report test temperature with your data and also ideally report some sensitivity study on possible parasitic effects such as those of temperature,

Also beware that polymer (such as the matrix, but also adhesives) age and that their stiffness changes with aging. However, the load history and temperature/moisture history also influence the aging characteristics. Therefore, simply performing the same test on the same undamaged part after aging may change the natural frequency even in the abscence of damage.

I think the idea of studying known damage states (that would be reresentative of real damage accumulation) to start with would definitely be advisable.

Best regards,

Laurent