One strategy for the survivability of wave energy converters(WECs) is to minimize the extreme forces on the structure by adjusting the system damping. In this paper, a neural network model is developed to predict the peak line force for a 1:30 scaled point-absorber WEC with a linear friction-damping power take-off (PTO). The algorithm trains over the wave tank experimental data and enables an update of the system damping based on the system state (i.e. position, velocity, and acceleration) and information on the incoming waves for the extreme sea states. The results show that the deep neural network (DNN) developed here is relatively fast and able to predict the peak line forces with a correlation of 88% when compared to the true (experimental)data. Then, the optimum damping for survivability purposes is found by minimizing the peak line force. It is shown that the optimum damping varies depending on the system state in each zero up-crossing episode.