We estimate the expected event rate of gravitational wave signals from
mergers of supermassive black holes that could be resolved by a space-based
interferometer, such as the Evolved Laser Interferometer Space Antenna (eLISA),
utilising cosmological hydrodynamical simulations from the EAGLE suite. These
simulations assume a $\Lambda$CDM cosmogony with state-of-the-art subgrid
models for radiative cooling, star formation, stellar mass loss, and feedback
from stars and accreting black holes. They have been shown to reproduce the
observed galaxy population with unprecedented fidelity. We combine the merger
rates of supermassive black holes in EAGLE with a model to calculate the
gravitational waves signals from the intrinsic parameters of the black holes.
The EAGLE models predict $\sim2$ detections per year by a gravitational wave
detector such as eLISA. We find that these signals are largely dominated by
mergers between $10^5 \textrm{M}_{\odot} h^{-1}$ seed mass black holes merging
at redshifts between $z\sim2.5$ and $z\sim0.5$. In order to investigate the
dependence on the assumed black hole seed mass, we introduce an additional
model with black hole seed mass an order of magnitude smaller than in our
reference model. We find that the merger rate is similar in both models, but
that the scenarios could be distinguished through their detected gravitational
waveforms. Hence, the characteristic gravitational wave signals detected by
eLISA will provide profound insight into the origin of supermassive black holes
and the initial mass distribution of black hole seeds.