Unstressed Fresh Water Erosion
The first experiment that was conducted was fresh water with no applied stress. From Fig. 3 the conditions that yielded the most erosion was at the maximum speed of 50 ms−1 and at droplet angle of 45° with a mass loss of 0.11%. From the wear map it is very distinguishable where the transition between negligible erosion to relatively high erosion occurs. This has been observed within the literature . This is theorised to be the threshold number of droplets on the surface of the material sufficient to weaken it significantly to initiate erosion. This behaviour was observed with the majority of all the experiments tested. This was also observed in the SEM images of the samples as the erosion was concentrated in the areas where the fibres were closest to the surface and hence the top layer of epoxy was the thinnest (Figs. 7 and 8).
In the first experiment, i.e. fresh water with no applied stress, the most erosive impact angle range was approximately 45° to 60°. This range of angles is interesting as classically ductile materials typically exhibit maximum erosion at relatively low angles of impact (20–30°) whereas with classically brittle materials the erosion increases continually until 90° is reached . This behaviour has been modelled using particle impacts; however the same behaviour has been observed using liquid impacts during these experiments.
The wear map from this experiment, Fig. 3, can also be visualised as a line graph shown in Fig. 9. From this graph it can be concluded that the material is neither experiencing fully ductile or fully brittle properties as the impact angle with the most erosion is 45°. This can be explained as G10 epoxy glass is a composite material involving a brittle glass fibre with a ductile matrix binding the fibres together. The surface features evident on the eroded samples are more in common with ductile gauging or ploughing type features rather than the fracturing or localised shattering that is evident in brittle materials.  As a result, the behaviour could be classed as quasi-ductile. These results agree well with the work of Rafee Abdulmajeed Rafee Ahamed et al.  where tidal erosion was studied. It is interesting that although this work investigated erosion of G10 composite in slurries of sea water and sand (SiC), the impact angle exhibiting peak erosion is the same; thus, analogous behaviour was observed albeit in different conditions.
Unstressed Salt Water Erosion
The second experiment that was conducted was with 3.5% salt water solution to simulate the effects of sea water on wind turbine blades under the same parameters to which the fresh water was tested. This was to determine if offshore wind turbines are eroding differently to those on shore. The Figs. 10 and 11 depict the erosion occurring to an extent where the glass fibres on the surface are exposed.
The maximum mass loss during this experiment was experienced at 50 ms−1 at 60° with a total loss of 0.077%; this can be seen in the wear map, Fig. 5. The difference in mass loss between the fresh water and salt water mass loss is not overly substantial as there will be some salt crystals remaining in the sample. From the wear map of the salt water testing (Fig. 5) a line graph can be constructed (Fig. 12). These results show neither fully “ductile” nor “brittle” erosion behaviour  as described in the literature in a similar fashion to the result exposed in fresh water.
Stress Testing in Fresh Water and Salt Water Erosion
The next tests that were conducted were a repeat of the fresh water and salt water experiments but under a static 3-point bend. This was to replicate the stresses which large wind turbine blades experience while in operation.
The maximum mass loss was at 50 ms−1 at 30 and 45° with a 0.12% mass loss, this can be observed in the wear map in Fig. 5. This mass loss can be seen in the SEM images in Figs. 13 and 14 as the top layer of epoxy is eroded to the point where the fibres closest to the surface are exposed. The erosion is accelerated in this region due to the surface being slightly rougher at this point and the ploughing mechanism of the water droplet seems to have more effect on the surface by removing material. This is also observed in the optical microscope images of the unstressed sample (Figs. 15 and 16) as the point at which the fibres are closest to the surface show a greater dark area. This possibly implies that there is excessive pitting and erosion in the material.
From the wear maps in Figs. 4 and 6 a slightly unusual pattern occurred over the different impact angles. Instead of peaking at one impact angle this experiment showed two peaks which was not as expected.
The manner in which erosion occurs on the material from liquid impacts changes depending on the impact angle, in the same fashion as a particle impact. At lower impact angles the mechanism is more abrasion dominant resulting in shearing of the top surface; this can result in significant mass loss to the material, as once the top layer of the matrix is compromised, the fibres which are more brittle are exposed. This is seen in an SEM image Fig. 14. It is postulated that the high surface stresses influence the resultant damage caused by the liquid impact on the material. As a result the classical ductile to brittle crossover graph  for the stress test would be altered.
At higher impact angles, the water droplet induces a much larger force perpendicularly onto the sample resulting in sub surface cracks and brittle/fatigue type wear, potentially leading to delamination . These two erosive mechanisms synergise to produce a distinctive pattern of erosion found in Figs. 3, 5, 9 and 12 which is consistent with the literature . However when the sample is stressed it varies from this pattern.
The stressed samples showed different surface features in the optical microscope images as a circular ridge formation; this was not present in the optical microscope images of the unstressed samples (shown in Figs. 15 and 16). It is postulated that this is the result of the stresses on the material. The feature can be observed in the Figs. 17, 18 and 19.
It is assumed these features are associated with plastic deformations of the top surface due to the impact of a water droplets which can be termed SICD (Surface Impact Circular Deformation). In Fig. 17 there is a crack in the top surface to the right of the SICD; this was assumed to result from the stresses which would have occurred during the formation of the SICD. This crack could explain the slightly higher magnitude mass loss in the stress tests compared to the unstressed tests as this crack is liable to increase the erosion rate resulting in the glass fibres being at an increased risk of erosion which would lead to considerably more mass loss.
SICDs were only found on the stressed samples; this was possibly due to the sample being held under conditions closer to its yield stress and the effect of any additional stress on the sample by a liquid impact has more chance to cause plastic deformation of the top surface, resulting in a greater visible distribution to the surface. The shape of the SCIDs can be described from the shockwave that the droplet creates on the sample. This has been previously been investigated and modelled. Figure 20 shows a contour plot of von Mises Stress in the Epoxy plate during a 140 ms−1 impact, using a cross-sectional view to detail the stress dissipation beneath the surface of the plate .
From stages 3 and 4 of this model, a shockwave ring has formed around the droplet replicating the SICDs, therefor this could be the reasoning behind the formation.
A comparison between an unstressed and stressed sample can be shown in the line graph (Fig. 21), illustrating 50 ms−1 using fresh water as the erosive medium.
From this graph (Fig. 21), the stressed sample experiences high erosion over a larger range of impact angles and experiences erosion past threshold at lower angles of impact. Therefore, it can be deemed that the stressed sample is more susceptible to erosion than the unstressed sample. It is postulated that the stressed sample is experiencing a larger crossover between abrasion and brittle/fatigue erosion which allows the top surface of the material to be likely impeded and have fibre exposure. This behavioural change when stressed is theorised to be due to the additional stress from the droplet impact causing plastic deformation (as all stresses are additive). The graph suggests the sample is now behaving in a more “ductile” than “brittle” manner due to the slight shift to the left. This change could be due to the polymer (which is the more ductile portion of the material) governing the behaviour rather than the brittle glass fibres. Experimental evidence to support this would be the presence of SICDs on the surface, with these features indicative of ductile behaviour due to the applied stress.
During the salt water testing of the pre-stressed samples, the same pattern of a wider range of impact angles experiencing more considerable erosion was observed. This suggests similar mechanisms were involved for both exposure media in the above study.