In railway planning, the timetabling step needs, as input, the train running times, which are calculated from a train dynamic model. Usually, this model determines the most energy-efficient train trajectory for a predefined time. However, this time may not correspond to the timetable-makers' needs. They should have the choice among a set of solutions, more or less energy-consuming. This study proposes a method capable of producing a set of alternative running times with the associated mechanical energy required. To this end, the authors' contribution is to set up an efficient evolutionary multi-objective algorithm builds a set of well-spread and diversified solutions which approximate a Pareto front. The solutions are all compromises between running time and energy-consumption, the two minimisation objectives concurrently optimised. Given that an evolutionary algorithm is strongly dependent on the initialisation phase, the efficiency of the algorithm is improved through a specific and original mechanism connecting multiple initialisations in cascade in order to accelerate the convergence towards the best solutions. A set of results obtained on randomly-generated instances is analysed and discussed.