Hydraulic turbines can undergo severe damage during operation, because of low quality water or detrimental flow conditions. Damage induces maintenance costs and power production losses, and can also endanger safety of installations. Hydropower plants operators and turbine manufacturers are interested in extending overhaul periods by reducing damage intensity and protecting turbine components with surface treatments, but accurate and reliable prediction of damage is however missing. The present work is related to the erosion arising from repeated impacts of high speed water droplets on specific parts of Pelton turbines. Indeed for high head Pelton units, the jet of water is composed of a liquid core surrounded by droplets. Observations show that regions of impact of these droplets exhibit specific erosion patterns. The aim of the work consists in understanding the corresponding erosion mechanisms through detailed numerical simulation (micro-scale) of the impacts of high speed liquid droplets on turbine components. These results will then be transposed at the machine level (macro-scale) in order to predict the damage along the life cycle of the turbine. When a high velocity water droplet with small diameter impacts a rigid surface, the ``water-hammer'' pressure due to inertial effect appears in the water droplet at the central contact zone, though the maximum pressure occurs on the envelope of the contact area and may be far higher than water-hammer pressure. The impact causes the traveling of a shock wave across the droplet, and lateral jetting occurs by compression when the wave front overtakes the contact area. Concerning the structure, erosion has been found to be due to fatigue cracking. First, material grains are weakened during an ``incubation'' phase. After a large number of impacts, micro-cracks emerge and lead to ejection or fracture of grains, what is called ``amplification'' phase. Numerical simulations are performed subsequently to understand and get a more detailed analysis. The entire simulation is modeled in explicit dynamics with a strong 2-way coupler which is energy conservative at the interface. The solid domain is computed by the finite elements method with Europlexus code, and the fluid one is discretized by Smoothed Particle Hydrodynamics with an Arbitrary Lagrangien-Eulerian description by an in-house software. Thanks to these simulations, the pressure peak on the contact surface can be found. Results are in line with literature and the impulse of the impact allows to locate the most loaded zones of the area. The fatigue-based mechanism is validated by observing the change of sign of hydrostatic stress. Finally, a post-processing erosion program developed with a simple damage criterion provides the location of the most eroded zones of the structure during a loading cycle. The stress range of the transient simulation is computed and gives a number of cycles to failure by a S-N curve for each Gauss point of the solid mesh. A number of cycles is chosen, which allows to make some elements eroded then removed from the mesh for further droplet impact simulations. This post-processing makes some strong assumptions for the number of cycles: every droplet falls with the same velocity, angle, diameter, density at the same place and water is cleaned between each impact. Moreover we assume that geometry does not change during the chosen number of cycles. These first results will be a strong basis for a sensitivity analysis on main impact parameters (droplets diameter and velocity). It is also planned to investigate the influence of a thin water layer set on the solid surface to mimic the wet environment, and a multi-layer material to take into account the coated surface of Pelton buckets.